Ultracapacitor Applications

Power and Energy Series 59
Ultracapacitor Applications
Ultracapacitor
Energy storage and in particular electrical storage of energy
has become a very talked about topic in circles ranging
from lay persons, in regard to hybrid and battery electric
Dr. John M. Miller was vice president of systems and
applications at Maxwell Technologies when this book was
written. He is currently with the U.S. Department of Energy, Oak
Applications
vehicles, to professionals, and certainly by legislators Ridge National Laboratory, National Transportation Research
Center. He is also founder and principal engineer of J-N-J Miller
and energy policy makers in government. But even
Design Services, P.L.C. Dr. Miller worked for 20 years in the
to professionals the distinctions between physical and automotive industry, leading several hybrid vehicle technology
chemical forms of electric energy storage are unclear programs including 42V Integrated Starter Alternator, ISG, for
and at times poorly understood, if at all. This book takes application into a SUV. He has been active in collaborations at
a critical look at physical storage of electricity in the the industry and government levels including the NSF funded
devices known collectively as electrochemical capacitors systems center for Future Renewable Electric Energy Delivery
and particularly as ultracapacitors. In its 12 chapters, this and Management, FREEDM. He was actively engaged in MIT’s
text covers ultracapacitor and advanced battery topics Consortium on Advanced Automotive Electrical and Electronic
Components and Systems and has served as Adjunct Professor
John M. Miller
with emphasis on clear understanding of fundamental
of Electrical Engineering at Michigan State University and at
principles, models and applications. The reader will Texas A&M University. Dr. Miller has authored over 170 technical
appreciate the case studies ranging from commercial publications, holds 53 U.S. patents, and authored or co-authored
to industrial to automotive applications of not only five books. He is a Fellow of the IEEE, Member of SAE, 2009
ultracapacitors, but of these power dense components in recipient of the IEEE Kliman Innovator award, and 2010 recipient
combination with energy dense battery technologies. of the IEEE Power Electronics Society distinguished service award.
Miller
The Institution of Engineering and Technology

www.theiet.org
978-1-84919-071-8
PBPO0590.indd 1 27/05/2011 16:35:40

IET POWER AND ENERGY SERIES 59
Ultracapacitor
Applications
PRELIMS 7 June 2011; 18:2:38

Other volumes in this series:
Volume 1 Power circuit breaker theory and design C.H. Flurscheim (Editor)
Volume 4 Industrial microwave heating A.C. Metaxas and R.J. Meredith
Volume 7 Insulators for high voltages J.S.T. Looms
Volume 8 Variable frequency AC motor drive systems D. Finney
Volume 10 SF6 switchgear H.M. Ryan and G.R. Jones
Volume 11 Conduction and induction heating E.J. Davies
Volume 13 Statistical techniques for high voltage engineering W. Hauschild and
W. Mosch
Volume 14 Uninterruptible power supplies J. Platts and J.D. St Aubyn (Editors)
Volume 15 Digital protection for power systems A.T. Johns and S.K. Salman
Volume 16 Electricity economics and planning T.W. Berrie
Volume 18 Vacuum switchgear A. Greenwood
Volume 19 Electrical safety: a guide to causes and prevention of hazards J. Maxwell
Adams
Volume 21 Electricity distribution network design, 2nd edition E. Lakervi and
E.J. Holmes
Volume 22 Artificial intelligence techniques in power systems K. Warwick, A.O. Ekwue
and R. Aggarwal (Editors)
Volume 24 Power system commissioning and maintenance practice K. Harker
Volume 25 Engineers’ handbook of industrial microwave heating R.J. Meredith
Volume 26 Small electric motors H. Moczala et al.
Volume 27 AC–DC power system analysis J. Arrillaga and B.C. Smith
Volume 29 High voltage direct current transmission, 2nd edition J. Arrillaga
Volume 30 Flexible AC Transmission Systems (FACTS) Y-H. Song (Editor)
Volume 31 Embedded generation N. Jenkins et al.
Volume 32 High voltage engineering and testing, 2nd edition H.M. Ryan (Editor)
Volume 33 Overvoltage protection of low-voltage systems, revised edition P. Hasse
Volume 34 The lightning flash V. Cooray
Volume 36 Voltage quality in electrical power systems J. Schlabbach et al.
Volume 37 Electrical steels for rotating machines P. Beckley
Volume 38 The electric car: development and future of battery, hybrid and fuel-cell
cars M. Westbrook
Volume 39 Power systems electromagnetic transients simulation J. Arrillaga and
N. Watson
Volume 40 Advances in high voltage engineering M. Haddad and D. Warne
Volume 41 Electrical operation of electrostatic precipitators K. Parker
Volume 43 Thermal power plant simulation and control D. Flynn
Volume 44 Economic evaluation of projects in the electricity supply industry H. Khatib
Volume 45 Propulsion systems for hybrid vehicles J. Miller
Volume 46 Distribution switchgear S. Stewart
Volume 47 Protection of electricity distribution networks, 2nd edition J. Gers and
E. Holmes
Volume 48 Wood pole overhead lines B. Wareing
Volume 49 Electric fuses, 3rd edition A. Wright and G. Newbery
Volume 50 Wind power integration: connection and system operational aspects
B. Fox et al.
Volume 51 Short circuit currents J. Schlabbach
Volume 52 Nuclear power J. Wood
Volume 53 Condition assessment of high voltage insulation in power system
equipment R.E. James and Q. Su
Volume 55 Local energy: distributed generation of heat and power J. Wood
Volume 56 Condition monitoring of rotating electrical machines P. Tavner, L. Ran,
J. Penman and H. Sedding
Volume 57 The control techniques drives and controls handbook, 2nd edition B. Drury
Volume 58 Lightning protection V. Cooray (Editor)
PRELIMS 7 June 2011; 18:2:38

Ultracapacitor
Applications
John M. Miller
PRELIMS 7 June 2011; 18:2:38

Published by The Institution of Engineering and Technology, London, United Kingdom
The Institution of Engineering and Technology is registered as a Charity in England & Wales
(no. 211014) and Scotland (no. SC038698).
† 2011 The Institution of Engineering and Technology
First published 2011
This publication is copyright under the Berne Convention and the Universal Copyright
Convention. All rights reserved. Apart from any fair dealing for the purposes of research
or private study, or criticism or review, as permitted under the Copyright, Designs and
Patents Act 1988, this publication may be reproduced, stored or transmitted, in any
form or by any means, only with the prior permission in writing of the publishers, or in
the case of reprographic reproduction in accordance with the terms of licences issued
by the Copyright Licensing Agency. Enquiries concerning reproduction outside those
terms should be sent to the publisher at the undermentioned address:
Michael Faraday House
Six Hills Way, Stevenage
Herts, SG1 2AY, United Kingdom
www.theiet.org
While the author and publisher believe that the information and guidance given in
this work are correct, all parties must rely upon their own skill and judgement when
making use of them. Neither the author nor publisher assumes any liability to
anyone for any loss or damage caused by any error or omission in the work, whether
such an error or omission is the result of negligence or any other cause. Any and all
such liability is disclaimed.
The moral rights of the author to be identified as author of this work have been asserted
by him in accordance with the Copyright, Designs and Patents Act 1988.
British Library Cataloguing in Publication Data

A catalogue record for this product is available from the British Library
ISBN 978-1-84919-071-8 (paperback)

ISBN 978-1-84919-072-5 (PDF)
Typeset in India by MPS Ltd, a Macmillan Company

Printed in the UK by CPI Antony Rowe, Chippenham, Wiltshire
PRELIMS 7 June 2011; 18:2:39

To JoAnn and Nathan
PRELIMS 7 June 2011; 18:2:39

PRELIMS 7 June 2011; 18:2:39
Contents
Preface xi
1 Types of ultracapacitors 1
1.1 Electrochemical capacitors 10
1.2 Symmetric types 17
1.3 Asymmetric types 24
1.4 Hybrid capacitors 30
Exercises 33
References 35
2 Ultracapacitor modeling 37
2.1 Electronic equivalent circuit model 42
2.2 Cell characterization methods and standards 58
2.2.1 EUCAR method 61
2.3 Simulation model validation 66
2.4 Capacitor–battery combinations 73
2.4.1 Passive parallel architecture 78
2.4.2 Active parallel architecture 81
Exercises 89
References 91
3 Power and energy 93

3.1 Specific energy and energy density 96
3.2 Specific power and power density 101
3.3 Ragone relationships 112
3.4 Ultracapacitors and batteries 114
Exercises 123
References 126
4 Commercial applications 129

4.1 Uninterruptible power supplies 129
4.2 Utility voltage stabilizer 135
4.3 Wind turbine systems 146
4.4 Photovoltaic systems 155
Exercises 158
References 161
PRELIMS 7 June 2011; 18:2:39

viii Ultracapacitor applications
5 Industrial application 163

5.1 Material handling trucks 165
5.2 Harbor cranes and rubber tired gantry cranes 168
5.3 Earth moving equipment 176
Exercises 185
References 187
6 Heavy transportation application 189

6.1 Battery electric bus 199
6.2 Hybrid electric bus 201
6.3 Shuttle bus 206
Exercises 208
References 211
7 Hybrid electric vehicles 213

7.1 Types of hybrids 219
7.2 Hybrid functions 223
7.3 Power assist hybrid 225
7.4 Plug-in hybrid 227
Exercises 232
References 234
8 Single-mode power split 235

8.1 Electronic continuously variable transmission 239
8.2 Ultracapacitor application in eCVTs 242
8.3 Drive cycle evaluations 246
Exercises 253
References 255
9 2-mode power split 257

9.1 Essentials of 2-mode eCVT 260
9.2 Operational modes of the EVT 263
9.3 Ultracapacitor application in 2-mode eCVT 269
9.4 Plug-in hybrid electric vehicle: Volt 272
Exercises 275
References 277
10 Life cycle testing 279

10.1 Leakage current influence 283
10.2 Reliability versus service life 286
10.2.1 Exponential reliability function 291
10.2.2 Normal reliability function 292
10.2.3 Log-normal reliability function 293
10.2.4 Weibull reliability function 294
PRELIMS 7 June 2011; 18:2:39

Contents ix
10.3 Mean service life 298

10.4 Comprehensive cycle life testing 302
Exercises 306
References 308
11 Abuse tolerance 311

11.1 Need for abuse testing 311
11.2 Overvoltage and overcurrent abuse 313
11.3 Insulation resistance and high potential 317
11.4 Vibration requirements 324
11.5 Ultracapacitor cycling exposure 325
Exercises 328
References 330
12 Future transportation systems 331

12.1 Future mobility systems 332
12.2 Wireless power transfer 333
12.2.1 Coaxial winding transformer 336
12.2.2 Inductive power transfer 337
12.3 Ultracapacitor applications in IPT 338
Exercises 343
References 345
Glossary 347
Index 351
PRELIMS 7 June 2011; 18:2:39

PRELIMS 7 June 2011; 18:2:39
Preface
This book addresses the need for a text on an energy storage system that treats the
fundamentals and applications of carbon-based electric double-layer capacitors.
Named the ultracapacitor, these double-layer devices are capable of kilofarad levels
of capacitance in single cells and possess sufficient energy to challenge batteries.
The book, however, is about more than the electrochemistry of ultracapacitors; it is
aimed primarily at practicing engineers, specifically applications engineers who are
responsible for the design-in, sizing, and customer follow-up of energy storage
installations. Energy storage in an electric double-layer capacitor, which defines
the interface between a conducting solid and an electrolyte, has been known for
over a century, but it was not until 1957 that H.I. Becker of GE patented a rudi-
mentary carbon device based on this phenomena. The ultracapacitor we know
today owes its design to Robert A. Rightmire, a chemist at the Standard Oil
Company of Ohio in 1962 who refined the carbon–carbon electrochemical capa-
citor. The rest, as they say, is history.
Applications for ultracapacitors range from small, thumbnail sized cells of
5–10 F used to hold up power in solid-state drives during flash-2-cache backup
through small electrolytic capacitor sized units of 25–150 F for various electronic
applications. Larger cell sizes such as D Cell ultracapacitors having greater than
300 F capacity are applied in industrial applications including wind turbine blade
pitch adjuster backup power modules and in some motive applications, for exam-
ple, electric bike and electric scooter. Still larger cell sizes such as the now standard
60-mm-diameter cylindrical cans range in capacity from 650 to more than 3000 F
and are applied in uninterruptible power supplies, bridge power, automotive
PowerNet stabilizing, and to facilitate automotive engine restart in idle–stop sys-
tems. Modules fabricated from large cell sizes are applied as heavy truck engine
cold cranking aides, as van, truck, shuttle bus and transit bus hybridization com-
ponents and in metro rail as energy recuperators. Application in rolling stock is
now expanding to include onboard energy recuperators and as trackside third rail or
catenary stabilizers. New applications are emerging every week with some pro-
mising applications being in the area of smart grid and rail to implement more
robust ancillary services for the utilities and to offset the need for additional sub-
stations along heavily traveled metro and light rail tracks.
Smart grid applications, for example, include the need for high rate charge and
discharge energy reservoirs to augment existing measures to regulate distribution
system frequency. The fast charge and discharge at high efficiency attribute of
carbon ultracapacitors means they can also be used to support utility voltage
PRELIMS 7 June 2011; 18:2:39

xii Ultracapacitor applications
regulation. These two applications require substantial real power in the megawatt
range for frequency support and even higher VAR ratings for voltage regulation.
Ultracapacitors can meet the challenge of utility energy storage applications over
10 s to 1 min duration and beyond that the need can be met with ultracapacitor and
battery combinations out to 8-h duration. These topics, along with their similar
application as high power caches in automotive applications, are afforded sub-
stantially coverage in this book.
The outline of this book was drawn from the courses taught by the author at
Maxwell Technologies from 2007 through 2010. For this I credit Mr David
Schramm, CEO, Maxwell Technologies, for initiating the program by assigning me
the task as Dean, Maxwell University. Early installments of the course were
developed in classroom style with lecture and examples. However, because of the
global nature of a company such as Maxwell Technologies, it quickly became
impractical to travel to the various sites and present materials in person, so a 2nd
edition of Maxwell University was initiated. For this round, the author elected to
revise the material and provide it on CDs in voice-over format. Some of those
materials can be found in example illustrations in many of the chapters in this book.
No undertaking of this sort can be executed without the encouragement and
assistance of many others. In particular, I wish to thank Dr Porter Mitchell, senior
scientist with Maxwell Technologies for reviewing Chapters 1–6 and for his helpful
critique of the electrochemistry topics. Dr Mitchell’s commentary on the remainder
of the book is gratefully appreciated. I also wish to acknowledge Mr David Wright,
Director of Applications Engineering with Maxwell Technologies for his review
and comments on the second half of the book dealing with some practical appli-
cations. But my special thanks goes to my wife, JoAnn, for bearing with me on so
many weekends and holidays when we could have been relaxing or doing nonwork
type activities like traveling.
I wish to thank everyone at the Institute for Engineering Technology involved
in the production of this book and especially to Lisa Reading, commissioning editor
with IET, and to the production staff.
John M. Miller
PRELIMS 7 June 2011; 18:2:39

Chapter 1
Types of ultracapacitors
Ultracapacitors, or to be technically correct, electronic double layer capacitors,

form a subset of the general category of electrochemical energy storage devices, in
particular, that of electrochemical capacitors (ECs) [1]. This chapter introduces
electrochemical energy storage devices including electrochemical couples or bat-
teries. To alleviate any concerns over nomenclature, the term ultracapacitor is a
more colloquial term for the symmetric, carbon–carbon, EC. Supercapacitor was
the original name given to this class of extremely high specific capacitance device,
but since NEC Tokin trademarked that name in 1975, during the early days of
EDLC development [2], it has been replaced with ultracapacitor in general usage.
Even today the term supercapacitor continues to be used for the class of asym-
metric, or carbon-metal oxide, ECs.
Invented in the U.S. at SOHIO (Standard Oil Company of Ohio) by Robert A.
Rightmire [3], the supercapacitor was viewed as an electrical energy storage
apparatus. This invention relates generally to the utilization of an electrostatic field
across the interphase boundary between an electron conductor and an ion conductor
to promote the storage of energy by ionic adsorption at the interphase boundary.
Electrochemical storage batteries and capacitors have been in existence for over
200 years, such as the Baghdad battery, the Volta ‘pile’ ca. 1800 to Ben Franklin
who coined the term ‘battery’ ca. 1848 [4]. The battery, as will be shown shortly,
stores energy in chemical bonds that follow reduction–oxidation (redox) reactions
in which mass transfer is involved. Capacitors store energy electrostatically in
electric fields, and for the class of electrochemical capacitors, between ions in
solution and a material. No mass transfer is involved; hence, there is no electro-
chemical wear out. Figure 1.1 illustrates the class of electrochemical energy storage
devices that are to be discussed in this chapter.
The field of electrochemical energy storage has been advancing rapidly over
the past two decades ever since the introduction of insertion chemistry batteries in
the early 1990s. This is a book on ultracapacitor technology so why delve into
batteries? We start this way because of the enormous amount of work done on the
electrochemical, Faradaic, energy storage cell that has directly benefited the class
of non-Faradaic storage devices, namely electrochemical capacitors. Let’s take a
quick look at lithium-ion battery developments [5]:
ch001 7 June 2011; 13:4:39

2 Ultracapacitor applications
Electrochemical energy storage
Battery Capacitor
Primary Secondary Electrostatic Electrolytic Electrochemical

(rechargeable)
Symmetric Asymmetric
Lead NiCd NiMH Li ion
acid
Aqueous Organic Aqueous Organic
electrolyte electrolyte electrolyte electrolyte
Most popular today Active research

Potential for
bulk storage
Figure 1.1 Taxonomy of electrochemical energy storage
● Insertion chemistry and intercalating cathode of LiCoO2 patented by John

Goodenough in 1980.
● Intercalating graphite anode patented by H. Ikeda at Sony in 1981.
● Li-ion battery principle conceived and developed by Asahi Kasei, Saga Univ.,
Japan.
● Li-ion battery commercialized by Sony 1991.
● Followed by A&T Batt Co. (JV of Toshiba Batt þ Asahi Kasei) in 1992.
● Early Li-ion cells had series safety issues:
* Lithium metal anodes form dendrites and powder deposits on recharge.
* Electrolyte reactions with lithium powder on anode.
* Basic electrolytes and separators did not provide any safety protection.
● Development in the 1990s into the early 2000s focused on safer electrolytes,
polymer separators, shut-down separators, electrolyte additives, and cell
overcurrent protection (PTC devices in lid).
Electrochemical couples, or cells that generate an electromotive force (emf ) during
the process of chemical reactions involving valence band electron transfers have
attendant mass transfer during discharge (primary cells) or charge and discharge
(secondary cells). It is this mass transfer that utilizes a good mass fraction of the
bulk of an electrode and gives electrochemical couples their high specific energy.
Electrochemical capacitors, on the other hand, rely on surface phenomena of
ion adsorption without mass transfer, hence are non-Faradaic processes that involve
conduction band orbital electron transfer of a salt in a solvent. Weak ionic bonds of
the salt constituent atoms or molecules break during solvation by the solvent with
an attendant conduction band electron transfer. It is this relative ease of ion
adsorption and desorption at a conductive surface at the solid–liquid interface that
gives electrochemical capacitors their high specific power.
For their part, batteries are constant voltage, high energy, electric energy sto-
rage devices, and electrochemical capacitors are basically voltage storage, high
ch001 7 June 2011; 13:4:39

Types of ultracapacitors 3
power, electric energy storage devices. The two types of electric energy storage
devices differ in that batteries store their energy in chemical bonds and capacitors
store energy in electric fields. Let’s take a look at the relative positioning of batteries
of different chemistries and ultracapacitors in what is commonly known as a Ragone
chart. The Ragone chart is a means to compare electric energy storage devices in
terms of specific energy and specific power, meaning energy and power normalized
to the device mass. In another format shown in Figure 1.2 the comparison can be of
specific energy (Wh/kg) versus energy density (Wh/L). This particular format is
valuable to show that evolution in lithium-ion technologies is evolving to lighter,
high specific energy (SE) and smaller, high energy density (ED) devices.
Primary lithium
3.6 V, 2.4 Ah
885 Wh/L
450 Wh/kg
n Panasonic
lutio 2008
300 e evo 6.6 Ah
anc 740 Wh/L
orm
Specific Energy (SE)
f
Per Lithium-air
200 theoretical
Li-ion 5,200 Wh/kg
Wh/kg
AA 11,140 Wh/L
Ni-Zn alkaline
primary Gasoline
100 Lead Ni-MH
@12,400 Wh/kg
acid
50 Ni-Cd
0
0 100 200 300 400 500 600 700 800
Energy Density (ED)
Wh/L
Figure 1.2 Ragone chart of electric energy storage devices: SE (Wh/kg), ED (Wh/L)
Electrochemical capacitors would be located in the area of approximately 10

Wh/kg SE and 15 Wh/L as shown in Figure 1.2, or between the origin and the lead-
acid box. Compared to the lead-acid battery, virtually all other battery chemistries
are lighter and smaller. The nickel-based chemistries such as nickel-cadmium
(NiCd), nickel-zinc (NiZn), nickel-metal-hydride (NiMH), and other combinations
have electrochemical couple potentials in the range 1.2–1.6 V. The low cell voltage
contributes to relatively low energy of the nickel-based cells. The introduction of
the lithium-ion cell broke this paradigm and pushed both SE and ED to high levels.
Even so, relative to the common alkaline AA cell the lithium-ion rechargeable cell
is about on par in both SE and ED. More recent lithium-ion chemistries are pushing
the envelope on SE and ED as seen in Figure 1.1. Compared to these commercial
and emerging cells the next advancement, that of lithium-air, offers the next
revolutionary step in SE and ED as noted with gasoline for comparison. However,
advanced batteries generally have SE that is only 20–40% or so; theoretically, a
practical Lithium-air cell may offer energy of approximately 1000 Wh/L. The
lithium-air chemistry has high marks in these areas because one electrode is air and
therefore not part of the cell’s mass.
ch001 7 June 2011; 13:4:39

Today, the trend to high power lithium-ion cells (and modules) is resulting in
these electrochemical electric energy storage devices trading away energy in the
quest for higher power as shown in Figure 1.3. As higher and higher power levels
are sought, the electrochemical cell of a given volume is filled progressively more
with current collector metals (copper anode and aluminum cathode foils in the case
of lithium ion) and less active materials to get higher surface activity for power.
The result is that as the cell specific power increases toward 20 kW/kg, its specific
energy decreases from >170 Wh/kg down to <50 Wh/kg. If this trend continues,
the battery will in effect become a supercapacitor.
Lithium-ion power versus energy including hybrid-cap Very high power

100 High power
High energy
Hybrid cap P/E = 100 P/E = 1
10 + P/E = 9
Specific power (kW/kg)
+ +
+ ++ +
+ + P/E = 9
+ P/E = 100
1
P/E = 1 True
energy Li-ion
0.1
0.01
0 25 50 75 100 125 150 175 200 225
Specific energy (Wh/kg)
Figure 1.3 Ragone relation for lithium-ion cells
The left–right diagonal traces in Figure 1.3 are contours of constant SP/SE, or
reciprocal time lines. For example, at P/E ¼ 100 the cell can discharge its energy in
0.01 time that of a P/E ¼ 1 cell. This is why power assist hybrid electric vehicles (HEV)
will use battery packs with P/E*15 while a plug-in hybrid electric vehicle (PHEV)
will have P/E*9 and a battery electric vehicle (BEV) will require P/E < 3 for high
energy. A true energy cell, one with thick electrodes such as laptop packs, will have
SE > 200 Wh/kg. A cell designed for high burst power will consequently have very thin
electrodes and SP > 1 kW/kg. The commercial ultracapacitor has SP > 1.5–2.5 kW/kg
with approximately 6 kW/kg at 95% efficiency (discussed in detail in Chapter 3).
With this background it is time to move ahead to the main topic, that is
ultracapacitor types. To begin, a short tutorial on capacitors is in order. Figure 1.4
illustrates the classical parallel plate electrostatic capacitor used in consumer
electronics, industrial systems and telecommunications, and many more applica-
tions such as paper, ceramic, polymer, and many other dielectric materials.
The most important point to remember about capacitors is the state equation
for charge, which for the passive, linear case is directly proportional to potential,
U(1.1). Capacitance itself is a geometric property of construction and materials as
noted in (1.2).
ch001 7 June 2011; 13:4:39

Separation, d
Energy
Voltage
+Q
+ V = (i0/C) t
Area, A
Time Voltage
(a) (b) (c)
Figure 1.4 (a) Parallel plate capacitor; (b) voltage versus time of capacitor;
(c) energy versus voltage of capacitor
i0
Q ¼ CU ¼ i0 t or U ¼ t ð1:1Þ
C
er e0 A
C¼ ð1:2Þ
d
According to (1.1), the voltage ramps linearly with time, as noted in
Figure 1.4b, when charge (or discharge) current has constant value. Capacitance
(F/V, from (1.1)) can be derived knowing the surface current density (C/m2), rs, on
a conducting plate of area, A, as Q ¼ rsA. From Maxwell’s laws of electro-
magnetics, we know that the divergence of electric flux, D, is the source of the
electric field, E, r(!D ¼ r). For any dielectric medium, the constitutive relation
plus dielectric polarization define the total electric flux, D ¼ e0E þ P, where
polarization P ¼ e0cE. Dielectric material susceptibility, c, is a measure of the
material contribution to total permittivity, e, and e0 is the permittivity of free space,
which is always present. What this means is that for a given potential, U, across the
parallel plate capacitor having charge separation distance, d, an electric field,
E ¼ U/d, will be set up that in turn elicits a total electric flux, D. Starting with (1.3)
and making the substitutions noted earlier, the capacitance can be calculated based
on total electric flux and the surface charge it corresponds to.
eU A D Q eA ðe0 þ e0 cÞA er e0 A
Q ¼ rs A ¼ DA ¼ ; !C ¼ ¼ ¼ ¼ ð1:3Þ
d U d d d
The relative permittivity, er ¼ (1 þ c)e0 ¼ ke0, accounts for the presence of
materials other than vacuum (air) within the charge separation space, and here ‘k’
or K-factor represents the polarization of the dielectric. Materials such as paper,
ceramics like barium titanate, and polymers contribute to K values much greater
than air. Some modified barium titanate ceramics can achieve k > 15,000.
Table 1.1 lists some common materials, their K-factor and breakdown voltage, Ubd.
The energy stored in an electrostatic capacitor, shown in Figure 1.4c, can be
derived by recalling the state equation for electrostatic charge, Q, from (1.1) and
noting that for any state equation relationship such as this the energy can be found
by integration of the state equation versus its extensive variable, in this case
ch001 7 June 2011; 13:4:39

Table 1.1 Properties of dielectrics
Material K-factor Ubd (MV/m)

Air *1 3
Teflon 2.1 60
Polyethylene 2.25 24
Paper 3.5 16
Pyrex Glass 4.7 14
Water (@20C) 80
Strontium titanate 310 8
Modified barium titanate *15,000 >3
Q q = f(u)
Wf
i
Q0 C < C0
dq
Wc
i
C0
0
Umx U
du
Figure 1.5 Illustration of the general case for capacitive charge versus potential
potential, U. For the case of a linear capacitor, its Q versus U function is simply the
linear relation for q(u) shown in Figure 1.5. For this more general case, the field
energy, Wf, is the area in the q versus u chart in the linear region, and is the portion
of input energy that is stored in the electric field. The co-energy, Wc, is a non-
physical energy measure and is not used.
q ¼ Cu; dq ¼ C du ð1:4Þ
ð
W ¼ W f ðqÞ ¼ u dq ð1:5Þ
Substituting (1.4) into (1.5) puts the field energy into the proper form where
the limits of integration are [0,Umx].
ð Q0 ð U mx
C 2 U mx 1
W f ðqÞ ¼ udq ¼ xðCdxÞ ¼ x ¼ CU 2mx ð1:6Þ
0 0 2 0 2
ch001 7 June 2011; 13:4:39

Equation (1.6) is the familiar expression for capacitor energy as shown in

Figure 1.4 c. For the linear case the calculation is Wf ¼ Wc, but in general this is not
the case. Consider the example to follow for the nonlinear case where q ¼ f(u)
exhibits a capacitance that decreases as potential is increased. This is typical of
ceramic types and represents dielectric saturation.
Example 1.1: Suppose the functional relation in Figure 1.5 is q ¼ f(u) ¼
Q0 sin(pu/2Umx) and we wish to find the field energy for this case. Note that this
relationship for charge is relatively close to the curve in Figure 1.5 when u/
Umx ¼ p/2 is at the maximum point of charge, Q0.

pu
q ¼ Q0 sin ð1:7Þ
2U mx

pQ0 pu
dq ¼ cos du ð1:8Þ
2U mx 2U mx
ð Q0 ð U mx
p
W f ðqÞ ¼ u dq ¼ xaQ0 cosðaxÞdx; where a ¼ : ð1:9Þ
0 0 2U mx
The integral in (1.9) is easily solved by referring to tables of integrals for

which the solution is found as
ðu u
1 ux
x cosðaxÞdx ¼ 2
cosðaxÞ þ sinðaxÞ ð1:10Þ
0 a a 0
Making the substitutions into (1.10), solving and substituting the result into
(1.9) reduces this expression for field energy, Wf (q), to

1 U mx
W f ðqÞ ¼ aQ0 2 ðcosðaU mx Þ 1Þ þ sinðaU mx Þ ð1:11Þ
a a
Note that the cos() term in (1.11) is zero and sin() term is unity. Therefore, the
field energy for a nonlinear capacitor exhibiting dielectric saturation becomes

2U mx Q0 2
W f ¼ U mx Q0 ¼ 1 U mx Q0 ¼ 0:363 U mx Q0 ð1:12Þ
p p
Equation (1.12) shows that a capacitor having a saturating dielectric stores less
energy for the same potential than a linear capacitor by the factor 0.363/0.5 ¼ 0.726
in this case. That is, if the dielectric were linear, it would store approximately 27%
more energy.
Example 1.2: The ultracapacitor has a more unusual relationship in its q(u) beha-
vior in that capacitance increases nonlinearly with potential. Consider the following
graphic showing that as time evolves under constant current charge, exhibiting
ch001 7 June 2011; 13:4:40

linear behavior until the charge potential reaches a threshold value that initiates
surface redox reactions between the electrolyte ions and the carbon surface (Tafel
relation). Surface redox produces a pseudo-capacitance effect (see Conway [1]).
Charging current in the nonlinear ultracapacitor that increases linearly then
becomes very nonlinear due to the onset of pseudo-capacitance has the form of
(1.13) where the exponential is the Tafel relation. Solution of (1.14) leads to q(u).
du 1 I 0 ku
I ¼C þ I 0 eku ; for which u_ ¼ þ e ð1:13Þ
dt C C
I þ I 0 kI 0
u_ ¼ þ u; where eku ð1 þ ku þ Þ ð1:14Þ
C C
Figure 1.6 shows that during discharge (I < 0) the capacitance of a production
3000 F cell starts initially at approximately 2770 F and increases with potential to
approximately 3300 F. This nonlinear behavior of capacitance means that for an
incremental potential increase the resultant charge accumulated is higher than it
would be in a linear dielectric capacitor. Charge recall is I dt ¼ C dU and for a
fixed increment in dU the dt increment is larger, taking longer at fixed current so
more charge is delivered. Note that for production cells the capacitance measure-
ment is standardized by international standards setting organizations to the value
measured during discharge.
BCAP3000
3800
3600
C 3400
Capacitance
3200
3000
B 2800
Potential
Onset of nonideal 2600

polarization, i.e., 2400
Faradaic reaction. 2200
1/C 2000
0
50
0
40
0
3
30
0
A 2.5
20
0
2
10
–1 0
1.5
00
–3 0
1
0
Time (at constant current i) Current

–4 0
0.5
–2
Voltage
0
00
0
00
I = C(dV/dt) + If
–5
(a) (b)
Figure 1.6 Potential versus charge current for nonlinear capacitor (ultracapacitor)
and actual C(U) for production cell [used with permission, Maxwell
Technologies]. (a) Potential versus current for nonlinear capacitor;
(b) C(u,i) of the production 3000 F ultracapacitor
Experiments performed for this cell during discharge have shown that its
capacitance can be very accurately modeled as:

u
CðuÞ ¼ C a þ C b tanh Ux ð1:15Þ
Ux
ch001 7 June 2011; 13:4:40

where Ca ¼ 2770, Cb ¼ 520, Umx ¼ 2.7, Ux ¼ 0.9 and u 2 ½0; U mx . The fit of (1.13)
to measured data for the 3000 F cell is shown in Figure 1.7. Note the exceptional fit
of (1.15) to experimental data (the points).
C(Uc) data and approximation f(Uc)
3500
3000
Capacitance
2500
2000
1500
1000
500
0
0 0.5 1 1.5 2 2.5 3
Voltage
Figure 1.7 Data and fit of nonlinear capacitance C(u)
The first step toward solving this example is to multiply (1.15) by the voltage
‘u’ then differentiate with respect to ‘u’ to obtain the quantity dq to get the
equivalent of (1.9). This is done in the following derivations:

u
qðuÞ ¼ CðuÞu ¼ C a u þ C b u tanh Ux ð1:16Þ
Ux
Taking the derivative of (1.16) leads to the differential quantity needed in the
calculation of field energy, Wf, of this highly nonlinear variation of charge with
potential. The resulting expression for differential charge, dq, becomes too com-
plex for closed-form integration.

u u
dq ¼ C a u þ C b tanh U x þ C b u sech 2
Ux ð1:17Þ
Ux Ux
ð U mx ð U mx
x x
Wf ¼ u dq ¼ x C a x þ C b tanh U x þ C b uðsech2 Ux dx
0 0 Ux Ux
ð1:18Þ
The solution to (1.18) is analytically difficult to obtain in closed form so a

numerical solver, such as Maple,1 is applied. The result for the numerical integra-
tion over the specified limits is given in Table 1.2.
Note that if this same ultracapacitor has a nominal capacitance, C0 ¼ Ca þ (1/2)
Cb ¼ 3030 F, it yields a field energy, Wf ¼ 1.104 104 J, which is approximately
12% lower than the production cell characteristic C(u) functional fit. Exercise 1.1
approximates this nonlinear case as a linear approximation to the tanh(x) variation
to show a very close agreement with the field energy calculated above.
1
Available in MathCAD software from MathSoft ver. 14 used here.
ch001 7 June 2011; 13:4:40

Table 1.2 Solution to the nonlinear charge versus potential of the

ultracapacitor in Example 1.2
Component of (1.18) to integrate Result of numerical integration

ÐU
W fa ¼ C a 0 mx xdx Wfa ¼ 1.0097 104
ÐU
W fb ¼ C b 0 mx x tanh Uxx U x dx Wfb ¼ 1.385 103

ÐU
2
Wfc ¼ 1.137 103
W fc ¼ C b 0 mx x2 sech Uxx U x dx
Wf ¼ Wfa þ Wfb þ Wfc Wf ¼ 1.262 104
1.1 Electrochemical capacitors

Electrochemical storage, whether redox or adsorption based, is subject to the
behavior of ions in an electrolyte. Ionic currents ebb and flow through a highly
porous separator into porous electrodes under the influence of electric fields, con-
centration gradients and, at times, thermal gradients. In the electrochemical capa-
citor (EC) the response is near instantaneous with current rate of change limited
only by the cell or module parasitic inductance, and at extremely short times by the
ion currents. However, the EC cell’s frequency response is strongly determined by
the ion kinetics of the electrolyte and ionic channels in the electrode film structure.
Thick electrode ultracapacitors composed of many layers of highly porous acti-
vated carbon particles, carbon black conductive agents, and polymeric binders have
more restrictive ionic channels than very thin electrodes of comparable density that
are composed of large surface area carbon particles possessing high surface rug-
osity. Figure 1.8 illustrates how the frequency response of an ultracapacitor is
influenced by surface rugosity and pore size characteristics. Adsorption is a surface
phenomena which means that by having its surface charge readily accessible the
a b
c
–Imaginary Z
d
b
c e
d
a
e
Real Z
Figure 1.8 Influence of electrode pore shape on ultracapacitor frequency

response [Courtesy JME]
Note: In the Nyquist plot, low frequency is to the upper reaches of Z00 and Z0 with increasing frequency
causing vector Z to move toward the origin
ch001 7 June 2011; 13:4:41

charge can be built up and discharged very fast. The Nyquist plot of imaginary, Z00 ,
versus real, Z0 , parts of total impedance, Z, shows that the transition of wide, open
surface pores of pure capacitive reactance into Warburg impedance characteristic
of diffusion limited response with increasing frequency. The more restrictive a pore
becomes, the more pronounced will ion kinetic limitations be on the dynamics of
the porous electrode to the point that electrode film Randles equivalent dominates
(semi-circular formations going into the region near the origin).
It is important to understand the behavior of constant phase elements [6]
because of their essential role in the dynamics of porous electrode structures. The
pore structure noted above and shown as traces ‘e’ and ‘d’ in Figure 1.8 has an
impedance function, Z, that is accurately modeled by a Randles equivalent with
Warburg impedance in the charge transfer resistance branch, Rct. Figure 1.9 is the
model for the modified Randles equivalent. Similar Nyquist response is obtained if
the Warburg impedance shown in Figure 1.9 is modeled in the double layer capa-
citance branch, Cdl. The similarities will be discussed in examples to follow.
Cdl
RS
Rct
W
Figure 1.9 Warburg impedance modified Randles equivalent of the porous electrode
ultracapacitor ( Rs * ESRdc of the cell; Cdl ¼ double layer capacitance;
Rct ¼ charge transfer resistance; W ¼ Warburg impedance)
To gain more insight into the behavior of electrochemical constant phase ele-
ments consider the general case of Zcpe, a constant phase element having angular
frequency raised to an arbitrary power b, where 0 b 1. In (1.19) the numerator
pffiffiffiffiffiffi
ffi
is an electrochemical conductance derived term having units of W= sec, and ‘s’
is complex angular frequency.
a
Z cpe ¼ ð1:19Þ
sb
For the specific case of b ¼ 0.5, Zcpe ¼ Zw, a Warburg impedance. For this
case, (1.19) can be rewritten as
að1 jÞ
Zw ¼ pffiffiffiffi ð1:20aÞ
w
In (1.20) the term a is given by the following:

RT 2 103
a ¼ pffiffiffi pffiffiffiffiffiffi ð1:20bÞ
2n2 F 2 A C 0 Dc
ch001 7 June 2011; 13:4:41

where R ¼ 8.314 J/(K mol) the gas law constant; F ¼ 96,485 C/mol, Faraday’s
constant; n ¼ 1, the number of electron exchanges involved; T ¼ 298 K, absolute
temperature; A (cm2) is the surface face area of the electrode; C0 ¼ 1.2 mol/L, the
2
electrolyte concentration; andpD c ffi(cm /s) is the ion diffusion constant. As noted
ffiffiffiffiffiffi
above, (1.20) has units of W= sec, which when divided by square root of angular
frequency results in Zw in Ohms.
The Warburg modified Randles equivalent of Figure 1.9 is discussed with
more specifics in Example 1.3.
Example 1.3: Consider the Warburg impedance modified Randles equivalent cir-
cuit for an electrochemical capacitor repeated here as Figure 1.10 and the decom-
position of the resultant impedance function, Z, as (1.21).
Z w ðwÞ ¼ Z 0 ðwÞ jZ 00 ðwÞ ð1:21Þ
Cdl
Rs
Rct Zw
W
Figure 1.10 The Warburg-modified Randles equivalent circuit for an EC
Analytically solving for the input impedance (1.21) of this equivalent circuit,
the following expressions are found for Z0 and Z00 .
pffiffiffiffi
0 Rs t2c w3 þ 2attc w2:5 þ 2a2 tC dl w2 þ atp w1:5 þ R0s w þ a w
Z ðwÞ ¼ ð1:22Þ
t2c w3 þ 2atc C dl w2:5 þ 2a2 C 2dl w2 þ 2aC dl w1:5 þ w
pffiffiffiffi
atpp w2:5 þ tc Rct w2 þ atw1:5 þ 2a2 C dl w þ a w
Z 00 ¼ ð1:23Þ
t2c w3 þ 2atc C dl w2:5 þ 2a2 C 2dl w2 þ 2aC dl w1:5 þ w
where
R0s ¼ Rs þ Rct
t ¼ Rs C dl
tc ¼ Rct C dl
ð1:24Þ
ts ¼ R0s C dl
tp ¼ t þ ts tc
tpp ¼ ttc tc ts
Using the values cited for (1.20) and taking the case of a 10 F symmetric
ultracapacitor having approximate parameters: Rs ¼ 180 mW, Rct ¼ 100 mW and
ch001 7 June 2011; 13:4:42

Cdl ¼ 10 F, all the parameters in (1.24) are defined. For the remaining parameters,
the Warburg impedance magnitude, use (1.20) to match the resultant Nyquist fre-
quency response to the analytically derived Nyquist function of (1.21) and its
components to yield the magnitude of a. The value of a is determined by the
ultracapacitor cell parameters and by iteration on (1.20) to find the diffusion
constant, Dc.
Following this procedure the results summarized in Table 1.3 are found for
three representative values of Warburg magnitude and the ion diffusion coefficients
resulting in these magnitudes. A constant phase element of Warburg magnitude
results in low frequency 45 behavior of Zw(w).
In Table 1.3, note the characteristic 45 phase angle in Bode phase at very low
frequency, a characteristic of Zcpe. Note the Nyquist response for a ¼ 0.002
and > 45 entry slope as the low frequency response enters the Randles semi-
circular zone, again, a characteristic sign of Warburg impedance.
Next, consider how the situation changes when the Warburg impedance,
Zw(w), is moved from the charge transfer resistance, Rct, branch to the double layer
capacitance, Cdl, branch. This analysis is performed in Example 1.4.
Example 1.4: For this follow-on example all the parameters and variables applied
in Example 1.1 are used, only the analysis for the Warburg modified Randles cir-
cuit model of a small ultracapacitor is changed to place Zw in the Cdl branch.
For this example, the components of Zw shown as (1.22) and (1.23) become
pffiffiffiffi
Rs t2c w2 þ attc w1:5 þ ða= w Þ þ R0s
Z0 ¼ ð1:25Þ
w2 t2 þ 1
pffiffiffiffi pffiffiffiffi
00 ðRs R0s Þtc w at2 w1:5 þ ðt tc Þa w ða= wÞ
Z ¼ ð1:26Þ
w2 t2 þ 1
Following the same procedure as in Example 1.3, the results are now sum-
marized in Table 1.4 for the three representative values of Warburg magnitude and
the ion diffusion coefficients resulting in those magnitudes.
There is similarity between the Nyquist responses for Zw in the Cdl branch
versus Zw in the Rct branch, but a distinctive change of character is also noted when
diffusion constant and Warburg magnitude values are the same.
It should now be clear to the reader that analytical model and representations
of the EC cell can be made and compared with laboratory test data so that a
determination of electrolyte characteristic parameters can be made, such as the
procedure above, to find ion diffusion coefficient.
Before proceeding to symmetric ultracapacitors, we conclude this section by
noting that ion current density, Jx, where x ¼ cation or anion, through a plane area
facing the electrode is a function of three forcing functions:
● Potential gradient
● Concentration gradient
● Temperature gradient
ch001 7 June 2011; 13:4:42

Table 1.3 Representative Nyquist and Bode plots of a small 10 F ultracapacitor with Zw in Rcf branch
Diffusion Dc ¼ 5.6 10–8 Dc ¼ 9 10–7 Dc ¼ 8.8 10–5

coefficient (cm2/s)
Warburg
pffiffi magnitude a ¼ 0.08 a ¼ 0.02 a ¼ 0.002
(W= sÞ
Nyquist for Warburg modified randles model Nyquist for Warburg modified randles model Nyquist for Warburg modified randles model
Nyquist 10 10 1
7.949 1.988 0.199
1 1
0.1
0.1
Z.2(i) Z.2(i) 0.1 Z.2(i)
0.01
0.01
–3 0.01
1 × 10
Imaginary part of Zw
–4 –3 –3
1.557 × 10 –4 4.284 × 10 9.266 × 10
1 × 10–3 1 × 10–3
ch001
1 × 10
0.1 1 10 0.1 1 10 0.1 1 10
0.182 Z.1(i) 8.1 0.181 Z.1(i) 2.259 0.181 Z.1(i) 0.479
Real part of Zw Real part of Zw Real part of Zw
Bode magnitude Zw Bode magnitude Zw Bode magnitude Zw

Bode magnitude 100 10 1
11.949 3.009 0.518
10
W(i) W(i) 1 W(i)
1
Magnitude |Zw|
Magnitude |Zw|
Magnitude |Zw|
7 June 2011; 13:4:42

0.181 0.181 0.181
0.1 0.1 0.1
1 × 10–4 1 × 10–3 0.01 0.1 1 10 1 × 10–4 1 × 10–3 0.01 0.1 1 10 1 × 10–4 1 × 10–3 0.01 0.1 1 10
1.013 × 10–4 ω(i) 10 1.013 × 10–4 ω(i) 1.013 × 10–4 ω(i) 10
Angular frequency Angular frequency Angular frequency
Argument Zw Argument Zw Argument Zw

Bode phase 50 50 25
44.461 41.345 22.554
40 40 20
30
30 15
Arg.W(i) 20 Arg.W(i) Arg.W(i)
20 10
10
Angle (deg)
Angle (deg)
Angle (deg)
0 10 5
–1.723 1.356 2.897
–10 0 0
1 × 10–4 1 × 10–3 0.01 0.1 1 10 1 × 10–4 1 × 10–3 0.01 0.1 1 10 1 × 10–4 1 × 10–3 0.01 0.1 1 10
1.013 × 10–4 ω(i) 10 1.013 × 10–4 ω(i) 10 1.013 × 10–4 ω(i) 10
Table 1.4 Representative Nyquist and Bode plots of a small 10 F ultracapacitor with Zw in Cdl branch
Diffusion coefficient Dc ¼ 5.6 10–8 Dc ¼ 9 10–7 Dc ¼ 8.8 10–5

(cm2/s)
Warburg
pffiffi magnitude a ¼ 0.08 a ¼ 0.02 a ¼ 0:002
(W= s)
Randles with Cdl + Zw branch Randles with Cdl + Zw branch Randles with Cdl + Zw branch
Nyquist 10 10 1
1 1
Z22(i) Z22(i) Z22(i) 0.1
0.1 0.1
Imaginary Zw
Imaginary Zw
Imaginary Zw
ch001
0.01 0.01 0.01
0.1 1 10 0.1 1 10 0.1 1
Z11(i) Z11(i) Z11(i)
Real Zw Real Zw Real Zw
Bode magnitude randles Cdl + Zw Bode magnitude randles Cdl + Zw Bode magnitude randles Cdl + Zw
Bode magnitude 100 10 1
10
W1(i) W1(i) 1 W1(i)
7 June 2011; 13:4:43

1
Magnitude |Zw|
Magnitude |Zw|
Magnitude |Zw|
0.1 0.1 0.1

1 × 10–4 1 × 10–3 0.01 0.1 1 10 1 × 10–4 1 × 10–3 0.01 0.1 1 10 1 × 10–4 1 × 10–3 0.01 0.1 1 10
ω(i) Z.1(i) Z.1(i)
Argument randles Cdl + Zw Argument randles Cdl + Zw Argument randles Cdl + Zw

Bode phase 50 50 25
40 20
40
30 15
Argw1(i) Argw1(i) Argw1(i)
20 10
30
Angle (deg)
Angle (deg)
Angle (deg)
Types of ultracapacitors
10 5
20 0 0
1 × 10–4 1 × 10–3 0.01 0.1 1 10 1 × 10–4 1 × 10–3 0.01 0.1 1 10 1 × 10–4 1 × 10–3 0.01 0.1 1 10
15
Z.1(i) Z.1(i) Z.1(i)

The question facing the industry today is what influence power electronic
induced high frequency ripple currents have on electrolyte and electrodes and
therefore on its service life. The behavior of EC cells in terms of electric potential
across the electrolyte (that is from one electrode carbon surface to the opposing
electrode’s carbon surface) has been analyzed according to Poisson–Nernst–Planck
theory [2]. The theoretical behavior of ionic current density under the stress of
potential and concentration gradients is given by (1.27). The parameters in (1.27)
are those covered in the above examples.
J c ¼ qmc C 0 ðrjÞ qDc ðrC 0 Þ ð1:27Þ
A similar expression to (1.27) can be written for the anion current density, Ja.
Figure 1.11 illustrates the passage of ionic currents through a surface element
representative of the separator cross-section area in an EC.
Jc
Ja
Figure 1.11 Ionic current passage through a surface (Jc is positive cation flow
and Ja is negative anion flow)
The character of ionic current obeys (1.27), such that immediately upon
application of a potential at the terminals of the ultracapacitor, such as will occur
when a constant current is applied, will result in a potential gradient across the
electrolyte of nearly linear slope with position across the electrolyte. Simulta-
neously, the ion concentration will be uniform in the uncharged cell at time t ¼ 0 þ .
Then, at a very short time interval, after t ¼ 0 þ , the ion concentration will quickly
shift the cations to the electrode having an excess of electrons in the carbon and the
anions to the electrode having a deficit of electrons in the carbon. The result, as
Figure 1.11 illustrates, is a flow of cations into an electrode and a counterflow of
anions out of the electrode. The concentration shift alters the potential gradient with
both working according to (1.27) to support the impressed electron current.
An on-going industry concern is what happens to the ultracapacitor cell when
the applied current contains a significant ripple component that occurs when the
ultracapacitor (or a battery) is interfaced to an application via a d.c.–d.c. or d.c.–a.c.
power converter. These concerns are addressed later in this book.
ch001 7 June 2011; 13:4:43

1.2 Symmetric types

Recall from earlier discussions that the ultracapacitor functions by electrochemical
adsorption of ions to a conductive surface, in this case, highly porous activated
carbon. This means that energy storage is physical, consisting of a layer of charged
ions in solution adsorbed within some small distance, a Debye length, from an
electron cloud in the carbon. This means that conduction band electrons in the
carbon are involved unlike chemical energy storage in which valence band electron
exchanges take place during reduction–oxidation (redox) reactions of the bulk
electrode materials. In the ultracapacitor, the electrolyte is essentially an ion
reservoir in which a surplus of ions is needed because the electrolyte conductivity,
hence its ionic contribution to resistance, varies according to state-of-charge
(SOC). Energy storage in the ultracapacitor is therefore proportional to electrode
surface area, whereas in a redox controlled battery the energy stored is proportional
to electrode mass.
Symmetric means that both electrodes in a carbon–carbon ultracapacitor are
fabricated identically, and in fact they are. However, once charged the impurities
contained in the activated carbon as residual ash react with the electrolyte giving
each electrode a potential set. If the ultracapacitor is then recharged in the reverse
polarity there will be an attendant irreversible loss of capacitance, but it will still
function as an energy storage device. In this same vein, materials of which the
ultracapacitor cell is fabricated from are also electrochemically important so that
side reactions do not occur between the physical container of the cell and its
electrolyte. A symmetric ultracapacitor will be made from high purity aluminum
positive termination materials and an aluminum alloy negative termination mate-
rial. The electrode current collectors are also made of aluminum that is etched for
better contact to the activated carbon electrode film (whether dry processed and
laminated or wet processed and slurry applied).
Symmetric ultracapacitors are also distinguished by the type of electrolyte sol-
vent. For example, sulfuric acid in water comprises an aqueous electrolyte, and an
organic salt in a solvent comprises an organic electrolyte. The organic electrolyte,
symmetric, ultracapacitor is our focus here. The organic electrolyte solvent can be
either propylene carbonate (PC) or acetonitrile (AN). The added salt may be tetra-
ethylammonium tetraflouroborate (TEATFB) where the cation [Et4N þ] is a qua-
ternary tetraethylammonium and anion [BF4–] is tetraflouroborate. These cations and
anions are solvated in solution and their properties are given in Table 1.5.
Table 1.5 Properties of common symmetric solvent and salts
Component Raw ion size, Solvated ion Molar mass, Solvated

d0 (nm) size, ds (nm) unsolvated (g/mol) mass (g/mol)
[Et4N þ] 0.67 1.30 116 485
[BF4–] 0.48 1.16 86.81 373.81
[AN] 0.45 0.45 41 –
ch001 7 June 2011; 13:4:43

The molar mass is calculated from the chemical formula for the respective ions
and solvent: [CH3CH2]4N þ ?mass [Et4N þ ], [BF4–] and [CH3CN]?[AN] as
summarized in Table 1.5. Typically, a solvated cation is surrounded by 9 AN
molecules and anion by 7 AN molecules. Some investigators claim that in pores of
2 nm or less, called micropores, the ions form strings or nano electric wires that
reach deep into a pore. Other investigators claim that as the pore size diminishes to
near the solvated ion size, the ion partially de-solvates and still fits into a micro-
pore. These contentions are the subjects of on-going research and in fact for
carbide-derived carbon(CDC), it has already been shown that nanoporous carbon
exhibits an anomalous capacitance [7].
In Figure 1.12 the core electrode assembly of a carbon–carbon ultracapacitor is
shown. The jelly roll consists of a pair of double-side coated electrodes, porous
paper separator and aluminum foil current collector extensions, or tab ends, on the
jelly roll for making electrical connection. The right hand pictorial in Figure 1.12
shows that a potential source depletes one electrode of electrons (positive elec-
trode) and deposits them on the opposing electrode (negative), creating an electron
excess and thereby setting up an electric field internally across the structure.
Figure 1.13 expands on this argument by showing the detail of the adsorption layer
and internal potential versus location.
(+) Terminal
Jelly roll
Paper
Film
Film Foil
(–) Terminal
Figure 1.12 Illustration of carbon–carbon symmetric ultracapacitor and

electrode detail
Figure 1.13 shows that an intense electric field exists across the charged
ultracapacitor from electron deficit (positive) carbon electrode to anion layer in the
electrolyte. This double layer structure is separated by an ion permeable and elec-
tron blocking separator material, typically porous paper, from the opposing elec-
trode. The opposing electrode, shown as cation accumulation layer and carbon with
electron excess, is the negative electrode. The ultracapacitor, therefore, consists of
two double layer capacitors, each with metallic connection on one ‘plate’ and liquid
connection on the adjacent ‘plate’. When the ultracapacitor charges the potential,
ch001 7 June 2011; 13:4:44

E E
d d
f(x) E(x) n(x)
Figure 1.13 Graphic showing ultracapacitor double layer charge concentrations,

potential and electric field
f(x) is nearly a diagonal straight line across the cell which then rapidly collapses to
sharp potential gradients across the double layer, also called Helmholtz layer or
compact layer of thickness ‘d’. The potential field across the separator is far shal-
lower due to the conducting nature of the electrolyte and another sharp potential
gradient at the opposite electrode’s Helmholtz layer. This charge separation dis-
tance called dc, or Debye length, is one characteristic of electrochemical capacitors,
effectively putting the ‘ultra’ in ultracapacitor. The charge accumulation at the
electrode surfaces in the highly porous activated carbon means that the separator
becomes ion depleting; thus, its resistance increases somewhat when fully charged
and replenishes again once fully discharged. In electrochemical capacitors based on
ion adsorption phenomena, the closest approach of the charge carrying ions to the
charged carbon is the Debye length [8,9]. For the symmetric carbon–carbon ultra-
capacitor the Debye length can be calculated knowing fundamental constants, the
electrolyte solvent permittivity, and electrolyte salt concentration, C0, according to
(1.28).
er e0 RT
d 2c ¼ ð1:28Þ
2F 2 C 0
For a 1 M electrolyte at room temperature and AN permittivity of 37.5, it is

easy to show that (1.28) calculates d 2c ¼ 44:18 1018 m, or dc ¼ 6.65 nm of
charge separation distance. It is instructive to carry out a simple example at this
point to reinforce the magnitude of this nanometer charge separation distance.
Example 1.5: Consider a symmetric ultracapacitor fabricated from activated car-
bon having a specific surface area, Sa, of 1610 m2/g. Also consider that the ultra-
capacitor is rated 3000 F at Umx ¼ 2.7 V and there are Mc ¼ 72 g of activated carbon
ch001 7 June 2011; 13:4:44

per electrode. Determine the electrostatic pressure exerted by the ionic charge layer
on the activated carbon containing the electron charge. The first step is to calculate
the total available surface charge of ion adsorption, A, and second using (1.28) in
the equation for electrostatic force, F.
A ¼ S a M c ¼ 1610ð72Þ ¼ 115; 200 m2 ð1:29Þ
Q ¼ CU mx ¼ 3000 ð2:7Þ ¼ 8100 C ð1:30Þ
Q2 ð8100Þ2
F¼ ¼
4per e0 d 2c 12:56 ð37:5Þð8:854 1012 Þð44:18 1018 Þ
¼ 3:56 1026 N ð1:31Þ
This is an enormously high force and the resultant electrostatic pressure will be
similarly extreme. Dividing (1.31) by (1.29) results in the ionic pressure due to
electric field at the ultracapacitor double layer. Note that when fully charged each
electrode holds off a potential of 1.35 V across the Debye length, or essentially
Umx/(2 dc) ¼ 2 1010 V/m. The electrostatic pressure at each electrode is therefore
F 3:56 1026
P¼ ¼ ¼ 3:09 1021 Pa ð1:32Þ
A 1:152 105
The electrostatic pressure given by (1.32) is 3.09 TGPa (terra (1012)

gigaPascal (109)). The reader is encouraged to work out Exercise 1.3 at the end of
this chapter to get a better appreciation for this scale.
The calculation of ultracapacitor capacitance is straightforward and follows the
classical formula given here as (1.33). Capacitance is a geometric property of
material surface, charge separation distance, and the permittivity of the medium
between the charges.
er e0 S a M c
C elec ¼ ð1:33Þ
dc
For the values cited in Example 1.5 the electrode capacitance given by (1.33) is
Celec ¼ 5751 F. What is not so clear is that for the cell, since two such double layer
capacitors are connected in series, that the cell capacitance, C, works out to one half
of Celec, or C ¼ 2875 F, but at twice the electrode potential, or 2.7 V. The charge at
each electrode, Q, is the same, but the energy of two electrode charges stacked up is
multiplied by a factor of 4, consult (1.30). It is easy to see that the energy stored in
both electrodes sums to total cell energy by considering each electrode to store a
packet of energy, W el ¼ 12 C el ðQ=C el Þ2 ¼ 14 ðQ2 =CÞ ¼ 12 W uc . The ultracapacitor,
therefore, stores two electrode packets of energy, one at each of the internal elec-
tronic double layer capacitors. The same applies to any number of series connected
cells, as the total energy of the string is the sum of the cell energies stored.
ch001 7 June 2011; 13:4:44

Ultracapacitor modeling is discussed in detail in Chapter 2. In this chapter, the

illustration in Figure 1.14 is sufficient to convey the essentials of ultracapacitor
modeling. Because an ultracapacitor consists of two double layer capacitors, one
per electrode, in series, the illustration captures this fact along with the presence of
electronic resistance, Re, and ionic resistance, Ri, as components of equivalent
series resistance (ESR) of the cell. Electronic resistance consists of conducting
material bulk resistance and all electronic interface resistances such as termination
post, aluminum lid and can, internal aluminum-to-aluminum welds, the current
collector foils, and their interface resistance to the carbon film. Add to this the
electronic resistance of the carbon-to-carbon particle interfaces and bulk resistance
and one has a complete picture of resistance to electron flow. The ion conducting
electrolyte and porous separator, along with ion current flow restrictions through
carbon electrode inter-particle channels and into pores, comprise the ionic resis-
tance. For example, the potential and concentration forcing functions to ionic cur-
rent density described by (1.27) define the kinetics of ionic current.
(+) Terminal
Jelly roll
Paper
Film
Film Foil
(–) Terminal
Re Ri Re
C(U) 3000 Farads/2.7 VOC C(U)
Figure 1.14 Illustration of symmetric ultracapacitor model
Figure 1.15 illustrates a method to identify the electronic and ionic partitions
of ESR in the ultracapacitor cell. In this figure, the cell is temperature soaked for at
least five thermal time constants and then characterized for capacitance and ESR.
The ESR measurements versus temperature are then curve fit to an approximating
function consisting of the Taylor series expansion of conductor resistance and an
ch001 7 June 2011; 13:4:44

ESR(T) for the 3000 F cell at BOL

6 × 10–4
5.4 × 10–4
4.8 × 10–4
R.ESR30(i) 3.6 × 10–4
Ohms
ESR.3000 2.4 × 10–4
1.2 × 10–4
0
0
−50 −20 10 40 70 100
–40 a(i), x 70
Temperature (°C)
Figure 1.15 Illustration of 3000 F cell ESR versus temperature
Arrhenius function defining ionic resistance as shown in (1.34).
ESRðT Þ ¼ b1 ESRðT 0 Þf1 þ gðT T 0 Þg þ b2 ESRðT 0 Þ eðk T =2ÞðT T 0 Þ ð1:34Þ
where T0 ¼ 293 K, g ¼ 0.007 (K–1) the temperature coefficient of the aggregated

aluminum and carbon constituents and kT ¼ 0.045 (K–1) the aggregate ionic
activation energy constituent. For the 3000 F cell ESR(T) response shown in
Figure 1.15, the electronic (b1) and ionic (b2) weight factors are 0.55 and 0.45,
respectively. This puts the resistive component split at nearly 50% for this parti-
cular size ultracapacitor cell. Shorter cells of the same diameter have greater
imbalance in the electronic to ionic ratio, with b1 in this case being a larger fraction
of b1 þ b2 ¼ 1. For example, in the 3000 F cell having a nominal ESRdc ¼ 0.29 mW,
the electronic component Re ¼ 0.1595 mW and ionic component Ri ¼ 0.1305 mW. An
example will serve to reinforce the materials just covered, especially the value of
ionic resistance for the 3000 F cell.
Example 1.6: In this example, the 3000 F cell discussed in Example 1.5 is used. For
this cell the separator has a face area, As ¼ 0.2474 m2, derived from the overall
electrode length and width. Take the separator thickness, ts ¼ 40 m, and the acti-
vated carbon electrode thickness of tc ¼ 140 mm. If the conductivity of a 1 molar
electrolyte is 56 mS/cm, then using the classical formula for resistance calculate the
ionic portion of resistance and compare to Ri above.
ts þ tc ð40 þ 140Þ106
Rbi ¼ ¼ ¼ 12:98 105 ð1:35Þ
ssl As ð56 103 Þð102 Þð0:2475Þ
The approximation made in (1.35) that the electrolyte layer thickness con-
tributing to ionic resistance consists of the separator thickness plus 1/2 the electrode
film thickness on either side of the separator is a very good match to measured data.
In this case, the estimated Rbi ¼ 0:13 mW is the same as obtained by temperature
measurements discussed above.
ch001 7 June 2011; 13:4:45

The voltage sensitivity of ultracapacitor capacitance was addressed earlier in

this chapter by noting that the double layer capacitance is strongly voltage depen-
dent and only somewhat current magnitude dependent, and virtually insensitive to
temperature, until temperature reaches below –30 C. Consider the ultracapacitor C
(U, I) function introduced earlier in this chapter. Capacitance increases with
potential because of the presence of pseudo-capacitive phenomena in the electro-
des, a topic to be discussed more in the following section. Figure 1.16, an expanded
view of the surface shown in Figure 1.6, shows the strong increase in capacitance
with potential and to some degree with current magnitude. Recall from (1.28) that
double layer capacitance Debye length, or charge separation distance, is inversely
proportional to electrolyte concentration and that concentration increases dramati-
cally at the Helmholtz layer as the cell charges. This charge accumulation results in
the Debye length decreasing and from (1.33) a proportional increase in capacitance
can be expected. Also, the variation of ultracapacitor capacitance with temperature
is masked by concentration changes, since concentration also decreases with
decreasing temperature to nearly offset the influence of temperature on
capacitance.
BCAP3000
3800
3600
3400
Capacitance
3200
3000
2800
2600
2400
2200
2000
0
50
0
40
0
3
30
0
2.5
20
0
2
10
–1 0
1.5
–2 0
0
–3 0
1
0
Current
–4 0
0.5 Voltage
0
–5 0
0
0
00
Figure 1.16 Ultracapacitor rated 3000 F measured capacitance as function of

potential and current
The behavior of C (U, I) for current |I | < 100 A is not well understood at this
time. There is the possibility of instrumentation resolution and error that must be
accounted for, since small voltage changes are being sampled over a fixed time
interval. The instrumentation computer has high resolution of time but is subject to
data acquisition limitations on potential. The potential itself is the ultracapacitor
terminal potential sampled 5–10 s after the extinction of charge or discharge cur-
rent. During these initial few seconds, the ionic distribution in the cell may still be
in flux and contribute to the discontinuity seen near zero current.
The final point to consider on the topic of symmetric ultracapacitors, now that
we are equipped with an understanding of double layer capacitance and cell ESR
fundamentals, is their RC product. It is seen from a variety of production cells of
ch001 7 June 2011; 13:4:45

the f ¼ 60 mm can diameter size and time constant t*0.65 s on average as shown
in Figure 1.17.
Capacity for tau = 0.65 s

1 × 104
Capacitance (F)
1 × 103
C(i)
C1
100
10
1 × 10–4 1 × 10–3 0.01
R(i), Rmx
ESR (Ohm)
Figure 1.17 Trend-line in ultracapacitor cell time constant, 60 mm OD

cylindrical (left to right points: 3000 F, 2000 F, 1500 F, 1200 F,
650 F, 350 F, 140 F)
The topics introduced in this section are expanded on in Chapters 2 and 3,

covering the modeling of symmetric ultracapacitors and detailed discussions of the
power and energy performance. These introductory chapters will lay the foundation
for subsequent discussions of applications and case studies of the carbon–carbon
ultracapacitor.
1.3 Asymmetric types
The taxonomy of electrochemical capacitors is described in Figure 1.1, where the

EC branch partitions into symmetric and asymmetric types. The previous section
covered symmetric, carbon–carbon, ECs in considerable detail. The asymmetric
ultracapacitor is ideally a battery-like, nonpolarizable electrode, such as a metal
oxide, that is paired with an ultracapacitor electrode, such as the carbon double
layer capacitor electrode. Representative companies active in asymmetric ultra-
capacitor work and commercialization activities are summarized in Table 1.6.
The Evans Capacitor Co. hybrid capacitor [10] can achieve cell level operating
voltages of 170 V by extending the formation time of the tantalum pentoxide,
Ta2O5, dielectric layer on the Ta anode plug. The rated voltage of this EC is given
as anode potential, Ua, minus cathode potential, Uc, or Ur ¼ Ua – Uc, where Uc is on
the order of 1.2 V. The amount of charge accumulated on anode and cathode is
identical, Qa ¼ UaCa ¼ UcCc ¼ Qc, and since Cc p Ca the anode potential will be
much higher than the cathode potential. Even at tantalum prices in the range of US
$500/kg, these components have found niche application in aerospace ground
systems because of high voltage and very short RC time constants (23 ms for
ch001 7 June 2011; 13:4:45

Table 1.6 Asymmetric (hybrid) electrochemical capacitor products
Company Chemistry Cell Specific Energy

potential (V) energy (J/g) density (J/cm3)
Evans þ Ta/Ta2O5/RuO2– 25 0.45 1.36
Capacitor Co.
Evans RuO2/H2SO4/RuO2 10 0.56 2.38
Capacitor Co.
ESMA þ NiOOH/KOH/C– 1.7 (0.8 V 36 53
<Unom<1.6 V)
ELINT þ PbO2/H2SO4/PbC– 1.7 36
ELINT þ PbO2/H2SO4/C– 2.1?2.33 86
aluminum hybrid, 1.7 s for tantalum hybrid, and 150 ms for symmetric RuO2). Their
tantalum-RuO2 hybrid product, for example, achieves high voltage from the tan-
talum pentoxide dielectric film and the extremely high double layer capacitance of
the negative electrode, an EC capacitor. The EC capacitance, Cc*100 mF/cm2 is
orders of magnitude greater than the tantalum anode capacitance, Ca. The overall
cell capacitance is therefore
1 1 1
¼ þ ð1:36Þ
C Ca Cc
JS Company ESMA [11] was among the first to develop asymmetric ECs for
heavy duty applications such as truck engine starting. The ESMA EC203 product,
for example, operates over a voltage window of 1.6–0.8 V and is rated at cell level
as 108 kF, 0.43 mW, RC ¼ 46.4 s, and is optimized for –30 C temperature. At
10 Wh/kg this asymmetric cell has roughly twice the specific energy of
carbon–carbon symmetric ECs.
ELINT is another manufacturer of hybrid capacitors, specifically nickel oxy-
hydroxide and carbon and lead oxide carbon types [12]. ELINT’s PbO2/H2SO4/PbC
utilizes a Faradaic electrode anode and a polarizable electrode exhibiting double
layer capacitance. The lead dioxide positive electrode is the same as a lead acid
battery electrode and operates at a potential of 2.0 V for long life. ELINT’s NiOOH/
KOH/C, as with ESMA, is one of the oldest and most studied of the asymmetric
ECs. The nonpolarizable anode is made from sintered nickel, the same as used in
nickel-cadmium cells. This asymmetric has good specific energy at 10 Wh/kg but
low specific power of only 200–600 W/kg. ELINT also manufacturers a PbO2/
H2SO4/C hybrid capacitor, one of the best in terms of quality for the price that
operates from 2.1 to 2.3 V/cell and delivers a specific energy of 24 Wh/kg but only
100 W/kg of specific power. A variant of this chemistry, the þ C/H2SO4/PbC–
utilizes a double layer carbon capacitor anode and redox cathode to operate at a
nominal potential of 1.3–1.4 V/cell.
Professor Burke [13] has studied hybrid and pseudo-capacitive energy storage
devices for many years and reported on work done at Telcordia on lithium titanate
ch001 7 June 2011; 13:4:45

hybrids, by IMRA on metal oxide organic hybrids, and by his university, UC-
Davis, on the þ C/H2SO4/PbC– chemistry that ELINT worked on. In his work, the
carbon electrode contributes to double layer capacitance and the lead dioxide
electrode is sized so that only a small fraction of its charge capacity is cycled
during operation. According to Prof. Burke, the cell potential (1.37) is composed of
the EDLC voltage swing plus the standard potential of the lead dioxide of which the
potential of PbO2 limits the maximum working potential of the cell to 2.25 V to a
minimum of 1 V.
U cell ¼ dU DLC þ U 0PbO2 ð1:37Þ
The UC-Davis experimental hybrid EC had an SE ¼ 68.4 J/g, an ED ¼

187 J/cm3, and 95% power capability of 5.7 kW/kg unpackaged at a working
potential of 2.25 V?1 V in sulfuric acid electrolyte. The lithium-titanate/acetoni-
trile/carbon hybrid cell operated from 2.8 V?1.6 V and exhibited SE ¼ 50 J/g,
ED ¼ 86 J/cm3, and 3.8 kW/kg power at 95% efficiency.
Example 1.7: Using the following symmetric ultracapacitor electrode parameters,
calculate the capacitance per unit surface area and compare this to the value noted
above for the Evans Capacitor Co. ruthenium dioxide cathode in their hybrid cell (i.e.,
100 mF/cm2).
For the symmetric carbon–carbon ultracapacitor having Sa*1610 m2/g,
Melec ¼ 72 g, and the electrode capacitance calculated in (1.33), one obtains
C elec C elec 5751 F

¼ ¼ ¼ 4:961 mF=cm2 ð1:38Þ
A S a M elec 1:159 109 cm2
The ratio of the RuO2 polarizable electrode capacitance density in the Evans
hybrid capacitor to the symmetric carbon ultracapacitor capacitance density is
therefore
100 mF 100 mF
¼ ¼ 20;157 ð1:39Þ
4:961 mF 0:00496 mF
The presence of adsorption pseudo-capacitance makes a very dramatic

comparison to that of near pure polarizable capacitance of a double layer
capacitance.
Before closing this section, it is worthwhile to examine adsorption pseudo-
capacitance from an electrochemistry view. The treatment of adsorption
pseudo-capacitance to follow is meant to introduce the reader to the topics
of pseudo-capacitive intermediates formed on conducting surfaces by reduction–
oxidation charge transfer processes [14]. The interested reader is encouraged to
read Chapter 19 of References 14 and 15 to gain useful insights into intermediates
beyond this almost superficial coverage derived from Reference 14.
Gileadi applies the Langmuir isotherm, or surface free energy under conditions
of constant temperature, to identify the fractional surface coverage, q (or surface
ch001 7 June 2011; 13:4:46

density), by a species for given conditions of bulk concentration, C, of the species

and its equilibrium constant of adsorption, K, as given by (1.40).
q
¼ KC ð1:40Þ
1q
where the equilibrium constant of adsorption, K, can be defined in terms of stan-
dard free energy as stated in (1.41) from Reference 14. The monolayer adsorbed
assumes the surface is homogeneous and that no lateral interactions of species
occur. Surfaces are inhomogeneous and more of a patchwork of varying degrees of
Gibbs energy (DG0 ¼ DH T DS (kJ/mol), free energy equals enthalpy change
minus entropy change, and if the result is negative the process is spontaneous). At
electrochemical equilibrium, the reaction quotient, Q?K, the thermodynamic
equilibrium is therefore constant.

D
DG ¼ DG0 þ RT lnðQÞQ!K ¼ 0
RT lnfKg ¼ DG0ads ð1:41Þ
Example 1.8: Compare the van der Waals separation distance of a univalent spe-
cies, rw, to that of the atomic spacing of metallic atoms on the surface of a metal,
Lm, given the density of atoms, nm ¼ 1.3 1015/cm2.
Solution: From the Lennard-Jones potential, the van der Waals force,
Fw ¼ dWw/dr, is defined as given in (1.42) and from this the radii at zero force,
F0, Fwmx, and Fwmn, can be extracted for a given species distance to a surface, rw.
dW w a1 b1
Fw ¼ ¼ 7 þ 13 ð1:42Þ
dr r r
For a1 ¼ 6 10–7 and b1 ¼ 1.2 10–3 and 3 Å < rw < 8 Å (where Å ¼ 10–10 m), the
force levels and distances are summarized in Table 1.7.
Take the force at the neutral point distance, rw ¼ 3.55 Å, and compare this to
the atomic spacing, Lm.
1 1
Lm ¼ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ 0:28 109 ¼ 2:8 Å ð1:43Þ
nm 1:3 1015
Table 1.7 Approximate van der Waals distance of ion from

conducting surface
Van der Waals force ( 10–11 N) Approach distance, rw

Fw ¼ 0 rw ¼ 3.55 Å
Fwmn ¼ max attractive (–) rw ¼ 3.94 Å
Fwmx ¼ repulsive at same magnitude (þ) rw ¼ 3.45 Å
ch001 7 June 2011; 13:4:46

For large molecules, as one would expect to have in ECs, Langmuir isotherm
parameters C (mM) and 1 < K < 25 are representative values. The surface density is
typically 0.02 < q < 0.98 and the constants applied in electrochemistry are:
● R ¼ 8.314 J/(kmol) the gas law constant
● F ¼ 96,485 C/mol Faraday’s constant
● T ¼ 293 K room temperature on Kelvin scale
For these large molecules, the expression (1.40) must be modified by the over
potential relative to surface state free energy (RT/F) given by (1.44) and re-arran-
ged to obtain (1.45) for surface density.

q
¼ KC eEF=RT ð1:44Þ
1q
KC eEF=RT
q¼ ð1:45Þ
1 þ KC eEF=RT
Take the standard potential of adsorption, E0, given by (1.46) and show that at
an over potential, h ¼ 0.23 V, transfer coefficient, a ¼ 0.5, and the Tafel current
representative of surface redox pseudo-capacitance will increase by two orders of
magnitude above its saturation value at the standard potential, i0. The Tafel relation
that was introduced in Example 1.2 is repeated here as (1.47).
RT 8:314ð293Þ
E0 ¼ ¼ ¼ 0:025 V ð1:46Þ
F 96; 485
0
i ¼ i0 eahF=RT ¼ i0 eah=E ð1:47Þ
It is easy to see from (1.47) that h ¼ 0.23 V by rearranging and setting the
exponential equal to 100 as follows:
0 ah
eah=E ¼ 102 or ¼ lnð102 Þ ¼ 2:3 logð102 Þ ð1:48Þ
E0
In electrochemical impedance spectroscopy (EIS) under slow cathodic sweep,
it is possible to locate the adsorption deposition potential by noting the voltage at
which Tafel current peaks occur. Adsorbed intermediates are formed during oxi-
dation of organic molecules in aqueous or organic electrolytes, provided the vol-
tage sweep increments are sufficiently slow to allow the redox processes to
proceed. To illustrate this consider a Faradaic charge, qF, that is deposited as a
fractional part of total surface charge existing in a monolayer, qc, such that
qF ¼ qqc, where q ¼ fractional coverage given by Langmuir isotherm (or by more
detailed Frumkin or Temkin isotherms).
The derivative of this Faradaic charge with potential leads to the definition of a
new differential capacitance, Cj, the adsorption pseudo-capacitance [14]. The
pseudo-capacitance exhibits capacitance behavior because it depends on charge
ch001 7 June 2011; 13:4:46

transfer, albeit via a redox process, and that when potential changes a current flows
until sufficient charge transfer occurs to equalize the free energy at which point the
current goes to zero. In a C–V plot, this behavior shows up as peaks in current
versus potential.
dqF dq
Cf ¼ ¼ qc ð1:49Þ
dE dE
Substitute (1.45) into (1.49) and simplify results to (1.50), with acknowl-
edgement of the procedure in Reference 14. In (1.50), the adsorption pseudo-
capacitance is derived from the large molecule Langmuir isotherm, CfL, which is
considered to give acceptable accuracy over the more linear range of q versus E.
( )
qF KC eEF=RT
C fL ¼ c ð1:50Þ
RT ð1 þ KC eEF=RT Þ2
The maximum value of (1.50) when q ¼ 0.5, again with acknowledgement due
[14] is:
qF RT
max C fL ¼ c and Emax ¼ lnðKCÞ ð1:51Þ
4RT F
At standard conditions for concentration, C ¼ 1, and for K, the equilibrium

constant of adsorption, as a parameter, the equilibrium potentials shown in
Table 1.8 are obtained by setting Emax ¼ E0f .
For more discussion on maximum value of adsorption pseudo-capacitance, see
Exercise 1.5. A model for adsorption pseudo-capacitance presence in an EDLC can
be represented as shown in Figure 1.18. In this representation, the cell electronic
resistance is modeled as Rs, a series element, and the double layer capacitance as
Cdl. The pseudo-capacitance Cf exists at the electrode surface in the presence of the
DLC so is modeled in direct parallel with Cdl along with a charge transfer resis-
tance Rct. The charge transfer resistance models the kinetics limited mass transport
phenomena and introduces a rate-limiting influence on the pseudo-capacitance
element. On a Bode plot, the presence of the pseudo-capacitance branch shows up
as a plateau at low frequency, where the frequency is determined by the adsorption
pseudo-capacitance kinetics, specifically how fast q in (1.49) can follow the
Table 1.8 Representative equilibrium potentials

for the case examined in Example 1.8
E0f ¼ 0 C¼1 K¼1

E0f ¼ 0:041 C¼1 K¼5
E0f ¼ 0:081 C¼1 K ¼ 25
ch001 7 June 2011; 13:4:46

potential versus frequency. At high frequency, the pseudo-capacitance branch

vanishes and Z00 goes to zero at Z0 ¼ Rs.
Rct Cj
Rs
Cd
Figure 1.18 Capacitor model having DLC and pseudo-capacitance branch
1.4 Hybrid capacitors
The final topic to be considered is the case of intercalating pseudo-capacitance for

the class of a particular hybrid capacitor referred to as the lithium cap. The presence
of intercalating pseudo-capacitance in purely carbon systems was studied by
Bakhmatyuk et al. [16], and they noted that when specific capacitance exceeds
*150 F/g, the pseudo-capacitive behaviors are exhibited. They noted that the max-
imum capacitance density obtained in carbon–carbon ultracapacitors is 25 mF/cm3, and
that for higher capacitance density in graphite the intercalating pseudo-capacitance
phenomena are observed. Before examining this particular hybrid capacitor, it is
worthwhile to start with the lithium-ion battery insertion chemistry. Kotz et al. [17]
fabricated an ultracapacitor having a graphitic, rather than activated, carbon electrode
so that the commercially available electrolyte salt, TEATBF4, is capable of inter-
calating as ions. It is known, and as already explained in the previous section, that
when cell potentials in the symmetric ultracapacitor approach 3 V and higher, the
adsorption pseudo-capacitive phenomena appear and become more prominent. By
replacing the nano-porous activated carbon with graphite, the authors contend that ion
insertion will take place as evidenced by graphite dilation on intercalation and con-
traction on ion de-intercalation. This indeed was the case as validated by measure-
ments of the electrode cell dimensions.
It is convenient to view the process of ion intercalation and de-intercalation by
examining the lithium-ion battery that is commercially available in mass quantity,
with global annual production in the range of 3B cells. The graphic in Figure 1.19
illustrates how lithium, present in the cathode as a metal oxide, along with con-
ductive agents and binders to retain electrode integrity, is oxidized on charging and
enters the electrolyte (LiPF6) as Li þ ions when the switch labeled ‘C’ is closed.
The potential source pumps electrons out of the cathode structure, releasing Li þ
ions into solution at the cathode through its protective surface electrolyte interphase
(SEI) shown as a coating. Simultaneously at the anode, since the electrolyte must
remain charge neutral at all times, a Li þ ion intercalates into the graphitic structure
via the anode SEI where the ion is reduced and remains trapped between three
carbon atoms above and three carbon atoms below its location in the graphite
planes.
ch001 7 June 2011; 13:4:47

R
e–
D
e–
C
(+) (–)
Cathode Anode
C = Charge
D = Discharge
= Ion
Figure 1.19 Lithium-ion insertion type chemistry
When the charge switch ‘C’ in Figure 1.19 is opened, a measurable mass will
transport out of the cathode into the anode, resulting in cathode structure change
and anode dilation as the graphitic planes are pushed apart by the presence of
trapped lithium. The charged lithium-ion cell will remain in this state for a con-
siderable length of time, years, as its self-discharge is very low. Next, consider
what happens when the discharge switch ‘D’ is closed. In this case, the potential of
the anode relative to the cathode, approximately 4 V in this case, will cause the
intercalated lithium to give up an electron, a redox process, exit the anode as a
lithium ion and enter the electrolyte. At the cathode side, there will be an attendant
intercalation of this Li þ ion into the cathode followed by reduction with one
electron entering into the cathode. The cathode, therefore, intercalates lithium on
discharge and the anode de-intercalates lithium. The chemistry of a particular
lithium-ion cell, nickel cobalt in this case, is given by the following pair of rever-
sible reactions taking place where charge ‘C’ and discharge ‘D’ directions are
labeled. In (1.52), the removal of ‘y’ electrons from the cathode by the potential
source delivers yLi þ into solution. Then, (1.53) applies for each electron transfer at
the anode, showing that the intercalating Li þ is reduced and trapped between the
six carbon atoms.
LiNi1x Cox O2 ye ! yLiþ þ Li Ni Co O ð1:52Þ

DC 1y 1x x 2
6C þ Liþ þ e ! C Li ð1:53Þ
DC 6
This short tutorial on insertion chemistry will be helpful in understanding the

lithium-capacitor. In 2006, Mr. Hatozaki, Subaru Technical Research Center of
Fuji Heavy Industries (FHI), introduced the lithium ion capacitor (LIC) [18]. The
ch001 7 June 2011; 13:4:47

LIC approach is to first pre-dope the graphitic negative electrode with lithium so
that a ready source of Li þ ions is available, and to construct an opposing electrode
of activated carbon to act as a double layer capacitor, much the same as discussed
for the RuO2 hybrid capacitor. Just as in this earlier case, the negative electrode
capacitance Cn is far greater than the positive electrode capacitance Cp and con-
tributes to the overall cell capacitance Ccell according to (1.54), where Mn and Mp
represent the mass of negative and positive electrodes, respectively.
1 1 1
¼ þ ð1:54Þ
C cell ðM n þ M p Þ C n M n C p M p
To understand the operating characteristic of the lithium capacitor refer to
Figure 1.20, which shows the potential of the negative electrode, lithium-doped
carbon material, having a potential of approximately 0.2 V relative to Li/Li þ and a
positive, EDLC, electrode having a potential of 3 V relative to Li/Li þ . The elec-
trolyte solution can be propylene carbonate (PC) or ethylene carbonate (EC) with
an LiPF6 salt. On charging, [Li þ ] ions from solution intercalate into the negative
electrode, leaving the [BF4–] anions to adsorb onto the positive electrode carbon
surface as the EDLC component. Discharge is somewhat different because of the
lithium pre-doped negative electrode. On discharge, lithium de-intercalates into
solution and neutralizes [BF4–] anions, maintaining charge neutrality of the elec-
trolyte, only the discharge can proceed further than in a normal EDLC because
excess [Li þ ] cations can enter the electrolyte.
<> – Electrolyte dissociation

4
UEDLC
2.5–2.7 V
3 +
ELIC
U(V)
1
–
ELIC
0 t
tchg tdch – EDLC
tdch – LIC
Figure 1.20 Operating principle of the lithium-capacitor hybrid cell
These phenomena are explained in the sketch of Figure 1.20 showing that the
lithium pre-doping basically biases the positive, EDLC, electrode up by 3 V where
after it still acts as a classical ultracapacitor except the energy storage, being a
voltage storage device, is much higher. For example, if the LIC is charged to 4 V
ch001 7 June 2011; 13:4:47

positive relative to the negative electrode and then discharged to 2.0 V, the deli-
verable energy from a 4400 F LIC is
3
W LIC ¼ C cell U 2mx ¼ 0:375ð4400Þð42 Þ ¼ 26:4 kJ ð1:55Þ
8
The typical 3000 F, 2.7 V ultracapacitor has a deliverable energy, Wuc ¼ 8.2 kJ,
which means the LIC is storing 3.2 times the energy of the symmetric ultra-
capacitor. Exercise 1.7 expands on this more.
It is also known that the transfer of [Li þ ] ion from the graphite negative electrode
through its SEI layer requires formation of a solvation shell on the [Li þ ] ion that must
dissipate during charging to intercalate the ion back into the carbon material. This
process of solvation and de-solvation tends to be a rate limiter to mass transport in this
type of hybrid capacitor. There is also need for excess lithium pre-doping because the
formation process of this hybrid capacitor results in some irreversible capacity loss of
the negative electrode, but not so much so that performance is impaired.
More recently, Okamura [19] announced the launch of the Power Systems Co.
Nanogate capacitor, a KOH-activated carbon that does not rely on electrode pre-
doping, but retains the characteristics of the EDLC. This proprietary ‘electric field
activation’ or formation cycle, the Nanogate capacitor symmetric electrodes, is a
distinguishing characteristic. The cell operates to 3.9 V and is claimed to function
via surface adsorption pseudo-capacitance only, not intercalation pseudo-
capacitance. An issue with this product is the high pressure build up in soft pack
cells that is accompanied by considerable electrode swelling.
The hybrid capacitor battery commercialized by Axion Power is one of the
more interesting developments of the recent past. The capacitor battery starts with a
lead acid battery, retains the PbO2 positive electrode and sulfuric acid electrolyte,
but replaces the lead with activated carbon on the negative electrode. This elim-
inates one of the biggest cycling and rate inhibitors of the lead acid battery –
sulfation of the negative electrode. The benefits of the PbC include higher rate
performance at partial state-of-charge, less grid corrosion, and much higher cycling
performance. The PbC hybrid is finding interest in ground vehicles for engine
starting and in utility applications for grid energy storage. The hybrid capacitor
battery commercialized by Axion Power is one of the more interesting develop-
ments of the recent past [20].
Exercises
1.1 Follow the procedure of Example 1.1 and represent the functional relation for
C(u) given in Example 1.2 to show that this approximation yields very good
agreement with the resultant energy found in Example 1.2. The values of
Example 1.2 apply to the present case
CðuÞ ¼ C 0 þ k u u and qðuÞ ¼ C 0 u þ 2k u u2 ðEx: 1:1:1Þ
Ans: W f ¼ 12; 620 J:
ch001 7 June 2011; 13:4:47

Compare this result to that of the same ultracapacitor, but for a constant
value, Cavg ¼ 2770 F þ 520/2, so Cavg ¼ 3030 F.
C 2 3030 2
Wf ¼ U ¼ 2:7 ¼ 11; 044 J ðEx: 1:1:2Þ
2 mx 2
Ans: The constant capacitance case results in a field energy that is 12% lower
than the nonlinear capacitance that increases with potential case.
1.2 Follow the procedure of Example 1.3, using all the same parameters
and variables, only carry out the analysis for the case having the Warburg
impedance element placed in series with the equivalent model series
resistance, Rs. In this case, one would think of the Warburg impedance
acting mainly in the separator rather than directly influencing the ion
currents into and out of the activated layers as was done for Examples 1.1
and 1.2.
Ans: An exercise left to the reader for a deeper understanding of EC
capacitors.
1.3 Using the approach of Example 1.5 for the electrostatic force exerted by one
electrode to ion layer in the carbon–carbon ultracapacitor calculated in (1.31)
as 3.56 1026 N, compare this to the force of atmospheric pressure exerted on
the entire surface of planet earth. Take the mean radius of the earth as 6371
km and compute the total force.
Ans: Fion/Fair ¼ 6.9 106, or the atmosphere force of some 7 million earths.
1.4 For the surface density of atoms given in Example 1.7, and given that the
species being adsorbed has unity valence, calculate the surface density of
adsorbed charge.
Ans: 0.21 mC/cm2.
1.5 Apply (1.51) to the case of qc ¼ 0.22 mC/cm2, a monolayer of large molecule
species being adsorbed and find the maximum value of adsorption pseudo-
capacitance, max{CfL}.
Ans: 2.18 10–3 F/cm2 ¼ 2180 mF/cm2.
1.6 Compare the value of pseudo-capacitance density of Exercise 1.5 to that
given by (1.38) in Example 1.7 to the value cited for an Evans Company
hybrid capacitor.
Ans: 2180 mF/cm2, 4.96 mF/cm2, and 100,000 mF/cm2.
1.7 A lithium-capacitor hybrid is rated 4.0 V and 4400 F. Calculate the Coulomb
and the Ah storage of the EDLC positive electrode. Refer to Figure 1.20 for
charge and discharge characteristics.
Ans: 17,600 C and 4.9 Ah.
ch001 7 June 2011; 13:4:47

References
1. B.E. Conway, Electrochemical Capacitors: Their Nature, Function, and
Application, Chemistry Department, University of Ottawa, Ottawa, Ontario,
Canada. Available online
2. J.R. Miller, ‘A brief history of supercapacitors’, Battery and Energy Storage
Technology, History of Technology series, Autumn 2007 issue
3. R.A. Rightmire, Electrical energy storage apparatus, U.S. patent 3,288,641,
November 1966
4. J. Schindall, ‘Concept and status of nano-sculpted capacitor battery’, Pre-
sented at 16th Annual Seminar on Double Layer Capacitors and Hybrid
Energy Storage Devices, Deerfield Beach, Florida, 4–6 December 2006
5. M. Yoshio, R.J. Brodd, A. Kozawa, Lithium-ion Batteries Science and
Technologies, Springer Science, NY, 2009
6. J. Lario-Garcia, R. Pallas-Areny, ‘Constant-phase element identification in
conductivity sensors using a single square wave’, Sensors and Actuators, vol.
A132, pp. 122–8, 2006. Available at: www.sciencedirect.com
7. J.B. Goodenough, ‘Basic research needs for electric energy storage’, Report
of the DOE Basic Energy Sciences Workshop on Electrical Energy Storage,
2–4 April 2007
8. A. McBride, M. Kohonen, P. Attard, ‘The screening length of charge-asym-
metric electrolytes: a hypernetted chain calculation’, Journal of Chemical
Physics, vol. 109, no. 6, 1998
9. M.M. Kohonen, M.E. Kaaraman, R.M. Pashley, ‘Debye length in multivalent
electrolyte solutions’, Langmuir, vol. 16, pp. 5749–53, 2000 [a publication of
the American Chemical Society]
10. D.A. Evans, ‘Tantalum HybridTM cell capacitor’, Presented at 13th Annual
Seminar on Double Layer Capacitors and Hybrid Energy Storage Devices,
Deerfield Beach, Florida, 8–10 December 2003
11. I.N. Varakin, A.D. Klementov, S.V. Litvinenko, N.F. Starodubtsev, A.B.
Stepanov, ‘New ultracapacitors developed by JSC ESMA for various appli-
cations’, Presented at 8th Annual Seminar on Double Layer Capacitors and
Hybrid Energy Storage Devices, Deerfield Beach, Florida, 7–9 December
1998
12. A.I. Belyakov, ‘Asymmetric electrochemical supercapacitors with aqueous
electrolytes’, Presented at ESSCAP’08, Roma, Italy, 6–7 November 2008
13. A. Burke, ‘Ultracapacitor technology: present and future’, Presented at 13th
Annual Seminar on Double Layer Capacitors and Hybrid Energy Storage
Devices, Deerfield Beach, Florida, 8–10 December 2003
14. E. Gileadi, Electrode Kinetics for Chemists, Chemical Engineers and Mate-
rials Scientists, Wiley-VCH Inc., New York, NY, 1993
15. A.K. Shukla, S. Sampath, K. Vijayamohanan, ‘Electrochemical super-
capacitors: energy storage beyond batteries’, Current Science, vol. 79, no. 12,
25 December 2000
ch001 7 June 2011; 13:4:47

16. B.P. Bakhmatyuk, B.Ya. Venhryn, I.I. Grygorchak, M.M. Micov, S.I. Mudry,
‘Intercalating pseudo-capacitance in carbon systems of energy storage’,
Reviews on Advanced Material Science, vol. 14, pp. 151–6, 2007
17. R. Kotz, M. Hahn, O. Barberi, F. Campana, A. Foelske, A. Wursig, et al.,
‘Pseudo-capacitive processes and lifetime aspects of electrochemical double-
layer capacitors’, Presented at 15th Annual Seminar on Double Layer
Capacitors and Hybrid Energy Storage Devices, Deerfield Beach, Florida,
5–7 December 2005
18. O. Hatozaki, ‘Lithium ion capacitor (LIC)’, Presented at 16th Annual
Deerfield Beach, Florida, 4–6 December 2006
19. M. Okamura, K. Hayashi, T. Tanikawa, H. Ohta, ‘The Nanogate-capacitor
has finally been launched by our factory’, Presented at 17th Annual Seminar
on Double Layer Capacitors and Hybrid Energy Storage Devices, Deerfield
Beach, Florida, 10–12 December 2007
20. E. Buiel, ‘Axion Power International, Inc.’, ibid
ch001 7 June 2011; 13:4:47

Chapter 2
Ultracapacitor modeling
This chapter summarizes activities performed at Maxwell Technologies Inc. that

were aimed at providing equivalent circuit models of the carbon–carbon symmetric
ultracapacitor for use in computer simulations. In particular, these models are for
users of ANSYS Ansoft Simplorer v.81 and Mathlab Simulink v.10 simulation
software packages. These are behavioral-level models based on laboratory char-
acterization of production cells. Each model consists of three components: (i) cell
parasitic elements, (ii) electrode dynamic representation of ESRdc and ESRac, and
(iii) main branch EDLC, C(U), and parallel resistance Rp representing leakage.
Before moving to cell parasitic elements, the electrode and main branch
components of the model are appropriate to begin our discussion of the model.
Figure 2.1 illustrates the constituents of ESR and C(U) in the ultracapacitor where
Ra_e is the anode electrode electronic resistance, Ri the electrolyte resistance, and
Rk_e is the cathode electronic resistance.
In this view of the ultracapacitor cell, it is straightforward to see that ESR is a
composite of electronic and ionic elements and that capacitance is the resultant of
anode EDLC and cathode EDLC as shown in (2.1) and (2.2), respectively.
ESRdc ¼ Ra:c þ Ri þ Rke (2.1)
1 1 1
¼ þ ; where Ca U ¼ Ck U ¼ C dlc (2.2)
C cell C a U C k U
It is important to keep in mind the interpretation of (2.1) that each constituent

to the cell ESR is in turn the composite of several more basic resistive elements.
Each of the electrode electronic resistances is in turn composed of bulk and inter-
face resistances. In Figure 2.2 an exploded view of a typical ultracapacitor cell is
shown. The anode electronic resistance, for example, consists of the stud or jove
post termination bulk resistance shown as the ‘can’ in Figure 2.2. This resistance
element then connects via an interference fit to the bottom current collector lid and
its bulk and interface resistance to the can. The bottom current collector lid in turn
is backside welded to the jelly roll aluminum tab ends that are shown in Chapter 1.
The next element of anode electronic resistance is the resistance of the current
1
Models are available for download from www.ansoft.com for users of Simplorer version 8.0 software.
ch002 7 June 2011; 13:51:39

Ra_e Ca_U Ri Ck_U Rk_e
Figure 2.1 Illustration of cell model constituents [Courtesy of Maxwell

Technologies]
collector lid to tab end aluminum foil welds, the axial spreading resistance of the
current collector electrode foil and its interface resistance, via a special surface
treatment, to the carbon granules of the electrode film. The final component of the
anode electronic resistance is the bulk resistance of the activated carbon granules
and its particle-to-particle interface resistance mitigated by conductive carbon
black additive. The same can be said of the cathode electrode; only its resistance
includes a two-stage top current collector and, finally, electrode termination
end lid.
BCAP3000 cell
Top current
Bottom current collector
collector
Lid
Can
Figure 2.2 Internal components of the ultracapacitor cell [Courtesy of Maxwell

Technologies]
ch002 7 June 2011; 13:51:39

Ultracapacitor modeling 39
For the 3000 F ultracapacitor, it takes approximately 20 coconut shells to

provide sufficient quantity of high-quality activated carbon, roughly 140 g, to
support the necessary 12,000 F of internal capacitance. This capacitance according
to (2.2) is split evenly between the two electrodes shown graphically in Figure 2.3 as
activated carbon film coated aluminum foils. The ionic resistance Ri in Figure 2.1
consists of the electrolyte, the separator, and the ionic current channels through the
activated carbon (AC) particles. In a 100-mm carbon film, there are roughly
20 layers of AC and the electrolyte must wick through all the available channels
until it reaches from the electrode film to separator interface back to the aluminum
current collector foil. It is because of this tortuous path that ultracapacitors exhibit
more of a transmission line behavior than other types of capacitors. It is also the
reason for considering multiple time constant response that was initially reported
by Miller et al. [1] detailing the three-branch MIT model [2] and its Cauer I and
Foster II network equivalents. In these three time constant network representations
of the ultracapacitor, the time constants range from seconds through minutes to
days to capture both short- and long-term response behavior.
Aluminum foil tab end

Carbon electrode film
Paper separator
Opposite electrode
Figure 2.3 Ultracapacitor internal rolled electrode structure showing aluminum

foil tab ends
A great deal of activity also went into transmission line approach to ultra-
capacitor modeling, and these methods did generate useful and accurate models but
were difficult to scale to arbitrary packs of N-cell strings and M-string parallel
branches. The computing industry has applied moment methods to parameter
identification of RC ladder networks that could be applied to ultracapacitors and in
fact were applied and found useful. Figure 2.4 summarizes the three time constant
transmission line model where each of the cell level parameters Ruc1 to Ruc3 and
Cuc1 to Cuc3 are determined by the moment method in proportion to the total cell
ESRdc and capacitance Ccell.
ch002 7 June 2011; 13:51:41

Ruc3 Ruc2 Ruc1 Rconn Lconn
+UO +UO +UO

Cuc3(U) Cuc2(U) Cuc1(U)
Rsd
Ultracapacitor cell scaleable to module
Figure 2.4 Ladder network ultracapacitor model with parasitic elements
The parasitic elements Rconn and Lconn in Figure 2.4 are needed to model the
cell interconnects when individual cells are connected into series strings. The
connection resistance, Rconn, models either the bolted or welded interconnect strap,
and Lconn its inductance along with the cell inductance. More will be said of these
model parameters in the next section.
The model of Figure 2.4 was in fact developed and reported on Reference [3] in
collaboration with the Ansoft Corporation. Figure 2.5 illustrates this three time
constant moment matched model as it appears in Ansoft Simplorer along with
parasitic elements for the cell interconnects of the 3000 F cell. Note the inclusion of
nonlinear table lookup functions for each of the branch capacitances. The resistances
in the ladder are proportioned according to their moment-matched values normalized
to the cell ESRdc according to the technique summarized in Reference [4].
External test Interconnects & lnd.

apparatus Rtran1 Rtran2
T1 T2 Rtran3
+ # #
~ V C C1 C2 # Rsd
I1
VM1 XY
~ XY
~ XY
~
XY_LINT1 XY_LINT2 XY_LINT3
3000 F cell scaled to module level
Figure 2.5 Moment matched equivalent circuit model of the 3000 F

ultracapacitor cell
Another modeling approach is to approximate the electrochemical impedance

spectroscopy (EIS) frequency response based Nyquist behavior exhibiting Warburg
impedance as an equivalent N-interleaved RC circuits in the time domain. This
approach yields very acceptable results for the ultracapacitor response but requires
the computation of 2N parameters. In the typical case of N ¼ 10 parallel RC circuits
in series, this represents considerable computation and unwieldy model to deal
with. In the frequency domain for this same method, the authors in Reference [5]
ch002 7 June 2011; 13:51:42

show that only four parameters are necessary, which can then be inverse trans-
formed to the time domain.
Surewaard and Tiller [6] apply the methods of Buller et al. in their computer
modeling of ultracapacitor where a cascade of N ¼ 10 RC elements is necessary to
adequately model the pore impedance of the electrode structure and electrolyte
influence. The N-RC circuits are modeled in Dymola, and Simulink is used to obtain
the parameter values against measured cell test data. The authors then use as test case
the ultracapacitor as power boost source for an electric power assist steering appli-
cation (EPAS) and find acceptable model results versus test data. Kotz et al. [7]
expand on this by looking at the temperature behavior of ultracapacitors using EIS.
The material and method discussed by Funaki and Hikihara [8] provide useful
insights into the sensitivity of capacitance with potential. In this paper, the authors
explore a modified Randles circuit but without introduction of Warburg impedance
and show good agreement with measured capacitance versus potential. The
importance of this work here is that the two time constant model adequately pre-
dicts the charge–discharge performance of ultracapacitor cells and packs and that it
can also accurately calculate the charge stored in the pack.
Before moving on to the model development of this chapter, it is important for
the reader to understand that modeling ultracapacitors requires an appreciation for the
fact that porous electrodes exhibit multiple time constant behavior, that pore structure
contributes to heating during cycling, that frequency response requires more than just
single time constant behavior and, lastly, that the process of ordering chaotic ion
distributions during charge and subsequent reordering during discharge does show
entropy change. The fact that reversible heat generation occurs during charge–
discharge of the ultracapacitor is known from accurate calorimeter measurements.
Entropy added to the ions is taken from the rest of the cell and removed from the ions
and returned to the cell on reversal. The fact that measurable temperature change is
seen is a consequence of entropy change (i.e., cooling on discharge) and that this
process is entirely reversible is proven [9]. Discussion of entropy change in the
EDLC during charge and discharge is beyond the scope of this book.
Component rating often comes up during discussion of how such values are
selected, for example, commercial resistors, capacitors, and inductors that are
marked by either color band code or numerical code of rating. The answer to this
question is that such component rating values follow a preferred number sequence,
for example, E6 or E12 series, in much the same manner as wire gauge or drill size,
as a geometric progression known as the Renard series. The foundation for this
series is the following, where i ¼ (0, 1, 2, 3, . . . , b 1) and b ¼ base (3, 6, 10, 12,
24, . . . ), and ‘Round’ means round to the closest integer (1, 10, 100, etc.).
Rði; bÞ ¼ 10i=b (2.3)
Rði; 10Þ ¼ Roundf10i=10 g ¼ 1; 2; 5; . . . (2.4)
For electronic applications, the base is generally b ¼ 6 or b ¼ 12 and so on. As an

example, consider the case E12 where b ¼ 12 and i ¼ {0, 1, 2, . . . , 11}, and start the
ch002 7 June 2011; 13:51:42

first value at 10. In this case, we get the familiar range of resistor (or capacitor) values
listed in Table 2.1 by multiplying R(i, 12) in (2.4) by 10 to obtain the E12 series
starting with value 10 – similarly for cases starting with 100 or 1000 and so on.
Table 2.1 Preferred values series for electronic components
E3 10 22 47
E6 10 15 22 33 47 68
E12 10 12 15 18 22 27 33 39 47 56 68 82
2.1 Electronic equivalent circuit model

The ultracapacitor equivalent circuit model developed and deployed for applica-
tions engineering use is shown as Figure 2.6. Unlike the moment-matched model of
Figure 2.5, this equivalent circuit model is capable of accurately capturing time and
frequency response behaviors of the symmetric carbon–carbon ultracapacitor. The
three main elements of the model noted earlier are clearly apparent and account for
parasitic (Rconn, Lconn), electrode dynamics (Rs, Cs, Rsa), and main branch energy
storage (C0, Uc) and charge leakage (Rp).
Rsa Cs
Rconn Lconn Rs
Uc
TLU: C(Uc) C0
Rp
Ultracapacitor cell with parasitics
Figure 2.6 Equivalent circuit model of the carbon–carbon ultracapacitor
The model development starts with identification of parasitic elements and pro-
ceeds with their extraction – in this case the cell interconnect resistance and inductance.
Ansoft Q3D Extractor is a 3D/2D quasi-static electromagnetic field simulation for
parasitic extraction of electronic components that relies on a method of moments and
finite element approach to compute the parameters of a structure. The tool also auto-
matically generates an equivalent subcircuit at nominal frequency or S-parameters for a
band of frequencies. This tool was used to assess the geometry of a common cell
interconnect, the aluminum strap shown graphically in Figure 2.7 and its Q3D para-
meter extraction of Rconn and Lconn in Figure 2.8. The interconnect strap holes lie along
the cell terminal post centerlines and also provide necessary cell-to-cell clearance dis-
tance for thermal and high-voltage considerations.
ch002 7 June 2011; 13:51:43

JAC (A/M)
4.1851e + 001
3.9237e + 001
3.6623e + 001
3.4009e + 001
3.1395e + 001
2.8781e + 001
2.6167e + 001
2.3553e + 001
2.0939e + 001
1.8325e + 001
1.5711e + 001
1.3097e + 001
1.0483e + 001
7.8695e + 000
5.2556e + 000
2.6416e + 000
2.7662e – 001
Figure 2.7 Cell interconnect strap geometry (left) and current distribution (right)
Ansoft corporation XY plot 2 Q3D Design strap

350.00
ACR (strap:source, strap:source), (μΩ)
Curve info
300.00 ACR (strap:source, strap:source)
Setup1: Sweep1
250.00
200.00
150.00
100.00
50.00
0
1.00 10.00 100.00 1000.00
Freq (kHz)
Ansoft corporation XY plot 1 Q3D Design strap

22.00
ACL (strap:source, strap:source), (nH)
Curve info
21.50 ACR (strap:source, strap:source)
Setup1: Sweep1
21.00
20.50
20.00
19.50
19.00
1.00 10.00 100.00 1000.00
Freq (kHz)
Figure 2.8 Interconnect parasitic resistance and inductance. Top: Rconn (35 mW at
1 kHz), bottom; Lconn (20.7 nH at 1 kHz)
ch002 7 June 2011; 13:51:43

In the complete model, parameter Lconn also contains the cell inductance that is
derived from the cell internal spiral wound jelly roll geometry approximated as a
set of concentric cylinders depicted in Figure 2.9.
rn
r
ri – 1
ri
r0
Figure 2.9 Cell inductance as function of radius r and length h
For the 3000 F cell, n ¼ 42 wraps of electrode foil, which are modeled in
Figure 2.9 as 21 concentric cylinder pairs with one group of 21 being anode elec-
trode and the second group of 21 being the cathode electrode. The current during
constant current steady state that enters one electrode will therefore split among the
21 concentric cylinders according to their length relative to total electrode length.
This means that the summation for electrode length, le, in (2.5) sums to a total
length L ¼ le meters of electrode.
42
rk ¼ r0 þ Dk; k ¼ 1!n (2.5)
X
n
lk ¼ 2prk ; le ¼ lk (2.6)
k¼1
lk 2pI 0
Ik ¼ I0 ¼ ðr0 þ DkÞ (2.7)
le le
The current in each cylinder will therefore proportion itself according to each
cylinder’s circumference relative to the total electrode length as noted in (2.7). This
effect is shown graphically in Figure 2.10 for a 3000 F cell having constant current
charge of I0 ¼ 200 A.
The magnetic flux due to a current in shell Ik ¼ (lk/le)I0 is given by (2.7a), an
expression that can be worked out by the application of Ampere’s and Gauss’s laws
ch002 7 June 2011; 13:51:43

3000-P ultracapacitor current distribution

by Shell # (at 200 A)
18
16
14
Current (A)
12
10
8
6
4
2
0
0 5 10 15 20 25 30 35 40 45
Shell #
Figure 2.10 Current by cylinder (or shell) of geometry in Figure 2.9
of electromagnetism for this geometry and the assumptions made regarding the
cylindrical shell approximation of a spiral winding.

m0 hl0 lk1 rkþ1
fk ¼ ln 2.7a
2p le rk1
The rationale for (2.7a) is that current flowing axially down the electrode
current collector foil can be treated as a current sheet, Ik 1, having magnitude of
the proportion of the inner shell circumference to the total circumference times the
total current, I0. A further approximation is that the adjacent shell, rk, collects this
current as the opposing electrode would. The flux, therefore, is taken to reside in
the annular volume between rk 1 and rk þ 1. These approximations simplify the
situation to make the calculations tractable. The flux of (2.7) divided by the current
producing it is the incremental self-inductance of that shell, Lk, given by (2.8).

m0 h rkþ1
Lk ¼ ln (2.8)
2p rk1
Ignoring mutual inductance between shells for the present, (2.8) can be sim-
plified by applying (2.5) and expanding the natural logarithm into its Taylor series
and neglecting higher-order terms.

m0 h 2D m0 h 2D
Lk ffi ln 1 þ (2.9)
2p rk1 2p rk1
The equivalent inductance is obtained by noting that the shell self-inductances

of (2.9) combine reciprocally as (2.10).
1
1
Leq ¼ (2.10)
L1 þ Ls þ þ L21
ch002 7 June 2011; 13:51:43

Example 2.1: The Maxwell Technologies 3000 F cell has the following approx-
imate parameters that will be used to evaluate its self-inductance using (2.9) and
(2.10): shell radii, r0 ¼ 4.897 mm, D ¼ 0.574 mm, h ¼ 0.11 m, and le * 2.25 m.
Calculate the equivalent inductance of the cell jelly roll and sum this to the value
obtained from Ansoft Q3D and given in Figure 2.8. Table 2.2 summarizes the steps.
Table 2.2 Inductance calculation by method of shells
Shell number Shell length Shell current Shell inductance Reciprocal inductance
0 30.786 5.111 509E08 1.97Eþ07
3 52.407 8.705 3.76E08 2.66Eþ07
5 66.832 11.102 3.21E08 3.12Eþ07
7 81.258 13.498 2.79E08 3.58Eþ07
9 95.684 15.894 2.48E08 4.04Eþ07
11 110.110 18.291 2.22E08 4.50Eþ07
13 124.535 20.687 2.02E08 4.96Eþ07
15 138.961 23.083 1.84E08 5.42Eþ07
17 153.387 25.480 1.70E08 5.88Eþ07
19 167.813 27.876 1.58E08 6.34Eþ07
21 182.238 30.272 1.47E08 6.80Eþ07
Sum: 1203.992 200.000 4.93Eþ08
Le ¼ 2407.985 Leq ¼ 2.03E09
Le, length; Leq, equivalent inductance.
Using this method gives a self-inductance value for the 3000 F cell as 2 nH.
Adding this to the value given in Figure 2.8 yields L3000 ¼ 22.7 nH.
This would seem a relatively small value of inductance compared with the
interconnect strap until one compares the self-inductance of embedded cylinders
with that of a precision leaf shunt that achieves very low inductance by splitting the
current into multiple shunt paths.
Benchmark values on inductance can be obtained by application of the
inductance of a thin cylindrical (shell) conductor for which the inductance is

m0 h 2h 3
Lthin shell ¼ ln (2.11)
2p rod 4
Substituting the values from Example 2.1 into (2.11) results in a shell induc-
tance of 11.6 nH. This may seem a more plausible approach until one compares the
self-inductance calculated by (2.11) with row 21 in Table 2.2 showing an outer
shell self-inductance of 14.7 nH. Because of this finding, the approach taken lead-
ing to (2.9) is deemed sufficiently accurate for the modeling purposes of the jelly
roll construction ultracapacitor.
The more discerning reader may wish to challenge this and compare these
results with more classical methods of inductance calculation, for example, the
ch002 7 June 2011; 13:51:45

self-inductance of a planar spiral such as the one used in printed circuit board
fabrication of inductances. For the planar spiral, the self-inductance will under-
estimate that of the jelly roll because in the planar spiral, the full current flows
through the entire spiral rather than only fractions of the total current as one sees in
the actual cylindrical ultracapacitor. Equation (2.12) is the self-inductance of a
planar spiral, and for the parameters listed in Example 2.1, this calculates out to
0.88 nH.
Lspiral ¼ 1:748 105 m0 pN 2 rod (2.12)
Another approach is to check this against the Wheeler formula for a solenoid
by approximating the jelly roll as a solenoidal winding approximation of the
42-layer jelly roll corresponding to N ¼ 21 turns and with radius rsol ¼ rod/2, where
rod ¼ 29 mm for the case given in Example 2.1. This method is listed here as (2.13),
and for the same parameters as above, it calculates to 0.29 nH, again significantly
underestimating the total inductance.
10pm0 N 2 r2od
Lsolenoid ¼ (2.13)
9rod þ 10h
Table 2.3 highlights the parameters of Maxwell Technologies line of large cell
products, ESR, capacitance, time constant, and inductance values. The thermal
parameters listed will be discussed shortly.
The next cell elements to consider deal with the electrode kinetics, Rs, Cs, and
Rsa where Rs ¼ ESRdc values specified in Table 2.3, and the lead network RC
values are computed as follows:
ESRdc ESRac
Rsa ¼ (2.14)
ESRdc ESRac
The value of Rsa ensures that under transient conditions, the model dynamic
performance approximates the real performance of the cell modeled. For any of the
ultracapacitor products listed in Table 2.3, the values for ESRdc and ESRac are used
to compute (2.14) for their model. In the event that ESRac is not specified or
omitted on a data sheet, its value can be approximated as 70% of ESRdc. Generally,
ESRac is taken as the Z0 resistance at the corner point on a Nyquist plot or the value
at f ¼ 1 kHz. To visualize this, the Nyquist and Bode plots for the equivalent circuit
model of the 3000 F ultracapacitor are shown in Figure 2.11 with frequency marker
points on the Nyquist plot over the range 10 mHz to 100 Hz. Note that the Nyquist
plot starts with a near vertical line at 10 mHz, representing the capacitive reactance
of the main branch 3000 F capacitance. This line intersects the real axis, Z0 , at
ESRdc and with increasing frequency trends to Z0 ¼ ESRac at f ¼ 100 Hz at the very
left on the chart.
ch002 7 June 2011; 13:51:45

Table 2.3 Parameter values of a line of ultracapacitor products [courtesy of Maxwell Technologies]
Cell
60 mm
60 mm 60 mm 60 mm
25 mm 33 mm 60 mm
ch002
Capacitance (F) 150 350 650 1200 1500 2000 3000
ESRac (mW) 8 2.2 0.60 0.44 0.35 0.26 0.20
ESRdc (mW) 14 3.2 0.8 0.58 0.47 0.35 0.30
Laxial (mm) 50 61 51.5 74 85 102 138
Elec h (mm) 22 33 24 46 57 74 110
Lconn (nH) 20 20 21.1 21.6 21.8 22.1 22.7
Icont (Arms) 9.1 22 54 70 84 106 127
Mass (kg) 0.035 0.063 0.20 0.30 0.32 0.40 0.545
7 June 2011; 13:51:45

Volume (dm3) 0.025 0.053 0.211 0.294 0.325 0.393 0.475
Rth ( C/W) 17.3 10.9 6.5 6 5.6 4.6 3.2
Cth (J/ C) 32 53 188 294 316 408 588
Temperature, OP 40–65 C 40–65 C 40–5 C 40–65 C 40–65 C 40–65 C 40–65 C
Temperature, Sto 40–70 C 40–70 C 40–70 C 40–70 C 40–70 C 40–70 C 40–70 C
Grey text values ¼ estimates.

Elec, electrode film width, h.; OP, operating; Sto, storage temperatures.
Im 0.20 m Nyquplotsel 0.29 m
5.31 m 5.31 m
10 m
20 m
40
m
79
0.16
m
0.32
0.03
0.23
0.1
0 0.00 m
0.20 m 0.29 m Real

Bodeplotsel
Gain
–48.00 –48.00
VM1.V
–56.00 –56.00
dB
–64.00 –64.00
–72.00 –72.00
10 m 0.10 1 10 100
f (Hz)
Phase
0 0
–18.00 –18.00
deg
–36.00 –36.00
–54.00 –54.00
–72.00 –72.00
–90.00 –90.00
10 m 0.10 1 10 100
f (Hz)
Figure 2.11 Nyquist (top) and Bode plots (bottom) for the 3000 F ultracapacitor
model
The Bode plot is amenable to locating the approximate corner point frequency, fc,
where the gain function slope intersects the horizontal portion of the magnitude plot.
This point occurs at roughly 45 , and the frequency is approximately 0.35 Hz. The value
for Cs used in the simulation that generates Figure 2.11 was calculated as follows:
1
wc ¼ ¼ 2pf c jf c ¼0:35 Hz (2.15)
tca
ch002 7 June 2011; 13:51:47

1 1
¼ (2.16)
tca ðRs þ Rsa ÞC s
0:455
Cs ¼ (2.17)
Rs þ Rsa
The frequency in (2.15) approximates the corner frequency in the 3000 F

cell Bode response. This corner point frequency in turn defines the time constant of
the electrode kinetics branch. For the 3000 F cell, Cs ¼ 511 F, which should corre-
spond to the rapidly accessible surface area of the electrode double-layer
capacitance.
The parallel resistance, Rp, to model leakage and self-discharge behavior is
modeled next. This parameter can be modeled using either of the two methods.
First method involves calculating the parallel resistance that corresponds to a vol-
tage decay of 4% of Umx when T ¼ 72 h and its corresponding energy loss in the
ultracapacitor. This is the most direct approach and is derived from actual cell
performance during self-discharge testing at room temperature. For the 3000 F cell,
using this method produces a parallel resistance of 2117 W (2.19) when the incre-
mental energy loss of the cell (2.18) is 857 J.
1 C 0 U 2mx
dE ¼ C 0 ðU 2mx ð0:96U mx Þ2 Þ ¼ ð1 0:962 Þ (2.18)
2 3

ðdU mx Þ2 T T¼72 h ð0:98 2:7Þ2 ð72 3600Þ
Rp ¼ ¼ ¼ 2117 W (2.19)
dE d¼0:98 857
The second method is based on an electrode metric of 2 mA/F of leakage that,

when modeled, yields a value for the parallel resistance of 450 W as noted in (2.20).

U mx 2:7 106
Rp ¼ ¼ ¼ 450 W (2.20)
I so C 0 I so ¼2 mA=F 2ð3000Þ
The dilemma posed by (2.19) and (2.20) is typical of modeling electrochemical

energy storage components. In this case the parallel resistance values are sig-
nificantly different because in the first approach ((2.18) and (2.19)), the evaluation
criteria is set 3 days after the cell is charged and allowed to exhibit natural decay.
This means that the initial value of leakage can be much higher than what it trends
to after several days. The second method yields a much lower parallel resistance
value because the evaluation criteria are applied for a new production cell that is
not exposed to voltage stress for any appreciable length of time.
Experiments on natural decay and leakage testing have been performed to
better characterize this parameter, and the finding is that Rp must also be made a
ch002 7 June 2011; 13:51:47

function of voltage and temperature exposure if the model is to be valid for

long-term voltage bias. Figures 2.12 and 2.13 highlight the distinction between
leakage, or voltage-pinned self-discharge and natural decay. In a leakage test,
the cell is charged to rated potential under constant current charge, then held at
this voltage under constant voltage control. The result is a current that expo-
nentially decays to near zero with time. Under constant voltage bias, the leakage
current eventually becomes nearly linearly decaying with time. A similar beha-
vior is found for natural decay, or self-discharge, where the cell is charged
to rated voltage under constant current, then allowed to discharge naturally
under open circuit conditions with only periodic or a single measurement at
T ¼ 72 h.
Ultracapacitor leakage current and function fit 25 F

1 × 10–3
1 × 10–3
8 × 10–4
6 × 10–4
I.lk
I.est
4 × 10–4
2 × 10–4
5.364 × 10–5
0
0 20 40 60 80
0 T.lk 73
Time (h)
Figure 2.12 Leakage behavior of the 25-F, 2.5-V ultracapacitor data versus
modeled response
Notice in Figure 2.12 that between 10 and 20 h under constant voltage expo-
sure, the ultracapacitor leakage current reduces from 1 mA to less than 15 mA, and
its decay thereafter is virtually linear.
Natural decay behavior is very similar to leakage current after 10 h as seen in
Figure 2.13, and the very linear behavior versus time is apparent. Leakage is
sometimes represented as a family of straight lines versus square root of time as
shown in Figure 2.14. This is characteristic of electrochemical phenomena that are
activation energy (redox) dependent. If the processes are dominated by diffusion
processes, the plots would be linear versus logarithm time as self-discharge shown
in Figure 2.15.
ch002 7 June 2011; 13:51:47

2.8 Self-discharge data and function fit

2.71
2.7
2.6
Voltage (V)
2.5
2.4
2.3 U.sd
2.216 U.est
2.2
0 20 40 60 80
0 T.sd 73
Time (h)
Figure 2.13 Natural decay behavior of the 25-F, 2.5-V ultracapacitor data versus
modeled response
100%
–40°C
80%
RT
60%
+65°C
40%
0 1 2 3 4 5
Time½ (days½ )
Figure 2.14 Characteristic of leakage current with temperature as a parameter
To illustrate the diffusion-limited behavior of natural decay, Figure 2.13 is

replotted versus log(time). It is seen in Figure 2.15 that self-discharge is indeed
linear versus log(time), which means that its functional representation is based on
exponentials.
The final element of the ultracapacitor model to be discussed is the main
branch double-layer capacitance, C0, which is both potential and current dependent.
ch002 7 June 2011; 13:51:48

Self-discharge data and function fit 25 F

2.8
2.701
2.7
2.6
Voltage (V)
2.5
2.4
2.3 U.sd
U.est
2.216
2.2
1 10 100
1 T.sd 73
Time (h)
Figure 2.15 Replot of self-discharge data versus log(time) where the grey scale
trace is measured data
Variability of C0(U) with potential is a first-order effect, whereas variability with

current, C0(U, I), is a second-order effect and many times can be omitted. Double-
layer capacitance C0(U) can be modeled either as a table lookup (TLU) function
based on measurements taken during cell characterization or as a functional rela-
tionship, C0(U) ¼ C0 þ kuUc, where Uc is the potential across the capacitor. The
more refined functional relation for C0(U) discussed in Chapter 1 and repeated here
as (2.21) is more appropriate for modeling use:

Uc
C 0 ðUÞ ¼ C a þ C b T anh Ux (2.21)
Ux
where Ux is the potential at the inflection point of the Tanh() term to fit the char-
acter of the measured capacitance variation, Cdn. Accepted practice in the industry
is to use the value of capacitance calculated during discharge, or downward value
Cdn, rather than capacitance measured during charge, or upward value Cup. Para-
meters fitting the 3000 F cell are listed in Table 2.4, and the functional fit to
measured data is shown graphically in Figure 2.16.
Table 2.4 Functional fit to double layer capacitance C0(U)
Intrinsic capacitance (F) Differential capacitance (F) Inflection point voltage (V)
Ca ¼ 2770 Cb ¼ 520 Ux ¼ 0.9
ch002 7 June 2011; 13:51:48

C(Uc) data and approximation of f(Uc)

3500
3000
Capacitance
2500
2000
1500
1000
500
0
0 0.5 1 1.5 2 2.5 3
Voltage
Figure 2.16 Functional fit C0( U) to measured data of 3000 F cell
The fact that C0(U, I) variability with current is second order for modeling
purposes is shown clearly by Figure 2.17, a plot of Cdn versus potential with current
as parameter. In this plot, the functional characteristic given by (2.21) and shown
graphically in Figure 2.16 is evident. In this graphic based on measured data for the
3000 F cell, the potential varies from 0 to 2.7 V and current from 10 to 500 A.
It is insightful to normalize ultracapacitor cell current to capacity to obtain its
carbon loading in terms of a mA/F metric. In this case 10 A/3000 F ¼ 3.33 mA/F,
100 A/3000 F ¼ 33.3 mA/F, and 500 A/3000 F ¼ 167 mA/F. In ultracapacitor appli-
cations, carbon loading of 70 mA/F is typical, and exposure to greater than 500 mA/F
enters the range of extreme carbon loading. Exercises 2.2–2.4 explore these topics
in more detail.
10 A 30 A 60 A 100 A 300 A 500 A

20 A 40 A 75 A 200 A 400 A
BCAP3000 Cdn versus voltage
3400
3200
3000
2800
2600
2400
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Figure 2.17 Variation of C0( U, I ) with current as parameter for 3000 F cell
Thermal considerations of EDLC capacitance have been discussed previously,

which led to the conclusion that C0(U, I, T) is influenced by temperature only
slightly and even then limited to very cold operating temperatures. But ESRdc is
strongly influenced by temperature and is essentially ESRdc (T) to be properly
modeled. The temperature-dependent functionality of ESRdc (T) is repeated here
ch002 7 June 2011; 13:51:48

ESR(T ) for 650 F cell

2 × 10–3
1.6 × 10–3
1.6 × 10–3
1.2 × 10–3
Ω
8 × 10–4 R.ESR(i)
4× 10–4 ESR.650
0
–50 –20 10 40 70 –100
–40 a (i), x 70
Temperature (°C)
Figure 2.18 Fit of (2.22) to the data for a 650 F cell
ESR(T ) for 3000 F cell

6 × 10–4
5.4 × 10–4
4.8 × 10–4
3.6 × 10–4
Ω
2.4 × 10–4
R.ESR(i)
1.2 × 10–4 ESR.3000
0
–50 –20 10 40 70 100
–40 a (i), x 70
Temperature (°C)
Figure 2.19 Fit of (2.22) to the data for a 3000 F cell
for completeness along with the functional fit to measured data for the 650 F and
3000 F cells (Figures 2.18 and 2.19 and Table 2.3).
ESRdc ðT Þ ¼ b1 ESRdc ðT 0 Þf1 þ gðT T 0 Þg

þ b2 ESRdc ðTÞfeðk T ðT T 0 Þ=2Þ g ð2:22Þ
where b1 and b2 are the electronic and ionic weight factors discussed earlier,
g ¼ 0.007 is the coefficient of resistance sensitivity to temperature, and kT ¼ 0.045
represents the Arrhenius factor of ionic temperature dependence. The factors for b1
and b2 are summarized in Table 2.5.
High-volume manufacturing of ultracapacitors requires stringent statistical
process control (SPC) so that product variability is maintained within set criteria.
ESRdc, for example, of a production sample must not exceed the data sheet speci-
fied maximum value. This means that cells will be shipped with ESRdc lower than
the specified value as shown in Figure 2.20. Similar is the case for capacitance, but
in this case the production sample must exhibit capacitances greater than the spe-
cified capacitance minimum as shown in Figure 2.20.
ch002 7 June 2011; 13:51:48

Table 2.5 Electronic b1 and ionic b2 weight factors for the

functional fit of the 650 F and 3000 F cells
Coefficient/parameter MC650 cell MC3000 cell

b1 0.65 0.55
b2 0.35 0.45
ESR(T0) 0.8 mW 0.28 mW
Cap freq 3000 P cell

250
Frequency
200
150
100
50
0
00
25
50
75
00
25
50
75
00
25
50
75
00
25
50
75
00
e
or
30
30
30
30
31
31
31
31
32
32
32
32
33
33
33
33
34
M
Capacity specification
3000 P ESR
200
180
Frequency
160
140
120
100
80
60
40
20
0
28
0
20
21
22
23
24
25
26
31
32
33
34
35
e
27
29
or
0.3
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
ESR specification
Figure 2.20 Illustration of quality control of production capacitance and ESRdc of

3000 F cells [Courtesy of Maxwell Technologies]
ch002 7 June 2011; 13:51:49

Example 2.2: Approximate the mean and standard deviation of the capacitance and
ESRdc shown in Figure 2.20 for the 3000 F ultracapacitor cell. Recall the cumula-
tive probability function and its density function shown here and apply (2.24) to the
data extracted. Note that nearly 100% (99.73%) of the distribution is contained
within (m 3s) in each of the charts to be constructed.
ðx
1 2
eðxmÞ =2s dx
2
Pðx xÞ ¼ FðxÞ ¼ pffiffiffiffiffiffiffiffi (2.23)
2ps 1
1 2
f ðxÞ ¼ pffiffiffiffiffiffiffiffi eððxmÞ =2s Þ
2
(2.24)
2ps
The solution to this example is given as a pair of charts, one for capacitance
and one for ESRdc of the 3000 F production cell (Figures 2.21 and 2.22).
Probability density of 3000 F production cell mm = 3117, sd = 27

0.4
0.369
0.3
Frequency
0.2
f(I)
x
0.1
0
0
3 × 103 3.05 × 103 3.1 × 103 3.15 × 103 3.2 × 103
3000 C(i), cap 3200
Capacitance (F)
Figure 2.21 Example 2.2 fit of capacitance data by the Gaussian function of
(2.24) (m ¼ 3117 F and s ¼ 27 F)
Probability density of 3000 F production cell mn = 0.25, sd = 0.014
0.5
0.4
Frequency
0.3
y
0.2 f.R(i)
y
0.1
0
0
0.2 0.22 0.24 0.26 0.28
0.20 R(i), ESR 0.3
Cell resistance (mΩ)
Figure 2.22 Example 2.2 fit of ESRdc data by the Gaussian function of (2.24)
(m ¼ 25 mW and s ¼ 0.014 mW)
ch002 7 June 2011; 13:51:49

In Example 2.2, the ESRdc data is more difficult to characterize unless

much higher sample sizes are used. In this example, the lot size was nearly 600
cells. For instance, the standard deviation of the 3000 F cell ESRdc is less than 6%
of the mean, a very tight distribution. The capacitance distribution is even tighter
since the standard deviation of the 3000 F cell capacitance is less than 1% of
the mean.
2.2 Cell characterization methods and standards

The equivalent circuit model of Figure 2.6 is now complete and ready to be used in
simulations. Parameter values for the components, obtained from cell character-
ization testing, are summarized in Table 2.6 for convenience and for later use in
system application simulations.
Table 2.6 Ultracapacitor equivalent circuit model parameter list
Capacitance (F) 150 350 650 1200 1500 2000 3000

Rconn (mW) 70 mW 70 70 70 70 70 70
Lconn (nH) 20 20 21.1 21.6 21.8 22.1 22.7
Rs (mW) 7.2 3.2 0.8 0.58 0.47 0.35 0.30
Rsa (mW) 7.200 3.200 2.400 1.823 1.371 1.011 0.600
Cs (F) 31.528 70.938 141.875 188.942 246.627 333.551 504.444
Rp (W) 13785.714 5514.286 2969.231 1608.333 1286.667 965.000 643.333
Voltage C(Uc) ¼ C(Uc) ¼ C(Uc) ¼ C(Uc) ¼ C(Uc) ¼ C(Uc) ¼ C(Uc) ¼
0.00 255.9 527.9 976.4 1199.9 1614.3 2415.5
0.50 280.1 570.8 1045.1 1303.8 1746.2 2604.8
1.00 311.8 632.4 1152.8 1437.4 1927.9 2877.3
1.50 339 688.4 1249.9 1559.8 2090.5 3122.7
2.00 355.3 722.8 1309.4 1640.8 2190.7 3288.3
2.50 360.4 728.4 1327.1 1661.1 2210.5 3378.9
2.80 360.1 718 1324.4 1640.1 2185.6 3420.7
C0 265.8 546.8 1001.9 1242.7 1671.6 2465.2
ku 39 73.9 133.2 169.3 219.8 374
The equivalent circuit of Figure 2.6 and simulation model in Ansoft Simplorer
is shown in Figure 2.23 with a six-step excitation waveform of current applied. This
current waveform is an accepted method of characterizing ultracapacitor cells and
includes provision for cell conditioning of nominally 6 h at room temperature to
equilibrate the cell to zero charge and equilibrium temperature.
Cell characterization methods are well established and published by several
standard setting and independent testing organizations such as IEC, EUCAR, and
others. The remainder of this section covers the more prominent characterization
methods used to extract capacity and ESRdc from test cells. Figure 2.24 illustrates
the application of IEC62391 [10,11] to the characterization of capacity and ESRdc.
ch002 7 June 2011; 13:51:51

BCAP3000 P cell six-step validation

of Li test data
Rsa Cs
Probe
200.00 − +
0.60 mΩ 504.4 F
0 DATA Rcon Lcon
–200.00 Rs
0 100.00 200.00 70 PΩ 20.7 nH
DATAPAIRS1 0.30 mΩ
+
t y
I1 V VM1 C0 − Rp
−
~
– 2174 + 431.2*C0.VF
643 Ω
Six-step cell voltage measured response

3.0
2.5
2.0
Voltage (V)
1.5
1.0
0.5
0
0 20 40 60 80 100 120 140 160 180 200
Time (s)
Figure 2.23 Equivalent circuit simulation model with applied excitation and
measured response of three cells (two of three cells are biased
slightly above zero)
In Figure 2.24 the cell under test is charged under constant current that yields
95% efficiency or better, the hold period according to IEC62391 is set for 30 min,
and the discharge is again the same level of constant current. Measurements during
the discharge interval, as noted in Figure 2.24 (top), include the voltage at specified
points, the current (which is regulated by the test equipment), and the time points at
each voltage measurement. On the basis of these measurements, the ESRdc and
capacitance per IEC method are given by (2.25) and (2.26), respectively. Cell
ESRdc is taken as the voltage step from the cell terminal voltage during open circuit
hold to the point at which a back projection of the discharge trajectory intersects the
vertical step in potential. This measure is taken in order to account for ionic
redistribution during the discharge interval that is the Rionic component of ESRdc.
dU 2
ESRdc ¼ (2.25)
1
ch002 7 June 2011; 13:51:51

Hold δU1
0.8 Umx δU2
Charge Discharge
0.4 Umx
T1
Terminal volts
2.85
VM1.V (V)
2.00
890.00 m
0 10 20 30
Figure 2.24 IEC characterization method for EDLCs. Top: Illustration of

measurement points. Bottom: Simulated discharge waveform
IT 1
C dn ¼ (2.26)
ð0:8U mx 0:4U mx Þ
The cell capacitance Cdn is calculated from the packet of charge, IT1, where
time T1 is accurately known from the test equipment applying the discharge cur-
rent. The voltage measurements, according to IEC62391, are specified as 80% to
40% of maximum cell potential. Note at this point that these voltage measurement
points will differ depending on what standard setting organization has released the
standard as we will see shortly. As pointed out earlier, the accepted convention is
to specify cell capacitance as the value calculated from the discharge condition.
Figure 2.25 helps clarify the IEC62391 standard by illustrating in simulation the
back projection technique and also how the model predicts the ionic tail occurring
after a change in current magnitude and the fact that C(U) is not a linear function
but more parabolic over this interval.
ch002 7 June 2011; 13:51:53

Terminal volts
2.85
IT1
C(U) = (U = Umx )
δU1 mx
VM1.V (V)
2.00
δU1
IT1
890.00 m
0 10.00 20.00 30.00
Terminal volts
2.85 δU2
δU2
ESRdc =
I
VM1.V (V)
2.60
2.40
2.31
1.40 2.00 4.00 6.49
Figure 2.25 Clarification of the IEC62391 back-projection technique
2.2.1 EUCAR method

The EUCAR method differs from the IEC approach in two ways: (i) the cell potential
is held at constant voltage during the hold, or dwell time, by the test equipment and
(ii) the voltage measurement points during cell discharge are modified. The proce-
dure is specified for room temperature conditions following cell conditioning and
initialization procedures (e.g., short-circuit and hold for 12 h at room temperature)
and with current magnitudes as in IEC62391 of <50 mA/F carbon loading for
charging and approximately 5 mA/F carbon loading during discharge.
ch002 7 June 2011; 13:51:53

Figure 2.26, extracted from Reference [12], illustrates the constant current
charge at 50 mA/F, followed by constant voltage hold for 30 s, then constant current
discharge at 5 mA/F. The voltage and time measurement points are specified on the
chart (Note: Rwv ¼ rated working voltage ¼ Umx).
Rwv
Voltage (V)
V1
V2
T1 T2
Time (s)
Determination of the capacitance value:
V1 = 0.6*Rwv dQ Idisch* (T2 – T1)

V2 = 0.3*Rwv C= =
dV V1 – V2)
LdiDLch = 5 mA/F
Figure 2.26 EUCAR test method for capacitance and ESR
ESRdc in the EUCAR method is the same as described for IEC62391, but
capacitance measurement is taken at 60% to 30% of rated voltage. The discharge
current is also much lower than that specified by IEC methods. The calculation of
capacitance for the stated measurement points is shown in Figure 2.26. Before
moving on to the IEC62576 test and characterization standard, the introduction of
two examples will help to quantify how charge and discharge current magnitudes
are selected and how limitations of data acquisition can influence the accuracy of
the measurements, especially the measurement of small voltages, of the order of
millivolts, during the calculation of ESRdc.
Example 2.3: An EDLC characterization procedure calls for application of con-

stant current that results in 95% or greater discharge efficiency (or charge effi-
ciency). Derive the magnitude of constant current and carbon loading that results in
95% cell efficiency in terms of the cell parameters and current application time, or
pulse time T for the 3000 F cell. For this cell, ESRdc ¼ 0.29 mW.
ch002 7 June 2011; 13:51:53

The method is best described by the following set of expressions, starting with
the definition of capacitor charge and energy efficiency, hd, on discharge [13].
Derivation of (2.28) is left as an exercise to the reader.
Q ¼ IT ¼ CU (2.27)
t
hd ¼ 1 2 (2.28)
T
2ESRdc C 1
ð1 hd Þjh¼0:95 ¼ ¼ (2.29)
T 20
T ¼ 40ESRdc C ¼ 40t (2.30)
The magnitude of constant current that results in 95% efficiency is calculated

by substituting (2.30) into (2.27) and carrying out the math. Caveat: for this
example, the cell is preconditioned to rated potential, Umx ¼ 2.7 V, and the dis-
charge condition is to terminate discharge at Umx/2. For this example, the constant
current used to characterize the 3000 F cell is given by (2.31) subject to the caveat
and the resultant carbon loading by (2.32):
CU mx U mx 2:7
I 3000 ¼ ¼ ¼ ¼ 116:3 A (2.31)
2T 80ESRdc 80ð0:29 103 Þ
I 3000 116:3
CL3000 ¼ ¼ ¼ 38:8 mA=F (2.32)
C 3000
According to 2.32, the characterization procedure would be to apply a carbon
loading of roughly 40 mA/F so that cell internal power dissipation is low, cell
internal heating is low, and parameter extraction is accurate.
Example 2.4: Calculate the test equipment data acquisition (DAQ) unit resolution
in bits when the full-scale voltage rating is 5.0 V for performing cell-level char-
acterization. Take the required DAQ accuracy needed as tenths of 1 mV (104 V).
Refer to Figure 2.25 bottom plot and note that the objective here is to find dU2 to
this level of accuracy for the 3000 F cell with the current magnitude found in
Example 2.2 applied.
U2 ¼ I 3000 ESRdc ¼ 116:3ð0:29 103 Þ ¼ 33:73 mV (2.33)
The result of (2.33) is that resolution to tenths of 1 mV is justified, particularly

if lower current levels such as 5 mA/F are used in testing as demanded by EUCAR
methods. This means that the DAQ must be capable of resolving 50,000 increments
of full-scale voltage (2.34).
ch002 7 June 2011; 13:51:53

U FS 5
2N ¼ ¼ ¼ 5 104 (2.34)
U LSB 104
To calculate the number of DAQ bits, N, one takes the logarithm of (2.34) and
solves. The procedure requires a change of base to simplify the calculations given
as log2 to log10.
log2 ð2N Þ ¼ log2 ð5 104 Þ (2.35)
log10 ð5 104 Þ 4:689

N log2 ð2Þ ¼ N ¼ ¼ ¼ 15:6 (2.36)
log10 ð2Þ 0:30103
Rounding (2.36) to the next highest integer means that the DAQ must have
N ¼ 16 bit resolution to have 100 mV accuracy. If the requirement were for 50 mV
resolution, the DAQ would require N ¼ 19 bits. Measuring such low-voltage signals
in the presence of high currents means that terminations, noise rejection, and fil-
tering are necessary to ensure stable ESRdc characterization values.
IEC62576 [14] is a recently released specification that was motivated by the
interest in use of EDLCs in hybrid electric vehicles, specifically in micro and mild
hybrid class. This IEC specification was the result of mainly automotive compa-
nies, academia, and some industry participants to put the characterization methods
of ultracapacitors into a more automotive application context.
Charge and discharge constant current magnitudes are specified in IEC62576
to be the values that are consistent with 95% charge or discharge efficiency (see
Example 2.2 for details). Hold time under constant voltage dwell per IEC62576 is
specified as TCV ¼ 300 s, and the voltage and time measurement points U1 and U2
shown in Figure 2.27 are 90% and 70%, respectively, of rated potential UR. These
measurement points are different from the previous IEC62391 and EUCAR meth-
ods. ESRdc calculated per IEC62576 is the same as previous methods (2.25) and
updated here as (2.37) for consistency with Figure 2.27.
dU 3
ESRdc ¼ (2.37)
Id
The backprojection technique of Figure 2.27 is referred to as the least squares

internal resistance method [14] to obtain the intercept from which dU3 is found.
For automotive applications, an earlier draft of the IEC62576 set the end-of-life
(EOL) criteria on ESRdc as 150% of initial value so that power at EOL was con-
sistent with hybrid electric vehicle pulse power requirements. In the final draft and
standard, these criteria were relaxed back to the previous EOL criteria of 200% of
initial value of ESRdc.
Capacitance characterization is based on the energy conversion capacitance
method and on the measurement of the discharged energy, W0, over the interval U1
ch002 7 June 2011; 13:51:54

Magnified figure
UR
ΔU3
U1
Voltage (V)
ΔU3
U2
TCV Time (s)
Key
UR Rated voltage (V)
U1 Calculation start voltage (V)
U2 Calculation end voltage (V)
ΔU3 Voltage drop (V)
TCV Constant voltage charging duration (s)
Figure 2.27 IEC62576 characterization method for ESRdc and capacitance
to U2 as given by (2.38). In this method, it is not the discharged amount of charge

that dictates the capacitance, rather the energy change measured versus energy
change computed from the terminal voltage. By doing these measurements at
current magnitudes characteristic of high energy efficiency, the error due to internal
resistance energy dissipation is minimized.
2W 0
C¼ (2.38)
ð0:9U R Þ ð0:7U R Þ2
2
Most of the standards discussed also specify methods to characterize maximum

power density of the cell, and information of this sort is always present on manu-
facturer data sheets for ultracapacitor cells and modules. IEC62576 specifies a
maximum power density method that relies on the internal resistance measured
according to (2.37) and stated here as (2.39). This is the matched impedance power
density, Pdm, method where M is the cell mass in kg.
0:25U 2R
Pdm ¼ (2.39)
ESRdc M
More will be said of power and energy metrics of ultracapacitors in Chapter 3.

For our intent here, it is sufficient to point out that a specification such as (2.39) is
essentially useless. The reason is that, unlike the battery, the EDLC is a voltage
storage device, which means that as soon as the power Pdm is extracted from the
cell, the voltage is no longer at rated potential UR. It means that a constant power
load at the value calculated in (2.39) could persist only instantaneously and there-
after a progressively lower matched impedance power level could be sustained. It is
ch002 7 June 2011; 13:51:54

more appropriate to use power levels at specified efficiency levels such as P95, or
P90, that do have meaning and practicality in applications.
There are other cell parameter characterization methods in use, principally at
independent testing labs, manufacturers, and universities. A steady-state method
derived from the EUCAR technique is used by Prof. Andrew Burke at UC-Davis, which
provides reliable ESRdc and capacitance values. The testing laboratory INRETS in
Paris, France, has developed techniques for cell characterization, cycling, and other
testing. Manufacturers generally use some variant of the standards discussed in this
section. The primary reason is that volume manufacturing is not amenable to ultra-
capacitor charge hold times of 30 s, 300 s, or 30 min. The six-step method in use by
Maxwell Technologies Inc. and discussed in the next section is one such method that
relaxes this hold time to 15 s so that end-of-line testing is sufficiently fast.
2.3 Simulation model validation

The criteria to apply when discussing model validation include the following set of
metrics:
● Capacitance must match the real cell as a function of potential (bias).
● Equivalent series resistance, ESR, must be consistent with specification sheet
values for ESRdc and ESRac and with frequency transition according to the
Nyquist laboratory data.
● ESR ¼ ESR(T) must be valid and consistent with laboratory test data.
● Parasitics must be accounted for.
● Natural decay must match real cell leakage at all temperatures.
● Stored charge must agree with real cell total stored charge. (Note here that
capacitance Cs does not contribute to stored energy as its initial condition and
steady-state value is set to zero potential.)
● Stored energy must match real cell total stored energy.
● Power performance of the model must agree with real cell power performance.
● ESR obtained from electrical simulation should be consistent with thermal
model predictions.
Consider a 650 F ultracapacitor cell response to a discharge–charge constant cur-
rent pair of pulses that are applied to its equivalent circuit. Figure 2.28 repeats the cell
equivalent circuit model and the simulated response to a 100-A current waveform.
For the same conditions as applied in Figure 2.23 for the 3000 F cell, a comparison
of the simulation model to measured data is summarized in Figure 2.29. In this figure,
the cell model included the complete C(U, I), ESR(T ), and parasitic elements to be as
complete as possible. The cell is conditioned and charged to half the rated potential,
then charged to rated potential and discharged per the six-step method, followed by two
complete full cycles of charge to Umx and discharge to 0 V. The objective of this vali-
dation is to compare the model response with that of the test cell.
Thermal characterization is the final step in our discussion of model validation.
One of the most important requirements for all energy storage systems is thermal
ch002 7 June 2011; 13:51:54

Rsa Cs
Rconn Lconn Rs
Uc
TLU: C(Uc) Co
Rp
Ultracapacitor cell with parasitics
Input current
100.00
AM1.I (A)
AM1.I (A)
−100.00
0 20.00 40.00
Terminal voltage
2.70
2.50
VM1.V (V)
VM1.V (V)
2.00
1.50
1.20
0 20.00 40.00
(a) Terminal voltage
2.77
2.70
VM1.V (V)
VM1.V (V)
2.60
2.50
2.44
28.91 35.00 39.16
(b)
Figure 2.28 Discharge–charge response of the ultracapacitor model for the

650 F. (a) Equivalent circuit, applied current waveform, voltage
response (Uc0 ¼ 2.7V) and (b) expanded view of cell voltage response
at termination of recharge pulse
ch002 7 June 2011; 13:51:54

Rsa Cs
− −
Lconn 0.60 mΩ 504.4 F
Rconn O\
− − Rs
70 PΩ 20.7 nH
Test C/C 0.30 mΩ
+
t y
I1 V VM1 C0 − Rc
−
~
CUL
Test C/C 643 Ω
t y V0 = 1.366
~
BCAP 3000 P cell six-step validation of lab test

data using the 3D table C(Uc, I)
Terminal voltage
2.24
C0-IWL
C0-IWL
2.00
C0c
C0c
2.50
2.29
0 250 500 600
Test curent
75
50
–50
0 250 500 600

(a)
Cell terminal voltage
2.8
2.5
2
VM1.V (V)
VM1.V (V)
1.5
1
500 m
0
–200 m
0 250 500 650
(b)
Figure 2.29 Validation of the 3000 F cell model with laboratory test data [Courtesy
of Maxwell Technologies Inc.]. (a) Full model of 3000 F cell (top) and
C(U, I) and 75-A current (bottom); (b) model response versus cell
response overlaid for same applied current waveform
ch002 7 June 2011; 13:51:54

management. Thermal management goes hand in hand with electrical and

mechanical design because electrical design and cell ratings dictate the heat
rejection during dynamic and steady state, whereas mechanical considerations of
package design, tolerances, materials, and exposure to simultaneous temperature
and vibration dictate the degree to which internally generated heat can be dissipated
to the module or pack case and from there to ambient. The situation is therefore one
of mutual dependence between thermal, electrical, and mechanical aspects of the
design. In this section, we look closely at the application of computer modeling
tools to this aspect of system design.
In Figure 2.30 a first-order representation of a cell string that comprises a
module is represented as cell ESRdc and capacitance, C, along with the thermal
parameters for each cell. In this representation, each cell contributes a thermal
capacitance, Cthx, arising from its mass, Mcell, and specific heat, cp, along with a
thermal resistance, Rthx, to ambient temperature, Ta. Thermal capacitance,
Cth ¼ Mcell cp, can be approximated using a specific heat for the ultracapacitor of
910 J/kg K. It turns out that, depending on cell axial length, Cth is closely calculated
using total cell mass multiplied by the specific heat of aluminum, a curious finding.
Figure 2.31 is representative of application scenarios that may be encountered
in practice where the cell pack is exposed to an arbitrary load profile of power or
Ic, Uc
Rth1
Cth1
Rth2
Cth2
Rth_NM
Cth_NM
Ta
Gnd
Figure 2.30 Module electrical and thermal design scenario
ch002 7 June 2011; 13:51:54

R1 Rsa Cs
E1 + Lconn Rconn 2.4 mΩ 141 F
-
D1 S1 Rs
C2NC1
47.4 V
R3 C2NC
0.07 mΩ 0.8 mΩ
I1 C Rp
#
S2 C1 + XY reduced
D2 V V x y 1.25 V
R2 10 μF
~
VM2 VM1 BCAP0650 cell mode 2.97 kΩ
0.01 V
E2 +
-
–47.4 V INTG1
I
FML1 EQU
Pd1: = Rs.R*Rs.I^2 H1
+ H RTH1 CTH1 + THM1
Pd2: = Rsa.R*Rsa.I^2
6.8 K/W
Pd3: = Rconn.R*Rconn.I^2 SUM1 188.57 Ws/K
Pd4: = Rp.R*Rp.I^2 299 K
Tamb +
Pd: = Pd1 + Pd2 + Pd3 + Pd4 Θ
Pin: = VM1 V*I1.1
299 k
Cell temperature
37.8
35
32.5
30
27.5
26
0 2 4 6 8 9.6
Thermal model
Rth = lab validated in passive air condition ~7°C/W
Cth = value extracted from lab thermal time constant
Figure 2.31 Illustration of an electrical–thermal interdependent model in

Simplorer
current and it must be sized to meet electrical specifications of power and energy at
EOL and not exceed the thermal design limitation of temperature rise <20 C at
EOL. This last requirement is particularly challenging because at EOL the cell
capacity has faded by 20% below its nominal value and ESRdc will be in the range
of 150–200% higher. With ESRdc a factor 2 higher than initial means that heat
rejection at EOL will be double that of beginning of life (BOL). This must be
accounted for during design-in and was also one of the motivators behind the
IEC62576 working group of the IEC.
ch002 7 June 2011; 13:51:54

One of the concerns that arise during both ultracapacitor design-in and thermal
performance has to do with current waveform duty cycle. For example, for the
same rms current, will the temperature rise of the cell be the same in steady state
regardless of pulse amplitude and pulse duration so long as its effective heating
value, the rms value, is the same? A simulation was executed to answer this con-
cern, configured as shown in Figure 2.31 and where the simulator state machine
controlled the current sources at appropriate current magnitudes for prescribed
pulse times by controlling switches S1 and S2.
For example, in Figure 2.32 the current waveforms tested have duty cycle,
d ¼ 0.9, 0.45, and 0.10 with correspondingly higher current magnitude so that each
possesses the rms value given by (2.40) and where I0 is the constant current mag-
nitude. The corresponding current magnitudes are I0 ¼ 47.4, 67, and 142.3 A.
pffiffiffiffiffiffiffi
I rms ¼ dI 0 (2.40)
Source current Source current Source current

47.40 67.00 143.00
20.00 25.00
I1.I (A)
I1.I (A)
50.00
I1.I (A)
0 0 0
–20.00 –25.00 –50.00
–67.00 –144.00
–47.5.00
0 20.00 40.00 60.00 80.00 120.00 0 20.00 40.00 60.00 80.00 114.00 0 20.00 40.00 60.00 80.00 114.00
Voltage Voltage Voltage

3 3 3
2.5
VM1.V (V)
2.5 2.5
VM1.V (V)
VM1.V (V)
2 2 2
1.5 1.5 1.5

1 1 1
0 20.00 40.00 60.00 80.00 120.00 0 20.00 40.00 60.00 80.00 114.00 0 20.00 40.00 60.00 80.00 120.00
Input power 650 Input power 650 Input power 650

154 265
130.00
50.00 100
Pin
Pin
0 0 0
Pin
–50.00 –100
–128.00 –152 –250
0 20.00 40.00 60.00 80.00 120.00 0 20.00 40.00 60.00 80.00 114.00 0 20.00 40.00 60.00 80.00 114.00
t t t
Figure 2.32 Response of ultracapacitor in Figure 2.31 to various duty cycle

current pulses of the same 45-Arms amplitude (waveforms for 90%
(47.4 Apk), 45% (67 Apk), and 10% (142.3 Apk) duty cycle) are shown
left to right
The peak currents applied in the simulation of Figure 2.31 cause the cell
temperature to ramp at increasingly faster rates during each pulse so that the
temperature profile has the appearance of a relatively smooth trace at 90% duty to a
sawtooth at 45% to that of a stair step at 10% duty. Figure 2.33 is the result of this
simulation for each of the specified duty cycle waveforms (Table 2.7).
ch002 7 June 2011; 13:51:54

Cell temperature Cell temperature

27.01 37.80
35.00
26.50 32.50
30.00
26.00 27.50
26.00
0 50.00 114.00 0 2.00 4.00 6.00 8.00 9.60

26.97 37.75
35.00
26.50 32.50
30.00
27.50
26.00 26.00
0 50.00 114.00 0 2.00 4.00 6.00 8.00 9.60

26.97 37.25
35.00
26.50 32.50
30.00
27.50
26.00 26.00
0 50.00 114.00
0 2.00 4.00 6.00 8.00 9.60
Figure 2.33 Comparison of 650 F cell thermal response to different duty cycle
current inputs at 45 Arms
Table 2.7 Summary of thermal response for various duty cycle inputs at 45 Arms
Duty Input Pulse Charge transfer, Temperature

cycle, d (#) current, I0 (A) time, td (s) Q (coulomb (C)) rise, dT ( C)
0.90 47.4 17.1 810.5 12.78
0.45 67 8.55 572.8 12.677
0.10 142.3 1.9 270.4 12.172
The essential conclusion from this thermal validation is that to a first-order

approximation, the cell temperature responds to effective rms current regardless of
duty cycle. This, in fact, may not be true if the ionic resistance, for some reason,
behaves differently to high-pulse amplitude and shorter dwell times. The high
current amplitude translates to a large amount of charge transfer in a short time and
ch002 7 June 2011; 13:51:55

high current density in the electrolyte and separator. What is interesting from
observation of Figure 2.32 is that even though the terminal arms are the same, the
waveforms with lower duty for the same rms current result in progressively lower
amounts of transferred charge and therefore lower voltage swing on the ultra-
capacitor. In this comparison work, the cell was conditioned to 1.35 V so that input
current charged and discharged the cell from that value. Note in Figure 2.32 that the
lower-duty cycle voltage response waveforms have progressively lower peak
amplitude and transition from sawtooth like to trapezoidal to near square wave in
shape.
2.4 Capacitor–battery combinations

There is growing interest in the combination of power-dense ultracapacitors with
energy-dense batteries at both the physical and package levels. In 2008 the Australian
Commonwealth Scientific and Industrial Research Organization (CSIRO) announ-
ced an advanced lead-acid battery that overcame one of the lead-acid battery tech-
nologies main limitations: fast recharge capability and extended operation at partial
state of charge (SOC). To accomplish this, CSIRO replaced half of the negative lead
plate electrode that was subject to sulfation during recharge with an activated carbon
EDLC electrode and named the product the UltraBattery. In the CSIRO design, the
negative electrode lead battery portion and the AC portion are tied in parallel at the
electrode in order to retain the best of both technologies. This product qualifies as
physical-level combination of battery and ultracapacitor since one of the battery
electrodes, the positive, is retained without change. CSIRO then tested their Ultra-
Battery in a Honda insight hybrid vehicle that was tested for 160,000 km without
battery replacement.
Furukawa Battery Co. in Yokohama, Japan, began volume production of
the UltraBattery and then licensed it to East Penn Manufacturing Company in
Pennsylvania in 2008. Axion Power International, also located in Pennsylvania,
developed a different version of the UltraBattery by completely replacing the
negative electrode with AC, which it dubbed the CapBat, and also began volume
manufacturing the product for application to military ground vehicles.
Figure 2.34 introduces the concepts of direct parallel architecture in which an
ultracapacitor cell or string is connected in tandem to a battery cell or string of
cells. This passive parallel configuration is at product level, rather than physical
level, so amounts to product integration of existing energy and power storage
technology components. An issue with the direct (or tandem or passive) parallel
configuration is that the relatively stiff voltage source battery effectively clamps
the voltage storage capacitor device, thereby severely limiting its contribution.
When making a direct parallel connection, the benefit decreases as one moves from
integration with lead-acid (Pb-acid) to nickel metal hydride (NiMH) to lithium-ion
(Li-ion). The reason is that advanced chemistry battery products such as NiMH and
Li-ion have already high cycling chemistries with good recharge capability and
have relatively low impedance.
ch002 7 June 2011; 13:51:56

Traction battery Contactor Ls-reactor
fixed bus
Traction battery
fixed bus
Rall
==
Energy management Battery
~300 V strategy ~300 V
Application
motor drive
DLC
‘ultracapacitor’ DLC
~300 V ‘ultracapacitor’
48 V to 125 V
Figure 2.34 Illustration of passive parallel (left) and active parallel (right)
combination configurations
Figure 2.34 also highlights measures to circumvent the voltage clamping

limitation of the passive parallel architecture. One method is to use a switch or
other means to connect the ultracapacitor to the battery branch when and if its
presence is required. In the active parallel configuration, no such switching is
necessary because the d.c.–d.c. power converter completely manages the energy
flows between the two components.
The presence of a line reactor (choke) in the active parallel architecture shown
in Figure 2.34 is not to limit current rates into and out of the battery but to restrict
exposure of the battery to ripple current leakage from the traction inverter d.c. bus
and also from the ultracapacitor d.c.–d.c. converter when in highly discontinuous
conduction boost mode. At the present, there is very little understanding of the
impact high-frequency ripple current has on electrochemical energy storage com-
ponents, particularly on their operational service life. There are studies underway,
but to date no data has been published; only the finding that if limits are placed on
peak-to-peak ripple current at the battery or at the ultracapacitor, that if adhered to,
will not significantly impact the battery life.
Recent publications address various aspects of ultracapacitor–battery combi-
nations. The authors in Reference [15] introduce the inductor input midpoint con-
verter (the half-H converter) as the converter of choice for interfacing an
ultracapacitor to a battery on a fixed d.c. link. In this study, the importance of energy
management strategy to regulate power flows between the power and energy com-
ponents is highlighted. The fundamental requirements on the converter for combi-
nation energy storage systems are summarized in Reference [16], and in Reference
[17] the topic is expanded to address application to plug-in hybrid electric vehicles
(PHEVs). It is well known that personal transportation vehicles have a 10:1 ratio of
peak power to average power so that a battery alone, especially in power assist
hybrids, will be stressed to high current rates and amplitudes such as 20–35 C rates.
Such high rates unduly stress the battery but are readily accommodated by an ultra-
capacitor power component. The authors in Reference [18] then address the
dynamics of vehicle drive cycle exposure on the energy storage system and conclude
that an ultracapacitor in combination with a battery can make a significant
ch002 7 June 2011; 13:51:56

contribution to reduced stress levels. With a combination architecture for energy

storage, the possibilities are considerable in regard to appropriate energy manage-
ment system (EMS) strategies, and the choice lies in what the vehicle drive cycle is
and the degree to which stress is reduced on the battery component. These topics are
addressed in detail in Reference [19], where the authors consider four very different
EMS strategies ranging from deterministic to predictive to frequency domain filter-
ing based. Environmental influence on the lithium-ion battery is discussed in depth in
Reference [20], including how the ultracapacitor combination can minimize thermal
stress by shifting battery heating to the ultracapacitor and, in fact, to minimize ESS
heating. The application of combination energy storage is very applicable to the case
of fuel cell hybrid vehicles as Chen et al. demonstrate in Reference [21]. This topic of
environmental limitations on lithium-ion, its poor power capability at cold tem-
perature, low rate charge acceptance near top of charge, and limited delta-SOC are
covered in Reference [22], where the conclusion is made that decoupled power and
energy provided by combination architectures go a long way to alleviate these lim-
itations in dynamic applications. A great deal of the findings made in References
15–22 are applied to a real-world situation in Reference [23], where a battery electric
vehicle (BEV), a Mazda Miata, is converted to electric drive and has a 28-kWh
lithium-ion ESS. Simulation of the system includes modeling of the lithium-ion
packs (200 Ah, 140 V), the ultracapacitor (26 F, 140 V, 78 Wh, 58 Wh deliverable),
and the interface d.c.–d.c. converter. The finding is that with only 58 Wh of useable
ultracapacitor energy, the thermal exposure of the lithium-ion is dramatically
reduced on Urban Dynamometer Drive Cycle (UDDS). Battery heating is reduced to
27% of its value acting alone over the UDDS cycle in this BEV. Table 2.8 sum-
marizes the findings for this study and shows the benefits of active parallel combi-
nation and of the necessity for high d.c.–d.c. converter efficiency and high
ultracapacitor efficiency. This is necessary because battery heating is shifted to the
converter and ultracapacitor and must be minimized in order for the ESS to have
lower overall thermal burden on its thermal management system.
Table 2.8 Energy storage system current magnitudes by component for battery
electric vehicle on Urban Dynamometer Drive Cycle
Mode Component Ipp Iavg Irms % %

(App) (Adc) (Arms) Reduction Heating
Battery alone Li-ion pack 344 11.8 41.8 – –
Active parallel Li-ion pack 158 10.9 21.7 48 27
Active parallel d.c.–d.c. converter 226 0.47 25.7 – –
Active parallel Ultracapacitor 361 0.47 38.3 – –
The main factor influencing battery heating is its rms current. When battery
rms current is reduced, it means the attendant internal heat generation is minimized
and shifted to the ultracapacitor branch of the active parallel combination according
to (2.41)
ch002 7 June 2011; 13:51:56


I 2rms batt % reduction 2
BHR ¼ 1 ¼ ð1 0:48Þ2 ¼ 0:27 (2.41)
I 2rms active 100
where BHR (battery heat reduction) is the overarching objective of the active
parallel combination.
Table 2.8 highlights the fact that the d.c.–d.c. converter used (bidirectional and
nonisolated) must be capable of high peak currents because the voltage swing of the
ultracapacitor under conditions of constant power output (or input) at the converter
high-voltage side is reflected to the low-voltage side, the ultracapacitor side in this
architecture shown as Figure 2.35, so that as its voltage decreases, the current
increases. The requirement therefore is that the converter should have an efficiency
greater than 97% over 20% to 90% of its working voltage range. The ultracapacitor,
for the power levels imposed, must have an efficiency of greater than 95% over its
full working voltage swing (Umx to Umx/2). If these requirements can be met, the
active parallel branch will have an operating efficiency of >92%.
Bidirectional
buck/boost converter
+ 50 V > Uin < 300 V;
Iin < 600 A
High side
100 V > Uo < 450 V;
filter +
Io < 150 A Low side
filter
− −
Enable
Controller
CAN
Figure 2.35 Schematic of the inductor input midpoint converter for active parallel
ESS application (low side is ultracapacitor and high side is lithium-
ion battery)
The discussion above and Figure 2.35 introduce a very fundamental concept in
capacitor–battery combinations – where should the converter go? This question is
answered in Reference [24], where all the possible configurations are evaluated by
simulation of a vehicle over a standard drive cycle. Table 2.9 is the summary of the
possible active parallel combinations of ultracapacitor and battery ESS. In this
table, the same vehicle discussed above is studied, the Mazda Miata BEV, equipped
with 28-kWh, 140-V lithium-ion battery and simulated with 58 Wh of useable
ultracapacitor energy.
There are two possible locations for the d.c.–d.c. converter: (i) on the ultra-
capacitor and (ii) on the battery. In the case of the converter buffering the ultra-
capacitor, there are two configurations: (i) ultracapacitor potential always less
than battery (i.e., d.c.-link voltage) and (ii) ultracapacitor potential always greater
than the battery (d.c. link). For the two cases of converter on the ultracapacitor, the
ch002 7 June 2011; 13:51:56

Table 2.9 Architectures for combination of ultracapacitor and battery
Architecture Robustness Cost Performance Overall
Up-convert
+ Fewer, large UC cells, 1 1 +
Battery
d.c. few conn’s
M
− a.c. High input current to Lower voltage Converter operates Best choice overall
converter semiconductor only when needed and considering PE
+ d.c.
Stable d.c. link High bandwidth technology advances
Ultra- Enable &
capacitor − d.c.
control
EMS strategy
Down-convert
+ More, smaller UC 0 1 0
d.c. cells, more conn’s
Battery M
− a.c. Lower input current High voltage Converter operates Too many interconnects,
Stable d.c. link semiconductor only when needed voltage management,
Ultra- + d.c. High bandwidth higher voltage UC
Enable &
capacitor − d.c. control system
EMS strategy
Converter on battery
More, smaller UC
–1 0 –
+
Ultra- d.c. cells, more conn’s
capacitor − M Converter Converter fault Requires ultrarobust
a.c. Highly dynamic d.c.
operational cannot be converter and high
link voltage
100% of time tolerated performance inverter
+ Difficult inverter
d.c.
Enable & Thermal Higher thermal controller and higher
Battery −
d.c.
PWM control
EMS strategy concerns burden current inverter switch
same energy management strategy is valid since the control strategy developed and
method implemented meant that boost and buck modes of this converter would
automatically switch depending on ultracapacitor voltage relative to d.c.-link
voltage. That in fact was implemented. For the battery, the EMS algorithm was
revised to accommodate a different objective function: shift dynamics to the
ultracapacitor.
More detail on this active parallel combination topic is provided in section
2.4.2. Let us close this section with some comments on the configurations shown in
Table 2.9.
● For the ultracapacitor voltage less than d.c. link (i.e., battery voltage), the
converter must up-convert to match the variable ultracapacitor voltage to the
fixed d.c. link. This means the converter operates in boost mode during vehicle
acceleration and buck mode during regeneration.
● The case for ultracapacitor voltage greater than the fixed d.c. link is just the
reverse. The converter must enter buck mode during vehicle acceleration and
boost mode during regeneration.
● With the converter on the battery, the situation changes dramatically in that the
d.c. link can no longer be at fixed potential but, instead, is a floating d.c. link.
The d.c.-link potential is the ultracapacitor voltage, and it varies in unison with
the loading. This means that additional burden is placed on the vehicle traction
inverter not only to meet all the demands of propulsion power but also to
simultaneously deal with a highly variable d.c.-link input voltage.
ch002 7 June 2011; 13:51:56

2.4.1 Passive parallel architecture

In this section, the direct parallel combination of ultracapacitor with lead-acid
battery is examined in detail. The lead-acid model used is available from ANSYS/
Ansoft as a library model with user input on several key parameters including
electrolyte specific gravity (SOC), internal resistance, Ah rating, and C-rate in
amperes. The ultracapacitor used in direct parallel combination is a
34S2P2000 F pack rated 87 Vmax, 118 F, 124 Wh, 93 Wh useable energy. The
battery is a VRLA 150 Ah, 36 cell, 76 Vmax, and 11 kWh that is used as the main
energy storage component in a BEV. In this case, the BEV is a 900-kg vehicle
(Af ¼ 2.2 m2, Cd ¼ 0.26), and the drive cycle being evaluated is the UDDS. The
UDDS cycle and BEV propulsion power requirement are summarized in
Figure 2.36. The BEV peak propulsion power is 20 kW during accelerating and
15 kW during braking. Propulsion power calculated for the UDDS cycle is shown
as ESS load power in Figure 2.37.
Urban dynamometer driving schedule, UDDS
60.00
50.00
40.00
Speed (mph)
30.00
20.00
10.00
0.00
0.00 200.00 400.00 600.00 800.00 1000.00 1200.00 1400.00 1600.00
–10.00
Time (s)
EV propulsion power UDDS cycle

25000.00
20000.00
15000.00
10000.00
Power (W)
5000.00
0.00
–5000.00 0.00 200.00 400.00 600.00 800.00 1000.00 1200.00 1400.00 1600.00
–10000.00
–15000.00
–20000.00
Time (s)
Figure 2.36 Battery EV drive cycle speed versus time and resultant propulsion
power
For this architecture, the objective is to determine how much battery rms
current reduction is possible with a direct parallel combination of ultracapacitor and
ch002 7 June 2011; 13:51:57

AM1 A Select: Battery only or

Rconn
ultracapacitor + battery
2.88 mΩ
– +
Battery Rsa
17.2 mΩ DATAPAIRS
LBATT_A
Rs I1
150 A Cs yt
V
~
50 mΩ + 17.6 F
Num_cells: = 36 V 6 mΩ
UDDS drive cycle for 900 kg
VM’ battery-EV load power
XY1
Rp C0 # yx
17 kΩ
~
72 V
34S × 2P × BCAP2000-P27
Load traction power (W)

20.60
DATAPAIRS1.VAL
–14.30
0 500.00 t 1.00 1.37
Figure 2.37 Simulation model for the direct parallel combination of lead-acid
battery with ultracapacitor
battery. The VRLA battery in a BEV is conditioned to full charge (sg (specific
gravity) ¼ 1.27) and internal resistance, Ri ¼ 50 mW, and cell rating is 2.1 V, 150 Ah.
It should be pointed out before going into detail on the ESS performance that this is a
large ultracapacitor module to be used in such a small passenger vehicle (900 kg),
so the results obtained should not be too surprising given that 93 Wh of useable
ultracapacitor energy is available to the 11-kWh lead-acid module.
The following set of simulation results summarize the performance enhance-
ment provided by the passive parallel configuration when the battery is lead-acid.
It must be pointed out at this stage that the drive cycle input power obtained
from a separated vehicle simulation over the given drive cycle is where the pro-
pulsion power as a function of vehicle speed, versus time, is derived. This
propulsion power file is then imported into the Ansoft/Simplorer simulation as the
power burden on the vehicle ESS. To accomplish this, the model of Figure 2.37 is
configured so that the load current source converts the command file power to a
current that is dependent on the instantaneous magnitude of the d.c.-link potential.
ch002 7 June 2011; 13:51:57

In this manner, the two components of ESS storage, the power and energy com-
ponents, each contribute their respective current to the total by satisfying Kirchh-
off’s current law at the node point on source I1. The response when both the VRLA
battery and the ultracapacitor pack are active is shown as Figure 2.39.
In conclusion for the passive parallel architecture, the following observations
can be made by inspection of Figures 2.38 and 2.39:
● Battery current is smoothed, has lower rates (dI/dt), and responds more to
average, steady-state demand than to dynamics.
● Ultracapacitor current is all dynamics, high rate, and essentially zero mean.
● Battery voltage, d.c.-link potential, has high impulse content and would present
considerable noise issues in a vehicle application.
● Ultracapacitor presence significantly reduces the battery voltage ripple,
resulting in a much smoother and lower rate voltage profile.
These facts are summarized in quantitative terms in Table 2.10 where the
peak–peak values, average (mean), and rms values of both current and voltages of
the ESS are listed. The percent reduction values show the degree to which the
passive parallel combination decreases a key metric. Table 2.10 shows significant
reduction in battery rms current and high rms current in the ultracapacitor, mainly
because of the lower impedance the ultracapacitor pack places in direct parallel
with the VRLA and thereby shunts dynamic currents that would otherwise flow
through the battery.
Battery UC current
358.00
200.00
Rconn.I (A)
AM1.I (A)
–134.00
0 500.00 t 1.00 1.37
Terminal voltage
108.00
100.00
VM1.V (V)
80.00
57.20
0 500.00 t 1.00 1.37
Figure 2.38 Direct parallel combination ESS used in small battery electric
vehicle application: battery only (72-V nominal d.c. link, 900-kg
vehicle, UDDS drive cycle)
ch002 7 June 2011; 13:51:57

Battery UC current
188.00
100.00
Rconn.I (A)
AM1.I (A)
–100.00
–190.00
0 500.00 1.00 1.37
t
Terminal voltage
82.90
80.00
VM1.V (V)
70.00
64.10
0 500.00 1.00 1.37
t
Figure 2.39 Direct parallel combination ESS used in small battery-electric

vehicle application: tandem mode (72-V nominal d.c. link, 900-kg
vehicle, Urban Dynamometer Drive Cycle)
Table 2.10 Summary of passive parallel ESS in a BEV driven on Urban

Dynamometer Drive Cycle
Mode Ipp Iavg Irms % Upp Uavg Urms %

(App) (Adc) (Arms) Reduction (Vpp) (Udc) (Vrms) Reduction
VRLA 488 25.9 59.8 – 50.5 77.3 10.16 –
only
Combo, 200 22.2 34.9 42 18.8 75.4 3.17 68
VRLA
Combo, 371 0.5 46.4 –
EDLC
2.4.2 Active parallel architecture

This section will complete our discussion on capacitor–battery combinations, and
for illustration the active parallel combination architecture studied in Reference
[24] and commented on in the previous section will be covered. Active parallel
means that the ultracapacitor component is no longer restricted by a fixed d.c.-link
potential that acts to clamp its power flow capability. Instead, the presence of the
matching d.c.–d.c. power converter ensures that the d.c.-link voltage remains
stable, while the ultracapacitor voltage experiences all the dynamics of the load
ch002 7 June 2011; 13:51:57

power. This combination only works as well as the energy management strategy
(EMS) developed to control power flows is capable of. For example, an EMS strat-
egy that follows frequency domain filtered load power will very readily reject an
average power while responding primarily to higher-frequency (>5 mHz) compo-
nents in propulsion power demand. The inclusion of a lower bound on frequency
response was noted to acclimate the reader to the fact that when addressing the
propulsion power demands of vehicles such as sedans or SUVs, the frequency
components of the dynamics are in reality low in magnitude but high in power level.
Another EMS strategy, for example, is one that restricts the operational power
bands over which the d.c.–d.c. converter is active. To limit the converter exposure
to low power demands that ordinarily should be handled by the battery, such as
maintaining adequate cruise power in a BEV, the converter would be inhibited until
some threshold on power or load current is crossed. Only then would the converter
be active in either boost or buck mode and its relative inefficiency at light loads is
avoided. Yet another EMS strategy is one that responds to the derivate of load
power (or load current) and makes decisions based on load power (or current)
amplitude, direction, and rate to control the d.c.–d.c. converter. The reader will
appreciate that there are many such algorithms for EMS that can be implemented to
obtain different performance and economy measures as may be desired.
What the EMS, however, must do is to ensure that the following conditions are
never violated:
● Converter input current does not exceed the maximum input rating.
● Ultracapacitor terminal voltage does not exceed its maximum rating.
● Ultracapacitor voltage remains within Umx, and Umn ¼ Umx/2 (or other
requirement).
● Ultracapacitor SOC remains within bounds: 25% < SOCuc < 100%, a resta-
tement of the preceding caveat.
● The foregoing constraints ensure that converter power rating is not exceeded,
and by extension, nor is its thermal rating.
The point made about the SOCuc window constraint is one of the more
challenging aspects of active parallel combination technologies, as it implies a
priori knowledge of what the future ESS power demand will be. For a standard
drive cycle, this is possible since future trends in power demand are known, but
implementing an EMS based on this knowledge would essentially be a cycle
beating strategy and one of limited to zero use in real-world applications. In reality,
the EMS strategy developed must be capable of tracking SOCuc and of constraining
it to within stated bounds. This last approach will be taken in the vehicle and ESS
evaluations that follow.
Figure 2.40 illustrates schematically the active parallel configuration having
the converter on the ultracapacitor, regardless of whether the ultracapacitor voltage
rating is greater than or less than the d.c.-link (i.e., battery) voltage, Ud. The energy
management system strategy controls the combination at a high level. The con-
verter control translates the strategy into duty commands to the S1–D2 pair for
ch002 7 June 2011; 13:51:58

Energy management control P(V)

Traction inverter
DT
Battery pack
Uc T Ud
Ultracapacitor Id
S1
pack L1 D1
Ic C2
C1 S2
D2
Figure 2.40 The d.c.–d.c. converter and ESS architecture to be evaluated
up-conversion boost mode when the ultracapacitor potential is less than d.c.-link
potential and to S2–D1 pair for buck mode in this same voltage scenario. Reverse
the switch–diode pair definitions for the down-conversion case.
Figure 2.41 is the equivalent of Figure 2.40 and represents the computer-
modeled version of the active parallel configuration. In this model, the ultra-
capacitor is modeled on the basis of the characterized parameters of the cell string
used, but with these parameters scaled from cell to pack level according to the
NSMPF rating of the pack. For example, N-cells in series per string and M-
strings in parallel define a pack. Recall that when scaling cells to modules/packs,
Cequiv ¼ Ccell(M/N) and ESRequiv ¼ ESRcell(N/M), and consequently, the time
constants are in fact constant, t ¼ CequivESRequiv ¼ CcellESRcell.
Charge–discharge Electrodynamics
A
& polarization Re, Ket, Cdle
Electrode Coulomb
Parasitics A d.c.–d.c. converter in
dynamics counter
state-space averaged
configuration Uc
Eb - e.m.f Eb(SOC)
C(U, I) Ah σ Drive cycle
Maxwell technologies functional Lithium current profile
energy management
pack
strategy SOCo
Application load
Ultracapacitor model
and dynamics
Figure 2.41 Simulation model of the ESS architecture for active parallel
The bidirectional, nonisolated, d.c.–d.c. converter is implemented in the

computer model as its state-space averaged implementation. The converter is
essentially a variable structure controlled circuit (VSC) that operates in discrete
configurations depending on the switch state. The averaging permits very fast
simulation of converters by modeling their behavior, not the individual switch
ch002 7 June 2011; 13:51:58

and diode events. The EMS strategy that controls the converter is always at the
heart of the implementation and often, if not always, the subject of intellectual
property of the manufacturer or applications designer. This is true of this book that
the converter and its model are shown, but no details of the governing EMS
are given.
The battery model shown schematically in Figure 2.41 is a heuristic descrip-
tion that accounts for electrochemical potential and its dependence on SOC,
the electrode kinetics (charge transport and EDLC effects) and the charge–
discharge polarization effects, and parasitic elements. Since most simulations over
drive cycles require significant charge transfer to/from the battery, its model is
configured to track SOC through a simple Coulomb counter. To a first-order
approximation, this provides sufficient accuracy for most simulations encountered
and certainly for all the examples to be covered in the following chapters of
this book.
The analysis of the Mazda Miata BEV discussed in the previous section is
expanded on in this section to include an active parallel ESS. In all the examples to
follow, the BEV has 28-kWh lithium-ion pack and 58 Wh of useable ultracapacitor
energy, regardless of cell configuration. The first architecture, shown pictorially as
Figure 2.42, is the up-convert case having the following features:
● Fixed d.c. link ? ultracapacitor up-converts to battery
● Can minimize d.c.-link voltage distortion
● Motoring mode means converter is in boost mode
● Battery current, converter disabled, Ib ¼ 68.7 Arms
● Converter enabled, Ib ¼ 44.4 Arms
● Battery Irms reduced by 35.3%
● Battery heating reduced by 58%
Ultracapacitor and battery voltages

168.50
+
d.c. 140.00
VM1.V (V)
VM2.V (V)
Battery M
− a.c.
120.00
+ d.c. 100.00
Ultra- Enable &
capacitor − d.c. EMS strategy 72.00
0 250.00 500.00 750.00 1.00 1.44
t
Figure 2.42 Up-convert case of active parallel ESS (Miata BEV over UDDS case)
(VM1, ultracapacitor voltage (grey); VM2, battery/d.c.-link voltage
(grey))
Case 2 of the converter on the ultracapacitor is the down-convert case, shown

pictorially as Figure 2.43, and has the following features:
ch002 7 June 2011; 13:51:59


286.00
+
d.c. 250.00
VM1.V (V)
VM2.V (V)
Battery M
− a.c. 225.00
200.00
+ d.c. 175.00
Ultra- Enable &
capacitor − d.c. EMS strategy 135.00
0 250.00 500.00 750.00 1.00 1.44
t
Figure 2.43 Down-convert case of active parallel ESS (Miata BEV over UDDS
case) (VM1, ultracapacitor voltage (grey); VM2, battery/d.c.-link
voltage (grey))
● Fixed d.c. link ? ultracapacitor down-converts to battery

● Can minimize d.c.-link voltage distortion via ultracapacitor power flow control
● Motoring mode means converter in buck mode
● Battery current, converter disabled, Ib ¼ 68.9 Arms
● Battery current, converter enabled, Ib ¼ 45.2 Arms
● Battery Irms reduced by 34.4%
Case 3 where the converter is on the battery in the down-convert case, shown
pictorially as Figure 2.44, has the following features:
● Floating d.c. link ? converter on battery
● Converter matches battery to variable d.c.-link voltage supported mainly by
ultracapacitor
● Extremely distorted d.c.-link voltage (grey) due to random power loading of
drive cycle. Battery current, converter disabled, ultracapacitor cannot support
load, loss of voltage control
● Battery current, converter enabled, Ib ¼ 44.8 Arms
● Battery Irms essentially same as before
The situation when the converter is on the battery is interesting in one respect:
the converter output power increases in reciprocal response to the d.c.-link tied
ultracapacitor as seen in Figure 2.44. That is, when the ultracapacitor power is
fading and is becoming less responsive in delivering the demanded power (low
SOCuc), the converter injects progressively higher amounts of current into the d.c.
link to help support the d.c. link and to head off collapse of the ultracapacitor
voltage. It is easy to appreciate that the small amount of ultracapacitor useable
energy, 58 Wh, is insufficient to support the vehicle propulsion power demand for
any significant length of time.
ch002 7 June 2011; 13:51:59


160.00
+ 150.00
Ultra- d.c. 140.00
VM1.V (V)
VM2.V (V)
M
capacitor − a.c. 130.00
120.00
110.00
+ d.c.
Battery Enable & 100.00
− d.c. EMS strategy 86.00
0 250.00 500.00 750.00 1.00 1.44
t
Figure 2.44 Converter on battery, down-convert case of active parallel ESS

(Miata BEV over UDDS case) (VM1, ultracapacitor voltage; VM2,
battery/d.c.-link voltage)
Let us summarize the salient features of the active parallel configuration. The
BEV is first modeled and simulated over the desired drive profile velocity versus
time V(t), UDDS cycle in the cases above, and from this simulation its propulsion
power, determined as the function P(V) given as (2.43), is applied to the ESS
simulator.

%gr
PðV Þ ¼ M V_ V þ gM V C rr þ sin tan1 V þ 0:5rair C d Af ðV V W Þ3
100
(2.42)
The simulator of Figure 2.41, according to the EMS employed, impresses a

current on the d.c. link in direct proportion to the vehicle power demand P(V) and
the d.c.-link potential, Ud. Figure 2.45 represents the BEV and its electric drive
system components.
Notice in Figure 2.46 that the two cases with fixed d.c.-link potential (con-
verter is on ultracapacitor) are running on the same EMS strategy. In this case, the
SOC variations are very similar and exhibit approximately the same limits. For the
floating d.c.-link case, the EMS strategy could have been designed to be more
aggressive and to utilize the ultracapacitor more, but this would have meant even
larger d.c.-link fluctuations in voltage.
To summarize this section, we conclude with a table illustrating the active
parallel architecture, the d.c.–d.c. converter output terminal power, and a fast
Fourier transform (FFT), of the converter output power to show how close the fixed
d.c.-link cases are and the difference between these and the floating d.c. link. One
can see by examining Table 2.11 that the fixed d.c.-link cases have very similar
converter power and FFT spectrums, but the floating d.c. link-case stands alone in
being different, and for the reasons cited.
The preferred architecture for active parallel ESS is the converter on the
ultracapacitor operating in the up-convert mode. A summary table has been
prepared to highlight the performance aspects of this particular architecture.
ch002 7 June 2011; 13:51:59

V, P(V)
ch002
IMC M/G
= ~
EMC
Id
Ud
ESS
7 June 2011; 13:51:59

Figure 2.45 Vehicle simulation for active parallel ESS case and representative vehicle (V, velocity (m/s); IMC, inverter-motor-
controller; EMC, electric machine controller; ESS, energy storage system; M/G, motor-generator component)
Ultracapacitor modeling
87
SOCuc comparison by arch

1.00
0.90
0.80
SOCuc magnitude
0.70
0.60
0.50
0.40
0.30
0.20
0.10
0.00
0 200 400 600 800 1000 1200 1400 1600
Time (s)
Figure 2.46 Illustration of SOCuc for three active parallel cases evaluated (grey
scales are converter on ultracapacitor, and converter on battery)
Table 2.11 Summary of active parallel ESS performance with focus on the
converter
ESS Converter output Fast Fourier

architecture power transform of
converter power
ESS architecture Converter output power FFT of converter power

FFT BEV Pd.c.out 134 V–150 V
Pd.c.out 134 V–150 V arch 100
+ 30,000 90
d.c. 20,000 80
Magnitude (%)
Battery M 10,000 70
− a.c. 0 60
Power (W)
50
–10,000 0 200 400 600 800 1000 1200 1400 1600
40
–20,000 30
–30,000 20
+ d.c. Enable & –40,000 10
Ultra- –50,000 0
EMS
capacitor − d.c. –60,000 0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
strategy Time (s) Frequency (Hz)
FFT 256 V–150 V arch

Pd.c.out 256 V–150 V arch
+ 30,000
100
90
d.c. 20,000 80
Battery M
Magnitude (%)
10,000
− a.c. 0
70
60
Power (W)
–10,000 0 200 400 600 800 1000 1200 1400 1600 50

–20,000 40
–30,000 30
Ultra- + Enable & 20

d.c. –40,000 10
EMS –50,000
capacitor − d.c. –60,000
0
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
strategy Time (s)
Frequency (Hz)
FFT conv2 batt arch

+ 50,000 Pd.c.out conv2 batt arch 100
d.c. 90
Ultra- 40,000
80
M
Magnitude (%)
capacitor − 30,000 70
a.c. 20,000 60
Power (W)
10,000 50
0 40
–10,000 0 200 400 600 800 1000 1200 1400 1600 30
+ d.c. Enable & –20,000 20
10
–30,000
Battery − EMS –40,000 0
d.c. 0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
strategy Time (s)
Frequency (Hz)
Table 2.12 contains the current and voltage performance metrics and shows that
with a converter on the lithium-ion pack (28 kWh) its peak currents, Imx and Imn,
are significantly lower than those for the battery acting alone in response to the
vehicle P(V) over a UDDS cycle. Battery average current is little changed, as
ch002 7 June 2011; 13:51:59

Table 2.12 Active parallel ESS for a battery-EV case having 28-kWh lithium-ion
pack and 58-Wh ultracapacitor
Component Imx Imn Ipp Iavg Irms Ub_pp Ub_rms Uc_pp Uc_rms
Battery only 208.2 136 344.3 11.77 41.8 17.4 2.5 – –
Active Battery plus
ultracapacitor
combined
via d.c.–d.c.
converter
Battery 133.7 24 157.9 10.9 21.7 8.75 1.5 – –
Conv out 96.7 129.4 226 0.47 25.7 – – – –
Ultracapacitor 156.8 204.2 361 0.57 38.3 – – 90.6 98.5
Conv, converter.
should be expected for a vehicle application, but most important, the battery Irms is
reduced by 48% of the rms current of the battery acting alone. This is a dramatic
reduction and amounts to shifting 73% of the battery heating out.
One may be tempted to compare Table 2.12 for the active parallel ESS case with
Table 2.10 for the passive parallel case and may notice that battery heating is reduced
by 73% (48% Irms reduction) in the active case and by 67% (42% Irms reduction) in the
passive parallel case and ask the question: So why introduce the complexity of a d.c.–d.
c. converter? The answer is this: Table 2.10 summarizes 93 Wh of useable ultra-
capacitor energy in direct parallel with 11 kWh of VRLA lead-acid pack, whereas
Table 2.12 summarizes 58 Wh of useable ultracapacitor energy in active parallel with
28 kWh of high-quality lithium-ion battery. The distinction is that in one case
approximately 2 per unit (pu) of ultracapacitor energy is placed in direct parallel with
approximately 1 pu of battery energy, whereas in the second case 1 pu of ultracapacitor
energy is placed in active parallel with 2 pu of a strong lithium-ion pack, and the
performance in the second case is still superior to that of the first (the VRLA case).
Exercises
2.1 Apply (2.4) and reference Table 1.1 to list the E12 component values for
precision resistors starting with 105 W.
Ans: 100k, 120k, etc.
2.2 Calculate the carbon loading of the 2000 F, 2.7 V carbon–carbon ultra-
capacitor that is used in an engine starting application at 600 A for 1 s.
Ans: 300 mA/F.
2.3 Calculate the current density through the separator that flows into the face
area of the carbon electrodes of the 3000 F ultracapacitor given a double-side
electrode film of dimensions 110 mm wide by 2.5 m in length. The terminal
current is 500 A. Let this current density be the face area density, JfA.
Ans: JfA ¼ 909 A/m2.
ch002 7 June 2011; 13:52:0

2.4 Calculate the current density internal to the carbon film of Exercise 2.3
given that its surface area A ¼ 115,000 m2, and assume a current of 500 A
into the ultracapacitor cell terminal. Let this current density be the carbon
area density, JcA.
Ans: JcA ¼ 4.35 mA/m2.
2.5 The conductivity of the electrolyte in the 3000 F cell in the presence of the
porous separator is taken as sel ¼ 4.54 mS/cm. For the face area current
density calculated in Exercise 2.2, compute the power dissipated in the
electrolyte of this cell given the electrode dimensions of Exercise 2.3 and
taking the separator thickness as 40 mm.
Hint: Calculate the separator volume (careful with units) and do the inte-
gration given by
ZZZ
J 2f A
Pi ¼ dV ðEx: 2:5:1Þ
ael
Ans: Pi ¼ 40 W.
2.6 Compare the value of power dissipation in the electrolyte found in Exercise 2.5
to the total power dissipation of the 3000 F cell for an applied current of 500 A.
Ans: Pd ¼ 72.5 W.
2.7 The bulk conductivity of a typical acetonitrile solvent-based electrolyte,
sel ¼ 56 mS/cm, is modified by the porosity of the separator approximately
as given by (Ex. 2.7.1) where r ¼ porosity of the separator, and let r ¼ 35%.
Use the value so calculated, and repeat Exercise 2.5 to obtain the value of
ionic power dissipation Pi.
ð1rÞ
ssep ¼ sel ðEx: 2:7:1Þ
Ans: With ssep ¼ 13.68 mS/cm, Pi ¼ 13.29 W (a factor 3 lower).
2.8 Derive the expression for EDLC discharge efficiency given as (2.28) under
constant current in terms of the cell time constant and dwell time of the
discharge pulse.
Hint: From an energy perspective, the output energy E0 is the difference
between stored energy Esto and dissipated energy Ed. Note that under con-
stant current conditions the internally dissipated power is constant with time,
and therefore, Ed varies linearly with time.
2.9 Calculate the maximum continuous current, Irms, for the 3000 F cell using
the thermal data provided in Table 2.3 given that the maximum permissible
dT ¼ 15 C.
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Ans: I rms ¼ ESRdTdc Rth ¼ 127
2.10 Calculate the pack energy ratings of a 34S1P3000 F ultracapacitor pack and
compare this to the energy of a 34S3P650 F cell pack. The pack voltages
ch002 7 June 2011; 13:52:2

are the same, assuming 2.55 V/cell maximum operating potential, and find the
carbon loading (CL) when the pack delivers 15 kW at 50% SOC. Apply the
lessons learned from this chapter on ultracapacitor cell to module scaling.
Ans: W3000 ¼ 92.7 Wh, W650 ¼ 80.4 Wh, CL3000 ¼ 115 mA/F, CL650 ¼
133 mA/F.
References
1. J.M. Miller, P.J. McCleer, M. Cohen, ‘Ultracapacitors as energy buffers in a
multiple zone electrical distribution system’, Global Powertrain Conference,
Crowne Plaza Hotel, Ann Arbor, MI, 23–25 September 2003
2. J. Schindall, J. Kassakian, D. Perreault, D. New, ‘Automotive applications
of ultracapacitors: characteristics, modeling and utilization’, MIT-Industry
Consortium on Advanced Automotive Electrical-Electronic Components and
Systems, Spring Meeting, Ritz-Carlton Hotel, Dearborn, MI, 5–6 March 2003
3. M. Rosu, J.M. Miller, U. Deshpande, ‘Parameter extraction for ultracapacitor
high power modules’, Power Electronics Technology Conference, Dallas,
TX, 30 October–1 November 2007
4. A.B. Kahng, S. Muddu, Optimal Equivalent Circuits for Interconnect Delay
Calculations Using Moments, Association for Computing Machinery, ACM
0-89791-687-5/94/0009, 1994
5. S. Buller, E. Karden, D. Kok, R.W. De Doncker, ‘Modeling the dynamic
behavior of supercapacitors using impedance spectroscopy’, IEEE Transac-
tions on Industrial Applications, November/December 2002
6. E. Surewaard, M. Tiller, ‘A comparison of different methods for battery
and supercapacitor modeling’, SAE Future Transportation Technology Con-
ference, Hilton Hotel, Cosa Mesa, CA, 23–25 June 2003
7. R. Kotz, M. Hahn, R. Gallay, ‘Temperature behavior and impedance funda-
mentals of supercapacitors’, Journal of Power Sources, vol. 154, pp. 550–5,
2006
8. T. Funaki, T. Hikihara, ‘Characterization and modeling of the voltage
dependency of capacitance and impedance frequency characteristics of
packed EDLC’s’, IEEE Transactions on Power Electronics, vol. 32, no. 3,
pp. 1518–25, 2008
9. J. Schiffer, D. Linzen, D.U. Sauer, ‘Heat generation in double layer capaci-
tors’, Journal of Power Sources, vol. 160, pp. 765–72, 2006
10. IEC62391-1, Fixed Electric Double Layer Capacitors for use in Electronic
Equipment – Part I: Generic Specification, IEC 40/1378/CD
11. IEC62391-2, Fixed Electric Double Layer Capacitors for use in Electronic
Equipment – Part II: Sectional Specification: Electric Double Layer Capa-
citors for Power Applications, IEC 40/1379/CD
12. EUCAR, Specification of Test Procedures for Supercapacitors in Electric
Vehicle Application, prepared by EUCAR Traction Battery Working Group,
April 2003
ch002 7 June 2011; 13:52:2

13. J.M. Miller, ‘Ultracapacitor efficiency: device in constant current and con-
stant power applications’, Bodo’s Power Magazine, pp. 30–32, 2008
14. IEC62576, Electric Double Layer Capacitors for Use in Hybrid Electric
Vehicles – Test Methods for Electrical Characteristics, ISO/IEC Directives,
Part 3, 2010
15. J.M. Miller, M. Prummer, A. Schneuwly, ‘Power electronic interface for an
ultracapacitor as the power buffer in a hybrid electric energy storage system’,
Published in Power Systems Design, Automotive Electronics Series Editorial
Article, July/August/September 2007
16. J.M. Miller, B. Maher, U. Deshpande, J. Auer, M. Rosu, ‘Requirements for a
d.c.–d.c. converter buffered ultracapacitor in active parallel combination with
an advanced battery’, Power Electronics Technology Conference, Dallas, TX,
30 October–1 November 2007
17. J.M. Miller, U. Deshpande, ‘Ultracapacitor technology: state-of-technology
and application to active parallel energy storage systems’, The 17th Interna-
tional Seminar on Supercapacitors and Hybrid Energy Storage Systems,
Deerfield Beach, FL, 10–12 December 2007
18. J.M. Miller, U. Deshpande, T.J. Dougherty, T.P. Bohn, ‘Combination ultra-
capacitor-battery performance dependence on drive cycle dynamics’, The
18th International Seminar on Supercapacitors and Hybrid Energy Storage
Systems, Deerfield Beach, FL, 8–11 December 2008
19. J.M. Miller, U. Deshpande, T.J. Dougherty, T.P. Bohn, ‘Power electronic
enabled active hybrid energy storage system and its economic viability’, The
24th IEEE Applied Power Electronic Conference, APEC’09, Marriott
Wardman Park Hotel, Washington, DC, 15–19 February 2009
20. J.M. Miller, M. Everett, P. Mitchell, T.J. Dougherty, ‘Ultracapacitor plus
lithium-ion for PHEV: technical and economic analysis’, The 26th Interna-
tional Battery Seminar and Exhibition, Broward Convention Center,
Ft. Lauderdale, FL, 16–19 March 2009
21. B. Chen, Y. Gao, M. Ehsani, J.M. Miller, ‘Ultracapacitor boosted hybrid fuel
cell’, IEEE Vehicle Power and Propulsion Conference, VPPC’09, Ritz
Carlton Hotel, Dearborn, MI, 7–9 September 2009
22. J.M. Miller, ‘Energy storage system technology challenges facing strong
hybrid, plug-in and battery electric vehicles’, IEEE Vehicle Power and
Propulsion Conference, VPPC’09, Ritz Carlton Hotel, Dearborn, MI, 7–9
September 2009
23. J.M. Miller, ‘Active combination of ultracapacitor-battery energy storage
systems gaining traction’, The 19th International Seminar on Supercapacitors
and Hybrid Energy Storage Systems, Deerfield Beach, FL, 7–9 December
2009
24. J.M. Miller, ‘Engineering the optimum architecture for storage capacitors’,
Advanced Automotive Battery Conference, AABC2010 Large EC Capacitor
Technology and Application, ECCAP, Omni Orlando Resort, Orlando, FL,
18–21 May 2010
ch002 7 June 2011; 13:52:2

Chapter 3
Power and energy
The evolution of symmetric electrochemical capacitor (EC) specific energy has

been incremental over the past 30 years, mainly because of the limits on cell
potential by the available electrolytes and purity of activated carbon. Referring to
Figure 3.1 it is evident that cell energy tracks closely to cell potential, which for
organic electrolytes has trended from 2.3 to 2.7 V at present. This averages to
20 mV/year in cell potential and clearly revolutionary change in cell potential will
not happen to this class of ultracapacitor other than what available materials can
provide. It is anticipated that cell potential will increase to 2.85 V, perhaps 3.0 V or
even 3.1 V at most, for high quality activated carbon. For example, Panasonic
commercialized the power ultracapacitor in the mid-1980s rated 470 F, 2.3 V,
and 3.9 mW, the increase in performance has been incremental with cell voltage
increase to 2.5 V in 1999 and to 2.7 V by 2006. Power evolution, on the other hand,
has been more dramatic and is projected to continue increasing with improvements
in materials, manufacturing process, and cell potential gains. One projection is that
specific power of ultracapacitors will reach 20 kW/kg by 2015.
There is growing interest in the use of electrochemical capacitors in auto-
motive systems because of their high power cycling capability, high efficiency,
robust performance over wide temperature extremes, and high lifetime energy
Ultracapacitor P&E evolution

30
Energy
Specific energy, power
25
Power
20 Voltage
15
10
0
1992 1996 2000 2004 2008 2012 2016
Year
Figure 3.1 Evolution of symmetric carbon–carbon ultracapacitor voltage, energy,

and power
Ch003 14 June 2011; 18:44:38

throughput. The following discussion treats these topics in more depth. In Figure 3.2,
the cycling capability of batteries and ultracapacitors is put into contrast as a function
of depth of discharge per cycle. For example, energy storage packs of battery electric
vehicles (BEVs) must deep discharge to nearly 90% for 1500 cycles in order to meet
10-year lifetime criteria. For a plug-in hybrid electric vehicle (PHEV) this increases to
4000 cycles at 70% SOC, and for strong hybrids that rely mainly on shallow cycling of
10% or lower, the number of cycles increases into the hundreds of thousands. A
microhybrid in contrast requires a power battery capable of greater than 600,000
shallow cycles, *2% SOC swing. The capability of the ultracapacitor shown on this
same scale in Figure 3.2 has 1 million cycles of 75% depth of discharge.
SOC variation by vehicle type

100
y = –13.67In (x) + 184.29
90
BEV
80
PHEV UC
70
60
SOC (%)
50
40
30
20
HEV
10
0
1000 10,000 100,000 1,000,000
Cycles (#)
Figure 3.2 Energy storage SOC window by electric vehicle type
Energy storage components must deliver adequate power at cold temperatures.

This is an essential requirement for vehicle applications and one that is difficult to
meet because of electrolyte performance at cold temperatures. Electronic resistance
at cold temperatures is not so much a concern because metals have positive tem-
perature coefficients that result in progressively lower total resistance as tempera-
ture decreases. This was already demonstrated by evaluating electronic resistance
using the Taylor series expansion of ESR(T) for the first two terms and noting that
the coefficient on the linear term is resistance sensitivity to temperature, g. For
ionic resistance the corresponding functional approximation is Arrhenius type,
where the coefficient on the exponential, kT, shows the multiplicative effect this has
versus temperature change, the sensitivity being one of negative temperature
coefficient. These effects are best illustrated by examining the electrolyte resistance
increase as temperature decreases. For example, in ultracapacitor cells and lithium-
ion cells (reference is made here to power lithium-iron-phosphate (LFP)) the rela-
tive proportions of electronic to ionic resistance of total ESR is approximately 40%
and 60%, respectively.
Figure 3.3 illustrates the representative electrolyte resistance change versus
temperature for the ultracapacitor and LFP cells. Both have ESR(T) normalized to
room temperature (20 C) as 1 per unit (pu) with vertical axis the relative change as
Ch003 14 June 2011; 18:44:39

Power and energy 95
ESR: Ultracapacitor and lithium battery

20
18 UC
LFP
16
14
Resistance (pu)
12
10
8
6
4
2
0
–40 –30 –20 –10 0 10 20 30 40 50 60
Cell temperature (°C)
Figure 3.3 Illustration of representative ESR( T) for ultracapacitor and LFP cells
temperature is lowered. Note: As temperature drops below 0 C the LFP resistance

is increasing dramatically faster than the ultracapacitor resistance, in particular of
ultracapacitors using acetonitrile (AN) electrolyte.
The implication is that the cold temperature power capability of the energy
storage component is dependent on its ESR(T). For example, if the ESR(T)
increases by a factor of 5 at 20 C for the LFP then its power capability is reduced
from 1C to C/5, a significant reduction. The ultracapacitor, on the other hand, may
see its ESR(T) increase to only 1.5 pu at this same temperature so that its cold
temperature performance is only modestly reduced.
Figure 3.4 is a more convenient means to represent the influence of tempera-
ture on energy storage component power capability. In this graphic, the power
performance of a mile hybrid vehicle LFP battery pack is contrasted to that of an
30
20
Discharge
15
Power (kW)
0
–40 –30 –20 –10 –5 0 5 10 15 20 30 40 50 60 70
Temperature (°C)
–5
Charge
–15
–20
Figure 3.4 Power performance of LFP and UC versus temperature (LFP is inner
set of traces; UC is outer set of traces)
Ch003 14 June 2011; 18:44:39

ultracapacitor module at power levels of 15–20 kW, the range of peak power
typical of mild-hybrid systems.
Notice in Figure 3.4 that once the LFP pack temperature drops below 0 C
to 5 C its discharge and charge power capability drops precipitously, as tem-
perature drops to the point by 20 C the pack will have less than 20% of its power
capability remaining. This is a serious performance and economy obstacle for any
type of hybrid electric vehicle, especially for PHEV and BEV, since these vehicles
rely so heavily on the battery for power. By contrast, the ultracapacitor discharge
and charge performance as shown in Figure 3.4 is basically constant over the wide
temperature range and charge or discharge performance is symmetrical. One can
say that combining the two technologies, battery for energy and ultracapacitor for
power, makes good sense. The term ‘eclectic principle’1 has been applied to such
synergistic combinations. The eclectic principle means:
● Decoupled power and energy.
● Power electronic control to manage energy flows.
● Battery cycling current shifted to the ultracapacitor for improved system life.
● Battery heating transferred to the ultracapacitor where it can be minimized by
higher efficiency storage and control.
Section 3.4 will deal more with combination technologies. For now, we focus
on ultracapacitor efficiency and discussion of specific energy and energy density.
3.1 Specific energy and energy density

Ultracapacitor data sheets generally will specify specific energy and some will
have energy density, but all data sheets list cell mass and dimensions that can be
used to calculate energy density. As a voltage storage device, the ultracapacitor has
specific energy (J/kg, Wh/kg) and energy density (J/L, Wh/L) given as (3.1) and
(3.2), respectively, where cell mass (M) and dimensions ( L) are used to calculate
volume.
CU 2mx CU 2mx
SE ¼ ðJ=kgÞ; ¼ ðWh=kgÞ ð3:1Þ
2M 7200M
CU 2mx CU 2mx
ED ¼ ðJ=LÞ ¼ ðWh=LÞ ð3:2Þ
2 Vol 7200 Vol
Putting (3.2) in terms of cylindrical cell diameter and length makes it more con-
venient to calculate cell ED, given cell diameter, f, and length, L, as shown in (3.3).
2CU 2mx CU 2mx

ED ¼ 2
ðJ=LÞ; ¼ ðWh=LÞ ð3:3Þ
pf L 1800ðpf2 LÞ
1
Term coined by Dan Coffey, San Diego Daily Transcript, 9 July 2010 as an energy storage system that
combines the best of power and energy components.
Ch003 14 June 2011; 18:44:39

Power and energy 97
Example 3.1: Calculate the SE and ED of the Maxwell K2 design ultracapacitor

rated 3000 F, 2.7 V that is manufactured in a f 60.4 L 138 mm cylindrical can
having a mass, M ¼ 0.51 kg. Give the gravimetric results in both Joules and Wh.
Solution:
CU 2mx 3000ð2:72 Þ CU 2mx 3000ð2:72 Þ

SE ¼ ¼ ¼ 21; 441 J=kg; ¼ ¼
2M 2ð0:51Þ 7200 M 7200ð0:51Þ
¼ 5:96 Wh=kg ð3:4Þ
For the energy density calculation, apply the stated cell diameter and length using
(3.3).
2CU 2mx 2ð3000Þ2:72 CU 2mx

ED ¼ ¼ ¼ 27; 655 J=L; ¼
pf2 L pð0:604Þ2 1:38 1800ðpf2 LÞ
3000ð2:72 Þ
¼ ¼ 7:68 Wh=L ð3:5Þ
1800ðpð0:604Þ2 1:38Þ
The preceding discussion on specific energy and energy density puts the cap-
ability of symmetric ultracapacitor cells in perspective. What is needed next is the
efficiency at which this energy can be extracted and replaced. Energy efficiency
under constant current conditions is considered next. During constant current
charging the DLC layers at each electrode accumulate charge at a constant rate.
Cell dissipation as internal heating is also constant because the current is constant.
This is true to the extent that internal heating is not causing significant change in
ESRdc. Consider the graphic of Figure 3.5 representing the cell capacitance (DLC)
and resistance (ESRdc) when the terminal condition is that of constant current
charge followed by constant current discharge of magnitude, I0.
Uc(t) Umx
R
(t/c)t
Umx/2
+Uco I(t)
Uc(t) i(t) U0(t)
C
Tc Td
Figure 3.5 Ultracapacitor under constant current charge (DLC capacitance, C, has
initial condition, Uco ¼ 0 at t ¼ 0þ, and Uc(t) is the internal potential of
the DLC as a function of time, R ¼ ESRdc and U0(t) is the terminal
voltage.)
Ch003 14 June 2011; 18:44:39

The graphic at the right in Figure 3.5 shows a quasi-square wave profile of
current having dwell Tc during charge and Td during discharge. If the charge and
discharge capacities are equal then Tc ¼ Td ¼ T. For constant current applied during
charge, the internal potential across the DLC ramps (nearly) linearly with time and
the terminal voltage shows resistive steps at the application of current and when the
current is extinguished. This effect is what appears when cell potential is monitored
during constant current charging. The two potential traces shown in Figure 3.5 obey
(3.6) for internal potential, Uc, and (3.7) for terminal potential, U0(t) (Figure 3.6).
q ¼ CU c ¼ I 0 t ð3:6Þ
I0
U 0 ðtÞ ¼ U c0 þ I 0 R þ t ð3:7Þ
C
Ic
I0
–I0
Charge Discharge
Uc(t)
Umx
(t)
U0 U0 (t
)
t
Tc Td
Figure 3.6 Illustration of charge–discharge behavior according to (3.6) and (3.7)
At this stage U0(t) in (3.7) is multiplied by terminal current, I0, to obtain the input
power during charging, Pi(t). Then, knowing input power the solution for energy
delivered to the cell is obtained by integrating Pi(t) (Figure 3.7) over the interval [0,
Tc] and solving the definition of energy efficiency, hc, as energy stored divided by
energy input. Equations (3.8)–(3.11) describe the method proposed, where energy
is given by the symbol, W (Figure 3.8).
I 20
Pi ðtÞ ¼ U 0 ðtÞI 0 ¼ U c0 I 0 þ I 20 R þ t ð3:8Þ
C
ðTc ðT C
I2
W i ðtÞ ¼ Pi ðxÞdx ¼ U c0 I 0 þ I 20 R þ 0 x dx ð3:9Þ
0 0 C
I 20 2
W i ðtÞ ¼ U c0 I 0 T c þ I 20 RT c þ T ð3:10Þ
2C c
Ch003 14 June 2011; 18:44:39

Power and energy 99
W sto U c0 I 0 T c þ ðI 20 =2CÞT 2c
hc ¼ ¼ ð3:11Þ
Wi U c0 I 0 T c þ I 20 RT c þ ðI 20 =2CÞT 2c
During charge the term Uc0 ¼ 0 in (3.11), so this expression may be resolved to
(3.12) showing that ultracapacitor energy efficiency under constant current charge
is given as the ratio of the cell RC time constant to the total charge time, T.

ðI 20 =2CÞT 2c 1 1
hc ¼ 2 ¼ ¼ ð3:12Þ
2 þ Þ þ
I 0 RT c þ ðI 0 =2CÞT c U c0 ¼0 cÞ
2 1 2ðRC=T 1 2ðT=T
c
The only difference from discharge efficiency is the sign of the dissipation term and
the form of the energy efficiency ratio. Derivation of the discharge efficiency under
Pi(t)
PC
Pd
t
Pd
PC
Figure 3.7 Illustration of charge–discharge power according to (3.9)
Stored Available
Ws(t)
WC Wa
Wdc Wdd
t
Dissipated Dissipated
Figure 3.8 Illustration of charge–discharge energy according to (3.17)
Ch003 14 June 2011; 18:44:40

constant current condition is given as (3.13)–(3.17) (Figure 3.8). Use I0Td ¼ CUc0
to go from (3.15) to (3.16) and simplify.
ðt
1
P0 ðtÞ ¼ U 0 ðtÞI 0 ¼ U c0 I 0 I 20 dx I 20 R ð3:13Þ
C 0
I 20
P0 ðtÞ ¼ U c0 I 0 t I 20 R ð3:14Þ
C
ðTd ðT d
I 20
W 0 ðtÞ ¼ P0 ðxÞ dx ¼ U c0 I 0 x I 0 R dx
2
ð3:15Þ
0 0 C

CU 2mx
W 0 ¼ CU 2mx RI 20 T d ¼ CU 2mx RI 20 T d ð3:16Þ
2C U c0 ¼U mx
ðCU 2mx =2ÞRI 20 T d t

hd ¼ ¼12 ð3:17Þ
ðCU 2mx =2Þ Td
Round trip efficiency in the constant current case is simply the product of charge
and discharge efficiency (3.18). An important point to adhere to in the application
of (3.12) and (3.17) is that the pulse times specified are the time required to fully
charge, Tc, and fully discharge, Td, the ultracapacitor by the applied constant cur-
rent. Only then does the ratio of cell time constant to the pulse time yield accurate
efficiency results (3.19).

1 t
hrt ¼ hc hd ¼ 12 ð3:18Þ
1 þ 2ðt=T c Þ Td

1 2ðt=T d Þ 1 2ðt=TÞ
hrt ¼ ¼ ð3:19Þ
1 þ 2ðt=T c Þ T c ¼T d ¼T 1 þ 2ðt=TÞ
The simplification given by (3.19) does not apply in general, since the cell
capacitance Cup does not equal Cdn, which means that Tc Td as explained in Chapter 2.
In conclusion, ultracapacitor efficiency under CC can be represented as the ratio of
its R C time constant to the pulse time, T. High efficiency demands long cycle
times T*40t, hence relatively low current. Conversely, a low time constant (low
ESRdc) yields higher efficiency (t?0, h?1).
Ch003 14 June 2011; 18:44:40

Power and energy 101
3.2 Specific power and power density

Power performance of the ultracapacitor is very important in every application
because the ultracapacitor is used where burst power is needed. Recently, the term
cyclable energy storage has been applied to the ultracapacitor to signify its high
efficiency under power cycling conditions. The most commonly used metric for
power is matched load power, PML, and corresponding specific power defined as
(3.20) and (3.21), respectively:
U 2mx
PML ¼ ð3:20Þ
4ESRdc
U 2mx
SPML ¼ ð3:21Þ
4ðESRdc ÞM
The use of matched load power as an ultracapacitor characterization metric is

discouraged because it has meaning only at the instant of discharge or charge,
before the potential can change. As soon as the potential stated in the two equations
above changes, the matched load value of power can no longer be sustained.
Therefore, when making measurements of power it is better to use the IEC defi-
nitions discussed earlier, or to use power levels that are tied to efficiency. By way
of comparison, the peak power level for batteries specified by the Society of
Automotive Engineers (SAE) [1,2] is stated in (3.22).
2
2 U oc
Ppk ¼ ð3:22Þ
9 Ri
The factor 2/9th in (3.22) is only slightly smaller than the 1/4th of (3.20), so the
peak power definition of a battery over 10 s bears very close resemblance to the
matched load power level, although in the case of a battery the terminal voltage is
assumed to drop to 2/3rd of open circuit potential, Uoc. The SAE Hybrid Battery
Task Force states that the purpose of this peak power test is to determine the
capability of the battery to sustain this peak power level for 30 s over its useable
discharge capacity range.
Pulse duration of 30 s applies to regenerative energy storage systems (RESS)
because a hybrid electric, plug-in electric, and battery electric vehicle must have
sufficient extended peak power in transient conditions to sustain acceleration on
uphill grade and for passing maneuvers. The testing regimes proposed in SAE
standards for RESS components and modules are consistent with the specifications
put forth in the U.S. Advanced Battery Consortium (USABC). Figure 3.9 shows
that standardized procedures for batteries apply constant current to characterize
battery peak power performance. In most applications for high burst power the
demand is for constant power operation. For example, the boosting mode of hybrid
electric vehicles is to provide engine power augmentation at constant power for
Ch003 14 June 2011; 18:44:40

Voltage
V(SOC)
Ohmic
IRC
V-drop
Polarization
V-drop
Current
O
t
I
Power
t
O
Figure 3.9 Battery module peak power waveforms per SAE J1798 for 30 s power
pulse (SAE J2758 redefines the pulse duration to 10 s)
lane change, passing, and grade maneuvers. Since voltage of the battery system is
relatively constant, especially for a power battery, the use of constant current is
acceptable. However, for an ultracapacitor ESS under constant power loading
neither voltage nor current is constant, in fact, both are highly nonlinear as we will
see in the discussion to follow.
+Uc0
Uc(t) U0(t)
C0
ic(t) P0(t)
Figure 3.10 Ultracapacitor under constant power loading
In Figure 3.10, the ultracapacitor is modeled using its simplest equivalent

circuit of linear capacitance, C0, and internal resistance, R. The ideal double layer
capacitance C0 has initial potential Uc0 and at t ¼ 0þ a constant power discharge
pulse, P0(t) ¼ P0, is applied. How the internal potential, Uc(t), and terminal
potential, U0(t), of EDLC respond is the topic of our analysis to follow. In this
derivation, the EDLC current during discharge, ic(t), is dependent on the rate of
change of the EDLC internal potential, Uc(t), and in conformance with its flow
circuit shown in Figure 3.10 is given as (3.23). Our next step is to apply Kirchhoff’s
Voltage Law (KVL) to the circuit in Figure 3.9, which results in (3.24).
dU c ðtÞ P0
ic ðtÞ ¼ C 0 ¼ ð3:23Þ
dt U 0 ðtÞ
Ch003 14 June 2011; 18:44:40

U c ðtÞ ¼ Ric ðtÞ þ U 0 ðtÞ ð3:24Þ
Since the output power is constant at P0, use the definition of ultracapacitor
current, ic(t), given by (3.23) in the expression for terminal voltage, U0(t) ¼ P0/ic(t),
and substitute these into (3.24). Then make the substitution from (3.23) and (3.25)
into (3.24) for EDLC internal potential.
P0 P0
U 0 ðtÞ ¼ ¼ ð3:25Þ
C 0 U_ c C 0 U_ c
P0
U c ¼ tU_ c ð3:26Þ
C 0 U_ c
With (3.26) we have an expression for how the EDLC internal potential varies
with constant power loading, P0. Multiply both sides in (3.26) by the derivative of
Uc to obtain the first order, second degree differential equation (3.27).
1 P0
U_ c2 þ U c U_ c þ ¼0 ð3:27Þ
t tC 0
Equation (3.27) can be written in terms ofpitsffiffiffiffiffiffiffiffitwo roots, l1, l2, as (3.28),
provided the EDLC internal potential U c ðtÞ > 2 RP0 , and the cutoff potential at
which constant power can be sustained during discharge conditions P0 > 0. The
roots of (3.28) are given as (3.29) and the full expression (3.28) is rewritten as
(3.30).
1 P0
U_ c2 þ U c U_ c þ ¼ ðU_ c þ l1 ÞðU_ c þ l2 Þ ð3:28Þ
t tC 0
v"ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
#ffi
u 2
Uc 1 u U 4P
t
c 0
l1;2 ¼ ð3:29Þ
2t 2 t tC 0
0 v"ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
#1
u 2
U u U P
ðU_ c l1 ÞðU_ c l2 Þ ¼ @U_ c þ t
c c 0 A

2t 2t tC 0
0 ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
v" #1
u 2
u
@U_ c þ U c þ t U c P0 A ð3:30Þ
2t 2t tC 0
Ch003 14 June 2011; 18:44:40

pffiffiffiffiffiffiffiffi
Let a ¼ 2 RP0 and u ¼ Uc so that the radical for the first root in (3.30) can be
simplified as follows:
v"ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
#ffi ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi
s
u 2 u 2
U u U P du u P0
U_ c þ
c
t c

0
¼ þ ð3:31Þ
2t 2t tC 0 dt 2t 2t tC 0
s
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi
du u u 2 P0 du u 1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
þ ¼ þ ½u2 a2 ð3:32Þ
dt 2t 2t tC 0 dt 2t 2t
Then simplify (3.32) to a total differential.

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
du u þ ½u2 a2 2t du
¼ ; dt ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð3:33Þ
dt 2t u ½u2 a2
Equation (3.33) for differential time can be re-written as (3.34) by rationali-

zation and the two terms that result for du can be set up for integration as (3.35).
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
dt u u2 a2
¼ 2 du du ð3:34Þ
2t a a2
ð ð ð pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
dt u u2 a2
¼ 2
du du ð3:35Þ
2t a a2
Before continuing with the anti-differentiation called for in (3.35), the initial
conditions on ‘u’ must be stated. By definition u ¼ Uc(t) and the initial value of Uc
(t ¼ 0þ) ¼ Uc0, so let u0 ¼ Uc0 ¼ Umx. When these definitions are applied, (3.35)
can be solved directly to the first term on the RHS, but not the second term since it
involves a variable minus a constant under the radical sign.
ðt ðu ð u pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
dx x t 1 x x2 a2
¼ ¼ t¼ 2
dx dx ð3:36Þ
0þ 2t 2t 0þ 2t u0 a u0 a2
ð u qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 ðu2 u2 Þ 1
t¼ 0 2 2 x2 a2 dx ð3:37Þ
2t 2a a u0
The RHS definite integral can be evaluated by making a further change of

variable, this time to u ¼ a cosh h and du ¼ a sinh h dh so that this integral can be
resolved to (3.38).
Ch003 14 June 2011; 18:44:41

ð u qffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiu !
1 1
x a dx ¼ 2 x ðx a Þ þa ln x þ ðx2 a2 Þ
2 2 2 2 2
ð3:38Þ
a2 u0 2a u0
ð u qffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 1 h pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi n pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffio
x2 a2 dx ¼ 2 u ðu2 a2 Þ þ a2 ln u þ ðu2 a2 Þ
a2 u0 2a
n
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffio
u0 ðu20 a2 Þ a2 ln u0 þ ðu20 a2 Þ ð3:39Þ
Substituting (3.39) into the RHS of (3.37) and simplifying results in the full
solution to the constant power discharge of the ultracapacitor.
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 ðu20 u2 Þ 1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
t¼ þ 2 u0 ðu0 a Þ u ðu a Þ
2 2 2 2
2t 2a2 2a
( pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi)
1 u0 þ ðu20 a2 Þ
þ ln pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð3:40Þ
2 u þ ðu2 a2 Þ
Multiply both sides in (3.40) by 2t and substitute back for u0 ¼ Umx, u ¼ Uc,
and a2 ¼ 4RP0 to yield
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
c0 2 C0
t¼ U Uc þ 2
U mx ðU mx 4RP0 Þ U c ðU 2c 4RP0 Þ
2
4P0 mx 4P0
8 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi9
<U mx þ ðU 2mx 4RP0 Þ>
> =
þ RC 0 ln qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð3:41Þ
: U c þ ðU 2 4RP0 Þ >
> ;
c
Example 3.2: Consider the case of a Maxwell Technologies 3000 F cell having
ESRdc ¼ R ¼ 0.29 mW, C0 ¼ 3150 F, and Umx ¼ 2.7 V. Let this ultracapacitor be
discharged under a constant power P0 ¼ 600 W until its internal voltage reaches
Uc ¼ 2.0 V. Note: The initial condition voltage Uc0 ¼ Umx ¼ 2.7 V.
Solution: The ultracapacitor time constant in this example is t ¼ RC0 ¼ 0.913 s and
pffiffiffiffiffiffiffiffi
the cutoff voltage criteria require that U c > 2 RP0 ¼ 0:834 V, which is satisfied
for Uc > Umx/2. When the stated values are substituted into (3.41), the result is that
t ¼ 8.94 s for Uc ¼ 2.0 V.
Simulation result is tsim ¼ 8.523 s, which is in very good agreement with the
calculated value. If the power level is cut in half, the simulated time becomes
tsim ¼ 17.168 s. For this same case of half power discharge, the analytical time,
t ¼ 17.56 s, is still in good agreement compared to a numerical solver (Figure 3.11).
Ch003 14 June 2011; 18:44:41

Internal volts at P0 = 600 W
2.7
2.4
C1.V [V] 2.2
C1.V [V]
2.0
1.8
1.6
1.3
0 2.5 5.0 7.5 10.0 15.0
Figure 3.11 Simulated voltage waveform, Uc(t) ¼ 2.0 V, when t ¼ 8.515 s for
P0 ¼ 600 W
The terminal voltage of the ultracapacitor under constant power loading can be
determined by subtracting the Ric(t) voltage from the Uc(t) value.
U 0 ðtÞ ¼ U c ðtÞ Ric ðtÞ ð3:42Þ
According to the results of Example 3.2, the values of both Uc(t) and ic(t) are
strong functions of power and time. Furthermore, for the transcendental equation
solved in this exercise it is not possible to obtain an explicit relationship for Uc(t),
and therefore, not for ic(t). The relation for cell potential and time is therefore
implicit and available analytically only over specified voltage intervals. Example 3.3
examines the waveforms for constant power discharge of the ultracapacitor.
Example 3.3: For the ultracapacitor in Example 3.2, use numerical simulation to
obtain the EDLC internal potential, Uc(t); the current, ic(t); and the terminal vol-
tage, U0(t). Consider taking the EDLC potential Uc(t) ¼ 1.35 V.
Solution: Ansys/Ansoft Simplorer V.7 is applied to this example to solve for the
specified variables as functions of time given a loading power of P0 ¼ 600 W
(Figure 3.12).
Current_ P0 = 600 W
Terminal volts P0 = 600 W
508.0
2.7
2.4 400.0
2.2
VM1.V [V]
300.0
I1.I [A]
VM 1.V
2.0
1.8 200.0
1.6
100.0
1.1
0
0 2.5 5.0 7.5 10.0 15.0 0 2.5 5.0 7.5 10.0 15.0
(a) (b)
Figure 3.12 (a) Terminal voltage U0(t) and (b)current ic(t) for constant power
discharge at P0 ¼ 600 W
Ch003 14 June 2011; 18:44:42

For this constant power discharge the ultracapacitor must initially discharge at
227.8 A, increasing nonlinearly to 508 A at 15 s to maintain the terminal power
constant. The initial voltage drop in Figure 3.13 is 66 mV increasing to 144 mV at
the termination of the power pulse (see Exercise 3.14). This difference shows up
clearly in Figure 3.13 as the discharge current increases and terminal voltage
decreases as given by (3.42).
Internal and terminal volts at P0 = 600 W

2.7
2.4
VM1. V [V]
2.2
VM1.1 [V]
C1.V [V]
C1.V[V]
2.0
1.8
1.6
1.1
0 2.5 5.0 7.5 10.0 15.0
Figure 3.13 Comparison of terminal voltage (bottom trace, dotted gray) with
EDLC voltage (top trace, solid gray)
It turns out that calculating the EDLC current, ic(t), and terminal voltage, U0(t),
under constant power loading is very straightforward from the procedure followed
in (3.23)–(3.41). The expression for circuit current (3.23) can now be solved
explicitly by substituting (3.33) for EDLC internal potential rate of change. The
result is a very compact form for ultracapacitor current, ic(t).
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
C0 C0
ic ðtÞ ¼ C 0 U_ c ðtÞ ¼ U c ðtÞ ðU 2c 4RP0 Þ ð3:43Þ
2t 2t
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
!
u 2
1 u Uc P0
ic ðtÞ ¼ U c ðtÞ t ð3:44Þ
2R 2R R
Substitute (3.44) into (3.42) to find the terminal voltage, U0(t), during constant
power discharge.
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
!
u 2
1 u Uc
U 0 ðtÞ ¼ U c ðtÞ Ric ðtÞ ¼ U c ðtÞ þ t RP0 ð3:45Þ
2 2
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
!
u 2
1 u U
U c ðtÞ þ t
c
DU 0 ðtÞ ¼ RP0 ð3:46Þ
2 2
Ch003 14 June 2011; 18:44:42

The voltage difference between EDLC internal potential and terminal potential
is U0 given by (3.46) where the negative sign of the result means a voltage drop.
Ultracapacitor efficiency under constant power discharge is a topic of con-
siderable interest in virtually all applications. For example, a shipyard crane or
rubber tired gantry (RTG) crane that move containers on and off ships and to and
from the docks using drayage trucks are constantly experiencing energy storage
system loads that are constant power (i.e., lifting a mass at constant force and at
constant velocity). Burke [3–5] discusses using power at a specified efficiency as a
better metric for making comparisons of energy storage systems and he derived an
approximate expression for efficiency under constant power by assuming a full
discharge from Umx to Umx/2 at the specified power efficiency Ph. Making the
approximation that the voltage is half, the specified voltage swing, or U ¼ 3/4 Umx,
results in his power at efficiency definition.
9 U2
Ph ¼ ð1 hÞ mx ð3:47Þ
16 R
Example 3.4: For the ultracapacitor treated in Example 3.3, the simulation was
modified to integrate the discharge power with time to get output energy and from
this the discharge efficiency. The ultracapacitor will discharge from Umx ¼ 2.7 V to
U0 ¼ 1.35 V. Find the EDLC internal potential, Uc(t), for this discharge and the
energy efficiency.
Solution: Figure 3.14 shows the energy efficiency of this case. The EDLC potential
Uc(t ¼ 13.44 s) ¼ 1.47 V when U0 ¼ 1.35 V at which point the trace in Figure 3.14
passes through h ¼ 0.95 (note the ordinate units are milli).
P0 = 600 W_energy efficiency

1.00
980.00
960.00
MUL1.VAL
MUL1.VAL
940.00
920.00
900.00
880.00
850.00
0 2.5 5.0 7.5 10.0 15.0
Figure 3.14 Ultracapacitor energy efficiency under constant power discharge of

600 W
Application of (3.47) to this example shows that Ph ¼ P95 ¼ 707 W when

h ¼ 95% and R ¼ 0.29 mW. For a more practical R ¼ ESRdc þ Rconn ¼ 0.34 mW, the
Ph ¼ 603 W, which is consistent with the case studies made in this chapter. Note
Ch003 14 June 2011; 18:44:42

that application of (3.41) to this example results in t ¼ 14 s when h ¼ 95% and

U0 ¼ 1.35 V.
To complete this example, calculate the energy stored, energy dissipated, and
energy output for a pulse duration of T ¼ tf – ti ¼ 13.02 – 0.21 ¼ 12.81 s (obtained
from simulation used to develop Figure 3.14).
C0 2 3150 2
W sto ¼ U mx U 2cf ¼ 2:7 1:472 ¼ 8078 J ð3:48Þ
2 2
W 0 ¼ P0 T ¼ 600ð12:81Þ ¼ 7686 J ð3:49Þ
W disp ¼ W sto W 0 ¼ 392 J ð3:50Þ
Using (3.48) and (3.49) the energy efficiency for this constant power discharge is
W0 7686
h¼ ¼ ¼ 0:9515 ð3:51Þ
W sto 8078
At t ¼ 13.02 s simulation time the efficiency shown in Figure 3.14 is h ¼ 0.9568,

which is reasonable in comparison to the value found in (3.51) due to round off
error and graphic trace function error. Computing dissipated energy is more dif-
ficult in constant power operation because the circuit current is highly nonlinear.
Dissipated energy can be approximated by a simple average of current over the
interval, T seconds, and doing so results in icTM ¼ (444.8 þ 227.8)/2 ¼ 336.3
from which Pdisp ¼ 32.8 W and Pdisp T ¼ 420 J, which is relatively close to the
value given in (3.50).
Another useful method to obtain constant power analytical results for the
ultracapacitor, or capacitors in general, has been developed and discussed by
Prof. Burke of UC-Davis Institute for Transportation Studies, http://www.ucdavis.
edu [6]. The treatment here will be mainly an overview of this useful approach to
illustrate the methodology of deriving an implicit relation between terminal voltage
and time. The approach loses the features of the calculus approach presented earlier,
especially the ease of obtaining circuit current and from it the terminal voltage drop.
Here, terminal voltage is the variable analyzed, given a stated starting condition such
as U0(t) ¼ Umx. The method starts as before, by the application of KVL to the circuit
repeated as Figure 3.15, and will be referred to as the method of variations.
R
+Uc0
Uc(t) U0(t)
C0
ic(t) P0(t)
Figure 3.15 Ideal ultracapacitor equivalent circuit under constant power loading
Ch003 14 June 2011; 18:44:42

The defining relationships and KVL are stated here as (3.52), where
R ¼ ESRdc, C ¼ C0, z ¼ U/Uc0, and 1> z >0.5 for the nominal voltage swing range.
We will use a variable change and define U ¼ U0 and P ¼ P0 to complete the spe-
cification.
ð
1 P
U0 i dt Ri U ¼ 0; i ð3:52Þ
c U
Rearrange (3.52) to put this into the z-factor form.

ð
P P dt
U0 U ¼ R þ ð3:53Þ
U C U
ð
U RP P dt
1 ¼ þ ð3:54Þ
U0 U 0 ðU =U 0 Þ C U 0 ðU =U 0 Þ
2 2
ð
v1 dt RP P
1z¼ þ v2 ; where v1 ¼ ; v2 ¼ ; z0 ¼ 1 v1 ð3:55Þ
z z U 20 CU 20
Differentiate (3.55) with respect to (wrt) z and simplify, making note that z ¼ z0
at t ¼ 0. The result is the total differential given as (3.56). When (3.56) has been
verified, integrate with respect to ‘z’ as a definite integral from z0 to z. The result is
given as (3.57).
v1
z dz þ dz ¼ v2 dt ð3:56Þ
z
1 2
ðz z2 Þ þ v1 ln z v1 ln z0 ¼ v2 t ð3:57Þ
2 0
Completing the process we substitute back for parameter z and arrange as an

implicit function of time, t, and the defined constants n1 and n2. The result is shown
here as (3.58) which is in a form similar to (3.41).
" 2 # ( )
1 2 U U=U 0
t¼ ð1 v1 Þ þ v1 ln ð3:58Þ
2v1 U0 ð1 v1 Þ2
Example 3.5: Repeat Example 3.2 using (3.58) and compare the time computed
here to the time found in that example.
Solution: For this example U0 ¼ 2.7 V, U ¼ 2.0 V, and z ¼ 2/2.7 ¼ 0.7407. For a
discharge power of P ¼ 600 W, R ¼ 0.29 mW, C ¼ 3150 F, t ¼ 0.913 s,
n1 ¼ 0.02387, and n2 ¼ 0.02613. Substituting these values into (3.58) results in
Ch003 14 June 2011; 18:44:43

t ¼ 8.4665 – 0.006011 ¼ 8.4604 s. In Example 3.2 t ¼ 8.94 s, which is in relative

good agreement. Comparison of constant power analytical method results when
U0 ¼ 2.0 V is given in Table 3.1.
The result is that the variational method appears to provide closer agreement
with the numerical solver when the voltage swing range is rather narrow
(z ¼ 0.7407). If U0 ¼ 1.35 V and all else is equal, the method of variations predicts
tvar ¼ 14.707 s for z ¼ 0.5 and the calculus method predicts tcal ¼ 14.03 s and from
the numerical solver tsim ¼ 14 s (Table 3.2).
Table 3.1 Comparison of constant power analytical method

results when U0 ¼ 2.0 V
Z = 0.7407 Calculus method Method of variations

Calculated time, tcal (s) 8.94 8.4604
Simulated time, tsim (s) 8.523 8.523
Time difference (s) 0.417 0.063
Table 3.2 Comparison of constant power analytical method

results when U0 ¼ 1.35 V
Z = 0.500 Calculus method Method of variations

Calculated time, tcal (s) 14.03 14.707
Simulated time, tsim (s) 14 14
Time difference (s) 0.03 0.707
It appears that the variational method does a better prediction job when the
constant power discharge pulse is a shallower discharge, or small voltage swing,
versus the calculus method for which the accuracy improves as the voltage swing
(i.e., SOC variation) is wide.
The interested reader is referred to the work by Verbrugge and Liu [7] that also
expresses the terminal voltage as a function of constant power and initial condition
on EDLC potential. The method is relatively involved but does resolve to a solution
that is very similar to (3.41) in this chapter, only in terms of terminal rather than
EDLC potential. The authors make an excellent contribution by showing that a
capacitor under constant power charge and discharge can be cast into a form that is
solvable using the transcendental Lambert function. A Lambert function is defined
by the expression of (3.59) and has value in calculating a response that involves a
discontinuous change such as the terminal voltage of the ultracapacitor at the
instant of constant power application.
z ¼ W ðzÞ eW ðzÞ ð3:59Þ
Ch003 14 June 2011; 18:44:43

Numerical software packages such as Mathsoft’s MathCAD and Wolfram

Research Mathematica have library models for the Lambert Function.
3.3 Ragone relationships
The derivation of constant power response leads directly to the foundation of the
Ragone relationship for energy storage systems of any variety, especially of electric
and electrochemical storage components. To see this we multiply both sides of
(3.41) by power, P0, to obtain the output energy, W0, as shown in (3.60).
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
C0 2 C0
P0 t ¼ U mx U 2c þ U mx ðU 2mx 4RP0 Þ U c ðU 2c 4RP0 Þ
4 4
8 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi9
<U mx þ ðU 2mx 4RP0 Þ>
> =
þ tP0 ln qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð3:60Þ
: U c þ ðU 2 4RP0 Þ >
> ;
c
When the power level is relatively low in comparison to matched load power,
the first three terms in (3.60) simplify to stored energy in the ultracapacitor. This is
a feasible edit when 4RP0 ? Umx so that the radicals are eliminated and the three
terms collapse as shown in (3.61). A further refinement was made by inverting the
argument of the natural logarithm term and changing the sign of its multiplier. This
was done to show complete agreement with the Ragone relation of (3.62) for output
energy, W0(P), in terms of stored energy, Wsto(U), and dissipated energy, Wd(U).
The functional relation of energy to power, therefore, has ultracapacitor voltage as
parameter.
8 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 9
>
< þ ðU 2c 4RP0 Þ > =
C0 2 U c
P0 t ¼ U mx U 2c tP0 ln qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð3:61Þ
2 :U mx þ ðU 2 4RP0 Þ>
> ;
mx
W 0 ðPÞ ¼ W sto ðU Þ W d ðU Þ ð3:62Þ
To better illustrate the Ragone relationship, (3.1) has been solved for the 3000 F
ultracapacitor for energy versus power for various discharge voltage levels and log–
log plotted. Since the Ragone relation is energy, W, versus power, P, the ratio of the
two is time – the constant time diagonal lines shown in Figure 3.16.
In Figure 3.16 the SE and SP are given as ordinate and abscissa, respectively.
The top trace for Uc ¼ Umx (top trace intersects the characteristic time line (black
diagonal line)), t ¼ 0.87 s, at SPML ¼ 11.6 kW/kg (or 6.28 kW cell level). At the
intersection of the Umx Ragone trace with characteristic time, the projection to the
abscissa is SPML and projection to the ordinate is SEML ¼ SE/2 ¼ (5.6 Wh/kg)/2 ¼
2.8 Wh/kg as shown in the plot. A set of diagonal time lines is drawn to represent
constant power operation at 8%, 25%, and 40% of PML, respectively. For example,
Ch003 14 June 2011; 18:44:43

Ragone for BCAP3000P270 at 2.7, 2.3, 1.8, 1.5 V, 1.35 V

100
30.539
E.sM27 (i)
.2s s 5s
21 6.2 3.5
3600
Euse (P) = Esto – Pdisp T= T= T= 7s
E.sM23 (i) 10
0.08 0.25 0.4 PML 0.8
t=
Specific energy (Wh/kg)
3600
E.sM18 (i)
CI CP
3600
1
E.sM15 (i)
3600
E.sM13 (i)
3600 0.1
T.τM (i)
0.045
0.01
100 1 × 103 1 × 104 1 × 105
100 P(i) 6.795 × 104
Specific power (W/kg)
Figure 3.16 Ragone plot of 3000 F ultracapacitor with discharge voltage level as
parameter
at 0.08PML ¼ 0.08(6280) ¼ 502 W the discharge time to Uc ¼ Umx/2 ¼ 1.35 V is

t * 17 s (caution: the circled points are somewhat right shifted, diagonal intersects
at t1.35 ¼ 7 s to t2.7 ¼ 24 s). Follow a similar procedure for the other power levels to
get a graphical representation of discharge time under constant power.
The appropriate view of the Ragone shown as Figure 3.16 is that the W0(P)
function is, in reality, a band representing SOC ¼ 100% to SOC ¼ 25%. Showing a
Ragone as a single line is not appropriate, unless the intention is to simply outline
the maximum envelope of the available energy.
Example 3.6: Refer to the Ragone in Figure 3.16 and for P ¼ 600 W calculate the
available energy, output energy, and dissipated energy when Umx>U>Umx/2.
Solution: The available energy is the first term on the RHS of (3.61) and
is calculated as (3.63). Discharged energy is the term on the LHS of (3.61)
and is calculated as (3.64). Finally, the dissipated energy is calculated as the
difference.
C 2 3C 2
W avail ¼ U mx U 2c ¼ U ¼ 8201 J ð3:63Þ
2 8 mx
W 0 ¼ P0 t ¼ 600ð12:75Þ ¼ 7650 J ð3:64Þ
Ch003 14 June 2011; 18:44:43

W d ¼ W avail W 0 ¼ 551 J ð3:65Þ
It is insightful to notice that the result of (3.65) can be shown to very

closely approximate the value of the second term on the RHS of (3.61), tP0 ¼
0.87(600) ¼ 522 J.
The interested reader is referred to Exercises 3.15 and 3.16 for extension
of Example 3.6 to the case of ultracapacitor modules. For example, an Nc cell
module will contain Nc Wavail_cell and the power will be Nc Ph_cell. In terms of a
Ragone relationship, both the E and P axes are scaled by Nc so the character of the
chart remains unchanged and time lines are unchanged since the ratio E/P is
constant.
3.4 Ultracapacitors and batteries

The interest in hybridized energy storage systems is growing very rapidly in recent
years [8–16]. For instance, Gonder et al. [8] make the case that a mild-hybrid
electric vehicle with a single 48 V ultracapacitor module (165 F, 8 mW) exhibits the
same performance as a 42 V NiMH battery pack over the UDDS drive cycle and
respectable performance over the aggressive US06 cycle. Verbrugge and Liu [9]
explore the requirements of ultracapacitor modules in direct parallel, or tandem
connection, with a lithium-ion battery for application to range extended vehicles.
Their findings are that using battery temperature as a surrogate for battery life (i.e.,
warranty), a tandem connection improves life. Schupbach and Balda [10,11] eval-
uate the requirements on d.c.–d.c. converter configurations and how energy man-
agement systems can be applied to actively manage power flows between the
battery and the capacitor in hybridized energy storage systems. Miller and Smith
[12] considered the case of ultracapacitor as distributed modules in the vehicle
electrical distribution system as means to stabilize and smoothen the PowerNet in
the presence of load disturbances and high power actuator demands. Distributed
power modules based on ultracapacitor as the power cache provide a new tool for
vehicle designers to deal with increasing electrical loads, the pressing demand for
clean power distribution and localized energy storage. Guidi et al. [13] expand on
earlier d.c.–d.c. converter work to show that a smaller converter can be used to
manage an ultracapacitor cell bank by appropriate cell sizing and converter sizing
as the means to reduce costs. Lee et al. [14] developed a novel switching
arrangement to combine ultracapacitor pack and battery pack into a hybrid energy
pack that utilizes energy management strategy to select the appropriate storage
component based on system demands. Most recently, Miller [15] and Miller and
Sartorelli [16] investigated the optimum architecture for d.c.–d.c. converter
location in combination technologies. Their finding is that interfacing the capacitor
via a d.c.–d.c. converter is best because it retains the fixed voltage d.c. link, is most
amenable to energy management strategy, and provides system level benefits
in idling the converter during periods of low power demand activity, such as a
Ch003 14 June 2011; 18:44:44

vehicle under highway cruise. The converter buffered ultracapacitor architecture

also expands on energy management strategy features by setting current thresholds
on the battery component for which only the battery is active, then battery plus
ultracapacitor, or ultracapacitor only.
There are two key motivators of hybridized electric energy storage. First,
ultracapacitors are very efficient high power caches that can benefit any type of
battery by cycling energy up to 10-s duration for a million-plus cycle thereby
exceeding the lifetime energy cycling of the battery. Second, by offsetting the
cycling component of battery power it is practical to realize increased operational
life, reduced current rates, and lower overall thermal burden (i.e., battery rms
current is a surrogate for operational life). These are very significant motivations to
take a very close look at combination technologies as the marriage of strengths,
optimized energy combined with optimized power, or better, decoupled power and
energy storage systems.
Table 3.3 summarizes the various converter locations in the ultracapacitor plus
battery combination storage system. When the d.c.–d.c. converter, taken as the half-
H converter architecture (more correctly, the mid-point converter), is located
between the capacitor and the d.c. link the performance is superior to the case of the
converter buffering the battery. Moreover, when the converter is on the capacitor
the pack can be configured to have voltage lower than, or higher than, the d.c.-link
rating under fixed d.c. voltage. This opens considerable flexibility in cell sizing and
series–parallel arrangement of cells to optimize reliability of the energy storage
pack.
Table 3.4 summarizes the d.c.–d.c. converter duty ratio control parameters, d1
for boost mode and d2 for buck mode of the inductor input midpoint converter. In
any combination energy storage technology, the two principle components must
adhere to Kirchhoff’s current law at the d.c. link and obey power invariance (in the
ideal case) at the converter input and output ports.
There are now several examples of demonstration and concept vehicles that
have hybridized energy storage systems, specifically, ultracapacitor plus lithium-
ion storage elements with a d.c.–d.c. converter for energy management. During late
2008, AFS Trinity displayed their XH-150 plug-in hybrid SUV near the AFS Tri-
nity Engineering Center in Livermore, CA. The XH-150 is so named because it
claims 150 mpg using the engine and hybridized energy storage system. The spe-
cifics of the energy storage system are not known, but assumed to be one or more
48 V ultracapacitor modules actively combined with a high voltage lithium-ion
battery pack. Another example is the Pininfarina B0 Blue Car that was displayed at
the 2008 Paris Auto Show. The B0 Blue Car employs a hybrid energy storage
system consisting of 30 kWh lithium-polymer battery and bank of ultracapacitors.
The vehicle as an estimated range of 153 mi, a maximum speed of 80 mph, and
battery life is estimated to be 125,000 mi due to the ultracapacitor power assist. In
fact, Bolloré and Pininfarina have teamed up to create a 50/50 joint venture to
produce a full electric vehicle to be marketed under the Pininfarina brand. Whether
this vehicle, shown in Figure 3.17, goes into limited production is still unclear.
Ch003 14 June 2011; 18:44:44

Table 3.3 Pareto evaluation of hybridized energy storage system converter locations
Architecture Robustness Cost Performance Overall

Up-convert þ 1 1 þ
+ Fewer, large UC cells, few Lower voltage Converter operates only Best choice overall and
Battery dc
– M connections semiconductor when needed considering PE
ac
technology advances
Ultraca- + dc Enable & High input current to High bandwidth control
pacitor – dc EMS strategy converter
Stable d.c. link
Down-convert 0 0 1 0
Ch003
+ More, smaller UC cells, High voltage Converter operates only Too many interconnects,
Battery dc
– M more connections semiconductor when needed voltage management,
ac
higher voltage UC system
Ultraca- + dc Enable & Lower input current High bandwidth control
pacitor – dc EMS strategy Stable d.c. link
Converter on battery 1 1 0 1
14 June 2011; 18:44:44

+ More, smaller UC cells, Converter operational Converter fault cannot be Requires ultrarobust
Ultraca- dc
pacitor – M more connections 100% of time tolerated converter and high
ac
performance inverter
+ dc Enable & controller and higher
Battery –
dc EMS strategy current inverter switch
Highly dynamic d.c.-link Thermal concerns Higher thermal burden
voltage
Difficult inverter pulse
width modulation
control
Table 3.4 Ideal converter input–output parameters under power

invariance
Converter Ultracapacitor Ultracapacitor Output

mode voltage, Uc current, Ic power, P0 ¼ UcIc

Buck d2Ud 1 UdI0
d2 I 0
Boost ð1 d 1 ÞU d 1
I
ð1 d 1 Þ 0
UdI0
Ic ¼ inverter input current, i0 ¼ converter output current, Ud ¼ d.c.-link voltage to

traction inverter, Il ¼ traction inverter input current.
Figure 3.17 Pininfarina B0 Blue Car at Paris Auto Show 2008
Example 3.7: It is insightful to cast energy storage systems in terms of real world
application examples. For this case assume a Nissan Leaf BEV having the speci-
fications listed in Table 3.5. We can use this example to contrast power and energy
for the case of an energy battery and compare it to that of a strong HEV battery. In
the strong HEV, such as the Prius, or Ford Escape and Mariner, or the GM 2-mode
vehicles that have battery P/E >15. For example, the Escape HEV has a 1.8 kWh
pack that is capable of 39 kW peak discharge power, for a P/E ¼ 39/1.8 ¼ 21.7.
(a) Compute the Nissan Leaf lithium-polymer battery P/E, (b) determine the Ah
rating, Cb, of the cells, and (c) determine the cell arrangement as Ns series and Mp
parallel, by cell Cb rating.
Solution:
(a) For this BEV case the lithium-polymer battery P/E ¼ 90/24 ¼ 3.75. This is
consistent with electric vehicles battery packs having 1 < P/E < 8.
(b) For E ¼ 24 kWh and given that there are 48 modules means each module is
500 Wh. Taking the lithium-polymer cell voltage at 3.5 V/cell, then for
500 Wh/module there will be four cells in series for 14 V, a nominal auto-
motive voltage, each cell is therefore rated Cb ¼ 500 Wh/14 V ¼ 35.7 Ah.
Ch003 14 June 2011; 18:44:45

Table 3.5 Specifications of Nissan Leaf BEV and lithium-polymer battery pack
Attribute Unit Value Attribute Unit Value

Length mm 4 445 All electric range, km/mi 160/100
AER, on US LA4
mode
Width mm 1 770 Max speed kph/mph 144/90
Height mm 1 550 Traction motor, kW/N m 80/280
power and torque
Battery system Laminated Battery kWh 24
lithium-ion capacity
Peak power kW >90 Specific energy Wh/kg 140
Specific power kW/kg 2.5 No. modules and mass #/kg 48/*200
Battery kW 50 For <30 min charge,
charger: 0–80% capacity
quick (19.2 kW, L2
charger)
Battery kW 1.94 200 Vac, less than
charger: 8 h to charge (L1
home charger)
(c) Suppose further that of the 48 modules in the full Leaf pack, there are Ns ¼ 24
in a string, so the pack voltage becomes Ub ¼ 24 14 ¼ 336 V. To meet the
energy target there must be Mp ¼ E/(Ub Cb) ¼ 2. The pack configuration is
therefore: 24S 2P 35.7 Ah.
If the pack is discharged at 35.7 A, which means a Cb /4 (i.e., C/4 rate) it will
deliver approximately 12 kW to the vehicle driveline, sufficient to maintain high-
way cruise speed for this size sedan. At this C-rate, the fully charged pack has a
deliverable energy of 70% dSOC window, characteristic of a BEV battery, then
70% of 24 kWh ¼ 16,800 Wh is available. This amount of deliverable energy, at
constant power of 12 kW is sufficient for 16,800 Wh/12,000 W ¼ 1.4 h of highway
cruising. Since 12 kW propulsion power is sufficient to propel this size sedan at
60 mph continuous, it, therefore, does meet the AER specification given in
Table 3.5 of 100 mi. For this AER ¼ 100 mi range and an energy consumption of
16.8 kWh, the energy usage per mile becomes, 16,800/100 ¼ 168 Wh/mi. This
appears very low, especially in light of the recently revised economy metrics on the
Chevy Volt PHEV, which states 340 Wh/mi.
To help clarify the power and energy situation of batteries, especially those
being developed for electric vehicles (BEV, PHEV, and REV), consider a battery
manufacturing plant with annualized capacity of 50,000 packs annual production
volume (APV), for this same capacity pack as in the Nissan Leaf of 24 kWh, the
upper bound on pack cost is $14,000. For the 48 module, 4 cells per module, and
total of 192 of the 35.7 Ah lithium-polymer soft pack cells, each cell represents
24,000 Wh/192 ¼ 125 Wh. Furthermore, one can take the cell fraction of pack cost
at 70%, in which case the cell cost turns out to be 0.7(14,000)/192 ¼ $51. This
Ch003 14 June 2011; 18:44:45

amounts to a specific cost, SC ¼ $51/125 Wh ¼ ¢40.8/Wh. This cell cost is right at

what the cost of laptop lithium-ion batteries are (*¢50/Wh), consistent with laptop
cells being energy cells, but small format, not large format cells as required for
vehicle traction applications.
Biden [17] in his update on the progress of the U.S. American Recovery and
Reinvestment Act of 2009 funding (*$787 billion) shows that vehicle batteries are
a top priority and substantial price reductions are projected. Figure 3.18 shows the
learning curve progress on electric vehicle battery price points. The near term
battery cost numbers are higher than what was stated here because Figure 3.18 is
not high volume pack production (50,000 APV), but perhaps 5000 packs per year.
What is interesting about Figure 3.18 is that the average consumption is given as
1 kWh/3 mi ¼ 333 Wh/mi, right at the value now advertised by GM for the
Volt PHEV.
Forecasted cost of a typical electric vehicle battery

$35, 000
$33, 333
$30, 000
$25, 000
$20, 000
Cost
$16, 667
$15, 000
$10, 000 $10, 000

$5, 000
$5, 000 $3,333
$0
2009 2012 2015 2018 2021 2024 2027 2030
Figure 3.18 BEV 24 kWh battery pack cost projection based on sedan getting
3 mi/kWh and 100 mi range
Figure 3.18 projects that this 24 kWh pack will cost $5000 in 2021, or a cell
cost of $18 which means a specific cost SP ¼ ¢14.4/Wh, well below anything seen
in the marketplace today for lithium-ion batteries. The real world operation of these
packs demands at least 4000 deep cycles, sufficient for 14 years of vehicle service,
or life of the vehicle. Temperature is the wild card in all this because lithium-ion
packs are notorious for poor power performance in cold climates. The temperature
limitation again points to the need for, and interest in, combination energy storage
technologies.
This chapter on power and energy will close with details on battery chemistries
of most interest in vehicle and transportation applications and then with a com-
parison to ultracapacitor attributes. Table 3.6 shows the major lithium-ion insertion
chemistry batteries of interest. These lithium-ion variants are distinguished by
cathode materials. For example, the Saft LNCA has a cathode mix that is 80%
Ch003 14 June 2011; 18:44:46

Table 3.6 Lithium-ion chemistries of interest in transportation applications
Lithium-ion type Risk Used in Life

LNCA HV, heat and flam- Mercedes 109
mable electrolyte S-Class BMW 7
leads to fire hazard Series Toyota and
w/o thermal Ford PHEVAzure
controls Dynamics van
LMnO (spinel) Spinel cathode can LG Chem and Spinel life suffers
release oxygen at Compact Power when Mn metal
higher SOC. for Chevy dissolves into
Require SRS and Volt PHEVEner1/ organic electrolyte
additives for life EnerDel for Th!nk resulting in capacity
fade
LFP Iron phosphate is not Daimler AG hybrid Good life and cell
(nanophosphate) very conductive so bus (Orions)Th!nk technology being
voltage is low EV’sMcLaren optimized for ultra
Formula One high power
KERS (introduced in
KERS)
LTO (lithium Low, use of LTO Protera custom High life and high
titanate) anode reduces shuttle bus cycling capability
possibility of since structural
thermal runaway. changes are
Graphite minimized
experiences 9%
volume change
due to structure
variations
nickel, 15% cobalt, and 5% aluminum (plus conductive agent plus binder). LNCA
is the most difficult chemistry to control at high SOC because it is exothermically
very reactive. The fundamental principle of insertion chemistry, known as inter-
calation, is that lithium ions can move into and out of the cell electrodes with ease
without causing molecular structure (and volume) changes to the electrodes. Che-
mistries that do incur electrode structural change are the ones that deteriorate more
rapidly and have issues meeting operation life requirements. It is also not prudent to
say that all lithium-ion chemistries are safe, in all these cells the electrolyte sol-
vents can burn, and without careful design and cell management all the ingredients
for combustion are present: high voltage (>4 V), fuel (anode carbon), oxygen
(liberated by cathode), and heat.
Table 3.7 expands on the attributes of lithium-ion chemistries with more detail
on their chemistry, electrical ratings, and benefits and issues.
In contrast to batteries, the ultracapacitor is power dense (>10 kW/L)
(Table 3.8). Lithium-ion batteries are energy dense (*350 Wh/L). The following
attributes apply to ultracapacitor and battery combinations.
Ch003 14 June 2011; 18:44:46

Table 3.7 Lithium-ion battery chemistry comparisons
Type Chemistry Ratings (PHEV) Benefits/issues

LCO LiCoO2 4.4 V, 140 mAh/g Cost and safety concerns
NCA LiNi0:08 Co0:15 Al0:05 O2 3.9 V, 180 mAh/g Safety concern, life and
performance improved operating
<3.9 V
NMC LiMn1=3 Co1=3 Ni1=3 O2 3.7 V, 200 mAh/g Better performance than NCA but
life remains an issue
LMO LiMn2 O4 4.2 V, 120 mAh/g Low capacity, good power, low cost
but life at temperature is an issue
LFP LiFePO4 3.4 V, 170 mAh/g High power from nanoparticle, good
safety, life is questionable
New AlF3 4.3 V, >200 mAh/g Surface coating of layered oxides
Table 3.8 Theoretical and practical properties of various batteries
System Negative Positive OCV Theoretical Theoretical Practical

electrode electrode (V) (Ah/kg) (Wh/kg) (Wh/kg)
Pb-acid Pb PbO2 2.1 83 171 20–40
Nickel- Cd NiOOH 1.35 162 219 40–60
cadmium
Nickel-metal MH NiOOH 1.35 178 240 60–80
hydride
Sodium- Na S 2.1?1.78 377 787 80–100
sulfur (2.0)
Sodium- Na NiCl2 2.58 305 787 80–100
metal
chloride
(300 C)
Lithium LixC6 Li1–xMO2 4.2?3.0 95 @ 380 150–200
ion M ¼ Co, (4.0) x = 0.6
Ni, Mn
Lithium Li VOx 3.3?2.0 340 884 150
polymer (2.6)
● Ultracapacitors and lithium-ion batteries are complimentary (power vs.

energy).
● Strengths and weaknesses are complimentary (power at low temperature,
energy at mid-temperature ranges).
● Ultracapacitors reduce or eliminate high rate discharge spikes on lithium and
thereby help to enhance SOC window and operational life (i.e., cell tempera-
ture reduction).
Table 3.9 puts the attributes of ultracapacitors and batteries into perspective.
Ch003 14 June 2011; 18:44:46

Table 3.9 Theoretical and practical properties of various batteries
Attribute Unit Ultracapacitor Lithium UC versus Li

Power density kW/L 10 3 þ
Energy density Wh/L 6 200

Cold temperature C <40 Approximately þ
20

Hot temperature C þ65 þ40 þ
Rate capability C/x >1800 <40 þ
SOC window % 100 /*50 þ
Efficiency @ % ?98% /*95% þ
40 C
Actively combining the ultracapacitor with a battery requires a bidirectional

d.c.–d.c. converter. The converter responds to EMS commands that in turn are
determined by some strategy that partitions battery and capacitor power flows all
the while acting to maintain the ultracapacitor within acceptable SOC bounds. For
example, the ultracapacitor component of the hybridized energy storage system
may strive for a nominal SOC0 ¼ 80% so that adequate margin is available for
accepting regeneration energy and sufficient reserve exists for boosting power.
Pconv
Id
Uc Ud
=
= P(V)
Ic IL
Resr Id. Ib
Rint Profile
Uc Uoc
Batt A
Ud
Uc EMS strategy control
Id. emd generator dynamic IL
SOC setpoint state machine
Figure 3.19 Practical implementation of ultracapacitor and battery in active

parallel
The real benefit of actively combining ultracapacitor and batteries is that the
deliverable energy of the battery may be increased. This means that BEV, PHEV,
and REV applications would make optimum use of the ESS pack energy. The
purchase price includes the cost of onboard stored energy and it is more than fair
for the customer to get the most use of energy storage purchased. Figure 3.20
illustrates how this may work out for combination technologies, and at this writing
there is experimental evidence that this is possible.
Ch003 14 June 2011; 18:44:48

Anticipated benefit of active ESS

SOC
100
Charge acceptance 90
85
Useable δSOC
energy with ultracapacitor
+ converter
30
Power at EOL
20
20
Lithium-ion only Lithium-ion + ultracapacitor
Figure 3.20 Hybridized ESS as means to enhance SOC window
Exercises
Exercises 3.1 to 3.9 refer to a large cell ultracapacitor rated C0 ¼ 3000 F, Umx ¼ 2.7
V, R ¼ 0.29 mW, and time constant t ¼ RC0 ¼ 0.87 s. Under constant current (CC)
charging, the ideal current source is placed directly in parallel with the ultra-
capacitor RC0 equivalent. The same for constant voltage (CV) charging, the ideal
voltage source, Vs, is placed directly in parallel with the RC0 equivalent of the
ultracapacitor. The ideal stored energy of this ultracapacitor, Esto ¼ 10,935 J.
3.1 Derive the analytical expression for circuit current, ic(t), and ultracapacitor
voltage, Uc(t) for CC and CV cases.
V s t=t h i
t=t
c ðtÞ ¼
Ans: icv e and V cv
c ðtÞ ¼ V s 1 e
R
I cc
c ðtÞ ¼ I c c ðtÞ ¼
c
icc cc
and V cc t
C
3.2 Using the results from Exercise 3.1, calculate the charge time under CV
condition using the ultracapacitor cell parameters specified. Consider the
ultracapacitor charged when Vc(t) ¼ 0.99 Umx.
t=t
c ðt ¼ T Þ ¼ U mx f1 e
Ans: V cv g ¼ 0:99U mx
T ¼ t lnð0:01Þ ¼ 4:6052t ¼ 4:0065 s
3.3 Using the result of Exercise 3.2, calculate the constant charge current that
must be applied to the ultracapacitor in order to fully charge it in the time
specified: T ¼ 4.0065 s.
Ch003 14 June 2011; 18:44:48

CU mx 3000ð2:7Þ
c ¼
Ans: I cc ¼ ¼ 2021:7 A
T 4:0065
3.4 Compute the source energy required to charge the ultracapacitor under
CV and CC conditions using the specifications and exercise results found
so far.

cc 2
Ic T
Ans: Es ¼
cc
þ I c RT ¼ 10;934:84 þ 4748:94 ¼ 15;683:8 J
cc
2C
ð
V s x=t V 2t h i
s ¼
Ecv e dx ¼ s 1 eT =t ¼ 21;651 J
0T R R
3.5 Compute the dissipated energy under both CV and CC conditions when the
charge time is specified as the CV charge time, T ¼ 4.0065 s.
3
d ¼ I cc RT ¼ ð2021:7Þð0:29 10 Þð4:0065Þ ¼ 4748:94 J
Ans: Ecc 2
U 2mx t h i
d ¼
Ecv 1 eT=t ¼ 10;934 J
2R
3.6 Using the result of Exercise 3.4, compute the ratio of source input energy to
stored energy for CV and CC conditions.
Ans: For the CV case the input energy/stored energy ¼ 21,651/10,935 ¼ 1.98.
For the CC case, input energy/stored energy ¼ 15,683.8/10,935 ¼ 1.434.
Therefore, the CC charge case should be approximately (1.98/1.434) ¼ 1.38
times more efficient.
3.7 Since the ultracapacitor under CV charging is not taken to 100% charge,
calculate the stored energy under CV when the charge is taken to 99% full.
ðT ðT
V 2s h x=t i
Ans: Ecv
sto ¼ Psto dx ¼ e e2ðx=tÞ dx
0 0 R

V 2s 1 1 2T =t T=t
¼ þ e e ¼ 10;717:4 J
R 2 2
3.8 Calculate the energy efficiency under CV and CC conditions for the specified
ultracapacitor during charging. Caveat, use Esto ideal for CC and Ecv
sto for the
CV case.
Esto 10;935
ch ¼
Ans: hcc ¼ ¼ 0:697
Esto þ Ecc
d 10;935 þ 4748:94
Ch003 14 June 2011; 18:44:49

Ecv 10;717:4
ch ¼ cv ¼ ¼ 0:5
sto
hcv
Esto þ Ed
cv
10;717:4 þ 10;935
3.9 What conclusions can be drawn from the comparison of CV versus CC
charging of the ultracapacitor?
Ans: The input energy under CV charging is far higher than under CC
charging, i.e., 1.38 times; consequently, the efficiency under CC charging is
higher, i.e., 1.394 times higher in fact. Also, the internal dissipated energy
under CV is 2.3 times higher than under CC charging. From these com-
parisons the recommendation is to always use CC charging, or at least
semicontrolled charging of the ultracapacitor.
3.10 Apply (3.22) to the case of a Saft VL12V, very high power lithium-ion cell
having Uoc ¼ 3.6 V, and internal resistance, Ri ¼ 0.55 mW. Suppose that a
P/E >100 is needed for 2 s (waveform of Figure 3.9), what is the available
peak power over a 2-s interval?
Ans: Ppk ¼ 5:75 kW

3.11 Following the same procedure of Exercise 3.10, consider a Valence Epoch
lithium-ion module with a specification of Uoc ¼ 12.8 V, a peak current,
Ipk ¼ 200 A for 10 s, and Ppk(10 s) ¼ 2.275 kW. Determine the module
internal resistance and its P/E ratio given that E ¼ 538 Wh.
Ans: Ri ¼ 16 mW and P=E ¼ 4:2

3.12 Apply (3.42) to the case of a 3000 F ultracapacitor having R ¼ 0.29 mW
internal resistance and compute the circuit current, ic (t ¼ 0þ), and for the
end of pulse, ic (t ¼ 8.94 s), when Uc ¼ 1.35 V as done in Example 3.3. The
constant power terminal condition is P0 ¼ 600 W.
Ans: ic ðt ¼ 0þ Þ ¼ 227:8 A and ic ðt ¼ 15:1 sÞ ¼ 497:64 A

3.13 Use the results of Exercise 3.12 in (3.42) to find the terminal voltage U0(t) at
these same time points.
Ans: U 0 ðt ¼ 0þ Þ ¼ 2:63 V and U 0 ðt ¼ 15:1 sÞ ¼ 1:206 V

3.14 Apply (3.46) to the ultracapacitor case discussed in Exercise 3.12 and find
the respective voltage drops during the constant power pulse.
Ans: DU 0 ðt ¼ 0þ Þ ¼ 0:066 V and DU 0 ðt ¼ 15:1 sÞ ¼ 0:144 V

3.15 Repeat Example 3.6 for the case of an Nc ¼ 18 cell ultracapacitor module
rated 48 V, 165 F, 8 mW and find the available energy in Wh, energy
Ch003 14 June 2011; 18:44:49

dissipated in Wh, energy output under constant power, W0, and efficiency
for the conditions specified (P0 ¼ 600 W).
18ð8201Þ 18ð551Þ
Ans: W avail ¼ ¼ 41 Wh; Wd ¼ ¼ 2:755 Wh;
3600 3600
W 0 ¼ N c P0 t ¼ 38:25 Wh;
W0 1
h¼ ¼ ¼ 0:933
ðW 0 þ W d Þ ð1 þ W d =W 0 Þ
3.16 For the ultracapacitor module specified in Exercise 3.15, calculate the Ph
power at the efficiency value obtained in Exercise 3.15 by application of
(3.47). Hint: Do not forget to scale parameters appropriately.
Ans: P93 ¼ 10,854 W, or Nc 603 W, which is in very good agreement with

P0 specified in Exercise 3.15 for cell power level at this efficiency.
References
1. SAE Hybrid Committee, Surface Vehicle Recommended Practice for Per-
formance Rating of Electric Vehicle Battery Modules, SAE J1798, SAE
Hybrid Committee, 2008. Available from http://www.sae.org [accessed July
2008]
2. SAE J2758, Determination of the Maximum Available Power from a
Rechargeable Energy Storage System on a Hybrid Electric Vehicle, SAE
Hybrid Battery Task Force of the Hybrid Technical Committee, Recom-
mended Practice, 26 June 2006
3. H. Zhao, A.F. Burke, ‘Optimum performance of direct hydrogen hybrid fuel
cell vehicles’, The 24th International Battery, Hybrid and Fuel Cell Electric
Vehicle Symposium & Exposition, EVS-24, May 2009
4. A.F. Burke, M. Miller, ‘The power capability of ultracapacitors and Lithium
batteries for electric and hybrid vehicle applications’, Journal of Power
Sources, 2009
5. A.F. Burke, M. Miller, ‘Testing of electrochemical capacitors: capacitance,
resistance, energy density, and power capability’, Electrochemical ACTA,
Nantes, FR, June 2009
6. A.F. Burke, ‘Electrochemical capacitors’, Handbook on Batteries, 2010
7. M.W. Verbrugge, P. Liu, ‘Analytic solutions and experimental data for cyclic
voltammetry and constant power operation of capacitors consistent with
HEV applications’, Journal of Electrochemical Society, vol. 153, no. 6,
pp. A1237–45, 2006
8. J. Gonder, A. Pesaran, J. Lustbader, H. Tataria, ‘Hybrid vehicle comparison
testing using ultracapacitors vs. battery energy storage’, SAE 2010 Hybrid
Vehicle Technologies Symposium, Double Tree Hotel, San Diego, CA, 10–11
February 2010
Ch003 14 June 2011; 18:44:49

9. M. Verbrugge, P. Liu, S. Soukiazian, R. Ying, ‘Electrochemical energy

storage systems and range-extended electric vehicles’, The 25th International
Battery Seminar & Exhibit, Broward County Convention Center, Ft.
Lauderdale, FL, March 2008
10. R.M. Schupbach, J.C. Balda, ‘35 kW ultracapacitor unit for power manage-
ment of hybrid electric vehicles: bi-directional dc–dc converter design’, The
35th IEEE Power Electronics Specialists Conference, PESC2004, Aachen,
Germany, 2004
11. R.M. Schupbach, J.C. Balda, ‘Comparing dc–dc converters for power man-
agement in hybrid electric vehicles’, IEEE International Electric Machines
and Drives Conference, IEMDC’03, Madison, WI, vol. 3, pp. 1369–74, 1–4
June 2003
12. J.M. Miller, R.M. Smith, ‘Ultracapacitor assisted electric drives for trans-
portation’, IEEE International Electric Machines and Drives Conference,
IEMDC’03, Madison, WI, vol. 3, pp. 1369–74, 1–4 June 2003
13. G. Guidi, T.M. Undeland, Y. Hori, ‘An interface converter with reduced VA
ratings for battery-supercapacitor mixed systems’, IEEE Power Conversion
Conference, PCC07, Nagoya, Japan, April 2007
14. B-H. Lee, D-H. Shin, B-W. Kim, H-J. Kim, B-K. Lee, C-Y. Won, et al., ‘A
study on hybrid energy storage system for 42 V automotive PowerNet’, IEEE
VPPC’06, Windsor, UK, September 2006
15. J.M. Miller, ‘Engineering the optimum architecture for storage capacitors’,
Advanced Automotive Battery Conference, AABC2010, Large EC Capacitor
Technology and Application, ECCAP, Session 3, EC Capacitor Storage
System Applications, Omni Orlando Resort, 18–21 May 2010
16. J.M. Miller, G. Sartorelli, ‘Battery and ultracapacitor combinations—where
should the converter go?’, IEEE Vehicle Power and Propulsion Conference,
VPPC2010, University, Lille, MEGEVH Network, Lille, FR, 1–3 September
2010
17. J. Biden, The Recovery Act: Transforming the American Economy Through
Innovation, Office of the Vice President of the United States, Update on
ARRA funding, August 2010
Ch003 14 June 2011; 18:44:49

Ch003 14 June 2011; 18:44:49
Chapter 4
Commercial applications
In this chapter our emphasis shifts from electrochemical fundamentals and

equivalent circuit modeling to the application of ultracapacitor products in com-
mercial systems. In the nontransportation-related applications to be discussed, the
operating voltage and power levels will be high, on the order of 100 s of kW to 10 s
of MW in scale. For example, a commercial uninterruptible power supply may
consist of a 900-V battery bank supported by an equivalent rated ultracapacitor
bank that is capable of fully supporting such high-power loads for 15 s to 15 min.
Details and examples will be presented to highlight the selection of appropriate
energy and power of the energy storage system to meet application goals and the
methodologies employed during a design-in phase.
4.1 Uninterruptible power supplies

An uninterruptible power supply (UPS) installation consists of an energy storage
unit, power converter, and automatic transfer switch that is used to protect sensitive
electrical loads. For example, a hospital, computer center, or banking center must
have reliable and uninterrupted power for critical equipment. The energy storage
portion of a UPS may be a battery bank, a flywheel, fuel cell, or other electrical
energy source that is available independent of the utility connection. In the case of
flywheel energy storage, the modular units are rated to supply an average power for
some critical application for a period of at least 15 s. In the event that a power
disruption occurs, such as momentary sag in utility voltage, short-term outage, or
longer-term loss of power, the UPS instantaneously transfers the critical load from
the utility to the backup energy storage. In many critical applications, a standby
engine powered generator is started during the short interval that the UPS supports
the load. Once started, the generator is available to pick up the critical load from the
UPS and to maintain a prescribed level of power until the utility service is restored.
A UPS has an efficiency of 96% owing to losses in semiconductor transfer
switches, mains contactors, line protection fusing, and matching transformers if
used. Once the UPS bypass switch is activated, the local energy storage is used to
provide load power via a power electronic inverter and generally a delta-wye
interconnection transformer. The power inverter has a nominal efficiency of 94% at
full load, dropping to 92% at 50% load, so thermal management is necessary in any
UPS installation. Dynamic performance of a UPS requires operation over the stated
ch004 7 June 2011; 13:20:46

power factor range within the kVA capability of the inverter, dynamic voltage
recovery to 5% of nominal line voltage within 50 ms and <3% total harmonic
distortion (THD). Additional criteria are listed in Table 4.1 for key parameters and
their tolerance [1,2].
Table 4.1 Eaton UT3220 220-kVA three-phase UPS system parameters
Parameter Symbol Nominal Tolerance

Line input voltage Uin (V) 480 þ10/15%
Input power factor PFin <0.8 lag Uncorrected
Output power factor PFo 0.4 lag ? 0.9 lead Inverter
controllable
Output voltage regulation Uo (V) 480 1%
Input current Iin (A) 265 150% Overload
capable
Output voltage Uo (V) 480 5%
Output voltage transient response dUo (V) Uo 5%
Output frequency fo (Hz) 60 þ/6%
Operating temperature range Top ( C) 0 C ? þ40 C
Battery voltage Ub (V) 408 204 lead-acid cells
Battery charger voltage Ufloat (V) 467 @2.29 V/cell
The UPS run time, trun, depends on the amount of energy storage system
deliverable energy, Wbd, and the average load power, Pavg, as stated in (4.1). A
typical UPS installation, for example, will require that trun ¼ 15 s at an average
power over this interval of Pavg ¼ 1 MW. A worked example will help clarify and
support these specifications.
W bd
trun ¼ ð4:1Þ
Pavg
For a typical lead-acid battery energy storage system, the total stored energy,
Wb, is given by the cell capacity, Cb, in Ah and the full string potential for Nc cells
as
W b ¼ N c C b U oc ð4:2Þ
where Uoc ¼ cell open circuit potential. To satisfy some battery life criteria, the
battery pack will not be discharged below 50% state of charge (dSOC), its dis-
charge rate will be limited so that self-heating is minimized, dmx ¼ 0.9 and the
interface inverter efficiency, hc ¼ 0.94. Therefore, the useable or deliverable energy
that can be supplied to the critical load supported by the UPS when its energy
reserve is that given by (4.2) becomes
W bd ¼ ðhI d mx dSOCÞW b ð4:3Þ
ch004 7 June 2011; 13:20:47

Commercial applications 131
A UPS installation having flywheel and battery storage is illustrated in

Figure 4.1, where three flywheel energy storage (FES) units, each rated 225 kVA,
are paralleled and connected to an additional battery storage rack to meet instal-
lation run time requirements.
Static switch
bypass
Utility
a.c.–d.c. d.c.–a.c. Application
Rectifier Inverter load
Bidirectional
converter
Battery
Motor-
×3 Parallel generator
bank
Flywheel
Figure 4.1 UPS installation having flywheel storage units in parallel with battery
bank
In the installation of Figure 4.1, the static bypass switch is normally ON and
the main UPS charging contactor is ON. The utility power, therefore, directly
supports the application load, while the flywheel and its converter remain in active
standby. Active standby means that the d.c.–a.c. inverter powered by the d.c. link
consisting of flywheel and battery bank is controlled to match the application load
voltage and frequency. Therefore, in the event of a utility outage, the main inverter
is active and can near-instantaneously pick up the application load power by taking
energy from the flywheel and battery bank.
In standby operation, the UPS shown in Figure 4.1 not only supports the
application load but also uses some utility power via the semiconductor rectifier to
support FES standby losses and battery bank trickle charging.
An alternative to flywheel energy storage is the ultracapacitor. In this case the
installation shown in Figure 4.1 can be modified to that shown in Figure 4.2 by
direct replacement of the flywheel with an ultracapacitor bank.
The particular UPS installation shown in Figure 4.2 consists of the series
connection of four heavy transportation modules, HTM125 s, manufactured by
Maxwell Technologies, and three such HTM strings connected in parallel. The
configuration is therefore a 4S3PHTM125. An example will help put this UPS
system in perspective by stating that the ultracapacitor bank must support the same
ch004 7 June 2011; 13:20:47

Static switch
bypass
Utility
a.c.–d.c. d.c.–a.c. Application
Rectifier Inverter load
Ultracapacitor
Battery
BMOD063–125 V
×3 Parallel bank
4s × 3p units
Figure 4.2 UPS having ultracapacitor and battery energy storage system
application load as the FES units shown in Figure 4.1, that is,
3 225 kVA ¼ 675 kVA for 15 s, or Wbd ¼ 10.125 MJ.
Example 4.1: Determine the deliverable energy of the UPS installation of Figure 4.2
given that the HTM125 modules have a capacitance of 65 F, an ESR of 14.8 mW,
and rated voltage of 125 V. Can the 4S 3P HTM125 replace three of the FES
modules shown in Figure 4.1, provided the HTM voltage swings over the full range
of Umx to Umx/2?
Solution: For this example we replace (4.2) with the expression for ultracapacitor
stored energy.
W uc ¼ 0:5C uc U 2mx ¼ 0:5ð3 65Þð4 125Þ2 ¼ 24:375 MJ ð4:4Þ
According to (4.4) the ultracapacitor stores sufficient energy to support the

required 10.125 MJ stated. To complete the solution, it is necessary to modify
the parenthetical term in (4.3) by noting that deliverable energy is determined by
the inverter efficiency and dSOC of the ultracapacitor. The major difference is that
for an ultracapacitor the rate capability parameter, dmx ¼ 1, since charge removal or
replenishment does not have the constraints of a lead-acid battery. Therefore, the
deliverable energy, using the results of (4.4), becomes
W ucd ¼ ðhI dSOCÞW uc ¼ ð0:94Þð0:75Þ24:375 MJ ¼ 17:18 MJ ð4:5Þ
At end of life (EOL), the deliverable energy given by (4.5) will have dimin-
ished by 25%, or to a value of deliverable energy of 12.88 MJ, sufficient to meet the
target energy.
ch004 7 June 2011; 13:20:47

As further clarification on the benefits of ultracapacitor versus flywheel energy

storage in a UPS, consider that a typical 850 V, 225 kVA flywheel with 15 s run
time, or deliverable energy of 3.375 MJ, has a standby loss of 250 W. The flywheel
energy storage system has a mass of 590 kg and a volume of 946 L (nearly 1 m3).
This is typical of units such as the Pentadyne GTX series rated for continuous
operation in a 20 to þ50 C ambient [3]. Using (4.1), the amount of energy stated
here corresponds to an average power of 225 kW.
For similar deliverable energy rating (3.295 MJ), it will require a 9S 1P
HTM125 string of ultracapacitor modules, each rated 125 V, 63 F, and 14.8 mW.
This group of ultracapacitor modules will have a total mass of 522 kg, a volume of
722 L, and only 1.37 W of standby loss (at 1.215 mA leakage at room temperature).
The ultracapacitor pack UPS has an operating temperature range of 40 to þ65 C
and can deliver 225 kW of power at % efficiency according to (4.6), taken from
Reference 4. For this calculation, the pack is rated Umx ¼ 1125 V, Co ¼ 63/9 ¼ 7 F,
and ESRo ¼ 9 (14.8) mW ¼ 0.1332 W.
9 U2
Ph ¼ ð1 hÞ mx ¼ 225 kW ð4:6Þ
16 ESRo
For the stated value of average power, (4.6) can be rearranged to (4.7) for
efficiency at this average power. The meaning of (4.7) is that the pack discharges
from rated voltage to half this voltage over the course of the power pulse. There-
fore, pack efficiency starts very high, >99%, and degrades to the value given by
(4.7) at the end of the pulse time. So, overall, the ultracapacitor efficiency is very
high in both discharging and charging.
16ðESRo Þ
h ¼ 1 Ph ð4:7Þ
9ðU 2mx Þ
Substituting into (4.7) for the HTM125 parameters shows that the 1125 V pack
will deliver Pavg ¼ 225 kW for trun ¼ 14.64 s, very close to the standard flywheel
product but with very much lower standby loss (1.37 W vs. 250 W), at an efficiency
of 95.7%. One clarification, the ultracapacitor modules are equipped with 24-V
cooling fans that are used only when the module temperature exceeds some cus-
tomer-defined threshold. In this case, the 24-V, 1.5-A, 36-W fans, a total of 2 per
HTM or 18 in total will present a load of 648 W. Since the efficiency of the
ultracapacitor at 225 kW is so high (i.e., >95%), the fans are mainly needed only in
warm environments. Figure 4.3 shows the HTM125 module for reference, and
Table 4.2 summarizes the attributes of the flywheel and ultracapacitor for UPS
applications.
Further insights into the application of ultracapacitors in a UPS system are
given in the exercises.
ch004 7 June 2011; 13:20:47

134
Ultracapacitor applications
ch004
(a) (b)
7 June 2011; 13:20:47

Figure 4.3 Ultracapacitor module. (a) HTM125 module and (b) HTM125 cell pack and interconnects [from Maxwell Technologies
HTM125, reproduced with permission]
Table 4.2 Comparison of ultracapacitor and flywheel application in UPS
Attribute Symbol Flywheel Ultracapacitor

storage storage
Deliverable energy Wdel (MJ) 3.375 3.295
Average power Pavg (kW) 225 225
Run time trun (s) 15 14.64
Rated voltage Umx (V) 850 1125
Operating efficiency at Pavg h (%) – >95.7
Operating temperature Top ( C) 20 to þ50 40 to þ65
Unit mass, total storage system Msys (kg) 590 522
Unit volume, total storage system Vol (L) 946 722
Standby power loss (Nominal/w-fans) Psb (W) 250 1.37/648
4.2 Utility voltage stabilizer

The introduction of energy storage into utility grid stabilization is becoming
essential. Mandates to introduce up to 33% renewable energy sources (RES) into
the grid by 2020 are driving the need for stabilization systems such as static volt
ampere reactive (VAR) compensators (SVCs), static synchronous series compen-
sators (SSSCs), static compensators (STATCOMs), and thyristor switched reactors
(TSCs) to become common place. This is especially true in the case of long
transmission lines from RES such as large wind farms and photovoltaic arrays that
are normally long distances from the consumers. Maintenance of voltage regulation
and frequency regulation in the presence of such RES penetration into the grid now
demand introduction of energy storage.
A novel approach to optimizing grid transmission capacity is to control the
energy flows between generation and consumption nodes as Kreikebaum et al. [5]
propose. This is a very good interim measure to curtail overloading certain sections
of lines in the national grid. However, as Makarov et al. argue in Reference 6 it takes
very substantial amounts of energy to perform ancillary services such as voltage and
frequency regulation at the shorter time scale to massive amounts of energy storage
such as pumped hydro or compressed air energy storage to impact the grid during
impending voltage sag. Even novel approaches such as vehicle-2-grid (V2G) where
large numbers of plug-in hybrid electric vehicles (PHEVs) are interconnected to the
grid and available for ancillary services to enabling low-voltage ride through (LVRT)
for 10 s of seconds are possible using the vehicle battery. This is what Rogers et al.
advocate in Reference 7. The same can be said for others who look more closely at
the impact of V2G on the vehicle battery, the need for robust communications for
timing and rate schedules and to ensure that the customer is not stranded [8].
To understand the role of energy storage on the grid, it is essential to first look at
power flows through a transmission line from generating station to a remote custo-
mer. In this case the line is assumed short, only 10 s of kilometers, so that its impe-
dance can be represented by line reactance only. The bus voltage at the customer
location is taken as the zero-phase reference. Figure 4.4 illustrates the case where the
ch004 7 June 2011; 13:20:47

generator voltage, Vs, at angle q, sources a line current, Is, through the line reactance,
Xs, to the remote load bus having reference potential, Vr, at zero angle.
Xs Customer loads
Is SVC
Vs ∠q
Vr ∠0
Figure 4.4 Power transmission from generating station to remote load
The total volt–ampere input at the generator, S ¼ P þ j Q, consists of real

power P and reactive power Q necessary to support the line and load reactance. It is
our intention here to show the impact of adding a static VAR compensator (SVR) to
the load bus to partially neutralize the generator burden of reactive power. To
accomplish this, we first analyze the Poynting vector, S, in terms of system voltages
and currents as follows:
S ¼ P þ jQ ¼ V s I s ð4:8Þ
where
pffiffithe
ffi conjugate of current is denoted by the asterisk and complex notation uses
j ¼ 1 and all quantities in (4.8) are vectors. In vector notation, the line current
(4.9) is calculated in the usual manner taking into account the magnitude and angle
of the respective quantities.
! !
! V s V r V s ffq V r ff0
I s ¼ ¼ ð4:9Þ
jXs X s ff90
Taking the conjugate of (4.9) and clearing the denominator results in the
expression for I s ,
V s ffðq þ 90Þ V r ff90

I s ¼ ð4:10Þ
Xs Xs
Substitute (4.10) into (4.8), and compute the total input volt-ampere (VA) to
the transmission line and its real and reactive power components.
V 2s ffð90Þ V s V r ffðq þ 90Þ
S ¼ V s I s ¼ ð4:11Þ
Xs Xs
V 2s V sV r
S¼ ð jÞ ðsin q þ jcos qÞ
Xs Xs

V sV r V sV r V2
¼ sin q j cos q s ð4:12Þ
Xs Xs Xs
V sV r
P¼ sin q ð4:13Þ
Xs
ch004 7 June 2011; 13:20:48


V 2s V s V r
Q¼ cos q ð4:14Þ
Xs Xs
From (4.13) it is clear that the real power that can be transmitted through a line
of reactance Xs up to the stability limit, q ¼ 90 , is proportional to the source and
load bus voltages and inverse with line reactance, Xs. Putting this into practical
perspective, note that the characteristic impedance of overhead transmission lines is
approximately 300 W. According to (4.14), the reactive power is strongly dependent
on the phase angle of the line current. When the line current angle approaches the
stability limit, the total VA throughput on the transmission line becomes all reac-
tive power. Therefore, some means is needed to compensate for line and load
reactances.
The simplest compensator is to insert a fixed VAR source to the utility load
bus. This is, in fact, done by adding a shunt capacitor at the load bus to compensate
reactive power and thereby limit line reactive power transmission. This action not
only reduces transmission line heating but also acts to maintain the load voltage
within regulation limits. The following examples 4.2 and 4.3 highlight the role of
adding capacitive energy storage to the utility network.
Example 4.2: In Figure 4.5, the supervisory control and data acquisition (SCADA)
blocks are used to monitor the generator and load side voltages, currents, and phase
angles. For the uncompensated case, the line current, Is, is given by the ratio of the
source voltage, Vs, to the line plus load impedances, Zs and ZL, respectively. In this
case, the line current is given asz
Vs 115kff0 115kff0
Is ¼ ¼ ¼
ðZ s þ Z L Þ ðð0:0149 þ 165Þ þ jð0:149 þ 96:1ÞÞ 190:983ff30:228
¼ 602ffð30:228Þ ð4:15Þ
SCADA
Load bus, (V) volts

WMs + AMs Rs Ls Fixed and switched loads
+ +
W A
+ 14.9 mΩ 395 µH 0 A + TCR
+ S1
WML +
Transmission line W Ccomp
+
Es 6 µF Rswl +
+
+ + VMs +
115 kV
− RL Fixed Lcomp + 330 Ω
0° + V V load
SCADA 165 Ω
+
0H Lswl +
Generation L2 Fixed 225 mH
source 255 mH capacitor
+ compensator Switched +
A A load A
AML AMcomp AMswl
Figure 4.5 Transmission line with fixed and switched load and fixed VAR
injection
ch004 7 June 2011; 13:20:48

As shown in Figure 4.6, the source current of 602 Apk lags the generator vol-
tage according to (4.15). When the switched load, Zsw ¼ (330 þ j84.8) W switches
in at t ¼ 71 ms, the source current increases to 920 Apk, representing the combined
loading of the fixed and switched load impedances. The line input starts at 64 to
5 MVA under fixed loading, then transitioning to 102 MVA during the period the
additional load is present.
Source voltage and current
1.15
500.00
VMs.V (V)
AMs.I (A)
–500.00
–1.15
0 25.00 50.00 75.00 100.00 125.00 150.00

t
(a)
Source power and VAR
102.00
80.00
WMs.P (W)
60.00
40.00
20.00
–5.00
0 25.00 50.00 75.00 100.00 125.00 150.00
t
(b)
Figure 4.6 Uncompensated transmission line with fixed and switched loads.
(a) Source voltage (115 kV) and current and (b) source input VA
( ¼ 64 MVA fixed to 102 MVA switched)
In Figure 4.7, the fixed load current (597 Apk, somewhat lower than (4.15)
predicts because the voltage source Vs * 114.9 kV) is the transmission line current
until the switched load is applied. At this point the additional load of 335.6 Apk
adds vectorwise to the fixed load current of 597 Apk. The vector math is carried out
in (4.16)–(4.18).
I L þ I sw ¼ 597ffð30:2Þ þ 335:6ffð14:4Þ ð4:16Þ
ch004 7 June 2011; 13:20:49

I L þ I sw ¼ ½597ð0:8643Þ þ 335:6ð0:9685Þ j½597ð0:503Þ

þ 335:6ð0:2486Þ ð4:17Þ
I L þ I sw ¼ 924:7ffð24:4Þ ð4:18Þ
Load and compensator current

607.50
100.00
AMcomp.I (A)
200.00
AMswl.I (A)
AML.I(A)
–200.00
–400.00
–600.00
0 25.00 50.00 75.00 100.00 125.00 150.00
t
Figure 4.7 Uncompensated network load currents of fixed, switched, and reactive
components
The result of (4.18) can be seen in Figure 4.6a, where 71 ms < t < 125 ms. The
meaning of (4.18) is that the generating station ‘sees’ an effective loading S ¼ V s I s
defined in (4.19) and demonstrating that a substantial reactive power of MVAR
must be output by the generator given by (4.20).
S ¼ 115kð924:7ffð24:4ÞÞ
¼ 106:34 MVAð0:911Þ j106:34 MVAð0:413Þ ð4:19Þ
S ¼ ð96:87 j43:92Þ MVA ð4:20Þ
The reactive power demand gives rise to high line current and additional
heating as noted previously, plus the requirement on the generator to provide this
reactive power via field control. The next example treats the case of adding fixed
compensation to the network in the form of a shunt capacitor.
As a follow-on example to Example 4.2, we consider the case of a fixed VAR
compensator at the customer load bus. Such compensators can be fixed capacitors,
STATCOMs or static VAR compensators. The simplest case to consider here is the
ch004 7 June 2011; 13:20:49

fixed capacitor for VAR compensation. The test case is the same as in Example 4.2,
but with a nonzero capacitive element.
Example 4.3: Referring to Figures 4.8 and 4.9c and vectorially adding the currents
result in a value for the transmission line current, Is. The addition is the same as
(4.16)–(4.18), so only the equivalent of (4.18) will be repeated here.
I L þ I sw þ I comp ¼ 597ffð30:2Þ þ 335:9ffð14:4Þ þ 260ffðþ90Þ
¼ 850:2ffð8:37Þ ð4:21Þ
SCADA
Load bus, (V) volts

+ +
W A
+ 14.9 mΩ 395 µH 0 A + TCR
+ S1
WML +
+
Es 6 µF Rswl +
+
115 kV + + VMs + Fixed + 330 Ω
− RL Lcomp
0° + V V load
SCADA 165 Ω
+
0H Lswl +
Generation L2 Fixed 225 mH
source 255 mH capacitor
+ compensator Switched +
A A load A
AML AMcomp AMswl
Figure 4.8 Transmission line with fixed VAR compensation at load bus (6 mF,
115 kV) capacitor
The source current is therefore 850 Apk at a lagging angle of 8.37 , nearly unity
PF. This result is seen in Figure 4.9b, the current trace when 71 ms < t < 125 ms.
The real benefit of the fixed VAR injection is again evident in Figure 4.3b, noting
the near complete overlap of line current and voltage waveforms. In fact, the actual
angle in this case of 6 mF (29.9 MVAR) injection is only 8.37 .
The result is that, for the same loads at the customer bus, the transmission line
current is reduced from 920 Apk found in Example 4.2 to 850 Apk found in this
example. This is a very substantial reduction of line current considering that the
same 100-MVA load power is supported.
Voltage regulation benefits of VAR compensators will be explored next, since
energy storage systems are necessary for their proper function. Grid voltage
regulation at the customer load bus can be understood by considering the role of
VAR injection at the regulated bus due to fixed capacitor VAR support as
demonstrated in Example 4.3, or with either a static VAR compensator (SVC) as
shown in Figure 4.4 or with a static synchronous series compensator (SSSC), or
ch004 7 June 2011; 13:20:49

Source power and variance

98.00
80.00
WMs.P (W) 60.00
40.00
20.00
–500.00
0 25.00 50.00 75.00 100.00 125.00 150.00
t
(a)
1.15
500.00
VMs.V (V)
AMs.I (A)
AMs.I (A)
10.0 m
0
–500.00
–1.15
0 25.00 50.00 75.00 100.00 125.00 150.00
t
(b)
Load and compensator current
610.00
400.00
200.00
AMcomp.I (A)
AMswl.I (A)
AML.I (A)
–200.00
–400.00
–600.00
0 25.00 50.00 75.00 100.00 125.00 150.00
t
(c)
Figure 4.9 Source power, line voltage and current, and compensator current.
(a) Input power (MVA), (b) source voltage and transmission line
current, and (c) load currents, IL, Isw, Icomp
with a static synchronous compensator, STATCOM. A STATCOM, Figure 4.10, is

a shunt compensator that relies on a current controlled power inverter to inject
currents at the appropriate phase to correct for VAR loading. A popular VAR
ch004 7 June 2011; 13:20:49

correction method is to use the SSSC shown graphically in Figure 4.11, consisting
of a transformer isolated power electronic inverter that is voltage-controlled vol-
tage source converter (VSC). Regardless of the approach taken, the results are
similar to those demonstrated in Example 4.3, namely that line current phase can be
dynamically compensated for unity PF, or close to unity in dynamic systems.
Vr
Rb Lb
Vsb
Ra La
Vsa
3 Customer
Rc Lc loads
Vsc
Vr
Lcb Lca Lcc
↑isb ↑isa ↑isc

+
+
Uchg Ud
Figure 4.10 Illustration of the STATCOM using CSC power inverter
Grid voltage at the customer bus without VAR compensation and with various
degrees, or types, of VAR support can be seen by working with (4.13) and (4.14).
Squaring both of these expressions and summing results in (4.22) for the reference
voltage in terms of P, Q, Xs, and Vs,
2 2
V2 V sV r
P þ Q s
2
¼ ð4:22Þ
Xs Xs
According to (4.22), the possible trajectories of real power, P, and reactive

power, Q, are limited by a circle of radius ðV s V r =X s Þ and offset in the Q-axis by
V 2s =X s . The transmission line reactance limits the voltage regulation capability at
the load bus. Consider a power law relation for P and Q at the customer bus,
according to (4.23), where P and Q have the voltage dependency shown. Further-
more, introduce into (4.22) the fixed capacitor compensation of per unit (pu)
ch004 7 June 2011; 13:20:49

inj
Vabc
Rs Ls
Customer
3 loads
Is
Vs
Transmission line
Vr
Ud
Ess
VSC inverter
Figure 4.11 Illustration of the SSSC using VSC power inverter
magnitude, Bc. With these modifications, (4.22) can be represented as (4.23), which
includes the power law dependency on bus voltage, Vr.
ðP0 V p X s Þ2 þ ½Q0 V q X s ð1 þ Bc X s ÞV 2s 2 ¼ ðV s V r Þ2 ð4:23Þ
With a fixed VAR compensation afforded by the fixed capacitor at the load bus
of pu rating Bc, the effect is to offset the load VARs, Q0, by a constant amount. See
Exercise 4.4 for more details on this point. Controllable VAR injection by STAT-
COM shown in Figure 4.10 or/and SVC as shown in Figure 4.11 is generally used
to dynamically adjust grid node voltage. To view this, a VARsvc term is added to
the reactive component in (4.23), resulting in the voltage expression (4.24).
ðP0 V p X s Þ2 þ ½Q0 V q X s ð1 þ Bc X s þ Bsvc X s ÞV 2s 2 ¼ ðV s V r Þ2 ð4:24Þ
From the perspective of the grid node voltage, the injected VAR support by
either a fixed capacitor or dynamically from a controlled VAR source is to dyna-
mically shift the center point of the circle in the P–Q plane so that load voltage, Vr,
can be maintained stable regardless of VAR loading.
Present SVCs utilize aluminum electrolytic capacitors on the d.c. link, Ud,
shown in Figures 4.10 and 4.11 as the energy storage system (ESS) component.
In the case of the STATCOM, the capacitor voltage is maintained by a charger.
Xi et al. [9] discuss the application of ultracapacitors as the energy storage element
in a static synchronous compensator rated 125 kVA, operating at 480 V and con-
nected to the grid point of common connection (PCC) at 12 kV via an interconnect
transformer. In this system, the STATCOM is voltage controlled with ESS rated
Ud ¼ 600 V.
ch004 7 June 2011; 13:20:49

In the remaining portion of this section, we explore the basic principles of grid
frequency and voltage regulation. It is well known that real power exhibits net-
work-wide influence on the grid frequency. Similarly, reactive power has a more
local effect and influences node voltages on the grid. To explore this further,
consider the phasor diagram of Figure 4.12 showing a source voltage, Vs, at some
angle from the reference node having voltage, Vr.
Vs
dV
f
Vr ΔV
Is
Figure 4.12 Phasor representation of grid voltage at reference node and supply
voltage, Vs
The transmission line voltage drop, DV , is due in large part to the need to
transmit higher than nominal load current due to line and load reactances as
demonstrated in Example 4.3. Equation (4.25) is the decomposition of the phasor
for line voltage drop and is the product of I s Z s where angle f is used in lieu of
angle q.
!
DV ¼ ðR þ jX s ÞðI s cosf jI s sinfÞ ð4:25Þ
s
Carry out the multiplication of (4.25), and put this into complex rectangular
form. Then multiply the result for line drop by the reference node voltage to derive
the expression for the influence of reactive power, Q, at the node.
!
DV ¼ ½R I cos f þ X I sin f þ j½X I cos f R I sin f
s s s s s s s s
ð4:26Þ
¼ DV þ jdV
R½V r DV ¼ V r Rs I s cos f þ V r X s I s sin f ð4:27Þ
J ½V r DV ¼ V r X s I s cosf V r Rs I s sinf ð4:28Þ
Two important conclusions can be drawn from (4.27) and (4.28). According
to (4.28), the transmission line voltage drop is driven by the reactive term since
Rs \ Xs. Similarly, the quadrature component of line voltage drop is due to
real power flow, as will be shown in the following derivations from (4.27)
and (4.28).
ch004 7 June 2011; 13:20:50

D X sQ
V r DV ¼ Rs P þ X s Q ffi X s Q!DV ¼ ð4:29Þ
Vr
D X sP
V r dV ¼ X s P Rs Q ! dV ¼ ð4:30Þ
Vr
The result is that reactive power impacts both components of DV , the direct
line drop and its quadrature component. Using (4.29), it is straightforward to put the
line drop in terms of the transmission line short circuit VA rating, Ssc, by normal-
izing DV to its node voltage, Vr. This result is shown as (4.31) where DV =V r is put
in terms of reactive power relative to the line short circuit rating.
DV X s Q Q
¼ 2 ¼ ð4:31Þ
Vr Vr S sc
A reactive compensator is therefore capable of exerting wide influence over

the node voltage; in many SVC and STATCOM applications, the range of PF
is approximately 0.9 < PF < 0.6, or 0.9 lag to 0.6 leading. To see how a
voltage regulation strategy can be implemented via the SVC or STATCOM, refer
to Figure 4.12 and (4.29) noting that V r ffi V s DV , making the substitution
and arriving at the approximate regulation function for node voltage for a VAR
source, Q.

Q
Vr ¼ Vs 1 ð4:32Þ
S sc
Figure 4.13 clearly illustrates that in order to shift the node voltage from some
dynamic operating point (V,Q) to the nominal point (Vo,Q0), the compensating
gain, kQ, is adjusted so that in the STATCOM, for example, its reactive current
magnitude, Iq is increased sufficiently to deliver leading PF current sufficient to
zero out |Q0 Q|. This, in fact, is how the system operates, and node voltage is
monitored via SCADA equipment, the STATCOM then injects an appropriate level
of compensating VARs, and the voltage is pushed back to its nominal value (see
also References 10 and 11 for additional insights).
Q-V droop
Vs
V0
V
dV kQ
f
Vr ΔV
Ir
Q0 Q
Figure 4.13 Implementation of grid voltage regulation using VAR source
ch004 7 June 2011; 13:20:50

4.3 Wind turbine systems

Wind energy is a fast-growing renewable energy resource across the globe with
over 121 GW installed capacity as of the end of 2008 and doubling roughly every
3 years [12,13]. Turbine ratings range from 1.5 to 7.5 MW and consist of geared
drives to asynchronous generators, typically a doubly fed induction generator
(DFIG) or a direct drive synchronous permanent magnet generator. Large turbines
have rotational speeds of only 11 rpm and typically have wind cut-in speeds of
3 m/s (*7 mph) and cut-out speeds of 25 m/s (55 mph). To put this into perspec-
tive, consider a direct drive 6 MW wind energy converter (WEC) consisting of
turbine, generator, and line interconnected power inverter. At the rotational speeds
given, the turbine shaft torque is 6 MNm (6 million Newton m)! Figure 4.14
illustrates a typical WEC research facility that is used to evaluate the performance
and economy of various manufacturer turbines.
Figure 4.14 Wind energy research facility of the Danish Wind Energy Research
showing turbines of 1.5 to 4.6 MW under test [courtesy of IEEE Joint
IAS/PELS/IES Danish Chapter]
For example, a Vestas 4.5-MW wind turbine designed for off-shore generation will
have a rotor diameter of 120 m and a hub height of 90 m, with gearbox and variable
speed pitch control to a high-voltage DFIG. The swept area of such a large rotor is
more than 1 hectare, or nearly three American football fields in size. Turbine sizes
continue to rise. In 2007, the installed base of wind turbines consisted of 45% in the
750 kW to 1.5 MW class, 47% in the 1.5 to 2.5 MW class, and 8% in the greater
than 2.5 MW class. Currently, the largest turbines being installed are 7.5 MW class
known in some circles as the 1 million barrel equivalent (MBE), with 10-MW
turbines on the drawing boards. This is getting to about as large as single pole
towers can withstand. The 7.5-MW turbine, for example, has a total mass of some
600 ton, with 200 ton being the generator atop the tower the height equivalent of
ch004 7 June 2011; 13:20:50

30 min
10 h/year
UH No power reduction
1h
10% reduction
UHF
35 min 3 min
Power Normal No requirement for
Voltage
reduction operation active power

UN 0%@49 Hz production
15%@47.5 Hz
ULF
1h
10% reduction
UL
47.5 49 50.5 51 53
Frequency (Hz)
Figure 4.15 Danish wind code for transmission networks
the Eiffel Tower. This notion of 1 MBE derives from a 7.5-MW WEC operating
continuously at rated output for its life of 25 years.
Wind turbine control requires a balance between real power output and net-
work frequency as well as the PCC reactive power injection and node voltage
regulation. An illustration from the Blaabjerg and Iov presentation [14] illustrates
one grid interconnection regulation on voltage and frequency and what the WEC
behavior must be. For example, if the grid frequency deviates above 50.5 Hz, there
is no demand for real power generation for 3 min. Rated output of the WEC is
desired if the network frequency is in the range of 49–50.5 Hz and the node voltage
is within the regulation range of ULF to UHF.
One caveat on WEC farms has to do with storm fronts moving across the
cluster of turbines. There have been instances of wind gusts above 55 mph on a
3000-MW wind farm that resulted in generation shedding of 3000 MW in just
minutes and the perturbation this has on the transmission network.
The mechanical power output from a WEC is proportional to the turbine swept
area, A, the wind speed, Vw, air density, r, and the turbine power coefficient, Cp,
also referred to as the Betz coefficient. Turbine power coefficient is a function of
turbine tip speed to wind speed, l ¼ rwm/Vw, where r is the rotor radius and wm the
angular speed. Typical three-blade (Danish concept) turbines have Cp * 0.45 when
4 < l < 5 ( Figure 4.15). Wind power is characterized by its statistical distribution,
typically derived from the Weibull probability function.
Pm ¼ 0:5rC p AV 3w ð4:33Þ
ch004 7 June 2011; 13:20:50

b1
b Vw
eðV w =aÞ
b
hðV w Þ ¼ ð4:34Þ
a a
For the wind turbine generator, the input shaft mechanical power (4.33) varies as
the cube of wind speed with proportionality factors due to air density (1.22 kg/m3)
and blade aerodynamic design. The wind speed probability density function (4.34)
gives the fraction of time the wind speed lies between V0 and V0 þ DV, or V0 < Vw
< V0 þ DV. The site specific parameters, a and b, are needed to characterize the
characteristic wind speed, a, and its shape about this value, b. For example, when b
is low the wind speed exhibits a flatter characteristic than when b is high. High
values of b characterize wind regimes having more variability about a character-
istic wind speed. For b ¼ 2 the Weibull distribution function becomes the Rayleigh
distribution that is typically used in WEC site selection. Figure 4.16 illustrates the
Weibull probability density function (pdf) for different values of the shape factor
when the characteristic wind speed parameter is set to 10 m/s velocity. This wind
speed is typical of most high wind areas. Notice that for b ¼ 1 the distribution is
fairly flat, while for b ¼ 6 it clusters about the characteristic wind speed. A value of
b ¼ 3.44 was intentionally used to show the approximate Gaussian distribution.
0.30 Weibull pdf for b = 1, 2, 3.44, 6 (a = 10 m/s)

0.22
h1 (k)
0.24 h2 (k)
h3 (k)
pdf of wind speed
0.18 h4 (k)
0.12
0.06
0
0
0 2.5 5 7.5 10 12.5 15 17.5 20
0.01 V (k) 19.81
Wind speed, V (m/s)
Figure 4.16 Example Weibull probability density functions: h1, b ¼ 1 is

exponential; h2, b ¼ 2 is Rayleigh; h3, b ¼ 3.44 is Gaussian; and h4,
b ¼ 6 is stronger Rayleigh distribution
There is considerable activity today to characterize and predict the wind

regime at test sites to better predict wind farm available energy. Predicting the wind
is a challenging proposition, so developers elect to space the turbines over a wide
area so that variability in the wind averages out over the turbine clusters. The U.S.
Department of Energy has made such surveys and documented the hourly power
ch004 7 June 2011; 13:20:51

output of a wind farm over 1 month. Figure 4.17 shows that the hourly energy
output is highly variable with a 550–50 MW swing during a day but a nearly con-
stant aggregate output over a month. The capacity factor in this case is approxi-
mately 40%, typical for a wind farm.
700
600
Each day is a different color
- Day 29
500
- Day 9
400
MW
300 - Day 5 - Day 26
200 - Average
100
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
–100
h
Figure 4.17 Hourly power output of a wind farm over 1 month (DOE EIA)
Energy storage for wind energy is now mainly for variable-speed wind turbine
blade pitch adjusters and on the d.c. link of DFIG rotor electronics for regulated
control of output P and Q levels. Figure 4.18 is a phantom view of a wind turbine
gear and generator assembly along with a view into one of the turbine blade
bolt circles showing the a.c. motor drive electronic boxes, motor, and energy
storage units.
Example 4.4: The ultracapacitor wind module shown in Figure 4.19 is used as the
energy storage backup module in a large turbine blade pitch a.c. drive unit rated
300 V. Assume that backup module is activated for 20 of rotation as the blade in a
three-blade turbine passes the tower shadow. Assume also that the turbine is spin-
ning at n ¼ 11 rpm and must supply a power burst at its maximum current rating.
How much charge must the module charger supply over the remaining 340 to
replenish the module SOC?
The methodology to address this design-in and sizing exercise is to first cal-
culate the energy withdrawn from the module, including losses, for each activation
of the module. For this application, there are Nm ¼ 300 V/75 V ¼ 4 modules in series,
each rated 94 F, 15 mW for a total of 23.5 F, 60 mW, Upak ¼ 300 V. The ultracapacitor
pack per a.c. drive, therefore, stores a total of Epak ¼ 0.5(23.5)(300)2 ¼ 1.057 MJ.
Next, compute the rotational characteristics of the turbine blade, wm, angular
velocity, blade tower shadow passing frequency, ftsh, and tower shadow passing
time, ttsh, over 20 mechanical degrees of rotation.
ch004 7 June 2011; 13:20:51

150
Ultracapacitor applications
ch004
7 June 2011; 13:20:51
Figure 4.18 Wind turbine in phantom view and close-up of blade pitch adjuster electric drives including ultracapacitor energy
storage units shown as light gray boxes [courtesy of Maxwell Technologies]
Nominal operating voltage Vd.c. 75

Maximum operating voltage Vd.c. 83
Surge voltage Vd.c. 86
Nominal capacitance F 94
Capacitance tolerance % +20/–0
ESR at RT m 15
Available energy Wh 55
Self-discharge, % in 30-day
% 50
12-h charge & hold, RT
Maximum current
5-s discharge to Umx/2 A 700
Life time, 75 V, RT h 150,000
Cycle life, 75 V to 37.5 V,
# 1,000,000
room temperature RT
Isolation voltage in 300 V
Vd.c. 1600
application
Operating termperature °C –40 to + 65
Figure 4.19 Wind turbine blade pitch adjuster energy storage module and
specifications
2p
wm ¼ n ¼ 0:1047ð11Þ ¼ 1:1518 rad=s ð4:35Þ
60
3wm 3ð1:1518Þ
f tsh ¼ ¼ ¼ 0:55 Hz ð4:36Þ
2p 2p

20 1
ttsh ¼ ¼ 0:303 s ð4:37Þ
120 f tsh
Therefore, the module discharges at Imx ¼ 700 A for ttsh s at 300 V for a dis-
charge energy of Edch ¼ 63,630 (J). This is only 63,630/[(0.75)(1.057 MJ)] ¼ 0.08,
or 8% of the module string deliverable energy. The power level, however, is high,
210 kW. The module power dissipation can be taken as constant over this short
duration pulse, or Pdisp ¼ I mx 2 ESRpak ¼ ð700Þ2 ð0:06Þ ¼ 29:4 kW, and dissipation
energy is Edisp ¼ Pdispttsh ¼ 8908 (J). The result is that the module string charger
must replenish Edch þ Edisp ¼ 72,538 (J). To provide the necessary recharge energy
at 300 V requires a charger rated Ichgr for tch ¼ trotttsh.
2p 6:283
trot ¼ ¼ ¼ 5:455 s ð4:38Þ
wm 1:1518
Edch þ Edisp 72; 538
I chgr ¼ ¼ ¼ 46:95 A ð4:39Þ
U pak ðtrot ttsh Þ 300ð5:15Þ
ch004 7 June 2011; 13:20:51

The previous example highlights the fact that the ultracapacitor backup module
may only experience shallow energy cycles while providing very highpeak power
pulses. In gusting winds when the wind turbine blade adjustment is more dynamic,
the module will be called up for more continuous duty and it will experience both
discharge and recharge from the turbine blade, discharge during motoring operation
and recharge when the blade is reverse rotated while the wind exerts a force on
the blade.
The use of energy storage in wind turbines other than for blade pitch adjus-
ters is to apply the ultracapacitor in the rotor converter circuit of a DFIG. Duan
and Harley [15], along with Abbey and Joos [16], discuss this possibility and
propose the use of a 1200-V, 2.3-F, 1.66-MJ ultracapacitor energy storage bank
on the DFIG rotor d.c. link. This ESS provides 833 A of supply current at 0.5 MW
to the rotor circuit for LVRT capability enhancement. To meet a LVRT require-
ment, the wind turbine must remain connected and deliver output throughout the
duration of the grid fault. Rotor energy storage must be sized to compensate for
stator voltage drop for the DFIG to continue delivering power to the grid. The
following section provides an analysis of the DFIG, and the goal is to show where
the ultracapacitor is applied to have the most benefit for LVRT.
Those familiar with Serbius drives will recognize the modern DFIG as its
power electronic enabled counterpart. In this system, the back-to-back power
inverters between grid and rotor are rated approximately 30% of DFIG output
power. This is one of the main benefits of the DFIG, along with operation at
either subsynchronous or supersynchronous speed. During normal operation at
super synchronous speed, the input wind power, Pwind ? Ps þ PREC and PESS > 0,
so that the ESS unit charges. At subsynchronous speed, the ESS discharges and
Ps þ PFEC ? Pgrid. Figure 4.20 illustrates the main components of the DFIG gen-
erating system and the power flows just mentioned. Not shown are the monitoring
and control circuits necessary to provide appropriate id (reactive power, Q) and iq
(real power, P) commands to the respective rotor side converter (REC) and grid side
converter (FEC).
To begin the DFIG transient model, we write the expressions for stator and
rotor voltages as functions of their respective currents and the machine flux.
Next, define the machine flux in terms of stator, rotor, and mutual inductances,
then substitute these into the voltage expressions and simplify. Easy enough
said, and engineering students with some electric machines background will
recognize the process. The analysis to follow will state the voltages (4.41) and
flux equations (4.42), then skip ahead to the matrix representation of stator and
rotor side voltages in the synchronous reference frame using the definitions and
relationships of (4.40) for the derivative operator, slip frequency, and rotor
transient inductance [17].

d L2
p¼ ; Sws ¼ ðws wr Þ; sLr ¼ Lr 1 m ð4:40Þ
dt Ls Lr
ch004 7 June 2011; 13:20:52

Grid
DFIG
Ps, Qs
Pr, Qr
Pg, Qg
REC FEC
Energy
d.c. link storage
Figure 4.20 Utility interconnected DFIG and power flows [from C. Abbey,
G. Joos, ‘Supercapacitor energy storage for wind energy
applications,’ IEEE Transactions on Industry Applications,
vol. 43, no. 3, pp. 769–76, 2007]
vds ¼ Rs ids þ plds wlqs

vqs ¼ Rs iqs þ plqs þ wlds
ð4:41Þ
vdr ¼ Rr idr þ pldr ðw wr Þlqr
vqr ¼ Rr iqr þ plqr þ ðw wr Þldr
lds ¼ Ls ids þ Lm idr

lqs ¼ Ls iqs þ Lm iqr
ð4:42Þ
ldr ¼ Lm ids þ Lr idr
lqr ¼ Lm iqs þ Lr iqr
Making the substitutions, and noting that the synchronous reference frame will
be denoted by the superscript ‘e’ on all variables results in the following FEC
(4.43) and REC (4.44) side voltages. The second term on the RHS of (4.43) is of
interest because it represents a pure oscillation source. This occurred because in the
stator voltage equations the resistance was neglected, so there is no damping. In a
more refined analysis, this deletion would not have been made. In addition, the
synchronous angular frequency becomes ws, as shown.

veds pLs ws Ls ieds p ws iedr
¼ þ Lm ð4:43Þ
veqs ws Ls pLs ieqs ws p ieqr
ch004 7 June 2011; 13:20:52


vedr Rr þ pLr sws Lr iedr p sws ieds
¼ þ L ð4:44Þ
veqr sws Lr Rr þ pLs ieqr m
sws p ieqs
Solve (4.44) for iedqr , and substitute this into (4.43), and rearrange for iedqs . The
notation here is that xedqs means the synchronous frame vector for stator (or rotor)
quantities. The procedure involves taking the inverse of the 2 2 matrix multi-
plying iedqr in (4.44). Following this methodology results in (4.45), the expression
for the stator (grid) side currents of the DFIG.
" #
ieds 1 p ws veds Lm iedr
¼ ð4:45Þ
ieqs Ls ðp þ w2s Þ
2 ws p veqs Ls ieqr
Substituting and using (4.40) for rotor transient inductance leads to

vedr Rr þ psLr sws sLr iedr
¼
veqr sws sLr Rr þ psLr ieqr
2 3
ðp2 þ sw2s Þ ðpws sw2s Þ
e
Lm 6
6 ðp þ ws Þ
2 2 ðp2 þ w2s Þ 77 vds
þ 4 ð4:46Þ
Ls ðsws pws Þ
2
ðp2 þ sw2s Þ 5 veqs
ðp2 þ w2s Þ ðp2 þ w2s Þ

vedr Rr þ psLr sws sLr iedr
¼
veqr sws sLr Rr þ psLr ieqr
2 3
ðp2 þ sw2s Þ ðpws sw2s Þ

6
Lm 6 ðp2 þ w2s Þ ðp2 þ w2s Þ 7 veds ¼ 0
þ 7
Ls 4 ðsw2s pws Þ ðp2 þ sw2s Þ 5 veqs ¼ ws lds ¼ const
e
ðp2 þ w2s Þ ðp2 þ w2s Þ
The control objective is to add an innovation to (4.46) shown as the second

expression that modifies the stator voltage vector, vedqs , so that when a grid fault
occurs, such as a three-phase short circuit, the DFIG ‘rides through’ the low-vol-
tage event that lasts perhaps 200 ms. During the time the fault persists and the stator
side voltage dips, meaning that vedqs decreases in magnitude, then the DFIG feed
forward controller removes the stator voltage feedback, and in its place incorpo-
rates a new control law that zeroes q-axis voltage and holds d-axis voltage to a
constant term proportional to d-axis flux linkage as shown in [4.46]. The second
term in (4.46) illustrates this innovation so that the rotor voltage, vedqr , is maintained
during the fault and the DFIG continues to generate for the duration of the grid
voltage sag. The innovation amounts to a decoupled rotor current regulator that has
immunity to grid voltage transients such as sag and surge. When the fault is cleared
ch004 7 June 2011; 13:20:53

at the distribution level, the DFIG controller returns to vedqs feedback mode and the
control law given as the first expression in (4.46) is effective. At this point, refer
again to the ESS size applied in Reference 16 that implemented a 2.3 F at 1200 V
ultracapacitor to provide 1.66 MJ of rotor energy support to ride through a grid fault
(strong voltage sag). The useable energy of this ultracapacitor ESS is therefore
1.25 MJ, sufficient to provide 0.5 MW for 2.5 s.
Frequency regulation is another of the grid ancillary services that can be
implemented with ultracapacitor (and with battery and with battery plus ultra-
capacitor combinations) energy storage systems. DeLille [18] describes the appli-
cation of 300-MVAR SVCs to compensate a grid frequency reduction from 50 to
below 47 Hz uncompensated by providing 5 MW of storage for 10 s to a 138-kV
line having 23 GW of loading. The uncompensated frequency droop is 3.5 Hz for
1 min, but only a 1-Hz drop for 10 s compensated.
4.4 Photovoltaic systems

Total solar isolation on earth is some 120,000 TW. By the end of 2007, there were
15 GW of installed grid connected photovoltaic (PV) arrays, and this increased
dramatically in 2009 to 21 GW of installed capacity. Germany, United States, and
Japan have some of the largest installations of PV arrays and account for 89% of
the global installations. Germany, for example, installed 3.8 GW of PV in 2009
alone.
During 2010, the IEEE Power and Energy Society (PES), in cooperation with
EnerNex, is putting together a smart grid information clearing house (SGIC) web
portal. A beta version is currently available from Virginia Tech University [19] that
is open for comment. For example, one project is the University of Nevada Las
Vegas, Integrated PV, Battery Storage, and Customer Products with Advanced
Metering, which is designed to realize an aggressive reduction in peak power
loading at the feeder substation level of 65%. To accomplish this goal, the
demonstration project will promote the design and building of 180 energy-efficient
homes. The energy-efficient homes will have 1.76–2.43-kW roof-mounted solar
panels, tankless water heaters, Energy Star appliances, low E-windows, and
advanced metering. Battery energy storage facilities will be located at the
substation.
Unlike wind energy, the application of energy storage in PV must be at the
panel or array output, and there are currently only demonstrations of using ultra-
capacitor storage. A good example to understand how PV-generated power fluc-
tuates is to examine the effects of an 8-MW PV installation in Alamosa, CO, as
reported by Mooney [20]. In this report, the effects of the PV array output on the
distribution feeder were monitored in a collaborative effort with Xcel Energy (the
utility) and Sun Edison (the PV integrator). Figure 4.21 shows the PV array that is
designed for fixed axis solar tracking (i.e., fixed elevation, fixed azimuth) for best
annual output considering cloud cover months. The output of this array over the
course of 1 day is shown in Figure 4.22, a day that started out clear and sunny and
ch004 7 June 2011; 13:20:53

deteriorated in the late afternoon to partly cloudy. The power variability due to
cloud passing is shown by the nearly 8-MW power fluctuations.
Figure 4.21 Solar PV array rated at 8 MW located in Alamosa, Colorado
Alamosa, CO – 5 min system output

September 4, 2008
900
800 PV system current
700
600
Current (A)
–81% change in 5 min.

500
400
300
200
100
0
4:48 7:12 9:36 12:00 14:24 16:48 19:12 21:36
Date: 9/4/08
Figure 4.22 Solar PV array power output for a representative day
The PV array power rises uniformly from approximately 6:30 to 9:00 a.m.,
leveling off at some 850 A of dc output, which it holds until approximately 4:45 p.m.
in the afternoon. Then, along with the setting sun, the influence of cloud coverage is
ch004 7 June 2011; 13:20:53

very evident, showing some 81% reduction in output in just a 5-min interval. It is
this variability in output that energy storage would help smooth out.
Example 4.5: For the solar panel output shown graphically in Figure 4.22, compute
the necessary size ultracapacitor energy storage to smooth out 5-min dropouts in
power. Assume that the output power drops by 81% in 5 min and resumes after a
drop in 5 min at the same ramp rate.
Solution: For the stated information, the PV array power can be taken as trape-
zoidal-shaped notch in output ramping down 6.48 MW over 5 min, holding at a
level of 1.52 MW, then returning to 8 MW over the next 5 min.
ð8 1:52Þ
EstoPV ¼ ðPr Pmn Þtrt ¼ MWh ¼ 1:08 MWh ð4:47Þ
6
Using available HTM125 ultracapacitor modules having 100 Wh deliverable

energy will take an array of Nm such units. Where the number of modules Nm
needed is found by taking the ratio of EstoPV to EdelHTM.
EstoPV 1:08 106

Nm ¼ ¼ ¼ 10;800 ð4:48Þ
EdelHTM 100
This is a great deal of modules, and given that one HTM has a mass of 58 kg
and volume of 85.82 L, this amounts to a total storage system mass and volume of
626.4 metric ton and 9.269 105 L (926 m3). To put this volume into perspec-
tive, consider a standard 530 container (80 80 600 530 ) with overall inside volume
of 102 m3. Using this metric, the ESS can be accommodated by nine such
containers.
To summarize this chapter for grid applications of energy storage systems1,

consider the following grid functions and their definition.
● Transmission curtailment (TC): Mitigation of power delivery constraint
imposed by insufficient transmission capacity. Energy storage is valued at
market electricity rates for the incremental renewable energy resource (RER)
delivered.
● Time shifting (TS): Firms and shapes RER by energy storage during off-peak
interval (6:00 p.m.–6:00 a.m.), supplemented by power purchased from grid
when RER is inadequate, and discharged during on-peak interval (6:00 a.m.–
6:00 p.m.). Such energy storage is valued at the market rates for time-shifted,
shaped energy.
1
Mears, ‘Application of Energy Storage to Enhance Wind Generation,’ DOE Energy Storage Peer
Review Meeting, 10–11 November 2004
ch004 7 June 2011; 13:20:53

● Forecast hedge (FH): Mitigates errors (shortfalls) in RER energy bid 3 h prior
to a 1-h delivery interval. Energy storage is valued at the incremental value of
RER delivered at market rates.
● Grid frequency support (GFX): Supports grid frequency during sudden large
decreases in RER over a 15-min discharge interval. Energy storage is valued at
the cost of alternative solutions.
● Fluctuation suppression (FS): Stabilizes RER generation frequency by sup-
pressing fluctuations (absorbing and discharging energy during short duration
variations in output). Energy storage is valued at the cost of alternative
solutions.
Exercises
4.1 Compute the standby power loss of the ultracapacitor UPS module with
specifications given in Table 4.2. Consider two cases, (a) for constant cell
leakage current determine the leakage current of the full voltage pack and
leakage in terms of mA/F and (b) for constant parallel resistance such that the
module voltage drops by half over a period of 30 days at room temperature.
Ans: (a) Use the state equation Qsb ¼ CoUop/2 ¼ IsdT where Uop ¼ 100 V for
the HTM125, Co ¼ 63 F, T ¼ 30 day (24 h/day) (3600 s/h) ¼ 2.592106 s
yields Isd ¼ 1.215 mA or 73 mA/F. (b) Requires solution of an exponential self-
discharge characteristic: Usd ¼ Uop{exp[T/tp]} ¼ Uop/2 yields tp ¼ 3.7395
106 s from which the total parallel resistance, Rp ¼ tp/Co ¼ 59.356 kW.
4.2 Using (4.7), calculate the exact efficiency of the ultracapacitor pack at a given
power level of 225 kW using the HTM125 parameters stated in section 4.1 at
BOL and EOL conditions.
Ans: At beginning of life (BOL), the ultracapacitor pack has an internal
resistance, Ri ¼ 9(ESRo) ¼ 9(14.8 mW) ¼ 0.1332 W. Therefore, h ¼ 0.958
(95.8%) at BOL. At EOL, the internal resistance specification is that Ri(EOL)

2(Ri(BOL)), so h ¼ 0.9158 (91.6%).
4.3 Expand on the Example 4.2 when the switched load has leading power factor
(PF). Consider the case of Zsw ¼ (330 j84.8) W. What will be the impact on
the line current during the time interval this load is switched in?
Ans: Best done in simulation as shown here for the source voltage and line
current. The result is a substantial correction of the line current lag angle to
near unity power factor as seen by the expanded trace in Figure 4.23b
(Exercise 4.3).
4.4 Suppose the fixed capacitor VAR compensation at the customer load bus in a
115-kV system is rated approximately 30 MVAR as demonstrated in Example
4.3. Suppose that the utility wishes to implement this fixed VAR compensa-
tion using a capacitor rated for operation at 4160 V. Compute the effective
ch004 7 June 2011; 13:20:54

SCADA
Load bus, (V) volts

+ +
W A
+ 14.9 mΩ 395 µH 0 A + TCR
+
WML + S1
+
Es 6 µF Rswl +
+
+ + VMs +
115 kV
− RL Fixed Lcomp + 330 Ω
0° + V V load
SCADA 165 Ω
+
0H Csw +
Generation L2 Fixed 31.3 µF
source 255 mH capacitor +
+ compensator Switched
A A load A
AML AMcomp AMswl
(a)
Source power and VAR
102.00
80.00
VMs.P (W)
60.00
40.00
20.00
–5.00
0 25.00 50.00 75.00 100.00 125.00 150.00
t
1.15
500.00
AMs.I (A)
VMs.I (V)
AMs.I (A)
10.00 m
–500.00
1.15
50.00 60.00 70.00 80.00 90.00 100.00
t (m)
(b)
Figure 4.23 Line current when a leading PF load is switched at the customer load
bus. (a) Transmission line example with switched leading PF load
and (b) power and current (expanded trace)
ch004 7 June 2011; 13:20:54

value of capacitance necessary at the interface transformer primary and

neglect leakage reactance.
Ans: Utility transformers generally have turns ratios less than 30:1, and this is
the case here. Hint: Set the specified VAR injection provided by the fixed
capacitor of 6 mF at the high-voltage bus equal to the VARs at the 4.26 kV
bus. The result is that capacitance at the 4.26-kV side must be 4.584 mF.
4.5 Apply the results of Example 4.2 for line total VA, |S| ¼ 106.36 MVA, of
which Q ¼ 43.92 MVA, or 1 pu VAR load. Show that the fixed capacitor
discussed in Exercise 4.4 amounts to 0.681 pu and that the SVC must inject
0.319 pu [refer to (4.24)].
Ans: Fixed capacitor VAR injection, Qc ¼ Bc X s V 2s ¼ 29:9 MVAR, and
29.9/43.92 ¼ 0.681 pu.
SVC VAR injection, Qsvc ¼ Bsvc X s V 2s ¼ ð43:92 29:9Þ MVAR ¼ 14
MVAR, or 0.319 pu.
4.6 Find the reactive current control magnitudes for the SVC shown in Figure
4.11 that is rated for total reactive power Qsvc ¼ 125 kVAR off a 480-Vac,
three-phase, VSC.
Hint: For this system assumepffiffiffi three-phase, wye connection, having phase
voltage of U ph ¼ ð480= 3Þ ¼ 277 Vrms and rated phase current of
I ph ¼ 125k=277
pffiffiffi ¼ 150 Arms . The d.c.-link supply nominal voltage is there-
fore U d0 ¼ ð 3 6=pÞU ph ¼ 648 Vdc , and neglect VSC real power control by
setting Id ¼ 0. Determine the VSC reactive current control variable, Iq, for a
nominal VAR injection, Qsvc ¼ 35 kVAR and dynamic VAR injection during
a three-phase fault of Qsvc ¼ 200 kVAR.
pffiffiffi pffiffiffi
Ans: I q ¼ 2Qsvc =3U ph ¼ 2ð35;000Þ=3ð277Þ ¼ 60 A when Qsvc ¼
35 kVAR, and Iq ¼ 340 A when Qsvc ¼ 200 kVAR.
4.7 Size the ultracapacitor needed in Exercise 4.6 so that it maintains a d.c. link
voltage of 600 V at the maximum Iq ¼ 340 A for a six-cycle, 100-ms fault
given the rated d.c.-link voltage is 648 V.
Ans: The ESS voltage drop drives the design requirement in this case. The
number of series connected ultracapacitors for a rated voltage, Ud0 ¼ 648 V,
is Nc ¼ 240 cells, each rated 2.7 V. For less than 48-V drop at Iq ¼ 340 A, the
maximum value of series resistance is 0.142 W. To meet this requirement
with production ultracapacitor cells of the symmetric carbon–carbon type and
nominal time constant of 0.8 s means that C0 ¼ 5.6 F at 648 V. Cell capaci-
tance is therefore Ccell ¼ NcC0 ¼ 1344 F, and the next available production
size is 1500 F with nominal ESR * 0.45 mW.
4.8 Prove to yourself that a 7.5-MW wind turbine generating continuously for
25 years produces the energy equivalent of 1 million barrels of oil.
ch004 7 June 2011; 13:20:54

Ans: Use the standard 42-gal barrel as reference, and take the lower heating
value (LHV) of petroleum.
4.9 Compare the relative rotor diameter of wind turbine generators (WTG) for the
following ratings (750 kW, 1.5 MW, 3 MW, 4.5 MW, 6 MW, and 7.5 MW).
Compute: (a) the rotor diameter for each and (b) the relative swept area values
between adjacent WTG ratings.
Ans: Apply (4.33), and use a Betz coefficient of 0.45, air density of 1.22 kg/m3,
and nominal wind speed of 12 m/s. Tabulate the calculations with (a) column
1 values for rotor diameter, Dr, and (b) column 4 for relative swept area values,
A(i þ 1):A(i) where i ¼ 0, 1, 2, . . . .
Dr (m) A (m2) Pm (kW) A(i þ 1):A(i)

25 490.87 232.84 N/A
45 1,590.43 754.40 3.24
65 3,318.31 1573.99 2.09
90 6,361.73 3017.60 1.92
110 9,503.32 4507.77 1.49
127 12,667.69 6008.74 1.33
142 15,836.77 7511.95 1.25
References
1. Three Phase Uninterruptible Power Supply Unity/I, Guide Specifications,
LTQ-1001C, product UT3220, December 1995. Available at: http://power-
quality.eaton.com
2. Three Phase Uninterruptible Power Supply Unity/I, Planning and Installa-
tion Manual, LTM-0356A, Best Power Technology, product UT3220,
December 1995. Available at: http://powerquality.eaton.com
3. Specifications on 225kVA Unit. Available at: www.pentadyne.com/site/fly-
wheel-ups/specifications.html
4. A. Burke, ‘Ultracapacitor technologies and applications in hybrid and electric
vehicles, Research report supported by the ITS-Davis STEPS program’,
International Journal of Energy Research, July 2009
5. F. Kreikebaum, D. Das, D. Divan, ‘Reducing Transmission Investment to
Meet Renewable Portfolio Standards Using Controlled Energy Flows’,
Innovative Smart Grid Technologies Conference, NIST Conference Center,
Gaithersburg, MD, 19–21 January 2010
6. Y. Makarov, P. Du, M.C.W. Kintner-Meyer, C. Jin, H. Illian, ‘Optimal size of
energy storage to accommodate high penetration of renewable resources in
WECC system’, Innovative Smart Grid Technologies Conference, NIST
Conference Center, Gaithersburg, MD, 19–21 January 2010
ch004 7 June 2011; 13:20:54

7. K.M. Rogers, R. Klump, H. Khurana, T.J. Overbye, ‘Smart-grid-enabled load

and distributed generation as a reactive resource’, Innovative Smart Grid
Technologies Conference, NIST Conference Center, Gaithersburg, MD,
19–21 January 2010
8. E. Pritchard, ‘Plug-in hybrid electric vehicle/plug-in vehicle (PHEV/PEV)’,
North Carolina State University Future Renewable Electric Energy Delivery
and Management (FREEDM) Webnair, 21 October 2009
9. Z. Xi, B. Parkhideh, S. Bhattacharya, ‘Improving distribution system per-
formance with integrated STATCOM and supercapacitor energy storage
system’, IEEE Power Electronics Specialists Conference, PESC2008, Island
of Rhodes, Greece, pp. 1390–5, 15–19 June 2008
10. S. Falcones, X. Mao, R. Ayyanar, ‘Simulation of the FREEDM green hub
with solid state transformers and distributed control’, Proceedings of the
FREEDM Systems Center Annual Review, Florida State University Con-
ference Center, Tallahassee, FL, 18–19 May 2010
11. R.K. Varma, ‘Elements of FACTs controllers’, IEEE Power & Energy
Society Transmission and Distribution Conference & Exposition, Ernst N.
Morial Convention Center, New Orleans, LA, 19–22 May 2010
12. The American Wind Energy Association. Available at: www.awea.org
13. The World Wind Energy Association. Available at: www.wwindea.org
14. F. Blaabjerg, F. Iov, ‘Power electronics and control for wind power systems’,
IEEE Power Electronics and Machines in Wind Energy Applications,
PEMWA2009, University of NE-Lincoln, 24–26 June 2009
15. Y. Duan, R.G. Harley, ‘Present and future trends in wind turbine generator
designs’, IEEE Power Electronics and Machines in Wind Energy Applica-
tions, PEMWA2009, University of Nebraska-Lincoln, 24–26 June 2009
16. C. Abbey, G. Joos, ‘Supercapacitor energy storage for wind energy applica-
tions’, IEEE Transactions on Industry Applications, vol. 43, no. 3, pp. 769–76,
2007
17. J. Liang, W. Qiao, R. Harley, ‘Direct transient control of wind turbine driven
DFIG for low voltage ride-through’, IEEE Power Electronics In Wind
Applications, PEMWA2009, University of Nebraska-Lincoln, 24–26 June
2009
18. G. DeLille, ‘Real-time simulation: the missing link in the design process of
advanced grid equipment’, EDF R&D Project RSSD (new ancillary services
for distribution grid operators and connection of distributed generation),
Department of EFESE (Economic and Technical Analysis of Energy Sys-
tems), internal research report, Site de Clamart, Clamart Cedex, France
19. Web portal platform for smart grid technologies, standards, rules and reg-
ulations, industry use case studies, public awareness and education. Available
at: http://www.sgiclearinghouse.org
20. D. Mooney, ‘Utility scale renewables: renewable and efficiency technology
integration’, Report NREL/PR-550-47146, National Renewable Energy
Laboratory, NREL, Denver, CO, 4 November 2009
ch004 7 June 2011; 13:20:55

Chapter 5
Industrial application
The industrial application of energy storage is a very broad topical area that cannot be
adequately treated in a short chapter. To convey the scope of such a diverse range of
applications, this chapter focuses on just three representative industrial areas:
(1) material handling trucks such as forklifts and front end loaders, (2) cranes and
hoists such as the rubber tired gantry crane used for container loading and unloading
in shipyards, and (3) earth moving equipment such as excavators and drag lines.
Before moving into the topical area of material handling trucks, a short tutorial
on the fuel cell (FC) is presented. It is important to understand the fuel cell energy
converter because it represents one of the main components of the propulsion
system in modern equipment. Figure 5.1 is the conceptual diagram of a fuel cell
showing hydrogen gas flowing to the anode (negative) electrode on the left, and
oxygen, or air, to the cathode (positive) electrode on the right. Hydrogen gas shown
as circles with dot (proton) in center enters the anode and in the presence of a
catalyst at the center membrane shed their electrons to an external circuit (not
shown), leaving the positively charged proton to migrate through the proton
exchange membrane (PEM) to the cathode side.
At the cathode side of the fuel cell electrode depicted in Figure 5.1, oxygen
(solid circles) enters and flows past the PEM where it combines with two protons
that migrated from the anode side and two electrons from the external circuit to
Figure 5.1 Fuel cell electrode structure, anode (–) on left, cathode (þ) on right;
water vapor and droplets are drained at cathode and electric potential
is generated across anode to cathode
ch005 6 June 2011; 16:55:5

form water. Air is forced through the cathode by an electric-driven compressor and
hydrogen enters under 2.5–3 atm pressure. The PEM operates in a nominal tem-
perature range of 65–80 C. The PEM is proton conducting and electron blocking so
that the cell is not internally shorted. Generally, the fuel cell plant operates at 48%
efficiency accounting for the power drain to operate the compressor motor and
auxiliaries. Fuel cells of 30–55 kW can ramp their output at 3 kW/s, but do have
ramp-up delays on the order of seconds, depending on the ambient temperature.
The water, as liquid droplets and vapor, is drained from the cathode and recircu-
lated to humidify the gas streams. The PEM fuel cell generates a potential
according to (5.1) representing the hydrogen to oxygen standard (Nernst) potential,
E0 ¼ 1.23, the activation potential, Uact, the Ohmic drop, UOhm, and the con-
centration potential, Ucon. The fuel cell has open circuit potential, UFC *1 V, at
very low currents. Fuel cell current (5.2) is directly proportional to total hydrogen
molar mass flow rate, dnH2 =dt and, Faraday’s constant, F, and number of cells, Nc.
Thermal power of the fuel cell (5.3) is computed as the difference between fuel
combustion power and electrical power output [1,2]. Where hydrogen fuel com-
bustion power is taken at its lower heating value, LHVH2.
U FC ¼ E0 U act U Ohm U con ð5:1Þ
2F
iFC ¼ n_ H ð5:2Þ
Nc 2
W_ th ¼ W_ comb W_ s ¼ n_ H2 LHVH2 U FC iFC ð5:3Þ
The thermodynamic reaction of hydrogen with oxygen in the fuel cell releases
energy depending on the lower heating value of hydrogen, 120.1 MJ/kg, and the
mass flow rate of hydrogen, dM H2 =dt. Heat flow from the stack is given as thermal
power in (5.3). Electrical power in (5.3) is found by substituting for cell potential
(5.1) and cell current (5.2), demonstrating that mass flow of hydrogen to the anode
controls the reaction rate and hence the available electrical power. It also provides
insight into why the FC response is relatively slow, and therefore one must account
for the delay in building up hydrogen flow through the anode to fuel the reaction.
This explains why ultracapacitors, or batteries, are used in combination with FCs to
provide a buffer during power transients that the FC itself cannot respond to.
Table 5.1 lists the properties of a representative PEM fuel cell that is suited to
operate in environments of 0–40 C, adequate for warehouses and lots.
Table 5.1 Hydrogenics HyPM12 fuel cell
Cell type PEM Peak power, Ppk kW 12.7

Number cells, Nc # 50 Working voltage, Ur V 37–58
Internal resistance, Ri mW 60 Maximum current, Imx Amx 350
Efficiency at peak power # 0.53 Response time, tr s <4
Unit mass, MFC kg 98 Unit volume, Vol L 154
ch005 6 June 2011; 16:55:5

Industrial application 165
5.1 Material handling trucks

All electric forklift trucks are becoming commonplace because of emissions and
health-related concerns of the indoor use of gasoline, diesel, natural gas, and butane
fuel. These earlier propulsion technologies and their hydraulic actuation systems
are giving way to electric drive and actuation. Electric material handling trucks can
be either battery electric or fuel cell electric designs. In this section, we consider the
fuel cell powered unit because the ultracapacitor makes a near ideal energy storage
component for a soft source such as a fuel cell. The FC voltage drops with
increasing load and the ultracapacitor voltage must drop in order to source current.
So, the two make ideal partners in a propulsion power plant and for lift actuation.
This was the main finding of Honda Motor company engineers who developed a
1350 F, 2.7 V ultracapacitor cell having an internal resistance of 2.5 mW packaged
in 80 cell, 200 V modules, two each for a 400 V FC hybrid vehicle [3]. In this
system the FC open circuit voltage was 380 V nominal and ranged from 216 to
432 V with the ultracapacitor pack operating at a nominal 77% state-of-charge
(SOC). Specifics of the ultracapacitor plus FC power plant and material selection
and development of the ultracapacitor cells are specified in References 4 and 5.
In keeping with all the specifics mentioned earlier, we consider here the case of
an FC powered forklift truck, shown in Figure 5.2a, that contains the HyPM12 FC
pack and ultracapacitors, and a representative drive profile, shown in Figure 5.2b.
Forklift truck propulsion current (A)

300
250
200
150
100
50
0
–50 0 5 10 15 20 25 30 35 40
–100
–150
–200
–250
(a) (b)
Figure 5.2 Fuel cell powered forklift truck, 72 V 30 V design. (a) Forklift truck
and (b) drive profile (discharge current is positive)
Initial sizing of the ultracapacitor for the vehicle and current profile, shown in
Figure 5.2, is to first compute the total charge deliverable over the cycle. Table 5.2
summarizes the properties of the load current profile shown in the figure.
Table 5.2 Properties of forklift propulsion current
Integral current, As 2507 Peak–peak current, App 460 rms current, Arms 85.5
Initial voltage, Uco 88 State-of-charge, SOC0 0.82 Cycle duration, s 40
ch005 6 June 2011; 16:55:6

For this application the cell size must be 3000 F due to the high rms current
rating should this profile be repetitive, which in a worst case it could be. Following
on with the sizing exercise, the pack must be capable of delivering a total charge of
Qdel ¼ I t ¼ 150(20) ¼ 3000 C from observation of the current discharge profile
shown in Figure 5.2b. For this design, the initial sizing calls for a 2S 1P 48 V
modules of 165 F, 48.6 V each. The pack is therefore rated at C0 ¼ 82.5 F,
Umx ¼ 97 V with an initial SOC0 ¼ (88/97)2 ¼ 0.82 and internal resistance of
2 ESRdc ¼ 2(8) ¼ 16 mW.
Next, a simulation model is built for the 2S 1P 48 V module configuration
having 80 Wh of deliverable energy, Qdel ¼ 3000 C ¼ C0(Ui – Uf). This relation-
ship, with the initial voltage at Ui ¼ 88 V, means that the final voltage should
remain greater than the minimum allowed, Umx/2 ¼ 48.5 V given in (5.4).
Qdel 3000
Uf ¼ Ui ¼ 88 ¼ 51:6 V ð5:4Þ
C0 82:5
This is very close to the minimum voltage for a good design, so in commercial
applications either a larger cell would be used (if available), or the pack would be
configured to a 2S 2P 48 V based on 1500 F cells for a deliverable energy of
110 Wh, substantially larger than the 2S 1P 48 V with 3000 F cells and 80 Wh
deliverable. For now, the case study relies on the 3000 F cell module for which
simulation results are shown in Figure 5.3.
To summarize the results for the forklift ultracapacitor-only ESS, the following
observations are made:
● Discharge current peaks at 266 A and charge current at 195.5 A (Figure 5.3c).
● Given SOC0 ¼ 0.82, the terminal voltage drops to 53.8 V, which is still above
the minimum threshold.
● Output power is 22 kW on discharge and 11 kW on charge (Figure 5.3b).
● Minimum SOC is 0.318 recovering to 0.377 after the regen pulse.
Losses in the ultracapacitor ESS pack are summarized in the simulation to quantify
those periods contributing most to energy loss ( Figure 5.4). As seen in Figure 5.4, the
losses peak at the extremes in current and reach values of 761 W at t ¼ 2 s the peak of
motoring current, 237.8 W at t ¼ 10 s, and 414.8 W at t ¼ 25 s the peak of regen cur-
rent. The integral of the power dissipation over the drive profile is Wd ¼ 1.64 Wh total.
The energy efficiency for this profile is h > 97.7%, which is very high, because peak
power of 22 kW is a small fraction of the 2S 1P 48 V module P95 value.
A similar procedure would be followed to design-in the lift actuator function. In
this case, the lift actuator is a regenerative electric drive that operates in much the
same manner as the propulsion drive discussed. Given the characteristics of the lift
actuator, the maximum design load, and the lift rate, the additional energy storage
could be designed-in. Also note that the ultracapacitor ESS for propulsion power
ch005 6 June 2011; 16:55:6

Rsa C1
2S × 1P × 48 V module 21.6 mΩ 14 F
Rconn Lconn
2.52 mΩ 2.3 PH Rs
Forklift drive
10.8 mΩ
profile + VM1 FML1 EQU XY1 C0
tY I1 V P0: = VM1·V*I1·I XY # Rp
88 V
SOC: = (C0·V/97)^2
23.2 kΩ
Pdisp: = Rconn·R*Rconn·I^2 + Rs·R*Rs·I^2 + Rsa·R*Rsa·I^2
Eff: = W0·VAL(W0·VAL + Wdisp·VAL)
W0 Wdisp
P0 I Pdisp I
119.4 0
– 0.2777 m 0.2777 m
(a)
Forklift terminal power (kW)
2S1P4Mod SOC
22.10
1.00
15.00
800.00
10.00 700.00
SOC
600.00
P0
5.00
500.00
0 400.00
–5.00 300.000
–11.40 200.00
0 10.00 20.00 30.00 40.00
0 10.00 20.00 30.00 40.00 t
t
(b)
Forklift terminal voltage Forklift terminal current
88.00 266.00
200.00
80.00 150.00
VM1.V (V)
75.00 100.00
Il.I (A)
50.00
70.00 0.00
65.00 –50.00
–100.00
60.00
53.8 –196.00
0 10.00 20.00 30.00 40.00 0 10.00 20.00 30.00 40.00

t t
(c)
Figure 5.3 Simulation results for the forklift truck example. (a) Simplorer
simulation model, (b) terminal power and SOC of ultracapacitor pack,
and (c) ultracapacitor ESS terminal voltage and current
only based on the 2S 1P 48 V module is not adequate to accommodate lift

operations without the FC providing either propulsion power or lifting power (12 kW
rating). In a more complete system simulation the FC would be modeled, including
its delay characteristics, and combined with the ultracapacitor ESS to determine
whether or not propulsion and lifting can be accommodated simultaneously [6].
ch005 6 June 2011; 16:55:6

Pack power dissipation

770.00
600.00
500.00
400.00
Pdisp
300.00
200.00
100.00
0
0 10.00 20.00 30.00 40.00
t
Figure 5.4 Power dissipation of ultracapacitor ESS used in forklift truck
5.2 Harbor cranes and rubber tired gantry cranes

The California Air Resources Board (CARB) data indicates that rubber tired gantry
(RTG) cranes are idle for 30% of the time, with idle periods of over 2-h recorded.
Elimination of idle times > 10 min would mean a saving of 120 gal of diesel fuel
and 1 metric ton of CO2 emissions would be avoided each year per crane on average.
For California ports only this nonidling regulation of harbor and RTG cranes would
eliminate 285 metric ton of CO2 each year and save an estimated 29,000 gal of diesel
fuel. Figure 5.5 shows a typical RTG (also called straddle crane) that is capable of
hoisting 90,000 pounds up to 18 ft. The inside width of the RTG is 39 ft. A diesel-
fueled generator provides power for the drive wheels and for the hoist mechanism. It
Figure 5.5 Rubber tired gantry crane with straddle carrier
ch005 6 June 2011; 16:55:7

is known that with highly cyclable energy storage such as ultracapacitors (and fly-
wheels), it is feasible that diesel generator emissions can be reduced by 65% and fuel
consumption reduced by 20–25% of nonregenerative designs. Furthermore, lift cycle
times can be reduced by 15% with the more responsive regenerative units.
Table 5.3 summarizes the equipment type, pollutants, and approximate fuel
costs of various harbor cranes in operation. Emissions result primarily from
extended idling times, typically 10–30 min with lower numbers of events at 1–2 h
and even longer.
Table 5.3 Port crane equipment and

emissions recorded1
Equipment type Engine power at idle (hp)

RTG 9.85
Side pick 3.3
Top pick 4.46
Pollutant Emission rate (g/hp-h)
PM 1.35
NOx 15.9
CO2 2562
GHG and fuel cost Values
CO2 (g/gal) 9860
Cost ($/gal) 2.87
1
DOE Energy Information Administration, 8/17/2009
Example 5.1: A regenerative traction drive is applied to an RTG to reduce diesel

emissions at a busy port. The electric drive system operates with a nominal 630 V d.c.
link that can swing between 520 V (minimum) and 745 V (maximum). Operation
above 745 V introduces a dump resistor to bleed off excess regenerative power, and
when performing below 520 V the ultracapacitor bank is not connected to the
inverter. For this application, determine the required number and configuration of
standard 48 V ultracapacitor modules to meet the power demands while remaining
within the specified d.c.-link voltage limits. The RTG, ultracapacitor ESS con-
tainer, and representative power profile are shown in Figure 5.6.
Solution: For the specified voltage range the number of 48 V modules,
Nm ¼ Integer{Umx/(48.6)} ¼ 15. With Nm modules in series the upper voltage
bound is satisfied. The nominal d.c.-link voltage represents an initial
SOC0 ¼ 0.715, with lower bound of 520 V being SOCmn ¼ 0.487.
The question is can a single string of 48 V modules of 3000 F cell make-up
meet the requirements. Each module is 165 F, 48.6 V, 8 mW. Therefore, a
15S 1P 48 V pack will have a full rating of C0 ¼ 165/Nm ¼ 11 F, Ud ¼
Nm(Umod) ¼ 745, Ri ¼ Nm(ESRmod) ¼ 0.12 W, and peak currents of > 1000 A are
feasible with this cell pack.
This example illustrates the benefit of simulation. Using the same model as
shown in Figure 5.3, modified for a 15S 1P 48 V, and configured to accept a
power input rather than current profile, the result will be exact representations for ESS
pack current, voltage, and other variables of interest. Since the equivalent circuit
ch005 6 June 2011; 16:55:7

(a) (b)
RTG propulsion power

60000
50000
40000
Power (W)
30000
20000
10000
0
0 5 10 15 20 25 30 35
Time (s)
(c)
Figure 5.6 RTG example [courtesy: Maxwell Technologies]: (a) RTG with
electric drive traction motors, (b) ultracapacitor pack, two of the 48 V
modules not shown, and (c) power profile smoothed
model solves for node voltage and branch currents, the approach taken is to have the
simulator divide the input power by the input node voltage, regardless of how
dynamic that voltage may be. This is done and the results are shown in Figure 5.7.
To summarize the simulation results for this example we have:
● Pack voltage remains within 745–520 V and given the pack initial condition it
is 630–569.6 V.
● Peak ultracapacitor pack current is 85.3 Apk, with an rms value of 20.8 Arms.
● Integral {I dt} ¼ 718.2 A-s and energy efficiency h ¼ 0.99.
● Total output energy W0 ¼ 118.5 Wh and total dissipated energy
Wdisp ¼ 1.14 Wh.
● Ultracapacitor SOC swing for the given initial condition on pack voltage is
0.585 < SOC < 0.715.
Example 5.1 does not consider replenishment of the ultracapacitor ESS, so two
scenarios are discussed to evaluate the benefits. First, the RTG diesel-driven gen-
erator idle output power, given in Table 5.3, is augmented for 1 and 10 min
ch005 6 June 2011; 16:55:7

RTG terminal voltage RTG terminal current

630.00 85.5.
620.00 70.00
60.00
I1.I (A)
610.00
VM1.V (V)
50.00
600.00 40.00
30.00
590.00 20.0.
580.00 10.00
0
568.00
0
00
0
.0
.0
.0
.0
.0
.0
5.
10
15
20
25
30
35
0
00
0
.0
.0
.0
.0
.0
.0
t
5.
10
15
20
25
30
35
t
RTG terminal power (kW) 15S1P4Mod SC
52.00 1
900.00
40.00 800.00
700.00
30.00 SOC 600.00
P0
500.00
20.00 400.00
300.00
10.00 200.00 0
00
0
0
.0
.0
.0
.0
.0
.0
5.
10
15
20
25
30
35
0
00
0
.0
.0
.0
.0
.0
.0
t
5.
10
15
20
25
30
35
t
Cycle energy dissipate Cycle energy efficiency
118.50 1
100.00
10.00 * W0.VAL
800.00
80
Wdisp.VAL
W0.VAL
600.00
Eff
60
40 400.00
20
0 200.00
0 20.00 35.00
0
00
0
0
.0
.0
.0
.0
.0
.0
t
5.
10
15
20
25
30
35
Figure 5.7 Results of RTG power profile simulation; terminal voltage, terminal
current; power profile from simulator 15S 1P 48V pack SOC;
output and dissipated energy, energy efficiency plot
recharging of the ultracapacitor ESS bank. The idle power of 9.85 hp (7.348 kW)
will be increased to some value to replenish the ESS within the time limits pro-
posed. Second, to accomplish this it is most feasible to use simulation, but a direct
analytical approach helps quantify the additional generation power level needed
and to base recommendations on these.
W UC ¼ 0:5C~0 ðU Þ½U 2i U 2f ¼ 0:5ð11:88Þ½6302 569:62 ¼ 430; 387 J ð5:5Þ
Based on (5.5) the additional power required from the RTG generator for 1 min
would be 7.17 kW and for 10 min, 717 W. Similarly a 50 kW diesel generator operating
at 7.34 kW at idle would benefit from a 7.17 kW load rather than a 717 W additional
ch005 6 June 2011; 16:55:8

load. The part load efficiency of compression ignited direct injected (CIDI) engines (i.
e., diesel engine) is much higher than that of naturally aspirated spark ignited engines
(i.e., throttled engines), so an additional light load changes the efficiency very little.
The goal of energy storage is nonidling to completely eliminate idle emissions and fuel
consumption as noted earlier. To conclude this discussion, the simulation carried out in
Example 5.1 is modified to include 60 s of 7.17 kW charging power to the ESS bank.
Figure 5.8 summarizes the benefits for comparison with Figure 5.7.
RTG terminal power (kW)

52.00 15S1P4Mod SOC
1
40.00 900.00
30.00 800.00
700.00
P0
20.00 SOC 600.00

500.00
10.00
400.00
300.00
–7.25 200.00
0 20.00 40.00 60.00 80.00 100.00 0 20.00 40.00 60.00 80.00 100.00
t t
(a)
RTG terminal voltage
631.50 RTG terminal current
85.50
620.00
610.00 60.00
VM1.V (V)
I1.I (A)
600.00 40.00
590.00 20.00
580.00
–13.00
568.00
0 20.00 40.00 60.00 80.00 100.00 0 20.00 40.00 60.00 80.00 100.00
t t
(b)
Power pack dissipation
Cycle energy dissipation
118.50
800.00
100.00
10.00 * W0VAL
600.00 80.00
Wdisp.VAL
W0.VAL
Pdisp
400.00 60.00
40.00
200.00
20.00
0 –1
0 20.00 40.00 60.00 80.00 100.00 0 20.00 40.00 60.00 80.00 100.00
t t
(c)
Figure 5.8 RTG power profile including charge replenishment. (a) Power (kW)
profile with regeneration @ 7.17 kW SOC (0.715?0.71499),
(b) terminal voltage (630?629.7 V) with regeneration. Current with
regeneration (approximately 12 A charge); (c) power dissipation
with regeneration energy discharged and absorbed, lost energy
ch005 6 June 2011; 16:55:9

Observations from Figure 5.8 that include ultracapacitor bank recharge at 7.17 kW
are the following:
● Recharge is under the condition of constant power (CP).
● Ultracapacitor pack voltage returns nearly identically to its initial condition
when the same charge is replenished as withdrawn. This is because the energy
loss is so low.
● Ultracapacitor pack SOC returns nearly identically to its initial value (SOC0).
● Current during CP recharge is a very shallow decreasing slope as the voltage
ramps up.
● Losses, and energy loss, are always positive.
● Energy balance is shown in the ultracapacitor energy accumulation graph.
To continue with the ESS charge replenishment topic discussed in conjunction with
Figure 5.8, consider that the electric traction drive of the RTG is the source of
regeneration power. The electric propulsion drive itself may be rated 50 kW to
perhaps 150 kW of peak power. Suppose that this RTG with a cargo container has a
mass, Mv ¼ 55,000 kg, and that it is moving at a speed of roughly 9 mph (4 m/s) and
the operator applies the brakes that in turn command the electric propulsion drive to
regenerate at a prescribed power level into the ESS in proportion to the brake pedal
depression and speed. The inertial power of the loaded RTG is then converted by
the electric drive to charging current to the ultracapacitor energy storage pack, plus
irreversible losses associated with rolling resistance and windage (5.6).
Pinertial ¼ Pbraks þ Proll þ Pasro ð5:6Þ
Pregen
Pinertial ¼ M v V V_ ; Pbraks ¼ ; Proll ¼ gC rr M v V ; Pasro ¼ C a V 3 ð5:7Þ
hdl
Pregen Ca 2
V_ ¼ þ gC rr þ V ð5:8Þ
hdl M v V Mv
Substituting the components of RTG power defined in (5.7) into the defining
relation (5.6) and simplifying results in the expression for RTG acceleration during
braking given by (5.8). According to (5.8), the RTG decelerates naturally due to
rolling resistance and aerodynamic drag, however insignificant these contributions
may be for a vehicle of such size. However, the regenerative braking power is
nontrivial and amounts to the controlling variable in this situation.
As an illustration, the RTG electric drive is taken as the cascade of four major
elements, the ESS, a power inverter with efficiency, hINV, a propulsion motor-
generator with efficiency, hmot, and the driveline itself having efficiency, hgear. The
ESS efficiency in this case is negligible based on the findings in Example 5.1. The
question is what level of deceleration can be accommodated to regenerate sub-
stantial power into the ESS and for how long? These questions are answered here
via a simulation of the loaded RTG.
ch005 6 June 2011; 16:55:9

At a regen power command, Pregen ¼ 50 kW, the braking power,

Pbrake ¼ 58.34 kW, is the amount of power that must be input via the RTG tires to
the ESS by discharging kinetic energy of the vehicle. This inertial power under
deceleration due to rolling resistance and windage is augmented by the RTG dri-
veline torque level set by the M–G torque command for generating mode. The
driveline efficiency factors modify the actual regeneration power command by
accounting for the inertial power that must be spent to deliver the charging power,
Pregen, plus the irreversible losses to rolling and windage. These efficiencies are the
product: hdl ¼ hgear hmot hINV ¼ 0.857. Figure 5.9 shows that at a 50 kW
regeneration setting the RTG decelerates from an initial V0 ¼ 4 m/s to approxi-
mately zero speed in T ¼ 5.18 s. That is, the RTG is decelerated under constant
power conditions. The reader may note the similarity of Figure 5.9b for speed, V,
and deceleration, V, _ to the case of ultracapacitor discharge under constant power
conditions for voltage, Uuc, and current, Iuc, covered in detail in Chapter 3.
FML1 EQU
V0: = 4 Simulation of RTG speed to regenerate power into ultracapacitor ESS
Pregen: = 50,000
Eff_dl: = 0.93 Cr_by_g Given: Driveline efficiencies:
eta_dl = 0.93
Eff_mtr: = 0.95 CONST eta_mtr = 0.95
Eff_INV: = 0.97 eta_INV = 0.97
Crr_g
Eff_rtg: = Eff_dl*Eff_mtr*Eff_INV
Pbrake: = Pregen/Eff_rtg Pregen MUL3 SUM1 INTG1 Velocity, V
Mv: = 55,000 CONST + +
Crr: = 0.02 4
Ca: = 6.1 Pmass SUM 5
Crr_g: = 9.802*Crr MUL1
Ca_M: = Ca/Mv MUL2
Pmass: = Pbrake/Mv UL: = 1000
LIMIT
DigViewSel1
LIMIT 1 CA_by_M
LL: = 0
Name Value CONST
Eff_rtg 857.00 m FCT_REC1
Pbrake 58.34 k Ca_M
Crr_g 196.04 m Rec
Ca_M 110.91 u
V0 4.00
(a)
Vehicle speed (Pregen = 50 kW) Vdot
4.00 14.60
12.00
3.00
INTG1.VAL
SUM1.VAL
10.00
2.00 8.00
6.00
1.00 4.00
2.00
20.00 m 196.00 m
0 1.00 2.00 3.00 4.00 5.18 0 1.00 2.00 3.00 4.00 5.18
t t
(b)
Figure 5.9 Illustration of RTG braking performance by simulation. (a) Simulation

diagram for the RTG and (b) RTG velocity, V, and deceleration, Vdot,
when Pregen ¼ 50 kW
ch005 6 June 2011; 16:55:9

When the regeneration power level is increased to 85 kW, the RTG, of course,
decelerates at a faster rate to accommodate the need to convert its inertial energy to
ESS charge energy at a much higher rate. Figure 5.10 illustrates the RTG velocity
and rate for this new condition on braking power, with driveline losses remaining
the same fraction as noted above.
Vehicle speed (Pregen = 85 kW) Vdot

4.00 26.20
3.00 20.00
INTG1.VAL
SUM1.VAL
15.00
2.00
10.00
1.00
5.00
0 196.00 m
0 1.00 2.00 3.00 3.52 0 1.00 2.00 3.00 3.52
t t
Figure 5.10 RTG braking performance by simulation for high rate ESS charging.
RTG velocity, V, and deceleration, Vdot, when Pregen ¼ 85 kW
It is easy to be misled by the convenient charts presented in Figures 5.9

and 5.10. That is because vehicle deceleration rates are very important to operator
comfort, cargo integrity, and operational safety of the equipment and environment.
Familiarity with operating a passenger car will help in understanding these points.
A common deceleration rate for a passenger vehicle is 0.15 g, or 1.47 m/s2 and brisk
braking on the order of 0.45 g, or 4.4 m/s2, and panic braking near 1 g, or 9.8 m/s2.
Table 5.4 summarizes the deceleration rates shown in Figures 5.9 and 5.10 in terms
of both g-level and absolute terms.
The total time for the RTG to brake to zero speed at the stated regeneration
power is shown in the bottom row of Table 5.4. The more important finding is that
in order to maintain the braking deceleration below 0.15 g, even power levels as
Table 5.4 Summary of RTG deceleration rates (m/s2) and g as

function of braking level
Time (s)/power Pregen ¼ 15 kW Pregen ¼ 50 kW Pregen ¼ 85 kW

level (kW)
T¼0 0.277 m/s2 0.0283 g 0.46 0.0469 0.65 0.0663
T¼1 0.283 0.0289 0.5 0.0510 0.74 0.0755
T¼2 0.289 0.0295 0.55 0.0561 0.94 0.0959
T¼3 0.298 0.0304 0.64 0.0653 1.5 0.1530
T¼4 0.309 0.0315 0.82 0.0837
T¼5 0.323 0.0330 2.0 0.2040
T¼8 0.42 0.0428 3.5
Total brake 10.1 5.18
time (Td)
ch005 6 June 2011; 16:55:9

high as 85 kW are tolerable up to a good fraction of the time it takes to reach zero
speed. For example, 0.15 g is possible at 85 kW up to 3 s when the time to zero
speed is 3.5 s. Beyond this point, the deceleration rate increases exponentially due
to imposition of constant power conditions by the electric drive system.
The amount of inertial energy contained in the loaded RTG having the para-
meters listed in Example 5.1 can be calculated by (5.9) to get Winertial ¼ 440 kJ
when the initial velocity V0 ¼ 4 m/s.
W inertial ¼ 0:5M v ðV 2i V 2f Þ ð5:9Þ
Compare (5.9) to (5.5) and note the striking similarity between the kinetic
energy of the vehicle with that of the energy stored in the ultracapacitor when its
voltage swings between two levels. The same is true of the RTGs inertial energy as
its velocity changes from an initial to final value. Exercises 5.3 and 5.4 expand on
these concepts and the solution procedure is recommended reading.
5.3 Earth moving equipment
The final industrial application of ultracapacitor energy storage to be considered is

the excavator. Such earth moving equipment ranges from the operator in turret with
boom and shovel to large surface mine drag lines. Here, the large drag line exca-
vator is analyzed as a case study.
Figure 5.11 shows the type and scale of excavator considered, the large earth
excavating industrial shovels that operate approximately 20 h/day, 365 days/year. The
ultracapacitor application in excavators, such as the system now in use by Komatsu [7],
is the energy storage for the shovel turret that is electric powered versus hydraulic
powered. The shift to electric powered turret in construction equipment provides a
dramatic reduction in diesel fuel consumption, as much as 25% overall for the Komatsu
PC200-8 excavator. In applications where the turret motion is more substantial, the
Figure 5.11 Illustration of large industrial shovel type excavator (Bucyrus)
ch005 6 June 2011; 16:55:10

hybrid solution using ultracapacitor ESS is verified to be 41% fuel reduction. This is
because the excavators’ upper structure turns more frequently and the electric drive
with energy regeneration to the ultracapacitor-only ESS is very efficient. The PC200-8
has an operating mass of 20,100 kg, which is approximately a 1 m3 bucket, and it is
capable of digging to 6.6 m below ground level. Its 6.7 L diesel engine is rated 110 kW
with 439 L/min main hydraulic pump. Hybrid units relax the hydraulic requirements
and replace this system with an electric generator and electric actuators.
The Bucyrus shovel, a much scaled version of the excavator just described,
such as the model 495HF mining shovel, shown in Figure 5.11, has a 30.6 m3
bucket that can load a 360 ton mining truck in three scoops at 120 ton payload each.
The unit has a drop height of 10.1 m and an overall operating mass of some 1380
(metric) ton. As with the Komatsu excavator, the goal is to electrify the turret so
that fuel consumption reduction on the same order is possible.
For a case study, consider the Bucyrus 495HF rope shovel, where the large turret
that includes the operator house, engine room with hydraulic pumps and generator,
boom and shovel is to be hybridized for electric drive of the turret. The system
operating voltage will be 960 V so that standard ultracapacitor modules, such as the
Maxwell heavy transportation module (HTM125), can be used. In this case, the
495HF demand power can peak at approximately 2800 kW during motoring opera-
tion and nearly 400 kW when decelerating. The application is very similar to the
smaller Komatsu excavator, only the scale in this application is extreme. This sys-
tem, therefore, makes an excellent case study for ultracapacitor module scaling.
Figure 5.12 is a representative operating profile of an electrified turret, which is
based on high pulse power to lift and accelerate the fully loaded bucket in the azi-
muth plane from excavation to mine truck as shown in Figure 5.11, then to decele-
rate, dump, and return the empty bucket to the work zone. According to this figure,
the ESS will discharge during high positive power peaks, with slower discharge on
ramps and some recharging during the reverse power flow just before t ¼ 30 s.
495HF terminal power (kW)

2.80
2.00
1.50
P0
1.00
500.00
–300.00
0 5.00 10.00 15.00 20.00 25.00 30.00

t
Figure 5.12 Representative power profile of the 495HF excavator (power in MW,
time in seconds)
ch005 6 June 2011; 16:55:11

The methodology followed is given below:

● Estimate the average operating power from the profile given (Figure 5.12 in
this case).
● Compute the deliverable energy for this profile as Wpack ¼ P0 T, where
P0 ¼ 1100 kW.
● Calculate the deliverable energy per HTM module (125 V, 63 F, 14 mW) ¼
100 Wh.
● Calculate the number of series modules, Ns ¼ Umx/(48 cells 2.55 V) ¼ 960/
122.4 ¼ Int{7.8}.
● Given that Ns ¼ 8 HTMs in series, then useable energy at EOL,
Wstring ¼ 0.75 Ns WHTM ¼ 600 Wh.
● Calculate the deliverable energy of the entire ESS pack, and using the deli-
verable energy per string compute the required number of strings to meet
energy at EOL. In this case, M ¼ P0T/Wstring ¼ (1,100,000 W 30 s)/
(600 Wh 3600 J/Wh) ¼ Int{33 MJ/2.16 MJ} ¼ 16 parallel strings.
● Given the estimated 8S 16P HTM ESS pack required, scale the HTM
parameters to this configuration of Ns ¼ 8, M ¼ 16 and build the model for
simulation of the representative power pulse.
Table 5.5 lists the HTM module parameters and scales these to an equivalent model
representative of the 8S 16P HTM configuration. In this table, resistances and
inductances are scaled by Ns/M, capacitances by M/Ns, and voltage by Ns. Cell,
module, and pack ESR C time constants are of course unchanged. The full
Table 5.5 Scaling the HTM module equivalent circuit parameters to

an 8S 16P HTM pack level
HTM125 model parameters Scaling 8S 16P HTM value

Rconn 3.36 mW Ns/M 1.68 mW
Lconn 3 mH Ns/M 1.5 mH
Rs 14.4 mW Ns/M 7.2 mW
Rsa 28.8 mW Ns/M 14.4 mW
Cs 10.5 F M/Ns 21 F
Rp 33 kW Ns/M 16.5 kW
HTM nonlinear C(U) Scaled to 8S 16P HTM
Uc (V) C0 (F) Scaled Upak (V) Cpak (F)
0 32.6 Uc Ns?Upak 0 65.2
C0 (M/Ns)?Cpak
24 49.1 192 98.2
48 56 384 112
72 58.6 576 117.2
96 61.6 768 123.2
120 65.8 960 131.6
134.4 69.3 1075 138.6
ch005 6 June 2011; 16:55:11

8S 16P HTM pack has a time constant only slightly larger than module or cell
time constants by the additional interconnect resistance of the modules.
Figure 5.13 shows the equivalent circuit model corresponding to the HTM
scaled to 8S 16P HTM for the ESS in the 495HF excavator. This figure also
lists the 8S 16P HTM terminal voltage and current when simulated over the
30 s of operating power profile shown in Figure 5.12.
Rsa C1
8S × 16P × HTM module
Rconn Lconn 14.4 mΩ 21 F 2.80
1.68 mΩ 1.5 PH 2.00

VM1 7.2 mΩ 1.50
+ FML1 EQU XY1 C0
16.5 kΩ
P0
tY I1 V P0: = VM1·V*I1·I XY
1.00
SOC: = (C0·V/960)^2 900 V Rp
Bucyrus
turrent Pdisp: = Rconn·R*Rconn·I^2 + R*Rs·I^2 + Rsa·R*Rsa·I^2 500.00
profile Eff: = W0(W0 + Wdisp)
–300.00
W0 Wdisp
P0 Pdisp 0 5.00 10.00 15.00 20.00 25.00 30.00
1 1 t
0.2777 m 0.2777 m
(a)
495HF terminal voltage 495HF terminal current
900.00 3.54
3.00
850.00
2.50
800.00
VM1.V (V)
2.00
I1.1 (A)
750.00 1.50
1.00
700.00
500
620.00 –480
0 5.00 10.00 15.00 20.00 25.00 30.00 0 5.00 10.00 15.00 20.00 25.00 30.00
t t
(b)
Figure 5.13 495HF excavator ESS schematic, power profile and terminal voltage
and current. (a) 495HF ESS equivalent circuit model for
8S 16P HTM and (b) terminal voltage (900 V?648 V) and
current (þ3.5 kA, 471 A) for the power profile
The power profile shown in Figure 5.12 results in Wpack ¼ 6872.6 Wh of deli-
verable energy expended by t ¼ 24 s, recovering approximately 6572 Wh when
t ¼ 30 s, accounting for the short-term regeneration power. The 8S 16P HTM
pack SOC ranges from SOC0 ¼ (900/960)2 ¼ 0.88 (t ¼ 0 s) to 0.455 (t ¼ 30 s).
The full pack power dissipation peaks at 108.3 kW at t ¼ 4.17 s when the pack
current voltage is 794 V and current is 3525 A. This calculates the total ESS
resistance of
Pdisp 108;300
ESRpak ¼ ¼ ¼ 8:7 mW ð5:10Þ
I 2pak 35252
Equation (5.10) is a back check on the data presented in Table 5.5, and
ESRpak *Rs þ Rconn ¼ 8.88 mW, which it does agree with. Pack SOC, efficiency,
ch005 6 June 2011; 16:55:12

and energy output and dissipated are shown graphically in Figure 5.14. The effi-
ciency over the full operating profile is approximately 97.8%.
8S9PHTM Mod SOC Cycle energy efficiency

1.00 1.00
900.00
800.00
700.00 950.00
SOC
Eff
600.00
500.00
900.00
400.00
300.00
200.00 850.00
0 5.00 10.00 15.00 20.00 25.00 30.00 0 10.00 20.00 30.00
t t
(a)
Pack power dissipation Cycle energy dissipation
109.00 6.90
6.00
80.00 5.00
Wdisp.VAL
W0.VAL
W0.VAL
60.00 4.00
Pdisp
3.00
40.00
2.00
20.00 1.00
0 0
0 5.00 10.00 15.00 20.00 25.00 30.00 0 5.00 10.00 15.00 20.00 25.00 30.00
t t
(b)
Figure 5.14 Excavator turret ESS dissipated power and energy output, W0. SOC
and overall efficiency (eff ¼ 0.978)
The appearance of relatively high power dissipation (*108 kW) is a matter of

perception, since the output power levels of the ESS are nearly 3000 kW. The total
energy discharged by t ¼ 24 s, Wpack ¼ 6872 Wh (24.74 MJ) represents an average
power, P0 ¼ Wpack/T ¼ 24,740/24 ¼ 1030 kW, very close to the initial sizing esti-
mate of 1100 kW used to design the 8S 16P HTM ESS system.
The ESS configuration discussed to this point is not sufficient for repetitive
operation, since the energy withdrawn from the ultracapacitor pack is not replen-
ished before a subsequent cycle starts. As discussed earlier, these excavators have a
large engine room with diesel-driven generators and the purpose of hybridizing is
to alleviate the diesel emissions and fuel consumption. The benefits of the ESS are
apparent when considering that without the ESS the engine-generator would need
to deliver the full 2.8 MW of power demanded, via some hydraulic system, or
electric drives. However, with energy storage discussed earlier it is possible to
downsize the diesel gen-set to 960 V and approximately 900 A, or 864 kW, a very
substantial reduction. Figure 5.15 illustrates the case when the gen-set delivers a
constant 900 A at 960 V to the turret drive motor(s). When this is done the ultra-
capacitor pack SOC returns to its initial value in time for the next turret cycle.
ch005 6 June 2011; 16:55:12

495HF terminal voltage 8S16PHTM Mod SOC

905.80 1
900.00
900.00
890.00
800.00
880.00
VM1.V (V)
870.00 700.00
SOC
860.00 600.00
850.00 500.00
840.00 400.00
830.00 300.00
816.50 200.00
0 5.00 10.00 15.00 20.00 25.00 30.00 0 5.00 10.00 15.00 20.00 25.00 30.00
t t
Figure 5.15 The 495HF excavator voltage and ESS SOC with engine-generator
support of 864 kW
One concern brought to light by Figure 5.15 is that with the excavator engine-
generator augmenting the ESS the ESS SOC swing is no longer the deep excursion
of 0.88–0.455 as in Figure 5.14, but now only 0.88–0.738. This means that the
ultracapacitor 8S 16P HTM pack is now oversized for the application. Since it
is also an operational strategy that the excavator system may include larger
regeneration power when hybridized, no further analysis will be made of ESS
charge restoration. Instead, we move to the topic of thermal design for such ESS
systems. This particular application may not be the most appropriate, but it will
illustrate the methodology and results.
For a thermal investigation, consider that the 495HF excavator is to operate
continuously at the profile shown in Figure 5.12, or some variant thereof. Using
Figure 5.12 as the exemplar and making this profile repetitive for the case of
engine-generator for charge balancing, the ESS voltage and power dissipation for
repetitive operation are as shown in Figure 5.16.
It is apparent from Figure 5.16a that the excavator turret load current is
unchanged from the earlier investigation, as it should be, since the power demands
remain unchanged. However, the voltage of the ESS does not make such deep
excursions as it would without engine-generator support. Power dissipation of the
module is still relatively high, 50.8 kW instead of 108 kW, and the dissipated
energy shown over four cycles is 255 Wh.
~ disp ¼ W disp ¼ 255ð3600Þ ¼ 7:65 kW

P ð5:11Þ
T 4ð30Þ
The rectified mean power of Figure 5.16b is PdispTM ¼ 7.653 kW, which is in
very good agreement with (5.11). Without going back into the design study, sup-
pose now that an 8S 9P HTM ESS pack is sufficient to support the turret peak
power conditions with engine-generator support. When the simulation model is
re-configured to this sizing, summarized in Table 5.6, the power dissipation is
altered because of the parameter changes, but the ESS will be exercised more
because of its lower deliverable energy.
ch005 6 June 2011; 16:55:13


909.00
900.00 3.36
890.00
2.50
880.00
VM1.V (V)
2.00
I1.I (A)
870.00
1.50
860.00
1.00
850.00
500
840.00
–360.0
821.5
0 20.00 40.00 60.00 80.00 100.00 120.00
0 20.00 40.00 60.00 80.00 100.00 120.00 t
t
(a)
Pack power dissipation Cycle energy dissipation
50.80 255.00
40.00 200.00
Wdisp.VAL
30.00 150.00
Pdisp
20.00 100.00
10.00 50.00
0 0
0 20.00 40.00 60.00 80.00 100.00 120.00 0 20.00 40.00 60.00 80.00 100.00 120.00
t t
(b)
Figure 5.16 Repetitive operation of the 495HF turret: power dissipation, terminal
voltage, and dissipated energy. (a) Terminal voltage and current for
repetitive case with engine-gen balancing and (b) dissipated power
and energy in the ESS with engine-gen balancing
For the 8S 9P HTM module the power profile is modified to better reflect
the motoring and generating power of a loaded turret when the electric drive
recuperates more of the available decelerating energy. This revised power profile,
generated to approximate actual operation, is shown in Figure 5.17 along with the
modified equivalent circuit model as per the values in Table 5.6.
Note that the new power profile applied to the 8S 9P HTM model in
Figure 5.17 is similar in motoring demand to the earlier profiles, but substantial
regeneration is now included to highlight turret braking maneuvers of the heavily
loaded bucket. The profile now has a duration of 40 s and is taken as repetitive. The
initial condition on ultracapacitor ESS has been raised to 950 V out of a maximum
960 V so that the SOC0 ¼ 0.979 and the minimum voltage is sufficient to hold
SOC > 0.25, the minimum design value. Figure 5.18 summarizes the simulation
results for ESS terminal voltage, current, and internal power and energy dissipa-
tion. Note that because the ESS ultracapacitor modules are now being exercised
substantially more than the earlier design, the power dissipation is increased as well
as the dissipated energy.
Application of (5.11) to Figure 5.18b yields a rectified mean power dissipation
of Pdisp ¼ 33.2 kW that is dissipated in the Ns M ¼ 72 HTM modules. Studies on
the Maxwell Technologies 48 V module found that it has a total mass, M48 ¼ 14 kg,
ch005 6 June 2011; 16:55:13

Table 5.6 Scaling the HTM module equivalent circuit parameters

to an 8S 9P HTM pack level
HTM125 model parameters Scaling 8S 9P HTM value

Rconn 3.36 mW Ns/M 2.98 mW
Lconn 3 mH Ns/M 2.67 mH
Rs 14.4 mW Ns/M 12.8 mW
Rsa 28.8 mW Ns/M 25.6 mW
Cs 10.5 F M/Ns 11.8 F
Rp 33 kW Ns/M 29.3 kW
HTM nonlinear C(U) Scaled to 8S 9P HTM
Uc (V) C0 (F) Scaled Upak (V) Cpak (F)
0 32.6 Uc Ns?Upak 0 37.6
C0 (M/Ns)?Cpak
24 49.1 192 55.2
48 56 384 63
72 58.6 576 65.8
96 61.6 768 69.3
120 65.8 960 74
134.4 69.3 1075 78
Rsa C1
8S × 9P × HTM module
Rconn Lconn 25.6 mΩ 11.8 F
Rs
2.98 mΩ 1.5 PH
+ VM1 FM1 12.8 mΩ

EQU XY1
I1 C0
tY V P0: = VM1·V*I1·I XY
29.3 kΩ
SOC: = (C0·V/960)^2 950 V Rp
Bucyrus Pdisp: = Rconn·R*Rconn·I^2 + R*Rs·I^2 + Rsa·R*Rsa·I^2
turrent
Eff: = W0(W0 + W0)
profile
W0 W0
P0 P0
1 1

2.80 Meg
2.00 Meg
1.50 Meg
1.00 Meg
P0
500.00
0
–500.00
–1.00 Meg
0
00
00
00
.0
.0
.0
0.
5.
0.
25
50
75
10
12
16
Figure 5.17 The 8S 9P HTM equivalent circuit model and modified power
profile for higher regeneration
ch005 6 June 2011; 16:55:13


952.00 3.83
3.00
850.00
800.00 2.00
VM1.V (V)
I1.I (A)
750.00
1.00
700.00
650.00 0
600.00
530.00 –1.43
0
0
00
00
00
.0
.0
.0
0
0
00
00
00
0.
5.
0.
.0
25
.0
.0
50
75
0.
5.
0.
10
12
16
25
50
75
10
12
16
t t
(a)
Power pack dissipation Cycle energy dissipation
225.00
1.48
200.00
175.00 1.20
150.00 1.00
Wdispp.VAL
125.00 800.00
Pdisp
100.00 600.00
75.00
400.00
50.00
200.00
25.00
0 0
0
00
00
00
0
00
00
00
.0
.0
.0
.0
.0
.0
0.
5.
0.
0.
5.
0.
25
25
50
75
50
75
10
12
16
10
12
16
t t
(b)
Figure 5.18 Summary of 8S 9P HTM performance in the 495HF excavator

turret over four cycles. (a) ESS voltage (950?530 V) and current
(3.818? 1.414 kA) of 8S 9P HTM under repetitive operation
and (b) ESS internal power dissipation (Pdpk ¼ 224 kW,
Pdisp ¼ 33.2 kW) and energy (1476.5 Wh)
and specific heat value, cp ¼ 1114 J/(kg W). The HTM has an approximate mass,
MHTM ¼ 58 kg, and a specified thermal resistance, Rth ¼ 0.046 K/W. The HTM
thermal capacitance can therefore be approximated by the specific heat of the 48 V
module and the mass of the HTM according to (5.12). Using the specified Rth and
calculated Cth the thermal time constant is approximately 2973 s.
C th ¼ cp M HTM ¼ 1114ð58Þ ¼ 64:63 kJ=K ð5:12Þ
Rth ¼ 0:046 K=W
Using the ESS pack dissipation power given earlier as Pdisp makes it possible to
now calculate the temperature rise characteristic of the ESS pack in the 495HF exca-
vator under continuous operation. In this case there are NHTM ¼ Ns M ¼ 72 HTM
modules, so the power dissipation in each is Pd ¼ Pdisp/NHTM ¼ 461 W. Temperature
as a function of time (5.13) varies exponentially during thermal charging of Cth until a
steady state is reached when HTM power dissipation Pd is in equilibrium with heat
ch005 6 June 2011; 16:55:14

dissipation to ambient temperature through the thermal resistance, Rth(5.12).
T ðtÞ ¼ T amb þ Rth Pd ½1 et=Rth Cth ð5:13Þ

The temperature rise with time given by (5.13) is plotted in Figure 5.19
according to the Rth Cth thermal time constant until thermal equilibrium is
reached at approximately 12,000 s, when the steady state temperature rise becomes
approximately 46–25 C ¼ 21 C. This steady state temperature rise, dT , given by
(5.14), is over the specification of 15 C, so the system controller will activate the
cooling fans provided on the HTM modules to bring the temperature within design
limits.
dT ¼ T ss T amb ¼ Pd Rth ¼ 461ð0:046Þ ¼ 21:1 C ð5:14Þ
HTM temperature T(t)

50
45
40
Temperature (°C)
35
30
25
20
15
10
5
0
0 2000 4000 6000 8000 10,000 12,000 14,000 16,000
Time (s)
Figure 5.19 HTM temperature rise of 8S 1P HTM in 495HF excavator
Exercises
5.1 A Hydrogenics HyPM12 fuel cell having the ratings given in Table 5.1 is
used in a material handling truck. (a) Compute the hydrogen mass flow
rate when the pack is outputting maximum power. Use the lower heating
value of hydrogen, LHVH2 ¼ 120:1 MJ=kg. (b) Compute the fuel cell cur-
rent at this power and the number of cells in the pack.
Ans: (a) Using data from Table 5.1 to compute dWcomb/dt in (5.3) results in
Ppk 12:7 kW
W_ comb ¼ ¼ ¼ 23:96 kW
hpk 0:53
Using (5.3) the hydrogen mass flow rate is
_ H2 LHVH2 4
W_ comb ¼ M !M_ H2 ¼ 23:96 ¼ 0:1995 g=s
120:1
ch005 6 June 2011; 16:55:15

(b) To compute the FC stack current and the molar flow rate of hydro-
gen, a cell potential of 0.67 V is used at peak power from which
iFC ¼ Ppk/UFC ¼ Ppk/(NcUcell) ¼ 12,700/(63 0.67) ¼ 300 A.
M
The molar flow rate of hydrogen is: n_ ¼ amuHH2 ¼ 0:1995
2:016 ¼ 0:099 mol=s
2
5.2 For the case of a rubber tired gantry crane, apply the results of the simulation
performed in Example 5.1 and show that the integral of ultracapacitor pack
current is consistent with the simulated voltage swing.
Ð
Hint: i dt ¼ 718:2 As and the voltage swing is: 569.6 V < Ud < 630 V, then
apply the state equation for the nonlinear capacitance times voltage as total
charge exchanged. Recall from earlier 630chapters that ultracapacitors exhibit

nonlinear C(U). In this case use C 0 ðU Þ ffi 11:762þ 12 ¼ 11:88 F
569:6
Ð
~ ¼ C 0 ðU Þ½U i U f ¼ 11:88ð630 569:6Þ ¼ 717:55 As ~ i dt ¼
Ans: Q
718:2 As:
5.3 Confirm that the RTG described in Example 5.1 has Winertial ¼ 440 kJ of
available kinetic energy.
Ans: Use (5.9) and V0 ¼ 4, Mv ¼ 55,000 kg to find that
W inertial ¼ 0 :5Mv ðV 20 V 2f Þ ¼ 440;000 J.
5.4 Expand Table 5.4 to include: (a) the total regen energy for each power level
and (b) the regen fraction of total kinetic energy for each power level.
Ans: (a) The kinetic energy returned to the ESS, not accounting for driveline
losses to transmit this power upstream in the driveline, is simply Pregen Td.
(b) To compute the regen fraction (RF) take the ratio of the Wregen/Winertial,
where Winertial is given in Exercise 5.3.
Table Exercise 5.4 RTG regen energy and regen fraction
Pregen (kW) 15 50 85
Brake time, Td (s) 10.1 5.18 3.5
Wregen (kJ) 151.5 259 297.5
Regen fraction (RF) 0.344 0.588 0.67
5.5 Provide a plausible explanation for the increasing electric fraction in Exercise
5.4 with regeneration power level.
Ans: When Pregen is low the RTG deceleration time is long, providing more
time for energy dissipation via rolling resistance of the tires. In all cases
windage loss is negligible since speeds are so low. However, as Pregen
increases a larger fraction of the available kinetic energy is converted into
stored electric energy because there is less time for rolling resistance to
consume it.
ch005 6 June 2011; 16:55:15

5.6 The large excavator, 495HF, unit simulated in Figure 5.14 has the following
performance over specified intervals.
636 V < Upak < 900 V; 0 < t < 24 s; and Wpack ¼ 6872 Wh of delivered energy
over this time interval.
What is the average ESS pack capacitance over this interval based on deliv-
ered energy?
Ans: Apply (5.5) to obtain the following:
2W pack ð3600Þ 49:48 MJ

C~0 ¼ ¼ ¼ 122 F
½U i U f
2 2
ð9002 6362 Þ
pak ¼ 768 V,
It is of interest that the midpoint of the stated voltage range: U
and inspection of the scaled values in Table 5.5 gives Cpak(768 V) ¼ 123.2 F,
which is in very close agreement with the average capacitance calculated
from the energy exchange.
5.7 Apply (5.19) to the case of Maxwell HTM modules used in an excavator
application for continuous duty. What maximum power dissipation is allow-
able before the fans are turned ON so that the temperature remains within
design targets?
Ans: dT < 15 C, therefore
dT 15
Pd ¼ ¼ ¼ 326 W
Rth 0:046
5.8 Given that the HTM module has an ESRdc ¼ 14.4 mW and the maximum
internal power dissipation calculated in Exercise 5.7 is used, calculate the
maximum rms current that can be tolerated without switching ON the
module fans.
rffiffiffiffiffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Pd 326
Ans: I rms ¼ ¼ ¼ 150 Arms
ESRdc 0:0144
Note: This value of rms current is the specification sheet value maximum
continuous current for the stated steady state temperature rise of the HTM
module.
References
1. W. Friede, M. Kammerer, N. Kodama, K. Harris, ‘Fuel cell hybrid minibuses
for niche applications’, The 22nd International Battery, Hybrid and Fuel Cell
Electric Vehicle Symposium & Exposition, EVS-22, Yokohama, Japan,
pp. 885–94, 23–28 October 2006
2. Z. Jiang, R.A. Dougal, ‘A hybrid fuel cell power supply with rapid dynamic
response and high peak-power capacity’, IEEE Applied Power Electronics
ch005 6 June 2011; 16:55:16

Conference, APEC2006, Hyatt-Regency Hotel, Dallas, TX, pp. 1250–5,

19–23 March 2006
3. M. Iwaida, N. Oki, S. Oyama, K. Murakami, M. Noguchi, ‘Development of
high power electric double-layer capacitor for fuel cell vehicle’, The 13th
International Seminar on Double Layer Capacitors and Hybrid Energy Sto-
rage Devices, Embassy Suites Deerfield Beach Resort, Deerfield Beach, FL,
pp. 165–77, 8–10 December 2003
4. K. Tamenori, T. Taguchi, A. Anekawa, M. Noguchi, ‘Application studies of
electric double layer capacitor system for fuel cell vehicle’, The 13th Inter-
national Seminar on Double Layer Capacitors and Hybrid Energy Storage
Devices, Embassy Suites Deerfield Beach Resort, Deerfield Beach, FL, pp.
178–88, 8–10 December 2003
5. K. Ikeda, K. Hiratsuka, K. Satoh, M. Noguchi, ‘Material development of
electric double layer capacitor for fuel cell electric vehicle and the newly
developed electric double layer capacitor cell’, The 13th International
Embassy Suites Deerfield Beach Resort, Deerfield Beach, FL, pp. 189–204,
8–10 December 2003
6. N. Omar, F. Van Mulders, J. Van Mierlo, P. VanDen Bossche, ‘Assessment
of behavior of supercapacitor-battery system in heavy hybrid lift truck
vehicles’, The 5th IEEE International Vehicle Power and Propulsion
Conference, VPPC2009, Ritz-Carlton Hotel, Dearborn, MI, pp. 962–5, 7–9
September 2009
7. Komatsu Corporate Communication Press Release, Komatsu Introduces
Worlds First Hybrid Excavator: Hybrid Evolution Plan for Construction
Equipment, 13 May 2008. Available on: www.komatsu.com
ch005 6 June 2011; 16:55:16

Chapter 6
Heavy transportation application
This chapter is a continuation of ultracapacitor application case studies with focus

on heavy transportation. Public transportation systems and vehicles in particular are
becoming more of a focus by regulatory agencies as a sector to promote energy
efficiency. Energy efficiency now has the connotation of energy security, offsetting
of imported oil, and moderation of climate change through reduced emissions. All
of these reasons mesh well with the trend to hybridize heavy transportation. Con-
sider for the moment transit buses, the type we are most familiar with as a city bus,
and one that is typically powered by a large CIDI engine burning diesel fuel. The
market drivers for hybridizing such buses are the following:
● Energy efficiency:
* Minimize petroleum imports.
* Recognize the volatility of fuel supply and price.
● Environmental concerns:
* Green house gases (GHG) and global warming.
* Necessity to regulate emissions.
● Fiscal responsibility:
* Fuel requirements, low sulfur, and so on.
* Lifecycle costs, ongoing maintenance costs.
● Societal impacts:
* Public health and diesel exhaust (polynuclear aromatics).
* Public concern for noise levels and noise abatement.
The hybrid transit bus provides these and more benefits at a cost. Today, hybri-
dization of public transportation is subsidized, so some of the costs to a manufacturer
can be offset. In the longer term, it will be necessary for manufacturers to absorb a good
portion of the hybridizing costs through process efficiencies, product communization,
and volume production. The benefits of hybridization can be summarized as
● Introduction of advanced technology drive systems, both engine and electric
drives.
● Fully controlled electric drive systems for acceleration and regenerative braking.
● Zero emission vehicle (ZEV) operation at city centers and where emissions are
prohibited.
ch006 8 June 2011; 14:40:48

● ZEV operation, that is, hybrid buses with substantial ESS capacity and battery
electric buses.
● Overall lower operating costs.
The propulsion system architectures of heavy transportation vehicles, buses, shut-
tles, and vans are now discussed. Figure 6.1 is the present situation, for the most
part, of diesel engine powered city bus with automatic transmission and mechanical
path driveline.
V
Conventional
mw
Peng Pdl ωw
Engine Trans FD
Com VSC
Figure 6.1 Conventional transit bus propulsion architecture (VSC, vehicle system
controller; Com, communications; Trans, transmission; FD, final
drive or differential)
In the conventional bus architecture, the engine is the main power plant pro-
viding propulsion power, Peng, via an automatic transmission to the driveline.
Accounting for losses in the transmission and torque converter, an amount of pro-
pulsion power, Pdl, or driveline input power is available to the final drive (FD) or
differential and then to the driven wheels. For straight ahead driving at vehicle
velocity, V, the engine supplies sufficient power, Pdl, to overcome friction, windage,
and grade at the driven wheels in the form of wheel torque, mw, at each axle and wheel
angular speed, ww. The basics of vehicle propulsion are that axle torque and angular
speed translate to wheel tractive effort, Ft, and vehicle speed, V, as follows [1]:
mw ¼ F t rw ð6:1Þ
V
ww ¼ ð6:2Þ
rw
mw ww
Pdl ¼ ð6:3Þ
ndl
F t V ¼ mw ww ð6:4Þ
where rw is radius of driven wheel in meters and hdl is efficiency of the dri-
veline, primarily of the FD gearing. A typical transit bus tire is specified as 275/
70R22.5, which means thread width is 275 mm, sidewall height is 70% of thread
width, and the rim diameter is 22.5 in. (571.5 mm). Unfortunately, the specification
ch006 8 June 2011; 14:40:48

Heavy transportation application 191
units are a mixture of metric and English, but that is a fact of automotive culture.
Given these tire specifications, the tire rolling radius, rw, is therefore
Dr þ 2H 571:5 þ 2ð0:7Þð275Þ
rw ¼ ¼ ¼ 0:478 m ð6:5Þ
2 2
Application of (6.5) in (6.1) and (6.2) permits the translational characteristics
of the vehicle to be recast as rotational quantities of torque and angular velocity.
Having made the conversion to the rotational system, it is straightforward to derive
the torque, angular velocity, and power levels throughout the propulsion and
energy conversion stages of the vehicle architecture shown in Figure 6.1 and sub-
sequent figures. Some examples will help illustrate the technique before moving
on to hybrid and battery electric architectures. Table 6.1 is a compilation of
Table 6.1 Transit bus parameters and electric drive system attributes

Curb mass, Mv kg 9600 Traction motor type – Induction
MG-A and MG-B
Passenger mass, kg 5400 MG-A peak power, kW 250
Mpass PMGA
Total vehicle mass, kg 15000 MG-B peak power, kW 150
Mtot PMGB
Tire rolling radius, rw m 0.478 Torque, stall MG-A, Nm 1086
mMGA
Efficiency final – 0.98 Speed, MG-A, MG-B, rad/s 345
drive, hfd wMGA
Efficiency transmis- – 0.97 Mass, MG-A, MMG-A kg 200
sion, hxm
Efficiency electric – 0.94 Mass, generator, Mgen kg 120
motor, hmtr
Efficiency inverter, – 0.96 Volume, MG-A, L 80
hINV VolMG-A
Bus aero coefficient, – 0.52 Volume, generator, L 65
Cd VolGen
Bus frontal area, Af m2 8 Coolant, MG ATF –
(automatic
transmission fluid)
Number pass seats, – 40 Power inverter rating, kVA 250
Nseats PINV

Number pass stand, – 30 Power inverter C 40 to þ60
Nstand temperature, TINV
Bus maximum kph 110 Power inverter coolant @<65 C
speed, Vwot ATF or WEG
Bus maximum m/s2 1.5 Power inverter comm. CAN
acceleration, a SAE J1939
Bus maximum m/s2 2 Power inverter mass, kg 90
deceleration, ab MINV
Bus maximum % 25 Power inverter L 130
grade, gr volume, VolINV
ch006 8 June 2011; 14:40:48

representative parameters for a transit bus that will be used in examples and exer-
cises of this chapter.
Example 6.1: Calculate the transit bus electric traction motor, engine-driven gen-
erator, and power inverter metrics for specific power (SP), power density (PD), and
torque density, specific torque (ST). Tabulate for ease of comparison (Table 6.2).
Table 6.2 Transit bus hybrid component metrics
Traction MG-A Generator Inverter

SP (kW/kg) 1.25 SP (kW/kg) 1.25 SP (kVA/kg) 2.8
PD (kW/L) 3.1 PD (kW/L) 2.3 PD kVA/L) 1.92
ST (N m/kg) 5.4 ST (N m/kg) 4.2
Solution: Example 6.1 makes it clear that electric traction motors, typically asyn-
chronous or induction type, compared with generators, typically permanent magnet
synchronous, are comparable in specific power and power density and specific
torque. Power inverters, because of their high continuous rating, do not have per-
formance metrics at the level of hybrid passenger cars, but are still very respectable
at 2.8 kVA/kg. The reader should be careful that electric machines deliver or
receive real power, hence kW at their shaft and inverters, because of switching and
variable displacement power factor, is rated in terms of apparent power, or kVA.
In the next example, the objective is to translate the acceleration characteristics
of the generic transit bus having parameters listed in Table 6.1 into driveline tor-
que, angular velocity, and power levels. For this example the engine, a Cummins
ISB, is rated Peng ¼ 184 kW at 2200 rpm, and it is desirable to translate this to
driven wheel tractive effort, Ft (N), and vehicle velocity, V (m/s and mph). Our
overall objective in setting up these examples is to build a representative micro-
cycle, or drive schedule, that will be applied in the three case studies of this chapter.
Example 6.2: The conventional transit bus of Figure 6.1 has the Cummins ISB diesel
engine noted above and the driveline parameters listed in Table 6.1. Given the trans-
mission ratio, gxm ¼ 4.75:1, and final drive ratio, gfd ¼ 2.83:1, calculate the following:
(a) Engine angular speed, rad/s, and torque, N m
(b) Driveline torque, angular speed, and power
(c) Wheel axle torque, angular speed, and power
(d) Wheel traction, Ft, vehicle speed, V, and propulsion power at wheels, P(V)
Solution: Apply (6.1) through (6.4) using the ratios stated and (6.5) as follows:
(a) Given that Peng ¼ 184 kW and neng ¼ 2200 rpm, then weng and meng are
2p
weng ¼ neng ¼ 0:1047neng ¼ 230:38 rad=s ð6:6Þ
60
Peng 184; 000
meng ¼ ¼ ¼ 798:68 N m ð6:7Þ
weng 230:38
ch006 8 June 2011; 14:40:50

(b) Given that gxm ¼ 4.75:1, then cascading (6.6) and (6.7) down the driveline
results in
mdl ¼ hxm gxm meng ¼ 0:97ð4:75Þ798:68 ¼ 3679:92 N m ð6:8Þ
weng 230:38
wdl ¼ ¼ ¼ 48:5 rad=s ð6:9Þ
g xm 4:75
Pdl ¼ mdl wdl ¼ 178:476 kW ð6:10Þ
(c) Given that gfd ¼ 2.83:1, so cascade (6.8) and (6.9) to the driven wheels to get
mw ¼ hfd g fd mdl ¼ 0:98ð2:83Þð3679:92Þ ¼ 10; 205:9 N m ð6:11Þ

wdl 48:5
ww ¼ ¼ ¼ 17:14 rad=s ð6:12Þ
gfd 2:83
Pw ¼ mw ww ¼ 174 kW ð6:13Þ
(d) Finally, given that rw ¼ 0.48, the wheel axle torque and angular speed are
translated to wheel traction and vehicle velocity and propulsion power.
mw 10;205:9
Ft ¼ ¼ ¼ 21;358 N ð6:14Þ
rw 0:478
8:19 rad=s
V ¼ rw ww ¼ 8:19 rad=s ¼ ¼ 18:3 mph ð6:15Þ
0:447
PðV Þ ¼ F t V ¼ 174:93 kW ð6:16Þ
Using the results of Example 6.2 and a simulation similar to that used in
Chapter 5, the traction force versus vehicle speed and traction force versus time can
be generated as shown in Figure 6.2.
Figure 6.2 for tractive force versus vehicle speed V F–V diagram shows that
the engine and driveline delivers constant tractive effort to the driven wheels given
by (6.14) up to a vehicle speed given by (6.15) and then the constant power region
is entered up to the cruise speed V ¼ 45 mph simulated is reached. At the cruise
speed, the accelerating power is throttled back to sustaining power of approxi-
mately 2.3 kW on level grade and no headwind. Figure 6.2b shows that the constant
power (CP) regime is reached when t ¼ 7 s and persists up to t ¼ 25 s when
V ¼ 45 mph. This information will be useful in constructing a drive schedule
microcycle to evaluate the case studies.
The series hybrid version of the transit bus is illustrated in Figure 6.3, where
the engine drives a generator having the specifications given in Table 6.1 and
electric power output, Pg. The generator voltage is taken as Ug ¼ 700 V, and this d.c.
link is the input to the power inverter having characteristics listed in Table 6.1.
ch006 8 June 2011; 14:40:51

Tractive force (N) versus velocity (mph)

22.50k
20.00k
LIMIT1.VAL 15.00k
10.00k
5.00k
2.30k
–5.00 m 10.00 20.00 30.00 46.00
VEL.VAL
(a)
22.50k
17.50k
LIMIT1.VAL
15.00k
12.50k
10.00k
7.50k
5.00k
2.30k
0 5.00 10.00 15.00 20.00 25.00 30.00 40.00
t
(b)
Figure 6.2 Transit bus propulsion system. (a) F–V and (b) Ft versus time
diagrams
Series-hybrid
mw
Pdl
Pg ww
Invertor
Engine Generator = MG FD
~
Com VSC Pe
ESS
Figure 6.3 Hybrid transit bus representative propulsion architecture
The power inverter accepts input d.c. power from the engine generator and
supplemental electric power from the energy storage system, Pe. The VSC role is to
manage the engine generator under torque control as well as the traction motor-
generator (MG), to meet vehicle propulsion power requirements. Series hybrid
means that no mechanical connection exists between the engine and the driven
ch006 8 June 2011; 14:40:51

wheels. The propulsion power flow is electric up to the traction MG. The benefit of
a series hybrid is that the engine generator can be optimized to its most efficient
operating point for low emissions and it can be stopped when power demands can
be met by the ESS alone.
Capacitor Inverter/converter
Motor
Generator
CNG–engine
Motor for auxiliary Auxiliary
Figure 6.4 Transit bus propulsion system component integration [2]
The ESS in many hybrid transit buses consists of ultracapacitor only storage,
or batteries. The authors [2] describe a city bus having the architecture shown in
Figure 6.3 in which the ESS consists of ultracapacitor only storage rated 1310 Wh
in a 378-V, 194-kg pack capable of 100-kW propulsion power. Figure 6.4 shows
the layout of this ultracapacitor-only bus developed by Nissan Diesel Corporation
and its ultracapacitor ESS. Nissan Diesel Corporation considered the benefits of
lithium-ion battery versus EDLC and found that 200 kg of EDLC would give the
same long-term benefit as 318 kg of lithium-ion battery. Mitsubishi Fuso Truck and
Bus Corporation [3] describe a series hybrid transit bus using a spinel chemistry
lithium-ion battery ESS that is designed to reduce fuel consumption by 50% and
NOx and HC emissions by 75% of that by a conventional diesel engine powered
bus. The lithium-ion battery in this hybrid bus is designed to operate between 60%
and 75% SOC. The ISE Corporation [4] discusses development of their gasoline-
electric hybrid bus using Cobasys nickel-metal-hydride battery. More recently, the
ISE Corporation elaborated on their choice of ultracapacitor energy storage for
hybrid electric transit bus applications [5]. In this update ISE Corporation com-
pared the ratings, service application whether constant generation mode, engine
stop-start mode or engine load following mode, to estimate ESS service life.
Table 6.3 summarizes the ESS type and ratings.
For the three operating modes of the engine and electric drive system of the
series hybrid transit bus, the ISE Corporation found that battery depth of discharge
(DOD) cycling had to be very low to meet service life demands, and even then,
service life of battery systems was low in comparison with an ultracapacitor ESS.
Table 6.4 summarizes the findings for the stated number of energy cycles per year.
For the engine stop-start and load following modes summarized in Table 6.4,
total energy cycles can be seen over the estimated service life, taking into account
the energy used per cycle, that the Zebra pack is rated for approximately 110,000
ch006 8 June 2011; 14:40:51

Table 6.3 Transit bus energy storage system types and ratings [5]
ESS type Rated Useable Peak Operating Manufacture

energy energy power temperature warranty
(kWh) (kWh) (kW) ( C) (year)
Zebra 60 53 95 35 to þ50 2.5–5
NaNiCl2
Cobasys 25 <20 340 10 to þ55 4–6
NiMH
Ultracapacitor 1 >0.75 >200 35 to þ65 2–5
Table 6.4 Comparison of battery with ultracapacitor ESS by series

hybrid engine mode
Constant generation mode @ 62,500 kWh/year

ESS type ESS energy % DOD per cycle Estimated service
life (year)
Zebra 60 1.3 1.4
Cobasys 25 2.6 3.4
Ultracapacitor 1 70 8.0
Engine stop-start mode @ 70,833 kWh/year

Zebra 60 1.4 1.3
Cobasys 25 3.4 2.8
Engine load following mode @ 29,167 kWh/year

Zebra 60 0.8 3.1
Cobasys 25 2.0 8.6
shallow cycles, the Cobasys NiMH for approximately 500,000 shallow cycles and
the ultracapacitor for 1 million deep cycles. This lifetime-cyclable energy is the
rationale used by ISE Corporation and others to select ultracapacitors for long-term
cyclable energy storage in heavy hybrid applications.
The same rationale applies to metro rail, subway, and light rail applications [6].
In this study the authors describe the issue facing light rail designers when deceler-
ating that trains enter a station and regenerate into the catenary or third rail, causing
overvoltage and what is referred to as regenerative brake failure of the power system
electronics. On-site evaluations found that a 1560-V line may be driven to over
1800 V initiating regenerative brake failure. Ratings of these ‘rolling stock’ appli-
cations are 1500 V, 400 A, which can be handled by 0.6 kWh of EDLC useable
energy storage. In one application a 570 cell string of 800 F ultracapacitors were
sufficient to store 0.28 kWh of useable energy on board the train, sufficient to absorb
regenerative brake energy and limit overvoltage of the supply.
ch006 8 June 2011; 14:40:51

The final hybrid architecture to discuss before moving on to case studies is the
series-parallel configuration shown functionally in Figure 6.5. The hallmark of
series-parallel architecture is the presence of two electric machines, motor-
generator MG-A and MG-B, of which MG-A is normally the traction motor of
single mode systems (a topic discussed in detail in Chapter 8) and MG-B the
starter-generator machine. In 2-mode electronic continuously variable transmis-
sions (eCVT) to be discussed in detail in Chapter 9, both MG-A and MG-B are
equally rated. For our purposes in this chapter, the series-parallel architecture is
important because it has both a mechanical power transmission path and an electric
power transmission path, hence, the name series (for electric transmission) and
parallel (for mechanical plus electric transmission). The transmission shown in
Figure 6.5 with both electric motor-generators integrated is also known as an EVT
for electrically variable transmission. Regardless of the structure, the key feature of
these propulsion architectures is the electric fraction, that is, what percentage of
total engine plus electric propulsion is electric and the capacity of the energy sto-
rage system.
Series-parallel
MG-A
Engine eCVT FD
Com VSC MG-B
Invertor
ESS = ~
Figure 6.5 Hybrid electric transit bus with series-parallel propulsion architecture
In Figure 6.5 the vehicle system controller (VSC) manages the operation of the
engine and EVT electric motor-generators for seamless ratio control between the
engine and driven wheels. Each of the various hybrid architectures discussed so far
is most adapted to particular routes and driving environments as summarized in
Table 6.5.
The entries in Table 6.5 help explain why a BAE Systems’ Hybridrive hybrid
Daimler Orion transit bus with lithium-ion battery is a good choice for operating
over city routes in Manhattan, whereas an ISE Corporation modified New Flyer
transit bus with ultracapacitor energy storage is suited to operation in Los Angeles
and the Allison 2-mode hybrid with battery storage is more suited to operation in
Portland. The fact remains that hybrid transit buses are route sensitive in part
because of the transmission type and in part owing to stop-start frequency of the
routes. Hybrids excel in stop-go low-speed driving.
ch006 8 June 2011; 14:40:52

Table 6.5 Transit bus hybrid propulsion system route

adaptation
Route/arch Level Hilly Metro Metroþ Cold

grade grade only highway climate
Battery Electric X X
Hybrid Configurations of Figures 6.3 and 6.5
Series þ battery X X X
Series þ ultracapacitor X X X X
S-P þ battery X X
S-P þ ultracapacitor X X X
Based on results of model and simulation of a transit bus, a microcycle of

power demand versus time is developed, as shown in Figure 6.6. In this power
versus time chart, the portions having linear ramp in power with time correspond to
constant torque (CT) operation of the drive train, whereas the cycle portions having
a flat profile in power represent constant power (CP) operation. Recalling the
F–V chart of Figure 6.2, a microcycle is defined in which CT persists for 3.5 s,
CP for 9 s, and cruise at CP for which the engine delivers the propulsion power, not
the ESS.
Transit bus microcycle

200
150
100
Power (kW)
50
0
–50 0 5 10 15 20 25 30 35 40 45 50
–100
–150
–200
Time (s)
Figure 6.6 Heavy hybrid representative microcycle (negative power is motoring)
In the following sections, the microcycle of Figure 6.6 will be modified

somewhat in the regeneration portion following the high brake regeneration
recovery power pulse in the area of t ¼ 40 s to approximate the engine being used to
replenish the ESS charge level and bring the pack back to nominal, SOC0. A transit
bus in revenue service can be expected to operate two shifts per day, 6 days/week,
and approximately 6000 h/year. A normal route will also mean 250 cycles/day of
stop-go events for loading and unloading at bus stops.
ch006 8 June 2011; 14:40:52

6.1 Battery electric bus

A battery electric bus of 11–12 m design and nominally 15,000 kg of loaded mass
can be expected to consume stored electrical energy at a rate of 1500 Wh/mi.
Table 6.6 lists the characteristics of a battery electric bus that is in revenue service
in the Asian market.
Table 6.6 Battery electric transit bus parameters
Attribute Definition Value Unit

Fully loaded mass Curb þ passenger (Mv) 16,000 kg
Frontal area L11.6 H3.2 W2.5 (Af) 8 m2
Maximum velocity Wide open throttle (Vwot) 91/56.5 kph/mph
Range All electric range (AER) With 400-Ah 210/130 km @ 40 kph
lithium-ion pack mi @ 24 mph
Range All electric range (AER) With 600-Ah 307/191 km @ 40 kph
lithium-ion pack mi @ 24 mph
Battery capacity With 600-Ah, 390-V pack (Cb) 234 kWh
Battery capacity With 400-Ah, 390-V pack (Cb) 156 kWh
Battery voltage Nominal voltage (Ub) 388.8 Voc, open circuit
Peak power Battery pack power (Ppk) 150 kW
Acceleration time 0–25 mph (tz25) 20.7 s
Figure 6.7 illustrates this class of transit bus, such as the type used for the
Beijing Olympics in August 2008 to transport contestants from Olympic village to
the competition areas of Beijing. The Beijing Olympic battery EV buses were
equipped with a fast exchange mechanism and novel robotized pack replacement
facility at the Transportation Authority garage, so spent packs could be exchanged
in less than 8 min.
Figure 6.7 Battery electric bus for 2008 Beijing Olympics with battery pack
compartments beneath floor [courtesy of JNJ Miller PLC]
ch006 8 June 2011; 14:40:53

Example 6.3: Using data from Table 6.6 for the battery electric bus, compute the
energy consumption per mile at 40 kph for both the 400- and 600-Ah lithium-ion
packs, assuming the packs are initially charged to 95% SOC and depleted when the
SOC is 25%. Capacity fade due to aging is not considered.
Solution: For this battery electric transit bus, the AER|400 Ah ¼ 130 mi for capacity,
Cb ¼ 156 kWh, and AER|600 Ah ¼ 190 mi for capacity 234 kWh. Call the energy
consumption rate, g, for which

dSOCðC b Þ 0:7ð156;000Þ Wh
g400 Ah ¼ ¼ ¼ 840 ð6:17Þ
AER400 Ah 130 mi 25 mph

dSOCðC b Þ 0:7ð234;000Þ Wh
g600 Ah ¼ ¼ ¼ 862 ð6:18Þ
AER400 Ah 190 mi 25 mph
The result is as one would expect, the same BEV transit bus with a larger-
capacity battery pack operated at the same speed will consume energy at the same
rate. At a higher speed, the bus will consume considerably larger amounts of energy
as discussed in the problems at the end of this chapter.
The range computed in Example 6.3 for the BEV transit bus assumes a level
route and no headwind. When grade is present, the AER numbers can drop sub-
stantially, a fact that all electric and hybrid electric bus manufacturers are well
aware of. Consider the case of the BEV having parameters listed in Table 6.6
negotiating a realistic gr ¼ 6% grade for z ¼ 1.5 mi along one segment of an overall
route. The transit bus battery pack will therefore be depleted by an increment of
energy determined by the change in potential energy of the bus. In this case, the
vertical height, h, ascended by the bus while traveling the slope distance, z, at the
stated grade is

%gr 6
h ¼ z sin tan1 ¼ 1:5 sin tan1 ¼ 1:5 sinð3:4336Þ
100% 100
¼ 0:08984 mi ð6:19Þ
Or, a vertical ascension, z ¼ 0.08984 mi (1609 m/mi) ¼ 144.55 m. This may not
seem like much of a vertical climb, but for a large bus it represents considerable
potential energy.
W P ¼ gM v h ¼ 9:802ð15;000Þð144:55Þ ¼ 21:253 MJð5903 WhÞ ð6:20Þ
The driveline efficiency as stated in Table 6.1 as FD transmission electric

motor power inverter equals htot ¼ (0.98)(0.97)(0.94)(0.96) ¼ 0.858 overall, not
counting ESS discharge efficiency. To supply this amount of energy, the ESS must
discharge by WP/htot ¼ 5903 Wh/0.858 ¼ 6880.7 Wh. This is a very substantial
amount of energy, even for a lithium-ion pack. For instance, for a 390-V ESS pack,
this amount of energy represents a capacity of 17.64 Ah, a very good fraction of a
standard 100 Ah cell just to negotiate one hill climb.
ch006 8 June 2011; 14:40:54

6.2 Hybrid electric bus

The situation for hybrid electric transit bus of CIDI engine (diesel or CNG fueled),
gasoline, or fuel cell is very similar. But for the case of an HEV bus, we now
consider the case of ultracapacitor-only ESS. In many of their conversion products,
the integrator ISE Corporation [4,5] retrofit New Flyer coaches with gasoline–
electric hybrid propulsion having ultracapacitor-only energy storage. In this sec-
tion, we consider the design challenge of ultracapacitors, specifically how to select
the nominal SOC0 level. This is an important consideration because the ESS must
have capacity above SOC0 to store regenerative brake energy and sufficient reserve
capacity below SOC0 for acceleration and other motoring events. Figure 6.8 is an
ISE conversion bus of the type to be evaluated here.
Figure 6.8 Hybrid transit bus having 280S 1P 3000F ultracapacitor ESS
(Ns ¼ 280 cells, 3000 F/cell in two packs, 760 V > Ud > 360 V,
Ppk ¼ 150 kW)
For the 3000 F ultracapacitor cell specified in the HEV bus of Figure 6.8, the
useable energy dWcell ¼ 2.27 Wh/cell and for the pack,
dW uc ¼ N s dW cell ¼ 280ð2:27Þ ¼ 635:6 Wh ð6:21Þ

For the amount of available energy given by (6.21), the energy capacity for
regeneration dWucþ ¼ dWuc/2 ¼ 318 Wh ¼ dWuc, the reserve capacity. Using this
information the initial SOC0 can be calculated. Let s ¼ Ur/Umx, and apply this to
dWucþ and dWuc, then solve for s as follows:
dW þ 2 2 2 2 2
uc ¼ ð1 s ÞU mx ¼ ðs 0:5 ÞU mx ¼ dW uc ð6:22Þ
pffiffiffiffiffiffiffiffiffiffiffi
2s2 ¼ 1:25; s ¼ 0:625 ¼ SOC20 ð6:23Þ
pffiffiffiffiffiffiffiffiffiffiffi
U r ¼ sU mx ¼ 0:625ð720Þ ¼ 569:2 V ð6:24Þ
ch006 8 June 2011; 14:40:54

From (6.22) the initial SOC0 ¼ 0.79, at which point the ultracapacitor pack is
biased at a rated voltage, Ur ¼ 569.2 V, or 570 V for a round number. At this
average voltage, the ultracapacitor pack retains 318 Wh of regenerative capacity
and has a reserve of 318 Wh for boosting. Another advantage of this operating
mode is that ultracapacitor cell potential is Urcell ¼ Umxcell ¼ 0.79(2.7) ¼ 2.13 V/
cell, which puts the cell voltage stress well in the range (i.e., < 2.4 V/cell) where
service life is maximized.
Table 6.7 highlights the essential parameters for the HEV transit bus that will
be used to evaluate its acceleration time and energy demand. In this example,
hybrid propulsion is capable of 150-kW peak power starting at a vehicle CT-CP
transition of V ¼ 18 mph (8.05 m/s), at which point the constant torque accelerating
phase ceases to deliver a constant force, Ft ¼ 18,633 N, to the driven wheels. Axle
torque can be found by using this value of traction force and the driven wheels
rolling radius, rw.
Table 6.7 Simulation attributes for 15,000-kg hybrid
transit bus with ultracapacitor-only ESS
Attribute Value Unit

Curb mass 9600 kg
Maximum passenger mass 5400 kg
Tires: 275/70R22.5, rw 0.478 m
Frontal area, Af 8 m2
Propulsion motor: 2x Siemens 150 kW
asynchronous into gear comb
Energy storage: 2 series UC packs, 0.636 kWh
140 cells each
Figure 6.9a shows the hybrid bus traction force, Ft, versus vehicle velocity, V,
for the ideal case of no driveline losses and simulated acceleration times from 0 to
25 mph. Corresponding vehicle speed and energy consumption are provided in the
charts in Figure 6.9b.
According to the acceleration results presented in the table in Figure 6.9a, the
amount of useable energy given by (6.21) is 318 Wh and the acceleration energy
required to accelerate from zero speed to 25 mph is 295 Wh, within the ESS cap-
ability limits for the SOC0 selected. Therefore, the bus engine should mesh with the
electric drive system in the neighborhood of V ¼ 20 mph to V ¼ 25 mph for a
seamless transition to engine power. When the transit bus approaches a bus stop,
the engine would shut down and the electric drive takes over regenerative braking.
During deceleration the electric drive acts as a generator and pulls 150 kW of
braking energy from the driven wheels and directs it to the ESS for storage. As
vehicle speed approaches zero and the electric motor-generator is not efficient, it is
shut down and the vehicle service brakes are engaged. This also means that service
brake friction pads see very little use in a hybrid or battery electric bus and need
little maintenance. It is now widely known that bus service brakes will last from
two to three times their normal service life in a hybrid bus.
ch006 8 June 2011; 14:40:54

Tractive forece (N) versus velocity (mph)

18.63 k
15.00 k
LIMIT1.VAL
12.50 k
10.00 k
7.50 k
5.00 k
2.20 k
–2.50 m 10.00 20.00 30.00 46.00
VEL.VAL
Metric Time Energy
Tz20 9.4 s 186 Wh
Tz25 12.4 s 295 Wh
Tz45 32.8 s 1078 Wh
(a)
Velocity, mph
46.00
VEL.VAL
30.00
20.00
10.00
–100.00 u
–2.50 m 10.00 20.00 30.00 40.00 47.60
t
Energy, Wh
1.33k
1.00k
ENERGY.VAL
800.00
600.00
400.00
200.00
–200.00 n
–2.50 m 10.00 20.00 30.00 40.00 47.60
t
(b)
Figure 6.9 Hybrid transit bus acceleration characteristics. (a) Ft by V

characteristic acceleration results tabulated and (b) velocity versus
time energy versus time
The hybrid transit bus having specifications listed in Table 6.7 is subjected to
the bus microcycle of propulsion power versus time given in Figure 6.6 for the case
of 280S 1P 3000 F cell pack, configured as 2 in series, 140-cell custom
ch006 8 June 2011; 14:40:55

modules. For the microcycle example, the ultracapacitor pack is charged to a level
above its 79% SOC0 value, to 680 V, because of the high initial discharge repre-
senting acceleration. Module time constant will be higher than cell time constants
because of additional resistance contribution from cell interconnect straps. The
simulation model accommodates these parasitic resistance effects. These facts are
made clearer by a short digression into cell and module characteristics provided in
Table 6.8 for the most current Maxwell Technologies K2 cell line and the modules
these cells are assembled into shown as Table 6.9. For example, the 3000-F K2 cell
shown in Table 6.8 has a time constant of 0.87 s, but when integrated into a 48-cell
module, the additional interconnect strap and end termination cables increase
this to 1.13 s by inspection of Table 6.9, column 3, row 3. Considering that the
time constants scale by the cell level ESR, then the 3000 F cell ESRdc ¼ 0.29 mW
has increased to ESRmod dc ¼ t/Ccell ¼ 1.13/3000 ¼ 0.377 mW. The astute reader
will recognize this value of interconnected cell ESRdc as simply ESRmod/
Ns ¼ 18 mW/48 ¼ 0.375 mW. This simple exercise illustrates very clearly the
need to reduce resistances as much as possible since added resistance not
only limits peak power achievable but also leads to higher internal power dissi-
pation and more difficulty in integrated thermal management systems. The
increase in ESRdc of the 3000-F component from cell level to module level means
Table 6.8 Standard large-cell product line (Maxwell Technologies)
K2 cell 650 F 1200 F 1500 F 3000 F

ESRdc ¼ 0.8 mW ESRdc ¼ 0.58 mW ESRdc ¼ 0.47 mW ESRdc ¼ 0.29 mW
t ¼ 0.52 s t ¼ 0.696 s t ¼ 0.705 s t ¼ 0.87 s
Table 6.9 Standard modules for bus, wind turbine, and train applications
8 cell, 48.6 V, 165 F 32 cell, 75 V, 94 F 48 cell, 125 V, 63 F

6.3 mW, 1.04 s 12.5 mW, 1.175 s 18 mW, 1.13 s
98 Arms, passive cool 50 Arms, passive cool 150 Arms, fan cooled
ch006 8 June 2011; 14:40:55

the added interconnect resistance, Rconn, used in the equivalent circuit model should
be approximately

ESRmod
Rconn ¼ ESRdc ¼ ð0:375 0:29Þ mW ¼ 0:085 mW ð6:25Þ
Ns
The hybrid bus ESS pack will have the parameter values listed in Table 6.10
that take into account the effect of interconnecting such a large number of indivi-
dual cells in a practical package. This is why it is critically important to make every
effort to reduce ESR, starting with cells and on up to full pack level. A simple
calculation of the pack time constant can be compared with that of the cells that
comprise the pack to show how well the design is being executed.
Table 6.10 Hybrid transit bus ESS pack parameters for
0.636 kWh at 720 V
Parameter Definition Calculation Value

ESRpak ¼ N s ESRmod
d:c: ¼ 280(0.375 m) ¼ 105 mW
Cpak ¼ Ccell/Ns ¼ 3000/280 ¼ 10.71 F
Ud ¼ 680 V
tpak ¼ 1.124 s
When subjected to the bus microcycle power profile, the hybrid transit bus
ESS pack voltage drops from the initial condition setting of 680 V to a minimum of
462 V, shown in Figure 6.10a, then recovers to 604 V after the regeneration power
Bus microcycle power Bus microcycle energy (Wh)
150.00 k 688.00
BusMicrocycle.VAL
100.00 k
600.00
INTG1.VAL
50.00 k
550.00
0k 500.00
–50.00 k 450.00
–100.00 k 400.00
332.00
–150.00 k
0 10.00 20.00 30.00 40.00 50.00 0 10.00 20.00 30.00 40.00 50.00
(a) (b)
Bus ESS voltage Bus ESS current
680.00 280.00
650.00 200.00
625.00
VM1.V (V)
600.00 100.00
I1.(A)
575.00 0
550.00
525.00 –100.00
500.00 –200.00
456.00
–312.00
0 10.00 20.00 30.00 40.00 50.00 0 10.00 20.00 30.00 40.00 50.00
(c) (d )
Figure 6.10 Hybrid transit bus ESS performance over the microcycle.
(a) Microcycle power, (b) ESS pack energy changes, (c) ESS pack
voltage (Ud0 ¼ 680 V), and (d) ESS pack current (812 A offset)
ch006 8 June 2011; 14:40:55

pulse. For the 150-kW peak motoring and regeneration power levels, the ESS pack
current reaches 309 A on discharge and þ277 A on charge. For energy mon-
itoring the initial energy level is set to correspond with an initial condition of
680 V, or 688 Wh. ESS pack energy following the motoring pulse drops to
333.6 Wh and recovers to 568.5 Wh following the regeneration pulse. The deficit,
119.5 Wh, is due to less regeneration energy returned compared with the energy
removed. Therefore, in the hybrid bus system, the engine generator would need to
replenish this energy at a power level commensurate with the driving route cruise
intervals. Suppose this interval is 90 s minimum, then 119.5 Wh can be replenished
by setting the generator output, Pg ¼ dWess/T ¼ 119.5(3600)/90 ¼ 4.78 kW–a very
modest sustained power level for a generator having 100-kW peak rating.
6.3 Shuttle bus

This section covers the topic of a class 4 truck, the E450 shuttle van. Shuttles of this
category are widely found as airport and hotel buses. Because of their heavy stop-
start and urban use, there have been numerous efforts over the years to reduce the
emissions of such shuttles and to electrify them. Figure 6.11 illustrates such a
shuttle bus, in this case one developed by the Ford Motor to operate on hydrogen in
a specially designed internal combustion engine. Hydrogen is stored in 5000-psi
(35 MPa) tanks to provide the E450 van a driving range of 150 mi. When this
vehicle was delivered to fleet customers in 2006, it realized 99.7% reduction in
exhaust CO2 emissions.
(a) (b)
Figure 6.11 Illustration of class 4 shuttle van and its Triton V10 engine designed
to operate on hydrogen. (a) E450 shuttle, hydrogen powered and
(b) Triton V10 6.8-L H2 engine
For the example shuttle bus having a parallel hybrid architecture drive train
with the parameters given in Table 6.11, the 400 Wh of deliverable ultracapacitor
energy is provided by a 10S 1P 48 V module set. The specifications of this
energy storage module are listed in Table 6.12 based on the same derivation of
ESRdc as done for the hybrid transit bus and HTM module. From Table 6.9, the
ch006 8 June 2011; 14:40:56

Table 6.11 Specifications for E450 shuttle bus

hybrid example
Attribute Value Unit

Curb mass 7200 kg
Maximum passenger mass 2800 kg
Tires: LT245/75R16, rw 0.387 m
Frontal area, Af 4.2 m2
Parallel hybrid architecture, 90 kW
electric drive power
Energy storage: 360 Vd.c., 0.400 kWh
10S 1P 48 V modules
Table 6.12 Hybrid shuttle bus ESS pack parameters

for 0.400 kWh at 486 V
Parameter Definition Calculation Value

ESRpak ¼ 10ESRmod
d:c: ¼ 10(6.3 m) ¼ 63 mW
Cpak ¼ Cmod/10 ¼ 165/10 ¼ 16.5 F
Ud ¼ 459 V
tpak ¼ 1.04 s
48-V module, consisting of Ns ¼ 18 cells of 3000 F cells with ESRdc ¼ 0.29 mW,
has increased to ESRmod dc ¼ t/Ccell ¼ 1.04/3000 ¼ 0.347 mW. The interconnect
resistance (6.25) is therefore determined as the difference between module level:
ESRmod/Ns ¼ 6.3 mW/18 ¼ 0.350 mW and the cell ESRdc ¼ 0.29 mW, or 0.06 mW.
For the E450 shuttle bus, the initial condition on ESS voltage is taken as the
maximum pack voltage of Ud ¼ 180 2.55 ¼ 459 V, which corresponds to the
maximum recommended working voltage of the 2.7 V cells for high service life, or
2.55 V/cell.
The energy of the shuttle bus ESS pack changes from the pack maximum energy
of 482.8 Wh to a minimum of 270 Wh, or a change of 212.8 Wh, well within the
available 400 Wh of deliverable energy of this pack. This means the pack voltage does
not reach Umx/2 ¼ 230 V and can be seen to reach 336 V minimum in Figure 6.12c.
Del Core [7] provided a comprehensive review on the status of hybrid transit
bus and comparison of hybrid bus to hybrid automobile at the 2010 Advanced
Energy Storage conference. Salient points of his comparison are listed as
Table 6.13 to summarize this chapter on heavy transportation.
Exercise 6.7 uses the energy cycles of the hybrid vehicle applications listed in
Table 6.13 in an effort to rationalize the life expectancy of the same battery, though
larger, used in the transit bus to that of the hybrid sedan. The reader is encouraged
to work through this exercise for an appreciation of how stressful such energy
cycling is on the energy storage system and what it may take to develop a battery
sufficiently robust for this duty.
ch006 8 June 2011; 14:40:56

Shuttle microcycle power Shuttle microcycle energy (Wh)

90.00 k 482.80
BusMicrocycle.VAL
60.00 k 450.00
40.00 k 425.00
INTG1.VAL
20.00 k 400.00
0k 375.00
–20.00 k 350.00
–40.00 k 325.00
–60.00 k 300.00
–90.00 k 270.00
0 10.00 20.00 30.00 40.00 50.00 0 10.00 20.00 30.00 40.00 50.00
(a) (b)
Shuttle ESS voltage Shuttle ESS current
459.00 238.00
440.00
100.00
VM1.V (V)
420.00
I1.(A)
400.00 0
380.00 –100.00
360.00
336.00 –258.00
0 10.00 20.00 30.00 40.00 50.00 0 10.00 20.00 30.00 40.00 50.00
(c) (d )
Figure 6.12 Hybrid shuttle bus ESS performance over the microcycle profile.
(a) microcycle power (90 kW peak), (b) ESS pack energy changes
(Wess0 ¼ 482.8 Wh), (c) ESS pack voltage (Ud0 ¼ 459 V), and (d) ESS
pack current (659.8 A offset)
Table 6.13 Comparison of hybrid transit bus to hybrid automobile [7]
Attribute Hybrid Hybrid

automobile – sedan transit bus
Curb weight (kg) 1590 13,600
ESS capacity (B, battery; U, ultracapacitor) (kWh) 1.5 B:14.6; U:1.0
Stop-start events per day 30 750
Energy delivered/absorbed per event (Wh) 50–100 500
ESS peak power (kW) 30 200
ESS nominal voltage and current (V)/(A) 273/110 600/333
Energy throughput per day (kWh-cycles) 1 B:13; U:189
Exercises
6.1 Compute the amount of stored energy in an ESS that is the equivalent of the
potential energy portion of the transit bus having parameters tabulated in
Table 6.1 when it executes a 1.5-mi route at a þ2% grade.
Hint: Compute the angle from the stated grade, then elevation from the given
distance and angle and finally the potential energy in Joules and Watt-hours.
ch006 8 June 2011; 14:40:57

Ans: a ¼ tan1 ð%gr=100Þ ¼ tan1 ð2=100Þ ¼ 1:146 is the grade angle.

For a distance on this slope, z ¼ 1.5 mi means that elevation h ¼ z sin(a),
so h ¼ ð1609 m=mileÞz sinðaÞ ¼ 48:26 m. Therefore, the electrical energy
in an ESS the equivalent of the potential energy is W pot ¼ ghM v ¼
9:806ð48:26Þð15;000Þ ¼ 7:098 MJ ¼ 1:97 kWh
6.2 For the energy obtained in Exercise 6.1, compute the equivalent mass of
various energy storage packs that this represents. Given that lithium-ion has a
specific energy, SE ¼ 126 Wh/kg fully packaged, NiMH has SE ¼ 63 Wh/kg
and the carbon–carbon ultracapacitor SE ¼ 4 Wh/kg.
Ans: Lithium-ion MLi ¼ Wpot/SELi ¼ 15.76 kg,
NiMH has MNiMH ¼ Wpot/SENiMH ¼ 31.27 kg,
Ultracapacitor has MUC ¼ Wpot/SEUC ¼ 492.5 kg.
6.3 Based on the findings of Exercises 6.1 and 6.2, what does this say about
various hybrids being route sensitive?
Ans: Any route having substantial portions of even moderate grades will be
very demanding of the electric energy storage system capacity, so the
burden of navigating grades falls back on the engine. Ultracapacitors
because of their very high cyclable energy throughput favor level, stop-go
routes.
6.4 Extend the findings of Example 6.3 to the case of a battery electric bus
operating over an urban route having posted 30 and 45 mph limits and assume
the bus will reach these speeds. Is it feasible to bracket the battery energy
consumption per mile at 1.5 < g < 2.0?
Ans: Yes, for higher speeds, the propulsion power incremental loading in-
creases at roughly V1.6, where Vavg * (30þ45)/2 ¼ 37.5 mph and (37.5/25)1.6 ¼
1.9. Therefore we have 1.9 (840) Wh/mi ¼ 1596 Wh/mi.
6.5 A hybrid transit bus operates two shifts per day, 6 days per week for a total of
5000 h/year. Find the total daily operating hours, and given that the bus
makes 250 stops a day, compute the average time between stops.
Ans: 3.3 min. Since top ¼ (5000/8760)24 ¼ 13.7 h, and Nsg ¼ 250/13.7 ¼ 18.25
stop-go per hour, the average time between stops is 60 min/18.25 ¼ 3.3 min.
6.6 Suppose the accessory power on average in battery electric transit bus is
Pacc ¼ 2.85 kW for electric drive air conditioning, lights, air brake compressor
intermittent use, and entertainment, then compute (a) the accessory energy
load per day for the conditions given in Exercise 6.5 and (b) the battery pack
cell capacity necessary to support this additional load.
Ans: (a) Wacc ¼ 39 kWh/day. This can be computed as
W acc ¼ hPacc itop ¼ 2:85 kWð13:7 hÞ ¼ 39 kWh:
(b) C bacc ¼ W acc =U b jU b¼ 390 V ¼ 39;000=390 ¼ 100 Ah cell
6.7 For the hybrid sedan and hybrid transit bus, battery data are provided in
Table 6.13, and assume that the hybrid sedan battery has a 10-year warranty,
ch006 8 June 2011; 14:40:57

then what would the corresponding warranty period be for the hybrid transit
bus if exercised in the same manner?
Hint: The hybrid sedan can be assumed to operate approximately 250 days/
year, and the hybrid bus 312 days/year (52 weeks in service, 6 days/week).
Ans: For this question the following tasks are carried out:
(a) Number of turns for the sedan battery over its estimated service life is
derived and tabulated.
(b) This same metric is applied to the battery in the transit bus to estimate
its service life.
Number of turns computed for hybrid sedan battery pack
Definitions Calculation Results

(1 kWh cycle/day) ¼ 2500 kWh cycles Over service life
(250 days/year) (10 years)
Battery capacity (kWh), Cb ¼ 1.5 At beginning of life
Number of turns @ 100% SOC ¼ 2500 kWh cycles/(1 1.5 kWh) 1667 turns
Number of turns @ 50% SOC ¼ 2500 kWh cycles/(0.5 1.5 kWh) 3334 turns
Number of turns @ 25% SOC ¼ 2500 kWh cycles/(0.25 1.5 kWh) 6667 ns
Number of turns computed for hybrid transit bus battery pack
(13 kWh cycle/day) ¼ 4056 kWh cycles/year At 1-year basis

(312 days/year)
Battery capacity (kWh), Cb ¼ 14.6 At beginning of life
Number of turns/year for the bus ¼ (4056 kWh cycles/year)/Cb ¼ 277.8 equivalents
(Number of turns @ 100% SOC)/ ¼ (1667 turns)/277.8 6 years
(number of equivalents/year)
The results presented in this exercise and cast as total battery energy turns at var-
ious SOC windows are done at nominal temperature (25 C) and are consistent with
documented work of battery industry experts. In this case, a chart of cycle cap-
ability of various battery types versus depth of discharge (i.e., SOC window) is
derived from work performed by Johnson Controls-Saft Advanced Power Solu-
tions. The lithium-ion curve (second from left) at 40% dSOC shows approximately
4000–5000 cycles of life. This would be sufficient to meet the necessary 3334 turns
at 50% SOC listed above. Of course, real-world environmental conditions are not
controlled laboratory conditions that the chart was developed for, so number of
turns in reality should be assumed to be well short of those in the graph
(Figure 6.13).
ch006 8 June 2011; 14:40:58

Advanced battery cycle life versus SOC window

100.00
90.00 NiMH
Depth of discharge (% DOD)
80.00 Li-ion
70.00 AGM
60.00 Li-FePhos
50.00
40.00
30.00
20.00
10.00
0
1.00E + 02 1.00E + 03 1.00E + 04 1.00E + 05 1.00E + 06 1.00E + 07
Cycles to wear-out (#)
Figure 6.13 Log-linear chart of cycle capability versus dSOC
References
1. J.M. Miller, Propulsion Systems for Hybrid Vehicles, 2nd edn., The Institu-
tion of Engineering and Technology (IET), Stevenage, United Kingdom,
2010
2. T. Kawaji, S. Nishikawa, A. Okazaki, S. Araki, M. Sasaki, ‘Development of
Hybrid Commercial Vehicle with EDLC’, The 22nd International Battery,
Hybrid and Fuel Cell Electric Vehicle Symposium and Exposition, EVS-22,
Yokohama, Japan, pp. 228–36, 23–8 October 2006
3. Y. Susuki, K. Yoichiro, T. Kondo, T. Moriva, S. Shiino, K. Mori, ‘Series
Hybrid Electric Drive System for City Bus’, The 22nd International Battery,
Hybrid and Fuel Cell Electric Vehicle Symposium and Exposition, EVS-22,
Yokohama, Japan, pp. 210–18, 23–8 October 2006
4. J. Goldman, P.B. Scott, ‘Modern Hybrid Electric Transit Buses – Research
Driving Development’, The 23rd International Battery, Hybrid and Fuel Cell
Electric Vehicle Symposium and Exposition, EVS-23, Anaheim, CA, pp. 1–4,
2–5 December 2007
5. P.B. Scott, J. Schulte, ‘Batteries in Heavy Duty Hybrid Electric Vehicle
Applications’, Presented at SAE 2008 Hybrid Vehicle Technology Sympo-
sium, San Diego, CA, 14–15 February 2008
6. Y. Sekijima, Y. Kudo, M. Inui, Y. Monden, S. Toda, I. Aoyama, ‘Develop-
ment of energy storage system for dc electric rolling stock applying electric
double layer capacitor’, 6th Committee Meeting on Vehicle Energy Storage
Systems, The Institution of Electrical Engineers Japan (IEEJ), University of
Tokyo, Tokyo, Japan, 2 November 2007
7. R. Del Core, ‘ISE Corp: innovative solutions for energy’, Presented at
Advanced Energy Storage 2010, Catamaran Hotel, San Diego, CA, 12–14
October 2010
ch006 8 June 2011; 14:40:58

ch006 8 June 2011; 14:40:58
Chapter 7
Hybrid electric vehicles
Ultracapacitors are now beginning to be applied in low-end hybrid electric vehicles

for primarily idle–stop feature [1]. In reality, an idle–stop system is not a true
hybrid electric vehicle, rather a microhybrid, since it applies no electric torque to
the vehicle-driven wheels. The PSA Peugeot Citroen system consists of a Valeo
iStARS (integrated Starter-Alternator Reversible System) that provides engine
stop–start function by way of the alternator belt. The iStARS delivers, on its own, a
15% fuel consumption reduction on the New European Driving Cycle (NEDC)
when integrated into 1.4 and 1.6 L HDi diesels. The market appears to be
approximately 1 million such units by 2012. This microhybrid represents two
industry firsts: (1) a belt-driven system for a diesel and (2) ultracapacitor for energy
storage.
Figure 7.1 illustrates the belt-driven microhybrid concept iStARS used in the
Peugeot that is designed for 600,000 warm restarts of the engine in <400 ms. The
alternator is rated 2.2 kW and can assist in engine cranking when the temperature is
below 25 C by being operated as a motor and provided a special belt tensioner is
used for reverse flow of torque to the engine.
The microhybrid features can be summed up as:
● Engine shuts off while the vehicle is in motion and below approximately 8 kph.
● In emergency braking, the engine remains running to provide hydraulic assist
to the brakes.
● If the battery SOC is low, the engine remains ON to charge the battery.
● If the temperature is above 30 C, the engine remains ON to sustain cabin air
conditioning.
● When the temperature is below 5 C, the engine stays ON for cabin heating.
● The two-cell ultracapacitor module is designed to aid the battery by boosting
its voltage during engine cranking so that vehicle electrical distribution system
voltage is stabilized.
Mild-hybrid implementations come in two varieties: (1) belt integrated starter
generator (B-ISG) or belt alternator starter (BAS) and (2) crankshaft mounted
integrated starter generator (ISG). The most recognized mild hybrid is the BAS that
GM put into production in the Saturn Vue Greenline hybrid as a 42 V system. In
this system, the alternator is a specially designed reversible system by Hitachi that
is rated 5 kW electrical power and > 4 kW mechanical power, 60 N m into the belt
ch007 7 June 2011; 17:59:19

1 e-Booster
1a Power Electronics
1b Ucaps 5V
2 Heavy Duty Battery 12 V 70Ah
3 Integrated Stop Start
4 Second generation Belt-driven starter-alternator
5 HDi Engine
Control
Power
Figure 7.1 Belt integrated starter alternator reversible system, eHDi [1], and
engine illustration (ultracapacitor module is approximated here as a
2S 1P 1200 F)
torque during warm restart. Figure 7.2 shows the integration of the 42 V BAS into
the Saturn Vue Greenline mild hybrid (now discontinued).
Highlights of the Vue Greenline mild-hybrid vehicle shown graphically in
Figure 7.2 are the following:
● 4T45E electronically controlled transmission, including electric drive oil pump.
● 2.4 L, in-line four-cylinder Ecotec engine.
● Reversible alternator, 5 kW electrical, 4 kW mechanical, 60 N m belt torque for
engine restart.
● NiMH 36 V battery (3S 1P 12V Cobasys modules) capable of 10 kW
power at 60% SOC.
The most recognized second type of mild hybrid is the Honda IMA (Integrated
Motor Assist) comprising a crankshaft mounted starter generator, separate power
electronics unit, and 144 V NiMH battery. The battery pack consists of 6.5 Ah cells,
ch007 7 June 2011; 17:59:19

Hybrid electric vehicles 215
Figure 7.2 Mild-hybrid vehicle, (top) Saturn Vue Greenline vehicle with NiMH
battery tray behind rear seat, (bottom) Hitachi reversible alternator
and power electronic control center
approximately 900 Wh stored energy, in C cell size. The Honda Civic 4th genera-
tion IMA is sandwiched between the 1.3 L iVTEC engine and the push belt con-
tinuously variable transmission (CVT). The engine is rated 68 kW peak and the
IMA is 15 kW electrical, 103 N m of torque motoring. The NiMH battery is air
cooled and mounted behind the rear seat. In this system, Honda realizes 50 mpg in
the Civic hybrid and tz60 ¼ 11.5 s. In recent work the DOE’s National Renewable
Energy Laboratory (NREL) has performed analysis [2] and vehicle demonstrations
[3] on the benefits of ultracapacitor energy storage in mild-hybrid electric vehicles.
Pesaran and Gonder [2] demonstrated that with ultracapacitors the Saturn Vue
ch007 7 June 2011; 17:59:19

hybrid could realize up to 20% fuel economy improvement in urban driving and
and concluded the following:
● Little fuel economy difference is noted between use of ultracapacitors and
lithium-ion batteries in a mild-hybrid vehicle.
● If cost is the same, and not taking advantage of lithium-ion’s energy reserve,
the ultracapacitors superior life and low temperature performance make them
more attractive.
Most recently, Gonder et al. [3] converted the Saturn Vue Greenline hybrid to
three energy storage system configurations: (1) with stock NiMH 36 V pack,
(2) with a single Maxwell Technologies 48 V, 165 F, ultracapacitor module, and
(3) with two of the Maxwell Technologies BMOD0165-P048 modules. Their main
finding from the demonstrator vehicle was that it performed as well with a single
48 V BMOD0165-P048 module as with the three stock NiMH modules. Their
conclusion was as follows:
● Ucaps possess excellent life and low-temperature performance.
● Ucaps have low long-term projected costs.
Dynamometer testing of a 2007 Vue Greenline mild hybrid using ultra-
capacitors in liu of a battery showed that cyclable energy of < 50 Wh is sufficient
for this class vehicle over the Urban Dynamometer Drive Cycle (UDDS) drive
cycle. For more aggressive driving, such as the US06 drive cycle, the Vue Green-
line with two each of the BMOD0165-P048 modules in parrallel provided < 70 Wh
of deliverable energy that shows a significant benefit, but data show that up to
150 Wh energy is needed for this drive schedule.
The reader is advised to consult current technical literature for updated
material on ultracapacitor applications in hybrid electric vehicles. This field is
evolving quickly, and the citations here are only representative of the wealth of
information on topic of ultracapacitor applications in HEVs. Lee et al. [4] propose
a hybrid energy pack consisting of valve regulated lead acid (VRLA) battery and
high power ultracapacitors in a 42 V PowerNet application. This topic is relevant
considering that General Motors went into production with a 42 V PowerNet BAS
hybrid in the Saturn Vue Greenline series. The hybrid energy pack demonstrated
here consists of a 36 V, 1540 Wh VRLA battery from Hyundai Enercell Co. Ltd.,
and an 18 cell, 5000 F/cell ultracapacitor 48 V module rated 278 F, 4.7 mW, and
77 Wh at 45 V. It was found that the hybrid energy pack only raised available
capacity by 5% but increased discharge efficiency by 19% compared to the VRLA
battery alone. Benson et al. [5] describe the application of ultracapacitors in a
Formula SAE (FSAE) hybrid race car. The vehicle is more of a GoKart that used
130 cells, 2.5 V, 2700 F units from Maxwell Technologies to provide a total of
325 V, 680 kJ (188 Wh). The vehicle propulsion motor was a Solectria AC55,
78 kW peak, 34 kW continuous with 240 N m torque and 122 kg mass. The vehicle
accomplished a 75 m run in 6 s, and the target is 4.2 s. The discrepancy is due to
wheel slippage and is understandable for such a low mass unit. The energy storage,
however, proved sufficient for two consecutive acceleration runs. Modeling of all
ch007 7 June 2011; 17:59:20

energy storage and driveline components was cited as a big factor in putting
together a competitive race car.
Gao et al. [6] describe in detail an electrical equivalent circuit model for the
lithium-ion battery that is suitable for the type of simulation noted above. The model
proposed includes nonlinear equilibrium potential, rate-dependent capacity, and
temperature influence on parameters. Model parameters are extracted from experi-
mental testing. Baisden and Emadi [7] discuss the benefits of combination technol-
ogies and how this energy storage system performs in a small HEV. The demonstrator
car was equipped with a 64 kW propulsion motor, Hawker Genesis 3 12 V, 26 Ah
lead-acid battery modules and Maxwell PC2500 (2.5 V, 2700 F) cells, 35 in series
(87 V). Fuel economy, acceleration, and gradeability benefits were significant.
Example 7.1: For the ultracapacitor ESS GoKart [5] discussed in this section,
assume that the operator is standard mass, 80 kg, and that the driveline is 47%
efficient overall. Compute the approximate total vehicle mass based on the infor-
mation and performance metrics realized. What is the mass of the vehicle minus
electric motor and driver?
Solution:

2d
d ¼ V i T þ 0:5aT ¼ 75 m
2
; a¼ ¼ 4:17 m=s2 ð7:1Þ
T ¼6 s T2
Assuming linear acceleration for this short distance, then using (7.1) and the
fact that initial velocity is zero, we find that the final velocity as the vehicle passes
through the Formula SAE measured distance is Vf ¼ aT ¼ 4.17(6) ¼ 25 m/s
(56 mph). Acceleration according to (7.1) is a ¼ 0.43 g. The expression for vehicle
kinetic energy can be applied to extract the total mass involved.
W ESS
K e ¼ 0:5M V V 2f ¼ h ¼ 0:82ð340 kJÞ ¼ 278:8 kJ; M V ¼ 892 kg ð7:2Þ
2
Subtracting OFF the stated electric motor-generator mass (122 kg) and one
standard passenger mass (80 kg) results in a mass for the engine, alternator, frame,
ultracapacitor modules, and all other essentials to yield
M bop ¼ M V M MG M pass ¼ 690 kg ð7:3Þ
Example 7.1 introduces another relevant point on the topic of vehicle electrical
distribution systems, namely, that engine driven alternators are necessary to maintain
a regulated and sufficient power distribution system. In the previous example the
ultracapacitor packs rated 325 V are replenished via a 42 V alternator and d.c.–d.c.
boost converter. Which also contribute to additional mass of the vehicle. To close out
this section, consider for the moment the BAS system in the Vue hybrid.
Example 7.2: A BAS reversible alternator such as the Hitachi unit shown in
Figure 7.2 is used to maintain the charge on a 48 V ultracapacitor module, or
ch007 7 June 2011; 17:59:20

multiple modules in parallel. Furthermore, assume that the 2.4 L Ecotec engine
requires 150 N m of crankshaft torque to perform a warm restart. For the BAS
parameters given in this section calculate: (a) the minimum belt ratio between
engine crankshaft and alternator pulley, (b) the approximate input current to the
power inverter and maximum current given a design margin of 160%, and (c) the
approximate torque constant of the reversible alternator.
Solution: This example is typical of the type of design-in applications that ultra-
capacitor modules may encounter in future mild-hybrid applications. In this parti-
cular case, the reversible alternator is stated to develop malt ¼ 60 N m of shaft
torque, so neglecting belt losses.
meng 150
g pr ¼ ¼ ¼ 2:5 : 1 ð7:4Þ
malt 60

5000
Pelec ¼ U d I d U dmin ¼32 V ¼ 5 kW; Id ¼ ¼ 156 A ð7:5Þ
32
malt 60
kt ¼ ¼ ¼ 0:385 N m=A ð7:6Þ
Id 156
a. The belt pulley ratio, gpr ¼ 2.5:1 as a minimum.

b. At its minimum potential during engine cranking [8], the power inverter input
current Id ¼ 156 Adc. At a 160% margin on inverter power semiconductor
ratings, the peak current of this unit must be Idpk ¼ 250 A.
c. The torque constant of the reversible alternator is kt ¼ 0.385 N m/Adc.
Exercises 7.1–7.3 expand on the concepts introduced in Example 7.2 to famil-
iarize the reader with some practical design requirements for BAS applications.
It is a fact that vehicle batteries are not sized based on engine cranking events
but to support emergency situations such as operating the emergency flasher front
and rear exterior lamps for 1-h duration. The exception may be for key starts under
cold climate conditions when engine cranking torque is high and duration can be
1.5 s. Hybridized vehicles depend on strategy starts, what we refer to here as warm
restarts, when the vehicle is performing idle–stop functions. Example 7.3 considers
this case for the Vue Greenline BAS, but it applies with slight modifications to the
PSA/Valeo iStARS system. The iStARS can perform a warm restart in 400 ms,
whereas the more capable BAS does the same on a larger displacement engine in
tstr < 300 ms.
Example 7.3: The 42 V BAS system has an input power, Pelec ¼ 5 kW for
tstr ¼ 300 ms. Given this information and the requirement to perform three con-
secutive restarts without recharging the ultracapacitor pack, then what minimum
size ultracapacitor cell is required? For this application, the ultracapacitor is sized
for end of life (EOL), at which point the capacitance is taken at 75% of initial, or
beginning of life (BOL) value.
ch007 7 June 2011; 17:59:20

Solution: To solve this type of design-in application, we must have a deeper

understanding of the range the ultracapacitor voltage is assumed to be confined to.
For this, Reference 3 is an excellent reference, and Figure 7.3 from this reference
will be used to set bounds on the ultracapacitor voltage swing. In this case, it can be
seen that the 48 V ultracapacitor module operates mainly between 45 and 38 V;
therefore, those limits will be applied to this example.
50 50
45 45
40 40
Fraction of time (%)
Fraction of time (%)

35 35
30 30 45 V = 2.50 V/cell
47 V = 2.61 V/cell
25 25
20 20
15 15
10 10
5 5
0 0
25 30 35 40 45 50 25 30 35 40 45 50
Voltage (V) Voltage (V)
Figure 7.3 Ultracapacitor voltage swing in a 42 V BAS application. (Left) 42 V

NiMH battery pack, (right) 48 V BMOD0165-P048 ultracapacitor
module, UDDS schedule
W elec ¼ 3Pelec tstr ¼ 3ð5000Þð0:3Þ ¼ 4500 J ð7:7Þ
2N s W elec 2ð18Þð4500Þ
C cell ¼ ¼ ¼ 371:8 F ð7:8Þ
U i 0:75 ð452 382 Þ0:75
U 2f 2
The capacitance specified at EOL (7.8) means that the cells must have at least 372 F
at BOL strictly for the purpose of engine warm restart.
7.1 Types of hybrids

The industrial definition of an electric hybrid vehicle is one having two power
plants, an engine and an electric source, both of which are capable of delivering
torque to the driven wheels. This is also why the low-end energy recuperation
systems such as boosting engine-driven alternator output during deceleration and
dual voltage systems that maximize alternator output are not true hybrid electric
vehicles since these are incapable of delivering boosting torque to the driven
wheels. Idle–stop systems, such as iStARS and BAS, are borderline systems and
are commonly called microhybrids.
ch007 7 June 2011; 17:59:20

Figure 7.4 illustrates how hybridization of the vehicle power plant contributes
to reduction in fuel consumption. The primary benefit of hybridization of any type
is the capability to recuperate a portion of already spent energy. Additional benefit
comes from not consuming fuel in the first place and that is where idle–stop
functions contribute. The zones of zero vehicle speed in Figure 7.4 are times when
the engine is turned-OFF to avoid fuel consumption while maintaining all vehicle
ancillaries and accessories from the electric energy storage system.
Drive cycle
Vehicle speed (V)
Electric
system
Engine Engine
Time (t)
Figure 7.4 Illustration of hybrid benefit: boosting on acceleration, recuperation

on deceleration
The overall effect can be summarized as shown in Figure 7.4 where the white
boxes represent energy that can be recuperated and shaded boxes represent energy
that is lost regardless of conditions. The end result is that only a fraction of avail-
able kinetic energy is recoverable due to irreversible losses of rolling friction and
aerodynamic friction.
Kinetic-accel Rolling friction Up-grade Aerodynamic

Energy ¼ þ þ þ ð7:9Þ
Kinetic-decel Rolling friction Down-grade Aerodynamic
Ultracapacitor electric storage can help because the nature of energy exchanges
in the vehicle environment depicted in Figure 7.4 are dynamic with constant cycling
of energy. Hybrid vehicle with battery energy storage demands battery technology
capable of supporting such high cycling at partial SOC and at high burst power.
Today, only NiMH, nickel zinc (NiZn), and lithium-ion are capable of meeting these
demands, and with lead acid chemistry only the absorbent glass mat (AGM) and the
newer PbCapTM by Axion Power [9,10] can fulfill the application. Brody [9] shows
that NiZn batteries at 1.6 V, versus 1.2 V for NiMH, are 25% smaller, 30% lighter
with 25% more power, and 25% lower cost ($/kWh) than NiMH. NiZn can operate
over the same temperature range as lead acid. In Reference 10, Edward Buiel of
Axion Power notes that by 2015 there may be 20 million microhybrids in service that
require highly cyclable energy storage systems. The lead capacitor hybrid is an
asymmetric cell in which the negative plate of a conventional lead-acid battery is
ch007 7 June 2011; 17:59:20

replaced with an ultracapacitor electrode to negate the detrimental sulfation effects

common in lead-acid batteries. The trademarked PbC battery is advertised to have
five times the cycle life of conventional lead acid, 80–90% round trip efficiency, 30%
less weight, and specific energy of 25 Wh/kg. This hybrid energy storage component
is also being advocated for application of the growing microhybrid market. But these
are still electrochemical cells that rely on redox processes that slow down in cold and
exhibit memory, capacity fade, aging, and impedance increase with decreasing SOC.
The carbon–carbon symmetric ultracapacitor is viewed as being superior to these
battery chemistries for short-term (<15 s) high power bursts. A Maxwell Technolo-
gies 3000 F K2 cell, for example, has 700 W of discharge power at >95% one-way
efficiency, an energy of 3 Wh, and a P/E ¼ 233. The battery of course can support its
power for longer durations.
Figure 7.5 graphically depicts the low end of hybrid architectures in which the
engine alternator is used to recover deceleration energy, to provide some torque aug-
mentation to the engine if reversible, and to operate at higher voltages if necessary. The
42 V BAS is one example of dual voltage operation where the vehicle electrical dis-
tribution system (EDS) remains at 14 V but the high power functions are at 42 V. This
is the class of hybrid vehicle that is poised to see upward of 20 million units in service
by 2015. It is also the architecture that requires a robust low-voltage battery such as
AGM, NiZn, PbC or ultracapacitor combination, such as ultrabattery.
At the next level of hybrid complexity, the architecture shown in Figure 7.5
morphs to the class of crankshaft mounted electric machines. For example, ISG
mild hybrid, power assist hybrid, and strong hybrids fall into this class.
Ultra-capacitor
Controlled regene-
Stop & rative braking Wheels
Dry clutch
start
Engine
(gas or diesel) Transmission
Tank
Power train management unit (PTMU)
Electric power
Drive power
Figure 7.5 Low-end hybridization: energy recuperator, idle–stop, BAS,

microhybrid categories
ch007 7 June 2011; 17:59:20

When the dry clutch shown in Figure 7.6 is omitted, the system is a conven-
tional ISG with motor-generator bolted directly to the engine crankshaft either in
place of the flywheel and ring gear or in the case of an automatic, around the torque
converter. The alternator can still be used to recuperate deceleration energy, but the
main recuperation path is via the driveline through the ISG to the ESS. In the case
of the Honda Civic IMA hybrid, the system operates with an NiMH battery at
144 V, but it could also operate with ultracapacitors as Gonder et al. [3] demon-
strated at NREL. The architecture shown in Figure 7.6 has been introduced into the
marketplace by GM in the Silverado pickup truck with 42 V ISG and 36 V AGM
battery packs. This particular vehicle could have been demonstrated using ultra-
capacitors just as NREL did for the Vue hybrid.
Ultra-capacitor
Controlled regene-
Stop & Dry clutch rative braking Wheels
start
Engine
Transmission
(gas or diesel)
Tank Electric motor High-voltage

power electronics NiMH battery
Ultra-capacitor
Power train management unit (PTMU)
Electric power
Drive power
Figure 7.6 Mid-scale and full hybrid architecture, ISG, series-parallel switching,
power split types
Table 7.1 summarizes the basic types of hybrid architectures and their func-
tionality. More will be said of hybrid functionality in Section 7.2. For now, the
distinction between types is one of functionality. According to Table 7.1, a micro-
hybrid can do some boosting and supply power consumers, for example, instant cabin
heating using a positive temperature coefficient (PTC) element in the air circulation
plenum. But it is really the mild hybrid that brings more functionality to boosting in
terms of not just augmenting, but boosting the vehicle during acceleration. Full, or
strong, hybrids have sufficient ESS to support some amount of electric only range,
ch007 7 June 2011; 17:59:21

Table 7.1 Hybrid architecture comparison
Hybrid- Full- Mild- Mini- Micro-

system hybrid hybrid hybrid hybrid
Function
Start-stop
Recuperation
Active boosting
Electric driving
Power consumer
Batteries + Batteries + Batteries + Boostcap

boostcap boostcap boostcap
for examples, 3 or 4 mi in the case of a PriusII. Mini-hybrid as used in Table 7.1

means simply that it is an energy recuperator system for opportunity charging.
7.2 Hybrid functions

Table 7.1 is a good transition into the discussion of hybrid functions. But, Table 7.1
is meant only to contrast the types of hybrid vehicles at a high level, the 30,000 ft
view. Getting closer to reality, we introduce Table 7.2 to contrast the functionality
Table 7.2 Quantification of hybrid vehicle functions
Power train Conventional Conventional Conventional Downsized Downsized Downsized

Electric M/G Belt ISG Belt ISG Belt ISG Crank-ISG Crank-ISG Offset ISG
14 V 42 V 42 V 42 V 150 V >300 V
Ancillaries Conventional Conventional Electric Electric Electric Electric
Battery Flooded VRLA, 30 kg VRLA, 30 kg NiMH, 20 kg NiMH, 40 kg NiMH, 60 kg
pb- acid, 25 kg
Functions:
Idle stop
Regen
Energy Mg’mt
Launch assist
ZEV
%FE benefit 3 7 10 30 35 <40
ch007 7 June 2011; 17:59:21

of the various hybrid functions by being more quantitative about their voltage level,
power level, and fuel economy gains. In this table, the functions listed by column in
Table 7.1 are listed by row in Table 7.2, and shaded to highlight the class of hybrid
the function pertains to.
According to Table 7.2, the move to hybridize impacts the vehicle engine size
and power train with a need to integrate a motor-generator into the power train.
Having done so, the introduction of idle–stop brings with it an attendant need for
electric drive or previously mechanically or hydraulically driven ancillaries. The
automatic transmission oil-pump, for example, that is needed to pressurize the
torque converter must be electric drive if the engine is shut down. This precludes
vehicle roll-back on grades, sustains hill holding, and eliminates backlash of the
driveline. Similar things can be said for the need for electric power assist steering,
electric drive engine water pump, and so on. The following can be said by
inspection of Table 7.2:
● Power train can admit downsized engine when electric torque is available.
● Motor-generator must be integrated into the engine and power train at an
appropriate voltage level and machine type for the speed range it must operate
over.
● Ancillaries remain conventional, or must become electric drive for no loss of
customer features.
● Battery must be an appropriate technology to meet robustness requirements,
cycle life, and sustained operation at partial SOC. Battery mass is shown
relative to percent fuel economy estimates.
● Hybrid functions:
* Idle–stop of the engine, once warmed up, to offset fuel consumption
* Regeneration of available vehicle kinetic energy to recuperate as much as
possible
* Energy management system (EMS) for optimal use of onboard ESS and
motor-generator
* Launch assist is the availability of sufficient electric motor-generator
power to provide significant boosting during acceleration, such as aug-
menting engine torque to 3000 rpm
* Zero emissions vehicle operation, basically a BEV mode, if sufficient ESS
capacity is available
● Percent Fuel economy gain for the class of hybrid is under consideration.
Moving from low-end microhybrid to strong hybrid produces substantial gains
in fuel economy over standard drive cycles.
Finally, for closure on hybrid functions it is useful to put metrics on the perfor-
mance aspects of various hybrid architectures. In the case of power assist hybrids,
the spider chart shown in Figure 7.7 is useful. In this figure, one sees that putting
quantifiable metrics on functions brings the benefits of hybridization into clearer
focus, or at least, clarifies what the expectations are of hybridization.
● Launch assist of a ¼ 0.45 g; refer to Example 7.1 on the GoKart that showed it
had a ¼ 0.43 g acceleration capability.
ch007 7 June 2011; 17:59:22

● Acceleration times, tz60 ¼ 11 s and tz85 ¼ 21 s, are representative of acceptable

performance.
● Maximum grade of 30% is a regulated requirement as is sustained speed of
90 kph on 6% grade.
● Wide-open-throttle (WOT) speed of 180 kph is again a customer expectation.
tz85 tz60
21 s 11 s
Launch
Maximum acceleration
grade 30% ZEV range: 45g
Towing capacity
WOT speed
Gradeability 90 kph 0% grade
for 20 min 6% 180 kph
Figure 7.7 Performance metrics for a power assist hybrid
7.3 Power assist hybrid
In this section, we examine the implications of Figure 7.7 in more detail, specifi-
cally to determine if the specifications noted are consistent, which in fact are not
consistent. The eight metrics shown are typical of the performance targets that are
set for vehicles, but the eight targets do not necessarily apply to one particular
vehicle. For example, a mild hybrid has no ZEV range, but it must meet maximum
grade and sustainable speed on a 6% grade by regulation. It may in fact have lower
tz60 just like the PriusII that can accelerate 0 ? 60 mph in t ¼ 8.9 s. Some basic
physics and an example will highlight the fact that meeting maximum grade and
simultaneously meeting WOT speed in a single speed power train is very difficult
to realize. This is in fact impossible to realize in an engine power plant vehicle but
not in an electric traction drive system if the appropriate electric machine tech-
nology is available.
Example 7.4: Consider a mid-size American sedan hybrid vehicle having a total
vehicle mass, Mv ¼ 1500 kg, and an electric drive system configured as shown in
Figure 7.6 with the electric motor-generator in the driveline. The performance
metrics of Figure 7.7 apply. (A) Compute the minimum driveline gear ratio given
that the vehicle-driven wheels have a dynamic rolling radius, rw ¼ 0.3 m, and assume
it must negotiate the maximum grade of 30%. (B) Compute this same gear ratio if the
vehicle is to meet the maximum acceleration requirement of a ¼ 0.45 g. (C) Finally,
ch007 7 June 2011; 17:59:22

compute the motor-generator maximum angular speed for this single speed driveline
when the vehicle is at its WOT speed, VWOT ¼ 180 kph (111.87 mph, or 50 m/s).
Assumption: The motor-generator is capable of maximum stall torque,
mMG ¼ 150 N m.
Solution: This example is typical of the design considerations given to design-in of
hybrid electric propulsion systems [8]. From physics, we know the force necessary
to hold a mass on an incline (the road grade in this example) is obtained by shifting
the x–y coordinate system to lie with x-axis along the incline. When this is done, the
force necessary to just hold the vehicle mass in place, hence the minimum gear
ratio in the vehicle driveline, is given as

1 30%
a ¼ tan ¼ 0:2914 rad;
100% ð7:10Þ
F t ¼ gM v sinðaÞ ¼ 9:802ð1500Þð0:287Þ ¼ 4219 N
The vehicle acceleration provided by the amount of tractive effort, Ft,
calculated in (7.10) is therefore 2.8 m/s2, or 0.28 g, which is not consistent with
Figure 7.7 metric that a sports vehicle would target.
Ft 4219 N
a¼ ¼ ¼ 2:812 m=s2 ð7:11Þ
M v 1500 kg
The previous derivations provide sufficient information to calculate the mini-
mum gear ratio to hold the maximum grade for this size vehicle.
grw M v sinðaÞ 9:802ð0:3Þð1500Þð0:287Þ
gr ¼ ¼ ¼ 8:448 : 1 ð7:12Þ
mMG 150
For a single speed driveline without gear, shifting the single ratio would mean
that at the high speed of the operating range the motor-generator must be capable of
spinning at an angular speed dictated by the fixed gear ratio.

V WOT 50
wMG ¼ g r ¼ 8:448 ¼1408 rad=s; so that
rw 0:3
ð7:13Þ
wMG 1408
nMG ¼ ¼ ¼13;448 rpm
0:1047 0:1047
For part B, repeat (7.11) to (7.13) for the case of a ¼ 0.45 g to yield
Ft ¼ 6616 N. Then making the substitutions yields the answer.
Ans: (A) gr ¼ 8.448:1 for 30% grade and nMG ¼ 13,448 rpm at VWOT
(B) gr ¼ 13.23:1 for 30% grade and nMG ¼ 21,000 rpm at VWOT
(C) nMG ¼ 13,448 rpm at VWOT for case (A) and nMG ¼ 21,000 rpm at VWOT
for case (B). Designing an electric machine for nearly 14,000 rpm is
possible and in fact this is what Toyota implements in the GS 450 h and
other Lexus hybrids. But for 21,000 rpm the demands are so stringent
that mass producing such an electric machine at nearly 100 kW is pro-
hibitive. See exercise 7.4 for more insight on this point.
ch007 7 June 2011; 17:59:22

Can ultracapacitors help in the case of power assist hybrids? Absolutely, and as
the architecture of Figure 7.6 shows, the fast charge and discharge capability of the
ultracapacitor means that vehicle boosting, such as launch assist and high power
from rapid braking, makes it ideal for the ESS. In 2005, BMW demonstrated their
Efficient Dynamics hybrid 5 that used ultracapacitor for launch boosting and
demonstrated what an additional 70 kW of boost could do when sustained for
nearly 7 s in an already high-performance vehicle. Figure 7.8 is representative of
that vehicle and the ultracapacitor installation. In this Efficient Dynamics concept
vehicle, the 1500 F ultracapacitors are mounted in the doorsill. For approximate
sizing of this HEV application refer to Exercise 7.6.
Figure 7.8 Extreme hybrid ultracapacitor demonstrator by BMW 70 kW, 660 N m

electric machine between engine and transmission as in Figure 7.6
7.4 Plug-in hybrid

The plug-in hybrid electric vehicle (PHEV) is the subject of considerable interest
and it represents a potential transportation mode that can offset imported petroleum
by shifting some of the driving distance of a vehicle to electric only range. Pre-
sently, the PHEV discussed here is the GM Volt shown in Figure 7.9.
ch007 7 June 2011; 17:59:22

Park and plug

General Motors is preparing to make the Chevrolet Volt,
a plug-in electric vehicle that can be recharged at home
Among the car’s systems:

Range extender
Small gasoline engine that
keeps the battery charged on long trips Power pack
Large battery than in
conventional hybrids
Electric motor
Dual-charge ports Plug-in ports
used to charge the battery
Utilities could use ports to borrow
power from car during peak demand
Figure 7.9 Volt PHEV concept (Detroit Auto Show Jan 2007) and phantom view
of power train
The most recent parameters available on the Volt are 16 kWh lithium polymer
battery pack shown in cutaway as the ‘T’ shaped installation in Figure 7.9 (bottom),
a 120 kW electric traction motor, and a 53 kW range extender internal combustion
engine, 1.1 L. During 2010, the all electric range numbers for the Volt were revised
to 3 mi/kWh (approximately 340 Wh/mi) and there were plans to introduce a
mountain mode in the vehicle of grades and hills. The transmission of the Volt is
believed to be a single mode eCVT (electronic continuously variable transmission),
to be discussed in Chapter 8. Some hints at the type of front wheel drive (FWD)
electronic variable transmission that could be used in the Volt were provided by
Mark Selogie in Reference 11, and reinforced by U.S. patent [12] that describes the
transmission in considerable detail.
So, how can ultracapacitors help in the case of a PHEV with very large battery
pack. The Volt pack, for example, has peak power of 136 kW, and a P/E ¼ 136 kW/
16 kWh ¼ 8.5. With a usable energy of 8 kWh, of which GM elects to use 63% SOC
and 340 Wh/mi, this Volt pack yields an AER of 29.6 mi (approximately 30 mi).
The fuel economy using a metric of 33.4 kWh/gge and 0.34 kWh/mi is FE ¼ 33.4/
0.34 ¼ 98.2 mpg where, the metric gallon gasoline equivalent (gge) is defined as
the electric energy equivalent of a U.S. gallon of commercial gasoline. Table 7.3
summarizes the attributes of the GM Chevy Volt PHEV, and Figure 7.10 shows the
simulation configuration studied for the case of battery plus ultracapacitor combi-
nation for this vehicle.
ch007 7 June 2011; 17:59:22

Table 7.3 Specifications of Volt PHEV
Vehicle Chevy Volt PHEV, 40 mi AER

specifications
Mass kg 1588 Air density kg/m3 1.2
Drag coefficient # 0.29 Gravity m/s2 9.81
Roll res kg/kg 0.0075 Pack volts V 335
Fit area m2 2.293 Pack energy kWh 16
Wh radius m 0.36 Battery Ppk kW 136
Roll res, Rolling resistance; Fit area, Frontal Area
V, P (V)
IMC ~ M/G
=
EMC
Id
Ud
ESS
New York city cycle driving schedule

Length = 598 s, distance = 1.18 mi, avg. speed = 7.1 mph EPA urban dynamometer driving schedule
90 Length = 1369 s, distance = 7.45 mi, avg. speed = 19.59 mph
80 90
Vehicle speed (mph)
Vehicle speed (mph)
80
70
70
60 60
50 50
40 40
30 30
20 20
10 10
0 0
0 10 20 30 40 50 60 70 80 90 10 11 12 13 14
0 50 100 150 200 250 300 350 400 450 500 550 600 0 0 0 0 0 0 0 0 0 00 00 00 00 00
Test time (s) Test time (s)
US06 or supplemental FTP driving schedule Japan 10–15 emission & fuel economy driving schedule
Length = 596 s, distance = 8.01 mi, avg. speed = 48.37 mph Length = 892 s, distance = 3.94 mi, avg. speed = 15.97 mph
90 90
80 80
Vehicle speed (mph)
Vehicle speed (mph)
70 70
60 60
50 50
40 40
30 30
20 20
10 10
0 0
0 50 100 150 200 250 300 350 400 450 500 550 600 650 0 100 200 300 400 500 600 700 800 900 1000
Test time (s) Test time (s)
Figure 7.10 PHEV simulation architecture for evaluation of

battery þ ultracapacitor (top) and drive cycles (bottom)
When the Chevy Volt parameters listed in Table 7.3 are applied to the simu-
lation program corresponding to Figure 7.10 (top), the drive schedule results for
New York City Cycle (NYCC), UDDS, and US06 for propulsion power required
are summarized in Figure 7.11. This figure also summarizes specific energy
requirements of this vehicle over the respective drive cycle. The summary shows
maximum vehicle speed by cycle, average speed, distance traveled, the average
ch007 7 June 2011; 17:59:23

propulsion power by cycle, and how much of the spent energy is recoverable
by cycle.
EV propulsion power NYCC cycle
40,000
30,000
20,000
Power (W)
10,000
0
0 100 200 300 400 500 600 700
− 10,000
− 20,000
− 30,000
Time (s)
(a)
EV propulsion power UDDS cycle
40,000
30,000
20,000
Power (W)
10,000
0
0 200 400 600 800 10000 1200 1400 1600
− 10,000
− 20,000
− 30,000
Time (s)
(b)
EV propulsion power US06 cycle
100,000
90,000
80,000
70,000
60,000
50,000
40,000
Power (W)
30,000
20,000
10,000
0
− 10,000 0 50 100 150 200 250 300 350 400 450 500 550 600 650
− 20,000
− 30,000
− 40,000
− 50,000
− 60,000
− 70,000
Time (s)
(c)
Drive cycles and volt PHEV results

Parameter Units NYCC UDDS US06
Vmx mph 27.2 56.7 80.3
Va mph 7.09 19.6 48.0
Dist miles 1.18 7.44 7.99
Pavg kW 0.81 2.1 9.88
Regen # 0.6 0.45 0.3
Energy/mi Wh/mi 282.6 193.6 293.6
(d)
Figure 7.11 Simulation results for Volt PHEV propulsion power: (a) propulsion
power on NYCC cycle; (b) propulsion power on UDDS cycle;
(c) propulsion power over US06 cycle; (d) summary of Volt
performance by cycle
ch007 7 June 2011; 17:59:23

Table 7.4 Summary of ultracapacitor þ battery combination in PHEV over UDDS cycle, Ubat ¼ 335 V
ch007
Units Base lithium only A 36 3000 F B 54 3000 F C 80 2000 F D 80 1500 F E 100 650 F
d.c.–d.c. Convertor %Uc swing – 80 85 90 94 100
Conv. Apk – 315 316 218 170 81
Ibb_rms – 26.3 38.6 45.2 37.6 25.8
Ultracapacitor Iuc_rms – 122.6 124.5 102.4 84.5 45.7
Lithium ion Ib_rms 42.3 18.9 10.7 17.7 22.6 27.2
Ib_avg 29.7 11.7 5.5 9.7 11.8 16.4
7 June 2011; 17:59:23

%Ib_rms/Ibase 100 45 25 42 53 64
Conv. Apk, Converter, Apk

Hybrid electric vehicles
231
The results shown in Figure 7.11 are surprising in two respects: First, the low-
speed NYCC with a maximum speed of only 27 mph consumes 282.6 Wh/mi,
nearly the same specific energy consumption per mile as the aggressive US06 cycle
with 80 mph top speed and 293.6 Wh/mi specific energy consumption. Second,
the slow NYCC cycle high regeneration fraction is due to high kinetic energy
recovery because the rolling and aerodynamic losses are low, whereas for the high
speed US06 cycle the kinetic recovery is low (0.3 fraction, (7.9)), the non-
recoverable losses are high from aerodynamic losses, but the decelerations are so
brisk that power levels are high (Figure 7.11c). The UDDS, or more average drive
cycle, shows low specific energy consumption but a respective regeneration frac-
tion of 0.45.
To evaluate how much an ultracapacitor pack can prolong the PHEV lithium-
ion battery pack life by reducing its operating temperature through reduction of
battery rms current is evidenced by the five cases summarized in Table 7.4 for an
active parallel combination. In this table, the ultracapacitor pack energy levels are:
● Case A: 36S 1P 3000 F, 108 Wh, Umx ¼ 97.2 V, 3.45 < boost < 6.89
● Case B: 54S 1P 3000 F, 162 Wh, Umx ¼ 145.8 V, 2.3 < boost < 4.6
● Case C: 80S 1P 2000 F, 160 Wh, Umx ¼ 216 V, 1.55 < boost < 3.1
● Case D: 80S 1P 1500 F, 120 Wh, Umx ¼ 216 V, 1.55 < boost < 3.1
● Case E: 100S 1P 650 F, 65 Wh, Umx ¼ 270 V, 1.24 < boost < 2.48
The best case for the active parallel combination of ultracapacitor and lithium-ion
battery in the PHEV is Case B, with 162 Wh of ultracapacitor energy and a d.c.–d.c.
converter boost gain range of 2.3–4.6. The boost ratio is just about centered on 3:1,
a very good location. The resulting reduction in battery rms current reflects this by
drooping to 25% of its battery acting alone case. Straddling this best case are Cases
B and D having 108 and 160 Wh of ultracapacitor energy. The lowest performing
case is E because of its low energy of only 65 Wh total. Case E, however, does the
best in fully utilizing the available SOC of the ultracapacitor by taking advantage of
100% of its voltage window.
The overall conclusion from this is that 60–160 Wh of total ultracapacitor
energy is sufficient for a PHEV with large lithium-ion pack to substantially reduce
battery stress and therefore improve its service life, hence warranty.
Exercises
7.1 A BAS reversible alternator provides Pm ¼ 4 kW mechanical power into the
engine belt during warm restart. Given that the reversible is still operating
within its constant torque (CT) region, at what speed can the engine be spun-
up to with this amount of power?
Pm 4000
Ans: neng ¼ ¼ ¼ 636:7 rpm:
0:1047malt 0:1047 60
ch007 7 June 2011; 17:59:24

7.2 For the belt : pulley ratio found in Example 7.2, and taking the WOT engine
speed as nWOT ¼ 6000 rpm, what is the sustainable speed requirement of the
reversible alternator? What are the implications for Autobahn driving?
Ans: nalt ¼ gprnWOT ¼ 2.5(6000) ¼ 15,000 rpm. The implications are that the
alternator rotor must be designed to sustain this high angular speed in a hot
environment, and the controller must be designed to provide adequate reg-
ulation in order to not overvoltage the ultracapacitor module.
7.3 To expand on Exercise 7.2 more fully, consider that the reversible
alternator has a voltage constant, ke ¼ kt (refer to Example 7.2), then what is
the alternator d.c. output voltage under a fault condition during Autobahn
driving?
Ans: U flt ¼ k e nWOT ¼ 0:385 V s=rad 6000 rpm 0:1047 rad=s=rpm
¼ 241:86 V:
This illustrates why the power electronic controller and field regulator for the
alternator must be robust to protect the system against uncontrolled genera-
tion under fault conditions.
7.4 The power assist HEV of Example 7.4 has sufficient ESS and a motor-gen-
erator of PMG ¼ 60 kW, then for the gear ratio found in part A what is the
base speed to which this MG can hold constant torque?
Ans: The MG base speed at which it transitions from CT to CP is
nMGb ¼ PMG/(0.1047 mMG) ¼ 60,000/(0.1047 150) ¼ 3820 rpm.
7.5 Considering the same vehicle of Exercise 7.4, use the base MG angular speed
calculated in Exercise 7.4 and the results of Example 7.4A to compute the
constant power speed range (CPSR) of the MG in this power assist hybrid.
Ans: CPSR ¼ nWOT/nMGb ¼ 13,448/3820 ¼ 3.5:1.
7.6 The BMW Efficient Dynamics demonstrator vehicle of ultracapacitor-boosted
acceleration consists of a crankshaft-mounted ISG rated 70 kW peak and
660 N m. Assume the ultracapacitor pack is capable of sustaining this power
for 6.5 s. (A) Given this information calculate the number of 1500 F cells
required, and (B) their configuration if the motor-generator input voltage is to
remain <550 Vdc.
Ans: (A) The required deliverable ultracapacitor pack energy, dWuc ¼
PpkT ¼ 70 kW(6.5 s) ¼ 455 kJ. A 1500 F cell as a deliverable
energy of dWcell ¼ 1.14 Wh, so Ns ¼ 400 cells.
(B) The configuration is 200S 2P 1500 F for this installation for
the specified voltage criteria.
ch007 7 June 2011; 17:59:24

References
1. W. Diem, ‘PSA Peugeot Citroen’s stop-start diesel future’, Automotive
World, 30 July 2010. Available at: www.AutomotiveWorld.com
2. A. Pesaran, J. Gonder, ‘Recent analysis of UCAP’s in mild hybrids’, The 6th
Advanced Automotive Battery Conference, AABC 2006, Baltimore, MD,
17–19 May 2006
3. J. Gonder, A. Pesaran, J. Lustbader, H. Tataria, ‘Hybrid vehicle comparison
testing using ultracapacitor vs. battery energy storage’, SAE 2010 Hybrid
Vehicle Technologies Symposium, Double Tree Hotel, San Diego, CA, 10–11
February 2010
4. B.-H. Lee, D.-H. Shin, B.-W. Kim, H.-J. Kim, B.-K. Lee, C.-Y. Won, et al.,
‘A study on hybrid energy storage system for 42V automotive PowerNet’,
IEEE 1st Vehicle Power and Propulsion Conference, VPPC2006, Haley
Conference Center, Windsor, UK, 6–8 September 2006
5. K.W. Benson, D.A. Fraser, S.L. Hatridge, C.A. Monaco, R.J. Ring, C.R.
Sullivan, et al., ‘The hybridization of a formula race car’, IEEE 1st Vehicle
Power and Propulsion Conference, VPPC2006, Haley Conference Center,
Windsor, UK, 6–8 September 2006
6. L. Gao, S. Liu, R.A. Dougal, ‘Dynamic lithium-ion battery model for system
simulation’, IEEE Transactions on Components and Packaging Technolo-
gies, vol. 25, no. 3, pp. 495–505, 2002
7. A.C. Baisden, A. Emadi, ‘ADVISOR-based model of a battery and an ultra-
capacitor energy source for hybrid electric vehicle’, IEEE Transactions on
Vehicular Technology, vol. 53, no. 1, pp. 199–205, 2004
8. J.M. Miller, Propulsion Systems for Hybrid Vehicles, 2nd edn., The Institution
of Engineering Technology (IET), Michael Faraday House, Stevenage, Herts,
United Kingdom, 2010
9. R. Brody, ‘Nickel zinc technology’, Powergenix Presentation to Advanced
Energy Storage, Catamaran Hotel, San Diego, CA, 12–14 October 2010
10. E. Buiel, ‘Battery requirements for micro-hybrid vehicles’, Axion Power
Presentation to Advanced Energy Storage, Catamaran Hotel, San Diego, CA,
12–14 October 2010
11. M. Selogie, ‘The GM 2Mode FWD hybrid system’, Presentation to the SAE
Hybrid Vehicle Technology Symposium, DoubleTree Hotel, San Diego, CA,
11–12 February 2009
12. B. Conlon, P.J. Savagian, A.G. Holmes, M.O. Harpster Jr., Output Split
Electronic Variable Transmission with Electric Propulsion using One or Two
Motors, U.S. Patent Publication 2009/008171 A1, 26 March 2009
ch007 7 June 2011; 17:59:25

Chapter 8
Single-mode power split
To discuss the application of ultracapacitors in strong hybrid vehicle systems that

are designed around power-split transmissions, it is important to first understand
the fundamentals of these components. The interested reader is referred to Refer-
ence 1 on electronic continuously variable transmissions (eCVTs), and to Refer-
ence 2 on electric machines and power electronic subsystems that constitute an
eCVT. The most basic of the power-split eCVTs is the single-mode system, such as
used in Toyota’s hybrid synergy drive (HSD), the basic propulsion architecture of
the Prius hybrid series and Lexus brands. A variant of the basic single mode is
adopted in the Ford Escape, Mariner, Fusion, and other hybrid vehicle products.
A second type of eCVT is the 2-mode power split that was introduced by GM, then
developed incollaboration between GM–Chrysler–Mercedes at their Troy, Hybrid
Development Center, Troy, Michigan, as a common component for application
across vehicle segments in each partner company. The overall business goal of
the company partnership is to realize benefits of scale in the manufacture of a
complex and very capable automatic transmission. Chapter 9 is devoted to the two-
motor electronically variable transmission (EVT), developed by GM for Alison
hybrid bus, but later it was used for Tahoe/Yukon and other large sport utility
vehicles (SUVs).
Figure 8.1 is a taxonomy of eCVTs showing that such infinitely variable ratio
power-transmitting devices can be realized in mechanical, hydraulic, and electric
domains. Details of all these types can be found in Reference 2, but our interest
here is the input-split variety of eCVT known as the single mode. This figure also
highlights the fact that single-mode systems are offered by Toyota and Ford Motor
companies.
The Toyota hybrid system (THS) is in fact the two-motor, single-mode system
under discussion, and the Ford hybrid system (FHS) is another type of single-mode
system. The core of the eCVT is the epicyclic, or more commonly planetary gear
set shown graphically in Figure 8.2. In this planetary gear set, three members are
free to move: sun, carrier with attached planet pinion gears, and ring (or internal)
gear. The planetary gear is essentially a mechanical summer, so the angular speed
relationship among the members is fixed by their radii, or number of teeth (8.1).
The base ratio, k, is defined as the ratio of ring gear radius to sun gear radius (same
as the ratio of number of teeth). Figure 8.2 is a full view of the planetary gear
design showing gear radii and angular speed of the elements followed by an
ch008 7 June 2011; 14:3:31

Continuously variable
transmissions CVT
Electric
Mechanical Hydraulic
eCVT
Push Toroidal Input Compound

Hydramatic Hyrdromechanical
belt roller coupled coupled
THS-I, II AHS-2, EVT
FHS eVT (Timken)
IVT (Renault)
Figure 8.1 Types of continuously variable transmissions
wR
R
UC wR
R
wC C wC
rC
rp wC
rx C
wR
rR wx wS S wS
S
Sun wC
Planet
C
R
Ring (inside and/or outside gear)
Figure 8.2 Planetary gear set showing carrier with four pinions, stick diagram
and functional representation
edge-on view of the gears in stick diagram form, where S ¼ sun, or central gear,
C ¼ carrier with attached planet gears, and R ¼ ring gear with internal teeth.
Angular speed relations of each planetary member are noted in the stick diagram.
For ease of representation, the ‘transistor-like’ block diagram is shown that will be
used in this book where the epicyclic gear is present.
rR #Ring teeth
k¼ ¼ ð8:1Þ
rS #Sun teeth
wS þ kwR ¼ ðk þ 1ÞwC ð8:2Þ
ch008 7 June 2011; 14:3:31

Single-mode power split 237
1
mS ¼ mR ; mR ¼ kmS ð8:3Þ
k
ðk þ 1Þ k
mC ¼ mR ; mR ¼ mC ð8:4Þ
k ðk þ 1Þ
mS þ mC ¼ mR ; mC ¼ ðmS þ mR Þ ð8:5Þ
To obtain the power flow in the planetary gear set, multiply each term in the
planetary gear angular speed relation (8.2) by the carrier torque, mC, to obtain
ðk þ 1ÞwC mC ¼ kwR mC þ wS mC ð8:6Þ
Then substitute from the relations (8.3)–(8.5) to obtain the expression for
power at each of the three ports of a planetary gear set, which should be obvious by
conservation of power in (þ) and power out ().
wC mC ¼ wR mR þ wS mS ð8:7Þ
Therefore, the epicyclic or planetary gear set is a power-splitting device that
partitions power flowing into one of its three ports to the adjacent two ports. For
example, in (8.7) a power into the carrier port splits into the ring and sun gear ports
according to their relative speeds as determined by (8.2). An example will help
clarify this situation.
Example 8.1: Using the illustration of Figure 8.2 for the planetary gear set, set up
the case for power flows in the planetary gear for an engine key start in a hybrid
electric vehicle that is parked. Determine the torque, angular speed, and power of
each element for this case.
Solution: Figure 8.3 defines the variables and planetary gear element conditions
present during an engine key start when the planetary gear carrier member is
connected directly to the engine crankshaft. It is given that the cold engine cranking
mR, PR
R wR ≡ 0
mC wC
PC
S mS, PS
wS
E1
R
Figure 8.3 Elementary planetary gear variables defined (fundamental ratio k)
ch008 7 June 2011; 14:3:31

torque needed is mC ¼ 300 N m and its speed must reach wC ¼ 30 rad/s. The con-
straint is that the ring gear, which is directly geared to the vehicle wheels, has an
angular speed condition of wR ¼ 0.
From (8.2) with wR ¼ 0, we have the definition of sun gear (electric motor)
input in terms of the carrier member (engine) angular speeds when the planetary
gear fundamental ratio is given as k ¼ 2.6.
wS þ k 0 ¼ ðk þ 1ÞwC ; wS ¼ 3:6wC ¼ 108 rad=s ð8:8Þ

The respective torque conditions, given that mC ¼ 300 N m, for engine crank-
ing are
1 300
mS ¼ mC ¼ ¼ 83:3 N m ð8:9Þ
ðk þ 1Þ 3:6
Similarly, the element power levels are
PR ¼ 0; PS ¼ mS wS ¼ 83:3ð108Þ ¼ 9 kW; PC ¼ mC wC ¼ 300ð30Þ ¼ 9 kW ð8:10Þ
The resisting torque offered by the planetary ring gear, necessary to transfer
cranking torque from the sun gear electric motor, is found to be
k 2:6
mR ¼ mC ¼ 300 ¼ 216:7 N m ð8:11Þ
ðk þ 1Þ 3:6
It is clear from this example that the sum of sun and ring tear torque levels
equals the torque on the carrier and that power flows from the electric motor on the
sun gear to the engine crankshaft on the carrier gear. The condition of zero angular
speed at the ring gear forces the power split to be all in favor of the carrier during a
key start condition.
Energy storage in the case of Example 8.1 must deliver a mechanical power
level of PS ¼ 9 kW to the sun gear of the planetary gear set. If the electric motor-
generator attached to the sun gear has an efficiency of 93% at this speed (108 rad/s,
or 1031 rpm) and the power electronic inverter has an efficiency of 97%, then the
ESS must deliver an electrical discharge power of
PS
Pe ¼ ¼ 9:98 kW ð8:12Þ
hMG hINV
The planetary gear operation in an eCVT is more complex than Example 8.1
would indicate. References 3 and 4 provide deeper insights into the eCVT that
some readers may find useful, although outside the focus of this book. The intent
here is to show that ultracapacitors have application in eCVTs because of the need
for high burst power levels and high power recuperation. Verbrugge et al. [5] ela-
borate on the application of ultracapacitors in range-extended electric vehicles
having an eCVT power train architecture. The next section explains the eCVT in
sufficient detail for the interested reader to understand this type of electronically
aided transmission.
ch008 7 June 2011; 14:3:32

8.1 Electronic continuously variable transmission

The electronic continuously variable transmission (eCVT) is necessary to under-
stand at a fundamental level to gain perspective on why the electric power trans-
mission path responds to vehicle and energy storage system demands in the manner
it does. This eCVT system is essentially an infinitely variable mechanical trans-
mission in which the variability is obtained through an electric variator. A variator
is a means to split input mechanical power into a mechanical power transmission
path and an electric power transmission path where the speed variability is
accommodated by the electric transmission path. The transmission ratio, or variator
gain, K, is defined as the ratio of input angular speed to output angular speed as
wi
K¼ ð8:13Þ
wo
Figure 8.4 puts the power-split transmission into perspective so that the true
influence of the back-to-back electric drives becomes apparent. In this figure, the
internal combustion engine (ICE) is connected to the planetary gear E1 carrier, and
this port is defined as the input (wi, mi). The ring gear connects directly to the
vehicle driveline and is taken as the output port of the variator (wo, mo), which is
also the mechanical shaft of the motor-generator (MG2) having output mechanical
power, P2 (w2, m2), and electrical-to-mechanical conversion efficiency, h2. The
variator circulation port, having mechanical power, P1, or sun gear of the planetary
gear E1, is given as (wS, mS) ¼ (w1, m1). From (8.13) the angular speed of MG2 is
1
w2 ¼ wi ð8:14Þ
K
m2 mR P0
P2
MG2 R m0
R2 w2
wR w0
wC mi
ICE C
mC wi
mS P1
S MG1
wS w1 m1 R1
R2
Pe
Pess = 0
ESS Ultracapacitor + battery
Figure 8.4 Core elements of the single-mode eCVT
ch008 7 June 2011; 14:3:32

Apply (8.2) to planetary gear E1, and simplify using (8.14) to obtain the input
angular speed of MG1 connected to the sun gear.

k
w1 ¼ k þ 1 wi ð8:15Þ
K
From (8.4), the MG1 input shaft torque is defined as
1
m1 ¼ mi ð8:16Þ
1þk
Taking the product of (8.15) with (8.16) defines the MG1 input mechanical
power that is available for energy conversion to output electrical power on the
interconnecting d.c. link of the eCVT.

k mi wi
P1 ¼ m1 w1 ¼ k þ 1 ð8:17Þ
K ð1 þ kÞ
The power-circulating function of the d.c. link is shown by the pair of equa-
tions in (8.18) for the link power and its conversion to shaft mechanical power of
MG2. In this analysis the motor-generator efficiencies are taken as constants, but in
general this is not the case, with motor-generator efficiency a function of speed–
torque operating point.
P2
Pe ¼ h1 P1 ¼ ; P2 ¼ ðh1 h2 ÞP1 ð8:18Þ
h2
Substituting from (8.17) into (8.18) yields the expression for MG2 mechanical
power in terms of the engine input power to E1.

k Pi
P2 ¼ ðh1 h2 ÞP1 ¼ ðh1 h2 Þ k þ 1 ð8:19Þ
K ð1 þ kÞ
Once the mechanical shaft power of MG2 is derived, it becomes easy to
compute the shaft torque of MG2 as

p2 k mi Kw2
m2 ¼ ¼ ðh1 h2 Þ k þ 1 ð8:20Þ
w2 K ð1 þ kÞw2
which simplifies to

k
m2 ¼ ðh1 h2 Þ K mi ð8:21Þ
1þk
It may not be obvious from analysis of (8.14)–(8.21) that the portion of ICE
input power that is split into electrical power is then recombined with the
mechanical fraction of engine mechanical power that flows to the planetary gear
ring. The ring gear is the mechanical power summation point where both ICE
power and MG2 power are combined.
ch008 7 June 2011; 14:3:32

To reinforce the developments in this section, consider Example 8.2 for the
eCVT in an HEV during acceleration from a low speed to a high speed, then pay
attention to the electric path power flows at the beginning and at the end.
Example 8.2: In this case, an HEV equipped with an eCVT is traveling at low
speed, Vi ¼ 18 mph (8 m/s), and the ICE is running at 2100 rpm (220 rad/s). The
vehicle then accelerates to Vf ¼ 55 mph (24.6 m/s). Compute P1, P2, and P0 for
initial and final speeds of the vehicle.
Solution: Let the planetary gear base ratio k ¼ 2.6, and both MG1 and MG2 effi-
ciency values be 0.93 ( ¼ inverter efficiency machine efficiency). Given a final
drive ratio, gfd ¼ 3.11, and driven wheel radius, rw ¼ 0.3 m, let the vehicle tractive
effort, Ft ¼ 2000 N for a realistic depiction of this HEV. On the basis of this
information, the eCVT output torque must match
F t rw 2000ð0:3Þ
m0 ¼ ¼ ¼ 193 N m ð8:22Þ
g fd 3:11
The output angular speed, w0, for initial and final vehicle speeds and variator
ratio are computed as

Vi 8
w0i ¼ g fd ¼ 3:11 ¼ 83 rad=s; K i ¼ 2:65 ð8:23Þ
rw 0:3

Vf 24:6
w0f ¼ g fd ¼ 3:11 ¼ 254:9 rad=s; K f ¼ 0:863 ð8:24Þ
rw 0:3
For these initial- and final-output angular speeds, the output mechanical power
for both end points is calculated as
P0i ¼ m0i w0i ¼ 193ð83Þ ¼ 16 kW ð8:25Þ
P0f ¼ m0f w0f ¼ 193ð254:9Þ ¼ 49kW ð8:26Þ
The MG1 input mechanical power split-off from the engine input power is cal-
culated by application of (8.17) to this case for both initial and final variator ratios.

k pii k 1
P1i ¼ kþ1 ¼ 1 Pii ¼ 0:7275Pii
K ð1 þ kÞ 1 þ k K K¼2:65
ð8:27Þ

k pif k 1
P1f ¼ kþ1 ¼ 1 Pif ¼ 0:163Pif
K ð1 þ kÞ 1 þ k K K¼0:863
ð8:28Þ
ch008 7 June 2011; 14:3:32

The MG2 output mechanical power to be combined with the engine input
power is calculated by the application of (8.19) to this case for both initial and final
variator ratios.
P2i ¼ h1 h2 P1i ¼ 0:932 ð0:7275ÞPii ¼ 0:63Pii ð8:29Þ
P2f ¼ h1 h2 P1f ¼ 0:932 ð0:163ÞPif ¼ 0:141Pif ð8:30Þ
Therefore, a fraction of engine input power (Pi P1) flows mechanically from
carrier to ring of E1, and the remaining fraction P1 flows electrically from MG1 to
MG2 to be combined with (Pi P1) as mechanical power P2 as follows:
P0i ¼ Pii P1i þ P2i ¼ ð1 ð0:7275 0:63ÞÞPii ¼ 0:902Pii ð8:31Þ
P0f ¼ Pif P1f þ P2f ¼ ð1 ð0:163 0:141ÞÞPif ¼ 0:978Pif ð8:32Þ
From (8.25) and (8.26) it is easy to verify that the initial power loss through the
eCVT variator is 1.57 kW at an engine power of 17.74 kW decreasing to 1.08 kW at
an engine power of 50.1 kW.
The previous example confirms that power splitting and losses in the eCVT are
strongly dependent on the variator, that is, on the ratio of engine operating speed to
vehicle speed reflected to driveline angular speed. No dynamics are considered in
the analysis, and should these be taken into account the benefit of having ESS
power available to absorb or deliver kinetic energy to rotational member inertias,
especially the electric machine rotors, would be apparent.
8.2 Ultracapacitor application in eCVTs
At present there are no applications of ultracapacitors in the eCVT propulsion

system architecture of hybrid electric vehicles. However, this application is open to
the introduction of ultracapacitors because of the highly dynamic nature of circu-
lating power in the eCVT, the need for ESS power to augment vehicle acceleration,
and the need for absorption of high power during vehicle braking. The benefits of
ultracapacitors as highly cyclable energy storage elements are clear but are yet to
be implemented because of cost reasons. There is also the question of integration
cost and package space for the d.c.–d.c. converter needed to buffer the ultra-
capacitor in a combination ESS. Schupbach and Balda [6,7] discuss design of d.c.–
d.c. converters for power management in hybrid vehicles using combination of
ultracapacitor and battery. Wang and Fahimi [8] discuss a high-efficiency d.c.–d.c.
converter for fuel cell vehicle applications, and Basu and Undeland [9] describe in
depth the need to constrain voltage and current ripple exposure of the ultracapacitor
in d.c.–d.c. converter buffering applications to improve their lifetime.
Figure 8.5 illustrates the eCVT architecture complete with active combination
ESS for high-performance power and energy supply. The system component
ch008 7 June 2011; 14:3:32

Wheels
Epicyclic
Inverter
274 V
MG2 R
ICE C FD
Battery
MG1 S
Ultra-
capacitor
Power split eCVT

Unbuffered ESS
Figure 8.5 The eCVT with active combination energy storage system
efficiencies discussed in Example 8.2 can be visualized more clearly from this
figure by noting that the electric machines will have efficiency values of 92–95%
and power electronic inverters at the system voltages used in HEVs will range from
96% to 98%, yielding the combined a.c. drive upper efficiency value of 93%.
The d.c.–d.c. converter shown schematically in Figure 8.5 must have an effi-
ciency of > 97% for such architectures to be viable in a vehicle application.
To visualize the single-mode operation during vehicle cruise, the sequence of
events from engine OFF vehicle electric launch to steady state cruise. In addition,
the torque and speed trajectories of the generator, MG1, and traction motor, MG2,
are illustrated. Note that the engine torque–speed trajectory starts at engine OFF,
then follows a ‘hockey stick’ path along its most efficient operating points as
shown in Figure 8.6. Generator MG1 starts at a high speed, and traction motor MG2
starts at zero speed and high torque for electric-only launch.
During BEV mode launch, the engine remains OFF and the vehicle enters its
initial acceleration phase on strictly ESS electric power during which time MG2
provides high drive torque to the wheels. This torque then falls to a low steady state
value reflecting power circulation mode, while the engine operates in its most
efficient mode during steady state cruise.
The engine map in Figure 8.6 is typical of any ICE:fuel efficiency contours
showing lowest brake specific fuel consumption (BSFC) at onion-peel chart top
plateau in g/km, plus the constant power hyperbola’s crossing these contours, and
finally the envelope of engine torque versus speed. General practice is to operate
the ICE below maximum torque at any speed to ensure reserve capacity for dri-
veability – meaning responsiveness. The electric machine torque–speed contours
are typical of wide constant power–speed ratio (CPSR) machines such as internal
permanent magnet (IPM) types [2]. The wide CPSRs of the IPMs studied in Figure 8.6
are apparent by noting the transition from CT to CP at approximately 1 krpm and
the onset of second breakdown at approximately 5 krpm, or CPSR ¼ 5:1. It is a
ch008 7 June 2011; 14:3:32

Atkinson engine map

Torque reserve
for driveability
(me–we) constrained
mapping points
115
Torque (N m)
0
20
210
90
220
0
230
55 kW
25
240
0
27
65 35 kW
35 20 kW
0
29
15 kW
10 kW
0 8 kW
0 1000 2000 3000 4000 5000
Engine speed (rpm)
Generator map Motor map
80 300
Electric subsystem
Torque (N m)
Torque (N m)
73 240 73
60 torque reserve for
200 82 power boosting
82
40 88 120 88
20 93 80 93
0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
Speed (krpm) Speed (kpm)
Figure 8.6 Engine and MG1 and MG2 operating maps leading to vehicle steady
cruise mode
requirement of eCVTs that the electric machines employed have CPSR value 5:1
or better CPSR and why the IPM is among the only choice of electric machine
to realize this. The IPM is also a high-torque electric machine, another
requirement of implementation in the eCVT.
BEV mode in the eCVT of Figure 8.6 is evident by following the power flow
from the ESS through the main traction inverter and through MG2 to the final drive
and wheels. In BEV mode the engine is OFF and MG1 is idling, even though
spinning at high speed because of the gear ratio, ring to sun of the planetary. In this
mode the angular speed of MG1 is wS ¼ kwR, which means it will be spinning
k ¼ 2.6 times faster than MG2 rotor speed. Example 8.3 helps explain the relative
speed ratings of the electric machines used in the eCVT and the engine operating
strategy at high speed, such as downhill cruise.
Example 8.3: The HEV with eCVT is operated at high speed on a downgrade.
(a) Given the angular speed limits of MG1 as w1 < 680 rad/s, w2 < 586 rad/s, what
is the wide open throttle speed, VWOT, of the vehicle such that MG2’s rotor is not in
ch008 7 June 2011; 14:3:33

overspeed? (b) At what speed, V2, on a downgrade must the HEV engine be started
so that MG1 is not in overspeed? (c) For faster speed on the downgrade, at what
vehicle speed, V2, must the engine be started and fixed at such that neither MG1 nor
MG2 rotors are in overspeed?
Solution: This is a real-world issue with single-mode eCVTs because the motor-
generators are on fixed gears from the driveline and vehicle speed translates
directly to their rotational speed requirements. For this example, the following
procedure applies:
(a) Compute VWOT based on the angular speed limit of MG2
rw w2 lim 0:3ð586Þ
V WOT ¼ ¼ ¼ 56 m=sð126mphÞ ð8:33Þ
g fd 3:11
However, with the engine off, (8.33) results in the rotor speed of MG1
increasing along with vehicle speed so that at VWOT its speed will be
wS ¼ kw2 lim ¼ 2:6ð586Þ ¼ 1524 rad=s w1 lim ð8:34Þ
(b) Therefore, the engine must be engaged at some speed intermediate to cut in,
V1 * 18 mph, the limit of BEV mode, and VWOT. This speed is determined by
application of (8.2) for the case wC ¼ 0 rad/s and is the maximum angular
speed of MG1 for engine OFF mode.
wS lim 680
wR2 ¼ ¼ ¼ 261 rad=s ð8:35Þ
k 2:6
rw wR2 0:3ð261Þ
V2 ¼ ¼ ¼ 25:18 m=s ð56:3 mphÞ ð8:36Þ
gfd 3:11
To avoid overspeed of MG1, the engine must start when the vehicle speed,
V ¼ V2, is approximately 56 mph on the downgrade. Last, we must verify that the
MG1 and MG2 are not in overspeed at faster speeds.
(c) The speed at which the engine ON must be fixed is computed using the fol-
lowing procedure, noting that wS lim ¼ w1, wR lim ¼ w2:
wS lim þ kwR lim ¼ ð1 þ kÞwset

C
ð8:37Þ
680 þ 2:6ð586Þ ¼ 3:6 wset
C ; C ¼ 234 rad=s ð2235 rpmÞ
wset
The engine must therefore be started and run at 2235 rpm to avoid over-
speeding of the electric motor-generators.
ch008 7 June 2011; 14:3:33

8.3 Drive cycle evaluations

From the preceding sections, the operation and limitations of the single-mode
eCVT for HEVs should be apparent. Figure 8.7 contains photos of single-mode
eCVT components that are useful to put this hardware into perspective. Referring
to Figure 8.7a, the generator MG1 is on the left and at the mechanical connection to
the engine. The installation is front wheel drive (FWD) with the planetary gear set
approximately centered in the eCVT, with MG2 on the right and a chain drive to
the final drive gears (i.e., the differential). Some of the final drive fixed gears are
visible in the cutaway of Figure 8.7a.
The power inverter is shown in Figure 8.7b, with d.c.-link filter capacitors
visible on the top and bus bars to the power connections and d.c.–d.c. converter also
visible. The two large plastic connectors (orange color in production systems to
signify high voltage) are the battery cable input. Output to the electric machines is
via the connector to the lower right in the picture.
The hardware illustration of a single-mode eCVT in Figure 8.7 has a NiMH
battery with prismatic cells of 1.2-V potential, 6.0-Ah capacity, and Nc ¼ 168 in
(a) (b)
(c) (d )
Figure 8.7 Hardware elements of the eCVT [photos courtesy of JNJ Miller PLC]
(a) MG1 and MG2 integrated into eCVT; (b) power electronic center,
inverters, and d.c.–d.c. converter; (c) main battery, NiMH prismatic
cells; (d) fully integrated system on L4 engine
ch008 7 June 2011; 14:3:33

series for 201.6-V pack for this particular installation. The cells are welded ter-
minations, air cooled, and fully sealed in the hermetic package.
Figure 8.7d gives a better view of the full system integration with four-cylinder
engine on the left that directly connects to the eCVT via a concentric shaft, with the
centermost shaft passing from engine crankshaft through the center of MG1 to the
planetary gear carrier. The sun gear connects to MG1 rotor and the ring gear
directly to MG2 rotor, as well as having provision to loop over the drive chain to
the final drive gearing. Sitting directly above the transmission is the power center
consisting of a pair of power inverters, one each for MG1 and MG2, and a d.c.–d.c.
converter to interface the high-voltage battery to the a.c. drive system electronics.
This function is explained in more detail in Figure 8.8, where the d.c.–d.c. con-
verter input inductor is shown at lower right with heavy bus bar interconnects.
(a)
500 V (2.5× boost)
Half bridge
boost converter
Boost inductor
250 A, 10 kHz
1500 µF S/A M/G
NiMH 600 V
6 Ah 10 kW 50 kW
201.7 V
233 Apk
1400 W/kg
(b)
Figure 8.8 eCVT d.c.–d.c. converter interface to ESS and power inverters.
(a) High-power d.c.–d.c. converter to buffer 201-V NiMH battery to
500-V electronics and (b) schematic of the ESS with d.c.–d.c. converter
ch008 7 June 2011; 14:3:34

To illustrate the performance of the eCVT and its ESS over a typical drive
cycle, the procedure outlined in Reference 10 is followed.
The power flow tally of Figure 8.9b shows that engine power, Pe, splits,
a portion, Pg, converted to electrical power by the generator and circulated by
MG1 to MG2, where it is augmented with battery power, Pb, to be recombined
with the mechanical fraction of engine power fed directly from carrier to ring
gear, Pe – Pg.
RX400h SUV hybrid
Pe + Pb
Pe – Pg FD
400 kW V8 R Pg + Pb
Pe
ICE C1 C2
MG1 S1 S2 MG2
109 kW generator Pg Dual Pg + Pb 123 kW motor
planetary
Pb
45 kW battery
Battery & converter
(a)
P Power split eCVT

Power flows
Pe + Pb
Total engine Pb Battery

Traction
Total drive power
Power
Pm motor
Pe + Pg Generator Pg
Pe
Engine
direct to
wheels
(b)
Figure 8.9 Camry-size HEV with single-mode eCVT. (a) Architecture based on
RX400h (Toyota Motor) and (b) power flow survey
ch008 7 June 2011; 14:3:34

For the vehicle to be simulated VWOT ¼ 112-mph N.A. and 146-mph Auto-
bahn. Table 8.1 contains the specifics of the vehicle and eCVT to be used.
Table 8.1 Camry-size HEV evaluated using the RX400h eCVT drive system
architecture (Figure 8.9a)
Attribute Symbol Unit Attribute Symbol Unit

Vehicle mass (kg) Mv 1412 Wheel dynamic rw 0.332
roll radius (m)
Frontal area (m2) Af 2.3 Final drive gear ratio gfd 4.05
Aero drag coefficient Cd 0.27 Planetary gear E1 ratio k1 2.6
Rolling resistance Crr 0.008 Planetary gear E2 ratio k2 2.478
coefficient
To perform a simulation, the engine-operating strategy must be defined as in

Reference 10 by selecting an ‘N’ by ‘V’ approach, that is, specifying the engine
speed, N, in relation to vehicle speed, V. When this is done and the hybrid vehicle is
simulated over a drive cycle, its performance can be analyzed. For our purposes
here, the Urban Dynamometer Drive Cycle (UDDS) and the aggressive US06 Drive
Cycle are used. As a first goal the occurrence of engine torque for the N/V strategy
employed is determined for the UDDS cycle, and this is illustrated in Figure 8.10.
In this simulation, one must focus on the engine strategy, N/V, showing the engine
torque–speed map trajectory to follow on demand in accordance with the road load
imposed. When this is done, the plot showing N/V is the engine ON occurrences
relative to vehicle speed, and the relative magnitude of the engine speed. It can be
seen that the N/V is very busy in the vicinity of V ¼ 12 m/s (26.8 mph).
UDDS drive cycle speeds UDDS N/V strategy

60 300
Engine speed (rad/s)
50
Speed (mph)
40 200
30
100
20
10 0
0 –5 0 5 10 15 20 25 30
–10 200 400 600 800 1000 1200 1400 –100
Time (s) Vehicle speed (m/s)
Engine map UDDS Engine torque occurrence UDDS

250 1200
Engine torque (N m)
Frequency M380
200 1000
800
150
600
100 400
50 200
0 0
0 50 100 150 200 250 0 55 100 130 150 170 190 200
Engine speed (rad/s) Torque (N m)
Figure 8.10 HEV strategy response to UDDS drive cycle
What is very interesting from Figure 8.10 is that the histogram of engine
operating torque clusters in the band of 100–130 N m, with a significant portion of
ch008 7 June 2011; 14:3:34

time at engine OFF. Next, the same vehicle is run on the aggressive US06 cycle,
and its response is shown as Figure 8.11.
US06 mph (m/s) N/V chart
90
80
Speed (mph, m/s)
70 500
Engine speed, m/s

60 400
50 300
40
30 200
20 100
10
0 0
–10 0 100 200 300 400 500 600 700 –5–100 0 5 10 15 20 25 30 35 40
Time (s) Vehicle speed (m/s)
Engine map US06 Engine torque occurence

250 500
Engine torque (N m)
450
200
Frequency/601
400
350
150 300
250
100 200
150
50 100
50
0 0
0 50 100 150 200 250 300 350 400 450 0 60 100 130 160 196 206 211
Engine speed (rad/s) Torque (N m)
Figure 8.11 HEV strategy response to US06 drive cycle

For the US06 drive cycle, the same engine N/V strategy is used, the vehicle is the
same but the engine ON response to N/V is now clustering about 30 m/s (67 mph)
because the US06 is a higher speed cycle and it has more events at high speeds.
There are however remnants of the lower speed N/V responses in the vicinity of
V ¼ 12 m/s (26.8 mph) as was the case for the UDDS. However, the US06 has a very
pronounced engine response at 67 mph. This bimodal response is apparent in a his-
togram plot of engine torque showing clustering not only about a torque of
100–130 N m as with UDDS but also about the higher torque of 196–208 N m.
This ‘busyness’ of engine ON–OFF is an area where the combination ESS
architecture can benefit because engine strategy today is to cycle the engine as much
as possible to offset undue cycling of the battery, thereby enhancing its life expec-
tancy and thus warranty. The application of ultracapacitor as power buffer can reduce
this engine ‘chattering’ by holding the engine OFF somewhat longer and still avoid
exposing the battery to excessive stress. This is all a topic of ongoing research.
One very important use of dynamic simulation of the HEV is to appropriately
size the motor-generators and to tune the planetary gear fundamental ratio. Charts
given in Figure 8.12 show the response of MG1 and MG2 for the drive cycles
discussed above with respect to their envelope torque–speed characteristics.
Figure 8.12 shows that MG1, the generator, operates predominantly in the first
quadrant as a generator to circulate power during delivery of power from engine to the
driven wheels. MG2, the main traction motor, however, is different: it operates pri-
marily in the second and third quadrants for forward motoring and forward generat-
ing, respectively. It is also interesting to see in Figure 8.12 the clustering of operating
torque–speed points, both motoring and generating, in the vicinity of 1000 rad/s.
It was not mentioned earlier, but these electric machines are designed for high-speed
operation, the generator for 11,000 rpm and traction motor for 14,500 rpm, as Toyota
Motor Co. did for the GS450h and other Lexus brand hybrids.
ch008 7 June 2011; 14:3:35

MG1 capability for kp2 = 2.478 (P = 30 kW)

80
60
40
Torque (N m)
20
–2000 –1500 –1000 –500 0 500 1000 1500 2000

–20
–40
–60
–80
Speed (rad/s)
MG2 capablity for kp2 = 2.478 (P = 61 kW)

200
1500
100
0
Torque (N m)
50
0
–2000 –1500 –1000 –500 0
–50
–100
–150
–200
Speed (rad/s)
Figure 8.12 MG1 and MG2 torque–speed with design values k1 and k2
(Table 8.1)
To illustrate the benefits of simulation on driveline tuning, the following chan-

ges were made to the architecture of Figure 8.9a compound (or dual) planetary. The
engine side input planetary of the common ring gear design was retained at k1 ¼ 2.6,
while the secondary side planetary, with grounded carrier, was modified from
k2 ¼ 2.478 – 2.1 in the first case and k2 ¼ 2.85 in the second case to determine the
benefits of having this parameter set below and above its design value, respectively.
Figure 8.13 shows the results in the torque–speed plane of MG1 and MG2.
When the k2 ratio is reduced from its design value as shown in the left-side plot
of Figure 8.13, the torque–speed scatter of MG2 increases to over 200 N m in
torque from approximately 160 N m and reduces in speed from over 1000 rad/s to
approximately 900 rad/s. This shows that with this ratio change, an induction
machine having approximately 2.7:1 CPSR would be capable of meeting the design
requirement on torque-speed, even if it could not meet the efficiency target of an
IPM. When operated to VWOT, the CPSR must increase to 4.2:1, well beyond what
an induction machine could deliver.
Lastly, when the k2 ratio is increased from design value as shown in the right-
hand side of Figure 8.13, the torque decreases to well below 150 N m and angular
ch008 7 June 2011; 14:3:35

MG2 capability for kp2 = 2.1 (P = 64.8 kW)

300
200
Torque (N m)
100
0
–2000 –1500 –1000 –500 0
–100
–200
–300
Speed (rad/s)
CPSR @72 mph = 2.7:1
CPSR @115 mph = 4.2:1
MG2 capability for kp2 = 2.85 (P = 64.8 kW)

200
1500
100
Torque (N m)
50
0
–2500 –2000 –1500 –1000 –500 –50 0
–100
–150
–200
Speed (rad/s)
CPSR @72 mph = 2.8:1
CPSR @115 mph = 4.3:1
Figure 8.13 MG1 and MG2 torque–speed with modified values k1 and k2
speed increases to well above 1200 rad/s, consistent with an increase in this ratio,
wS ¼ k2wR. These evaluations demonstrate the clear advantages of system simu-
lation of complex electromechanical architectures.
To summarize this chapter, the chart of Figure 8.9b is repeated here for sum-
mary comments. Referring to Figure 8.14, the survey of power flows in an eCVT,
the following can be said:
● A portion of engine power in the eCVT is transmitted directly to the vehicle-
driven wheels mechanically at minimal transmission loss.
● The remaining, and variable, fraction of engine power is converted to electrical
power and first transmitted electrically to the fraction motor-generator and
there re-converted to mechanical power to propel the vehicle-driven wheels.
ch008 7 June 2011; 14:3:35

● The electrical transmission path of the eCVT can be augmented with electrical
energy stored onboard in the energy storage system. This onboard energy
reservoir can both deliver and absorb high levels of transient power.
● Therefore, the total power at the vehicle-driven wheels is the sum of peak
engine power, assuming zero losses in the mechanical and electrical trans-
mission paths, and peak ESS power.
P Power split eCVT power flows
Pe + Pb
Total engine Pb Battery

Traction
Total drive power

Power
Pm motor
P e – Pg Generator Pg
Pe
Engine
direct to
wheels
Figure 8.14 Survey of eCVT power flows for review
The Prius hybrid introduction, in fact, was a good example of this power aug-
mentation by onboard electric energy storage. The 2004 Prius has a 1.5-L, 57-kW
gasoline engine and 21-kW peak power NiMH battery. The total driveline power is
therefore 78 kW of peak power. It is interesting to note that this hybrid is designed
with a 29-kW generator, MG1, and 50-kW traction motor, MG2. The engine can
therefore deliver a maximum of 29 kW electrically to MG2 that can be augmented
by 21 kW from the battery for a total MG2 power of 50 kW, the design value – all
consistent with the power flow tally of Figure 8.14.
Exercises
The exercises in this chapter refer to the single-mode eCVT shown schematically in
Figure 8.15, where MG1 connects to the planetary sun gear and MG2, the main
traction motor, to the ring gear with engine connected to the planetary carrier. The
fundamental ratio, ring to sun, of the planetary gear is k ¼ 2.6, the final drive (FD)
ratio is gfd ¼ 3.11, the wheel dynamic rolling radius is rw ¼ 0.3 m, and the ESS d.c.
link potential is Ud ¼ 274 V.
8.1 For the case stated in Example 8.1, assume that the engine must spin up to
wC ¼ 30 rad/s in T ¼ 300 ms. Compute the discharge energy at the ESS
terminals for this condition.
Ans: WESS ¼ PeT ¼ 9980(0.3) ¼ 2994 J (*0.83 Wh).
ch008 7 June 2011; 14:3:35

Wheels
Epicyclic
Inverter
274 V
MG2 R
ICE C FD
Battery
MG1 S
Ultra-
capacitor
Power split eCVT

Unbuffered ESS
Figure 8.15 Single-mode eCVT with combination ESS

8.2 Using (8.27) calculate the value of variator ratio, K, that results in the cir-
culation power equal to 75% of the engine input power of the single-mode
eCVT given that k ¼ 2.6.
Ans: Rearrange (8.27) to get

P1 k 1
¼ 1 ¼ 0:75
Pi 1þkK
\ K ¼ 2:89
8.3 Using the driveline values provided in Example 8.2 for a single-mode eCVT-
equipped HEV, (a) compute the driveline angular speed, w0, corresponding to
this value of K, and (b) compute the vehicle speed, V, for this same case.
Ans: (a) w0 ¼ wKi ¼ 2:89
220
¼ 76:125 rad=s
(b) V ¼ rwgw0 ¼ 0:3ð76:125Þ
3:11 ¼ 7:34 m=s ð16:4 mphÞ.
fd
8.4 To limit the level of electric circulation power in the eCVT-equipped HEV,
the vehicle is operated as a battery electric vehicle (BEV) during vehicle
launch. In this mode the energy storage system (ESS) supplies traction power
via the MG2 inverter and electric machine with engine OFF and MG1 idle.
Compute the maximum grade the vehicle of Example 8.2 can hold if the
value of MG2 torque is 300 N m and the vehicle mass, Mv ¼ 1100 kg.
Hint: Compute the longitudinal force, Fl, necessary to hold the vehicle in
place on the grade and equate this to the vehicle wheel tractive effort, Ft.

Ans: F l ¼ gM v sin a ¼ F t ¼ fdrw 2 ; so a ¼ sin1 gMfd v r2w ¼ sin1 0:29
g m g m
Converting this slope angle in radians to % grade results in the final answer.
gr ¼ 100 tanðaÞ ¼ 100ð0:298Þ ¼ 29:8%
ch008 7 June 2011; 14:3:35

8.5 For the tractive effort computed in Exercise 8.4 and the BEV vehicle speed
calculated in Exercise 8.3, compute the ESS power that must be supplied to
MG2 to sustain this speed on grade.
Ans: From Exercise 8.4
F i ¼ gM v sin a ¼ 9:802ð1100Þsinð0:294Þ ¼ 3124:5 N and
V ¼ 7:34 m=s; so PðV Þ ¼ F l V =h2 ¼ 3124:5ð7:34Þ=0:93 ¼ 24:7 kW
8.6 The ESS battery component shown in Figure 8.5 is a NiMH module rated
Ud ¼ 274 V, Cb ¼ 6.5 Ah, and SOC0 ¼ 0.6. If the available SOC window is
dSOC ¼ 0.2, then how far can the vehicle drive at the grade found in Exercise
8.5 at the steady state power level determined in Exercise 8.5? Call this range
the all electric range at grade, AERgr.
Ans:
dW ESS dSOCU d C b 0:2ð274Þð6:5Þ
AERgr ¼V T ¼V ¼V ¼7:34 ¼381mð0:24kmÞ
pðV Þ pðV Þ 24;700
Comment: The ESS battery at partial SOC and the vehicle in electric mode at
maximum grade result in a very limited AER of only a quarter kilometer. On
level grade, no headwind, this same amount of deliverable ESS energy would
be sufficient for approximately 1 mi of AER at 356 Wh/mi.
8.7 Show that although different in design, the Lexus RX400h hybrid with 3.3-L
V6 engine has power flows consistent with Figure 8.14 when the ratings of its
major components are as follows:
● Engine power, Pe ¼ 155 kW
● Battery peak power, Pb ¼ 36 kW
● MG2 peak rating, PMG2 ¼ 123 kW
● MG1 peak rating, PMG1 ¼ 109 kW
Ans: Pe þ Pb ¼ 155 kW þ 36 kW ¼ 191 kW peak driveline power, and if all of
MG1 power is split from the engine and routed to MG2, we have that
Pe PMG1 ¼ 46 kW flows mechanically to the driveline to be augmented by
PMG2 ¼ PMG1 ¼ 109 kW and Pb ¼ 36 kW to get 46 þ 109 þ 36 ¼ 191 kW total
driveline power.
References
1. J.M. Miller, ‘Hybrid electric vehicle propulsion system architectures of the
eCVT type’, IEEE Transactions on Power Electronics, vol. 21, no. 3,
pp. 756–67, 2006
of Engineering and Technology (IET), Michael Faraday House, Stevenage,
ch008 7 June 2011; 14:3:35

3. B. Conlon, P.J. Savagian, A.G. Holmes, M.O. Harpster Jr., Output Split
Motors, U.S. patent publication 2009/008171 A1, 26 March 2009
4. K. Ahn, S. Cho, W. Lim, Y. Park, J.M. Lee, ‘Performance analysis and
parametric design of the dual-mode planetary gear hybrid powertrain’,
Proceedings of the Institute of Mechanical Engineering, Journal of Auto-
mobile Engineering, vol. 220, Part D, pp. 1601–14, 2006
5. M. Verbrugge, P. Liu, S. Soukiazian, R. Ying, ‘Electrochemical energy
storage systems and range-extended electric vehicles’, The 15th International
Battery Seminar and Exhibit, Ft. Lauderdale, FL, March 2008
6. R.M. Schupbach, J.C. Balda, ‘Comparing dc–dc converters for power man-
agement in hybrid electric vehicles’, IEEE International Electric Machines
and Drives Conference, IEMDC2003, vol. 3, pp. 1369–74, 1–4 June 2003
7. R.M. Schupbach, J.C. Balda, ‘35kW ultracapacitor unit for power manage-
ment of hybrid electric vehicles: bi-directional dc–dc converter design’, The
35th Annual IEEE Power Electronics Specialists Conference, Aachen,
Germany, pp. 2157–63, June 2004
8. S. Wang, B. Fahimi, ‘High efficiency and compact dc–dc converter for high
power fuel cell system’, IEEE Power Electronics Society Newsletter, vol. 19,
no.3, pp. 14–19, June 2007
9. S. Basu, T.M. Undeland, ‘Voltage and current ripple considerations for
improving lifetime of supercapacitors used for energy buffer applications at
converter inputs’, The 13th European Power Electronics and Applications
Conference, EPE2009, 8–10 September 2009
10. J.M. Miller, ‘Overview of hybrid vehicle drive train system designs and
manufacturing constraints’, Presentation to the Advanced Power Electronics
and Electrical Machines (APEEM) Motor and Magnet Workshop, Ames
Laboratory, U.S. Department of Energy, Iowa State University, Ames, IA,
4–5 April 2006
ch008 7 June 2011; 14:3:36

Chapter 9
2-mode power split
Possibly one of the most important innovations in electrically variable transmis-

sions (EVTs) has been the introduction of the electric 2-mode by General Motors–
Allison for application in transit buses [1]. This innovation went on to become the
hybrid technology vehicle of choice in the General Motors–DaimlerChrysler–
BMW Hybrid Development Center collaboration during the era 2004–2008. A major
advantage of the 2-mode is the uniform rating of the two electric machines needed to
implement the electric variator function shown functionally in Figure 9.1 [2]. The
basic architectures are familiar by now to those who read over Chapter 8.
Engine Engine
Power split device Power split device
Electric motor- Electric motor-

generator generator
Electric energy Electric energy

storage system storage system
Electric motor- Electric motor-

generator generator
Final drive to vehicle wheels Power split device
Final drive to vehicle wheels
Input split Compound split

(a) (b)
Figure 9.1 Electrically variable transmission (EVT) single-mode versus

2-mode architecture. (a) Single-mode configuration and (b) 2-mode
configuration
Ch009 8 June 2011; 16:6:9

The single-mode power split eCVT is recognized by the power splitting device
located at the engine input port and is therefore characterized as an input split
architecture. Similar to the feature of all single-mode eCVTs, the system has a
single mechanical point, that is when power is transmitted mechanically. A 2-mode
eCVT in comparison has two, or more, mechanical points at which the electric path
power transmission is zero. The 2-mode is recognized by the presence of a power
splitting device at both input and output ports. It is possible to have a pair of power
splitting devices, or planetary gear sets at one port, but that configuration is not
germane to this discussion. For our purposes, it is important to understand the
operating characteristics of the 2-mode and how ultracapacitor electric energy
storage can benefit such systems.
It has been pointed out in Chapter 8 on the eCVT that in a single mode the
engine input power can split as high as 75% into an electric transmission path
depending on the variator ratio, K. This is not the case in the 2-mode as both
electric machines can be equally rated. Referring again to Figure 9.1, the function
of the electric variator path is to absorb the output-to-input speed variations so that
the input power plant can operate at speeds independent of the output speed,
without need of step-ratio gear shifting. For example, Figure 9.2 illustrates a con-
ventional 6-speed automatic transmission manufactured by Allison for use in sport
utility vehicles (SUVs) and buses.
In the 6-speed automatic transmission depicted in Figure 9.2, left to right, we
have torque converter consisting of impeller, recuperator, and turbine, then input
Figure 9.2 Conventional automatic transmission, Allison 6-speed
Ch009 8 June 2011; 16:6:9

2-mode power split 259
shaft of the transmission showing four clutch packs and three planetary gear sets,
and the output shaft with parking gear and pawl assembly. Hydraulic controls and
circuitry are contained in the base of the unit. The vehicle benefit of more gear
steps in a transmission is throughput efficiency, a 6-speed unit typically being 10%
more efficient than a 4-speed unit. This makes a dramatic difference in overall fuel
economy and is the reason why some manufacturers are moving to 7-speed and
higher automatic transmissions. In reality, these smaller step ratio-shifting trans-
missions begin to approximate a continuously variable transmission (CVT).
Automotive manufacturers require that next generation technologies have the
same form and fit as the technology being replaced, and this is evident in the case of
vehicle transmissions as one can see by comparing the conventional automatic trans-
mission of Figure 9.2 with that of the single- and 2-mode eCVTs shown in Figure 9.3.
In Figure 9.3a, the Toyota Motor Lexus 600 h eCVT is shown with its 165 kW MG2
and 125 kW MG1, which together with a 5-L V8 engine output a total of 327 kW to the
driveline and the GM 2-mode eCVT used in SUVs and high line vehicles.
The point to note about these electronically controlled transmissions is the
relative size of MG1 and MG2. In the single mode, the electric machines are always
Figure 9.3 Comparison of single-mode eCVT, Lexus LS600 h on top and 2-mode
by GM (bottom)
Ch009 8 June 2011; 16:6:9

different size and in the 2-mode they are identical or nearly identical in ratings. The
cutaway of the GM 2-mode shows both electric machines MG1 and MG2 as being
the same size [3]. This is a distinct advantage of the 2-mode system.
For the eCVTs to make sense in a vehicle application they must perform as well,
or better than the conventional technology, which in this case is the 6-speed automatic.
Table 9.1 summarizes the specifics of the 6-speed that a single or two mode must
compete with in addition to approximately the same form and fit and mass of
approximately 130 kg. In Table 9.1, maximum engine speed is taken as 5000 rpm.
Table 9.1 Characteristics of the Allison 6-speed and max vehicle speed by gear
when gfd ¼ 4.11
Gear (gx) Ratio (K ¼ wi/w0) V @wimax (mph) wi without shift rps (rpm)
1 3.10 29 167 (1595)
2 1.81 50 287 (2741)
3 1.41 65 373 (3562)
4 1.00 91 522 (4986)
5 0.71 128 735 (7020)
6 0.61 150 861 (8223)
REV 4.49 –
The maximum vehicle speed in Table 9.1 is calculated based on a vehicle with
tire rolling radius, rw ¼ 0.32 m, and the gearing specified. For this set of conditions,
the eCVT type transmission without step ratio gear shifting would require an
electric traction motor that would be exposed to the angular speed values listed in
the right-hand column of Table 9.1. The ratio of maximum vehicle speed reflected
to the transmission input angular speed, wi, to minimum vehicle speed gives the
gear shift ratio coverage, gsrc, of the conventional transmission and the value to
which the eCVT must respond to for the same performance. In this case, gsrc ¼ 861/
167 ¼ 5.16:1 and this must be equated to the CPSR of the MG2 in the single-mode
transmission, again demonstrating that IPM type electric machines are necessary.
The fourth column values in Table 9.1 are calculated using (9.2).
rw wimax
Vx ¼ ð9:1Þ
g x g fd
Vx
w ¼ gfd ð9:2Þ
rw
9.1 Essentials of 2-mode eCVT
The 2-mode system discussed in earlier paragraphs is now in series production in

passenger vehicles and of course for several years in Allison transit buses.
Figure 9.4a illustrates the 2-mode eCVT integrated to a V8 engine along with its
power converter package mounted above the air intake manifold. Note the same
Ch009 8 June 2011; 16:6:9

GM 2-mode hybrid electric powertrain
(a)
Engine
Battery control unit
Planetary gearset A
Planetary gearset B
engaged friction
Two selectively
clutches
Internal Electric Electric

Clutch
combustion motor motor

engine A B
Output shaft Input shaft Output shaft
Dual-mode electronically
variable transmission
(b)
Figure 9.4 GM 2-mode powertrain assembly (top) and functional diagram

(bottom). (a) Fully integrated 2-mode and (b) functional diagram of
the 2-mode (MGA ¼ MG1, MGB ¼ MG2)
relative sizes of the motor-generators; in this picture, the MG1 stator is cutaway
showing the rotor and portions of the input planetary gear set. Interconnect cables
for inverter inputs to each motor-generator are shown in the lower right at the
junction box feed through assembly.
Ch009 8 June 2011; 16:6:10

The functional diagram, redrawn from Figure 9.1b, shows that in addition to
planetary gear sets there are two or more clutches required in a 2-mode to compress the
wide operating speed range encountered in vehicle service. This has been discussed in
earlier paragraphs as gear shift ratio coverage, and in the 2-mode this requirement is
relaxed because gsrc > 5:1 is not necessary since the electric machine speed ranges are
reusable. Figure 9.4 also shows the motor-generator power electronics (inverters) as
part of the engine control unit with ESS input from a battery. The eCVT of 2-mode
design, also referred to as an EVT, will nominally operate at 330 Vd.c. from the ESS.
Figure 9.5 is the full function diagram of the 2-mode EVT showing the con-
centric design of electric machines, MG1 and MG2, along with clutch packs CL1,
CL2, and CL3 for mode changing. The two planetary gear sets, shown as epicyclic
E1 and E2, are the input and output power splitting devices, respectively. An
engine engagement clutch, CL1, is necessary for BEV mode when MG2 is used as
the main traction motor. In this mode, as with reverse gear, vehicle motion is BEV
mode with propulsion power coming from the ESS.
PM/G
R2
R1 CL3
MG1 MG2
C1 C2
CL2
ICE CL1
wo gfd
we wi S1 wvi S2 wo
FD
C1 CL2
C2
R1 CL3
R2
E1
E2
Mode 1 Mode 2
E1 Differential Differential
E2 Torque multiplies Differential Neutral
CL1 1 1
CL2 0 1 0
CL3 1 0 0
Figure 9.5 EVT functional diagram, two planetary and three clutch designs with
mode selection table
The following characteristics of this EVT can be seen in Figure 9.5:

● Engine connects to E1 ring gear.
● Both E1 and E2 carriers are connected together and to the final drive
(differential).
Ch009 8 June 2011; 16:6:10

● Both MG1 and MG2 are the same rating in torque and speed, and always
connect to sun gears.
● Clutch packs CL2 and CL3 toggle, that is, they are mutually exclusive.
● Planetary set E2 is either a power splitter or torque multiplier, depending on mode.
When CL1 ¼ CL3 ¼ 1 and CL2 ¼ 0, the EVT is in the input split, or single-
mode operational phase. In this mode the engine drives the ring of E1 rather than
carrier as in a single-mode eCVT, but the generator, MG1, still connects to E1 sun
gear. Since CL3 is engaged, the output planetary set, E2, is a torque multiplier with
grounded ring gear enabling the torque produced by MG2 to be multiplied for high
traction effort low-speed operation [4].
When CL1 ¼ CL2 ¼ 1 and CL3 ¼ 0, the EVT is in its output split, or com-
pound mode of operation. In this mode both E1 and E2 are power splitters, and both
the engine and motor-generators contribute to vehicle traction effort. This is
nominally the high-speed mode of an EVT. The following section describes the
operation of both modes in depth leading to an introduction into how ultracapacitor
augmented energy storage can benefit the EVT system.
9.2 Operational modes of the EVT
The two operational modes of the EVT introduced in Section 9.1 are now examined
in detail to reveal the speed and torque regimes of the two electric machines, MG1
and MG2. Figure 9.6 shows that this mode leaves MG1 connected only to E1’s sun
gear while CL3 grounds the ring gear of E2. In this analysis, we let the fundamental
ratio of E1 be k1, and for E2 it is k2. Nominally, both k1 and k2 may have the same
numerical value, but subscripts will be retained for identification. Note that the
analysis to follow the depiction of planetary gear sets is ‘upside down’ from pre-
vious usage. This is done merely for ease of representation and does not affect the
analysis outcome.
SOC
Cb Battery
EVT in mode 1 Uoc
CL1 = 1, CL2 = 0, CL3 = 1 V
Pb
Wheels
wg’ Pe wg’
mg’ mg’ rw
wa’
Jg Jg gf2m ma’
gz2f
gi2g S1 MG1 w i’ MG2 S2 Ja
mi’
Ji gm2i
we’ C1 FD
C2
me’ Intermediate shaft wfd’
Je ge2g CL2 gr2m mfd’
ICE R1 R2 Jfd
Counter shaft wr2’
CL1 wr1’ CL3
mr2’
mr1’
Jr1 Jr2
E2 – torque
E1 – differential
multiplier
Figure 9.6 EVT in mode-1, low-speed input split mode
Ch009 8 June 2011; 16:6:10

The analysis of the EVT in input split mode is very similar to the procedure
followed in Chapter 8 for the single-mode eCVT with the exception that the engine
inputs to the ring of E1 rather than to the sun gear of E1. Referring to Figure 9.6,
the following definitions and derivations for angular speeds and torques are
determined where subscript ‘1’ applies to MG1 and ‘2’ applies to MG2:
w1 ¼ k 1 wi þ ð1 þ k 1 Þw0 ð9:3Þ
wi
w2 ¼ ð1 þ k 2 Þw2 ; and K¼ ð9:4Þ
w0
Computing the MG1 input shaft power, P1 ¼ m1w1, in terms of input torque,
mi, and angular speed, wi, using the relations given in (9.3) and (9.4) results in

1 1
P1 ¼ þ 1 Pi ð9:5Þ
K k1K
Power circulating on the d.c. link, Pe, is generated by MG1 only when the ESS
contribution, Pb ¼ 0, and the respective a.c. drive system efficiencies are h1 and h2
so that the resultant shaft power on MG2 is

1 1
P2 ¼ h1 h2 þ 1 Pi ð9:6Þ
K k1K
The torque on MG1 and MG2 mechanical shafts is

1 1 1
m1 ¼ mi ; and m2 ¼ h1 h2 K 1 mi ð9:7Þ
k1 k1 1 þ k2
Let the fraction of input mechanical power, Pi, which flows mechanically
through the 2-mode be Pm, then the following relations lead to the complete
expression for output power:
P i ¼ P m þ P1 ; and P0 ¼ Pm þ P2 ¼ Pm þ h1 h2 P1 ð9:8Þ
Substituting for Pi in (9.8) and P1 from (9.5) back into (9.8) results in

1 1
P0 ¼ 1 ð1 h1 h2 Þ þ 1 Pi ð9:9Þ
K k1K
According to (9.9) the output power of the 2-mode while in input split mode is
both a function of variator ratio, K, and motor-generator efficiencies. If the a.c.
drive system efficiency is 100%, then (9.9) predicts that the power throughput will
be 100% as expected.
Ch009 8 June 2011; 16:6:11

Example 9.1: Calculate the first mechanical point of the 2-mode EVT. This is the
point at which the power flow through the electric path is zero and all input power
(i.e., engine power) flows mechanically to the output.
Solution: To compute the value of K at the mechanical point the expression in (9.5)
is set equal to zero. This results in a value of K corresponding to the first
mechanical point of
k 1 þ 1 2:6 þ 1
K¼ ¼ ¼ 1:385 ð9:10Þ
k1 2:6
At this value of K the EVT input angular speed relative to its output angular
speed is
wi ¼ Kw0 ¼ 1:385w0 ð9:11Þ
Figure 9.7 describes the EVT in mode 2, the high-speed operating mode. In
this configuration the clutch pack, CL2, CL3, toggles so that MG1 is now engaged
with both the sun gear of E1 and the ring gear of E2. In mode 2, MG2 remains
connected to E2’s sun gear. Electrical power, Pe, can circulate between MG1 and
MG2 as needed to affect the appropriate variator ratio, K, such that engine speed is
independent of vehicle speed. In mode 2, the equations for E1 remain the same as
(9.3) but the expressions for E2 speeds and both m1 and m2 torque levels change
dramatically.
SOC
Cb Battery
EVT in mode 2 Uoc
V
CL1 = 1, CL2 = 1, CL3 = 0 Pc
Wheels
wg’ Pe wg’
mg’ mg’ rw
Jg wa’
Jg gf2m ma’
S1 MG1 S2 gz2f
gi2g w i’ MG2 Ja
mi’
we’ Ji gm2i
C1 C2 FD
me’ Intermediate shaft wfd’
Je CL2 gr2m mfd’
ge2g
ICE R1 R2 Jfd
CL3 Counter shaft wr2’
CL1 wr1’ mr2’
mr1’ Jr2
Jr1
Figure 9.7 EVT in mode-2, compound split, high-speed mode
In mode 2, the EVT angular speeds for MG1 and MG2 are given by (9.12)
and (9.13), respectively.

1 k1
w1 ¼ þ k 1 wi ð9:12Þ
K K
Ch009 8 June 2011; 16:6:11


1 1
w2 ¼ þ k1k2 1 wi ð9:13Þ
K K
The corresponding torque expressions at E1 and E2 are
1
m1 ¼ k 2 m2 þ mi ð9:14Þ
k1
w1
m2 ¼ h1 h2 m1 ð9:15Þ
w2
The procedure here is to compute the angular speed ratios in (9.15) using
(9.12) and (9.13), then substitute the resulting expression into (9.14) to obtain MG1
torque in terms of the EVT input torque. Then back-substitute the value derived
into (9.15) to obtain MG2 torque. With this complete, it will be possible to compute
the power flows in the electric variator without ESS contribution. It should be
obvious that ESS power, Pb, can be added to electric circulation power, Pe, at will
and without disruption of the engine to wheels power splitting. So, taking the ratio
of MG1 to MG2 angular speeds results in
w1 1 ðK 1 ð1=k 1 ÞÞ
¼ ð9:16Þ
w2 k 2 ðK 1 þ ð1=k 1 k 2 ÞÞ
Now, substitute from (9.16) into (9.15) to find MG2 torque m2.
h1 h2 ðK 1 ð1=k 1 ÞÞ
m2 ¼ mi ð9:17Þ
k 1 k 2 ½ðK 1 þ ð1=k 1 k 2 ÞÞ h1 h2 ðK 1 ð1=k 1 ÞÞ
Substitute from (9.17) into (9.14) and simplify.
h1 h2 ðK 1 þ ð1=k 1 k 2 ÞÞ
m1 ¼ mi ð9:18Þ
k 1 ½ðK 1 þ ð1=k 1 k 2 ÞÞ h1 h2 ðK 1 ð1=k 1 ÞÞ
The mechanical points for the EVT in mode 2 can be obtained directly from
the numerator of (9.17) and (9.18). It will be shown that the first mechanical point,
or node, in the EVT mode 2 is identical with the mechanical point found in mode 1,
the input split mode. Call the mechanical points (nodes) as l1 and l2 for which
1 1 þ k1
l1 ¼ 1 þ ¼ ð9:19Þ
k1 k1
1 k1k2 1
l2 ¼ 1 ¼ ð9:20Þ
k1k2 k1k2
Next, define the root of the m2 numerator as l3 so that the torque expression
can be simplified in terms of its mechanical points. This helps in understanding the
Ch009 8 June 2011; 16:6:11

2-mode EVT because the relationship of mechanical points in the output torque
expression will be evident.
h1 h2 ðl1 l2 Þ
l3 ¼ ð9:21Þ
h1 h2 1
Following the same procedure as in the single-mode eCVT, the total output
torque for the EVT in mode-2 is obtained as the sum of MG1 and MG2 torque
levels written in terms of input torque, mi. The result can be greatly simplified by
making substitutions from (9.19) to (9.21) into (9.17) and (9.18) resulting in (9.23)
after some algebra.
m0 ¼ ð1 þ k 1 Þm1 ðk 1 k 2 1Þm2 ð9:22Þ
l1 ðK l2 Þ h1 h2 l2 ðK l1 Þ
m0 ¼ ð9:23Þ
ð1 h1 h2 ÞðK l3 Þ
When the speed ratio, K, of the EVT equals either of the mechanical nodes, the
circulating electric power is zero and in between the two nodes the circulation power
can be relatively low. This is one of the strong merits of the EVT. The MG1 power,
P1, is obtained from (9.12) and (9.18), which after simplification comes out to
ðK l1 ÞðK l2 Þ
P1 ¼ Pi ð9:24Þ
Kð1 h1 h2 ÞðK l3 Þ
As an example, compute the MG1 normalized power flow for a variator ratio
0.35 < K < 3.1. Figure 9.8 shows the trajectory of MG1 input power for this range
of K when the base ratio of E1 is 2.3 and for E2 it’s 2.6. Note that the mechanical
P1 versus K (k1 = 2.3, k2 = 2.6)

3.000
2.500
2.000
1.500
P1
1.000
0.500
0.000
0 0.5 1 1.5 2 2.5 3 3.5
–0.500
K
Figure 9.8 Illustration of electric variator normalized power P1 relative to K
Ch009 8 June 2011; 16:6:11

nodes are approximately 0.83 and 1.43 calculated by substituting the k1 and k2
values into (9.19) and (9.20), respectively.
A second very important factor to consider on inspection of the MG1 power
flow is that when P1 ¼ 1 it means 100% of the input power from the engine flows
through the electric path and when P1 ¼ 0 it means 100% of the engine power flows
through the mechanical path (refer to Figure 9.1b). In addition to these facts, it must
be borne in mind that positive P1 means power flows into the MG1 shaft and MG1
is therefore in generating mode. Conversely, when the mechanical power flow, P1,
is negative MG1 is motoring and delivering additional torque into E1. On single-
mode systems when the generator reverses its function and delivers motoring
power into the power split device, it is said to operate in negative split, meaning
that the MG1 lugs the engine into a lower speed higher efficiency operating point
via reversal of the d.c. link power flow, Pe.
The final point to consider regarding the electric variator power flow, shown in
Figure 9.8, is that power flow between the mechanical nodes is relatively low,
approximately 0.15 pu, so that operation between these points is at very good
efficiency.
Note the equivalent gear shift ratio coverage, gsrc, of the EVT in Figure 9.9. In
this figure, the variator gain, K, ranges from 0.5 to 2.5, while circulating power
P1 < 1 pu. This ratio, not surprisingly, is 5:1, and more than equivalent to the
6-speed automatic transmission with its gsrc ¼ g1/g6 ¼ 3.1/0.61 ¼ 5.08:1. The EVT,
therefore, has the ratio coverage of a 6-speed automatic.
A big advantage of the EVT is that the location of the mechanical nodes can be
adjusted to some degree by design, and their spacing can also be adjusted by tuning
the E1 and E2 base ratios. To see this effect the values of k1 and k2 are shifted down
to lower values, but not below a critical point of k ¼ 1.6 for a planetary gear.
P1 versus K (k1 = 1.7, k2 = 1.8)

1.400
1.200
1.000
0.800
0.600
P1
0.400
0.200
0.000
0 0.5 1 1.5 2 2.5 3 3.5
–0.200
–0.400
K
Figure 9.9 The EVT of Figure 9.8 with different E1 and E2 base ratios (K > l1
during vehicle launch to K < l2 for highway cruise)
Ch009 8 June 2011; 16:6:11

With the new values of k1 and k2, we see two dramatic changes in P1 versus K:
(i) The entire curve shifts down into the motoring quadrant with motor power
reaching 0.2 pu. (ii) The mechanical points are now spaced farther apart, l1 ¼ 1.588
and l2 ¼ 0.673, so that gsrc ¼ 2.85/0.4 ¼ 7.1, an even wider ratio coverage, more in
line with a 7-speed automatic transmission. The point is that an EVT can be ‘tuned’
to realize mechanical points that can correspond to vehicle operating speeds and the
spacing can be adjusted to realize high efficiency in the EVT over a wide range of
vehicle operating speeds.
9.3 Ultracapacitor application in 2-mode eCVT
The benefit of ultracapacitors in the 2-mode eCVT, or EVT for that matter, is the
same as for the single-mode system discussed in Chapter 8. These benefits include
the following points:
● Highly cyclable energy storage for rapid dynamics of electric variator power
flows.
● High peaking power, both recuperating and boosting, to cancel MG1 and MG2
inertial effects.
● Fast power boosting via MG2 to the vehicle driveline for aggressive maneuvers.
● Enhanced life of the battery component of the ESS.
● Lower rms currents in the battery component of the ESS.
● Less heating of the battery component, minimized by the ultracapacitor.
● Enhanced performance of the ESS at cold temperatures.
Figure 9.10 shows the functional schematic of an EVT with power electronic
converters dedicated to MG1 and MG2, the main traction battery operating at high
300 V
=
=
Battery
Ultracapacitor = =
Inverter Wheels
~ ~
S1 MG1 MG2 S2
C1 C2 FD
CL2
ICE R1 R2
CL1 CL3
Epicyclic Epicyclic
input Compound split, EVT output
Unbuffered ESS
Figure 9.10 Illustration of an ultracapacitor augmented ESS in the EVT
Ch009 8 June 2011; 16:6:12

voltage and the buffered ultracapacitor. With power converter on the ultracapacitor,
the d.c.-link voltage can be held fixed for highest performance.
Design aspects of the ultracapacitor plus battery-active parallel combination
has been dealt with in considerable detail in earlier chapters. Additional details can
be found in References 5–7 that focus primarily on how to minimize the necessary
electronics, hence system cost. The methods range from switchable configurations
to buffering only a portion of the ultracapacitor pack and thereby making use of a
lower rated power converter. These concepts are but a sampling of the high levels
of interest in the integration of ultracapacitor power cache alongside an energy
optimized battery component in a vehicle ESS.
Figure 9.11 is offered as a summary of the EVT showing in clearer fashion the
electric variator function and its electronic content. Coordination of MG1 and MG2
via their attendant power electronic converters has been the subject of hundreds of
patents over the past two decades. It goes without saying that control aspects of the
eCVT and EVT (or 2-mode) are subjects of intense industrial and research
laboratory interest.
Power electronics
M/G1 Vφ, Iφ
wvi
S1 mvi
Control electronics
E1 Controller, Comm.
R1 C1 Gate drives, Power supply
wi wo
PM/G
mi mo
C2 R2
E2
S2 Power electronics
wvo
mvo
M/G2 Vφ, Iφ
Control electronics
Controller, Comm. Vb
Gate drives, Power supply Pb
Energy storage system
(ESS)
Figure 9.11 Full schematic of the EVT
Ch009 8 June 2011; 16:6:12

In Figure 9.11, the EVT output torque, m0, and angular speed, w0, are applied
via a final drive to the driven wheels resulting in a vehicle speed, V, and a pro-
portionate amount of tractive effort. The electric variator input torque, mi, and
angular speed, wi, come from the engine and are essentially independent of vehicle
speed to the limits of the variator speed ratio, K. Electric motor-generators, MG1
and MG2, along with their respective planetary gear sets, E1 and E2, constitute the
speed variator that decouples engine speed from vehicle speed so that overall power
transmission can be optimized. Figure 9.12 illustrates graphically how the MG1
and MG2 rotational speeds, shown as rpm’s Nm1 and Nm2, vary with vehicle speed,
V. Also shown is the engine speed, assumed in this particular case to be a diesel
engine that ranges in speed, Ne, from 1800 to 2400 rpm regardless of vehicle speed
where the vehicle is taken as an Allison Hybrid System (AHS) in a large SUV.
AHS-2 EVT
N (krpm) 10
Nm1
8
6
Nout
4
2 Ne
–10 0 10 20 30 40 50 60 70 80 90 100
–2 V, mph
–4
–6
–8
Nm2
–10
Rev
Vs1 Vshift Vs2 Vs3 Vs4
Forward
Figure 9.12 EVT MG1/Nm1 and MG2/Nm2 angular speeds in rpm versus vehicle
speed V in mph
Note that reverse motion has both MG1 and MG2 rotating in reverse direction.
Then at vehicle launch from zero speed, the speed of traction motor MG2, oper-
ating in the input split mode (also BEV mode, if necessary) increases from zero
speed to some higher speed that aligns with a mode shift point, Vshift. At the same
time during mode 1, input split mode, the engine speed was ramped to maximum
and then held steady at vehicle speed, Vs1. When the mode shift occurs, EVT clutch
pack, CL2–CL3, toggles and MG2 is able to reuse a good fraction of its positive
rotational range well into its negative rotational range. All the while the speed of
MG1 can be at different slopes relative to vehicle speed. Figure 9.12 is a plot of
(9.3) and (9.4) for the case K > l2 to K ¼ l2 as the vehicle is launched from zero
speed, then a plot of (9.12) and (9.13) for the case of K< l2 to K< l1 as the vehicle
accelerates to highway speeds. Note that at zero vehicle speed the K-value goes to
infinity and represents points well to the right in Figures 9.8 and 9.9. The dis-
continuity in Nm2 is the mode shift point.
Ch009 8 June 2011; 16:6:12

9.4 Plug-in hybrid electric vehicle: Volt

This chapter will close with an update on the most recently introduced eCVT, the
GM Volt plug-in hybrid electric vehicle, also known as a range extended vehicle
(REV) [2]. The Chevy Volt consists of electric drive components similar in rating
to those of the earlier GM EV1 electric vehicle, as shown in Figure 9.13. In the
Volt a 53 kW, 3-cylinder gasoline engine is the range extender that drives a 53-
kW electric generator to replenish the batteries and also to augment electric
propulsion power. The main propulsion system consists of a 120-kW front wheel
drive electric motor, a 16-kWh lithium–polymer battery and attendant
subsystems.
53 kW
3-cylinder
1-L genset
Plug-in wiring harness
16-kWh Li-ion battery pack

Cooling duct
120-kW electric motor
Figure 9.13 Chevy Volt REV chassis illustration
For the analysis to follow, the main propulsion system in the Volt REV is assumed
to consist of a 1.4-L Ecotec engine driving a 53-kW generator. The propulsion unit
can be assumed to be a series–parallel switching arrangement having both series-
HEV and parallel-HEV characteristics. The eCVT can operate in one of four modes
depending on the activation of three synchronous clutches, CL1, CL2, and CL3 in a
2-motor eCVT architecture. Power splitting is performed via a single planetary set,
E1, with fundamental ratio, k1.
The Volt range-extended vehicle is designed to operate as a battery electric
vehicle (BEV), during low-speed driving (V< 30 mph) and in engine augmented
higher speed driving as a range extended vehicle (REV) up to 70 mph. For condi-
tions when the ESS state-of-charge (SOC) is less than SOCmin, the engine is
engaged as a battery charger via MG2. For high load conditions, such as climbing
grades or high-speed travel, the system operates as a 2-motor eCVT with both MG1
and MG2 providing propulsion power (Figure 9.14).
Ch009 8 June 2011; 16:6:12

(a)
rw
CL3 CL2 CL1
ICE MG2
ω2 ωi R
S2
= gfN
~
Pe FD
= C ωo
~
S1
Ud Pb
MG1 S
ESS ω1
E1
k1
(b)
Figure 9.14 (a) REV engine and (b) functional schematic
When the REV is in mode 1 (Table 9.2), the low-speed BEV mode, the angular
speed of MG1 is given in terms of K as

1 k1
w1 ¼ þ k 1 wi ð9:25Þ
K K
1 þ k1
l1 ¼ ð9:26Þ
k1
Ch009 8 June 2011; 16:6:12

Table 9.2 REV clutch engagement map
Mode CL3 CL2 CL1 Description

1 0 0 1 BEV, low-speed mode, ESS power to wheels via MG2
2 0 1 0 BEV, high-speed mode, ESS power to wheels via MG2 þ MG1
3 1 0 1 REV, ESS charging mode, SOC < SOCmin
4 1 1 0 REV, high-speed or combined mode, MG2 þ MG1 propulsion
It is immediately clear that (9.25) and (9.26) are identical to the MG1 angular
speed relations for a single-mode eCVT, or a 2-mode EVT during input split mode.
The same can be said of its mechanical point speed, l1, at which point MG2 speed
is released from zero by clutch, CL1’s release.
Example 9.2: Suppose the eCVT on the Volt REV has k1 ¼ 2.5. Assume also that
the input and MG1 angular speeds wi ¼ w1 ¼ 340 rps. (a) Determine the output
angular speed to the final drive. (b) Given a final drive ratio gfd ¼ 2.16 and
rw ¼ 0.32 m, compute the vehicle speed. (c) Let the E1 synchronous clutch pair
CL1 and CL2 toggle to transition from high-speed REV to low-speed REV mode in
0.15 s and given a MG2 rotor inertia J2 ¼ 0.02 kg m2/rad, and MG1 rotor inertia
J1 ¼ 0.04 kg m2/rad, compute the ESS power pulse necessary to ensure that no jerk
is noticeable on the vehicle driveline. (d) How much of this switchover energy
needs to be supplied or absorbed by the ESS?
Solution:
(a) For this case, the output angular speed is
w1 þ k 1 wi 340 þ 2:5ð340Þ
w0 ¼ ¼ ¼ 340 rad=s
1 þ k1 3:5
rw w0 0:32ð340Þ
(b) V¼ ¼ ¼ 50:37 m=s ð112:7 mphÞ
gf d 2:16
(c) Refer to the chart in Figure 9.15 that shows the angular speed transitions of
MG1 and MG2, the constant torque levels during the transition, and their
respective power profiles.
w w1 = 3.5 ω0 m
m1 = J1w1 P1
M
w0, w1
340 w0 0 t t
wi 0+ G T 0+ T
m2 = J2w2 P2
wi = 0
0+ T t
Figure 9.15 Dynamic transitions in the REV eCVT motor-generators during shift
Ch009 8 June 2011; 16:6:13

Dw1 ð1190 340Þ

w1: ¼ ¼ 5667 rad=s2
Dt 0:15
Dwi ð0 340Þ
wi : ¼ ¼ 2267 rad=s2
Dt 0:15
m1 ¼ J 1 w_ 1 ¼ 0:02ð5667Þ ¼ 113:3 N m
mi ¼ J 1 w_ i ¼ 0:04ð2267Þ ¼ 90:7 N m
P1 ¼ m1 ðw01 wT1 Þ ¼ 113:3ð1190 340Þ ¼ 96 kW
Pi ¼ mi ðw0i wTi Þ ¼ 90:7ð0 340Þ ¼ 30:8 kW
Therefore, MG1 requires 96 kW of motoring power to accelerate, 30.8 kW

of this circulates from decelerating MG2 to zero speed in generating mode.
The difference must therefore be supplied by the ESS, which in this instance
amounts to
Pb ¼ P1 Pi ¼ 96 30:8 ¼ 65:2 kW
(d) Activation of the synchronous clutch packs, and assuming none of the
switchover inertial power dissipation occurs in the clutches, means that the
ESS must supply an amount of energy equivalent to
W ESS ¼ 0:5Pb T ¼ 0:5ð65:2kÞ 0:15 ¼ 4890 Jð1:36 WhÞ
Example 9.2 highlights a significant aspect of eCVT systems used in HEV,
PHEV, and REV architectures: mode shifting is accompanied by rapid transitions
in angular momentum of rotating elements of the driveline and the attendant power
excursions cannot be ignored. For example, if the ESS did not supply the inertial
power demand of the REVs eCVT during a synchronous shift, this energy would
have to come from the driveline resulting in a noticeable deceleration jerk and
therefore completely unacceptable driveability.
Here again is an illustration of how the fast power transfer capability of the ultra-
capacitor can help, in this case, by either supplying or absorbing inertial power tran-
sitions. Requiring the battery to supply the 65-kW power pulse found in Example 9.2 on
a regular basis is not good for battery longevity. Energy is far more than the nominal
2000 mF of d.c. link capacitance can supply. Exercise 9.5 treats this particular case.
Exercises
9.1 Using (9.10) compute the fundamental ratio, k1, of the EVT such that its
variator ratio, K, corresponds to the value listed for gear number 3 in Table 9.1.
Ans: k1 ¼ 2.4
Ch009 8 June 2011; 16:6:13

9.2 Use the value of k1 calculated in Exercise 9.1 and the corresponding value of
K in (9.5) to compute the MG1 input power.

Ans : Mechanical point P1 ¼ 1:4171
þ 2:4ð1:417Þ
1
1 Pi ¼ 0
9.3 Knowing that planetary gear sets have nominal base ratio of greater than 1.6:1
and typically less than 3:1, then evaluate the EVT power flow of MG1 dis-
cussed in detail in Section 9.2 to quantify the resulting performance for the
following two cases:
(a) k1 ¼ 1.7 and k2 ¼ 2.9
(b) k1 ¼ 2.9 and k2 ¼ 1.7
Show in the results in chart format and state the mechanical points, l1, l2, and
gsrc that result for each case.
Ans: (a) l1 ¼ 1.588, l2 ¼ 0.797, and gsrc ¼ 5.5:1
(b) l1 ¼ 1.345, l2 ¼ 0.797, and gsrc ¼ 4.75:1
P1 versus K (k1 = 1.7, k2 = 2.9) P1 versus K (k1 = 2.9, k2 = 1.7)
3.000 3.000
2.500 2.500
2.000 2.000
1.500 1.500
P1 P1
1.000 1.000
0.500 0.500
0.000 0.000
–0.500 –0.500
0 1 2 3 4 0 1 2 3
K K
9.4 Refer to Figure 9.12 and for an EVT having k1 ¼ 2.3 and k2 ¼ 2.6 with
mechanical points as given in section 9.3 as l1 ¼ 0.833 and l2 ¼ 1.435, then
tabulate the angular velocities of MG1 and MG2 at these nodes.
Ans: In both mode 1 and mode 2, the MG1 rotational speed relative to input
speed is given as

1 k
w1 ¼ þ k 1 wi ð9:27Þ
K 1
But, the MG2 rotational speeds are different in mode 1 and mode 2

1 k2
w2 ¼ þ wi ðmode 1Þ;
K K
ð9:28Þ
1 1
w2 ¼ þ k1k2 1 wi ðmode 2Þ
K K
The MG1 and MG2 speeds can be tabulated by mode at each node.
Ch009 8 June 2011; 16:6:14

Mode 1 2
K ¼ l1 w1 ¼ 0 w1 ¼ 0
w2 ¼ 2.509 w2 ¼ 2.509
K ¼ l2 w1 ¼ 1.663 w1 ¼ 1.663
w2 ¼ 4.223 w2 ¼ 0
The first mechanical point occurs when MG1 speed is zero and the second
mechanical point when MG2 speed reaches zero, after its mode 1 to mode 2
reversal.
9.5 For the REV described in Example 9.2 a d.c.-link voltage Ud ¼ 400 V and
Clink ¼ 2000 mF are assumed. Suppose further that the d.c. bus voltage can
only droop dUd < 20 V during the power pulse that occurs during an eCVT
synchronous shift. For how long can the d.c. link capacitor support a power
pulse of Pb ¼ 65.2 kW?
Ans: The power pulse has the triangular waveshape shown in Figure 9.15 and
an energy, WESS ¼ 4890 J, and
W C ¼ 0:5C link ðU 2di U 2df Þ ¼ 0:001ð4002 3802 Þ ¼ 15:6 J
t Pb t 2 4
PðtÞ ¼ Pb ; W ðtÞ ¼ ¼ WC ¼ t ¼ 8:5 ms
T T 2
Therefore, relying on the power inverter d.c. link capacitors is not acceptable
to handle these levels of inertial power flows.
References
1. A.G. Holmes, M.R. Schmidt, Hybrid Electric Powertrain Including a 2-mode
Electrically Variable Transmission, U.S. patent 6,478,705, issued
12 November 2002
of Engineering Technology (IET), Stevenage, Herts, United Kingdom, 2010
3. B. Conlon, P.J. Savagian, A.G. Holmes, M.O. Harpster Jr. Output Split
Motors, U.S. patent publication 2009/008171 A1, 26 March 2009
4. D. Zhang, J. Chen, T. Hsieh, J. Rancourt, M.R. Schmidt, ‘Dynamic modelling
and simulation of 2-mode electric variable transmission’, Proceedings of the
Institution of Mechanical Engineers, vol. 215, Part D, pp. 1217–23, 2001
5. J.R. Miller, ‘Capacitor/battery load-leveling of hybrid vehicles without the
use of active interface electronics’, The 6th International Seminar on Double
Layer Capacitors and Similar Energy Storage Devices, Deerfield Beach, FL,
pp. 1–13, 9–11 December 1996
Ch009 8 June 2011; 16:6:14

6. G. Guidi, T.M. Undeland, Y. Hori, ‘An interface converter with reduced VA

ratings for a battery-supercapacitor mixed systems’, IEEE Power Conversion
Conference, PCC2007, Nagoya, Japan, pp. 936–41, 2–5 April 2007
7. A.W. Stienecker, T. Stuart, C. Ashtiani, ‘A combined ultracapacitor-lead
acid battery energy storage system for mild hybrid electric vehicles’, IEEE
Vehicle Power and Propulsion Conference, VPPC2005, Paris, France,
pp. 350–55, 7–9 September 2005
Ch009 8 June 2011; 16:6:14

Chapter 10
Life cycle testing
Life cycle evaluation of ultracapacitors is the subject of considerable industrial

interest, especially in terms of elevated temperature conditions and power cycling.
The authors of References 1 and 2 have investigated ultracapacitor parameter
changes due to electrical and thermal stress and have explored their aging char-
acteristics. Ultracapacitor life can be evaluated under two different characterization
methods: (1) power cycling where the unit under test is charged from zero to its
rated voltage, then cycled from rated to half-rated voltage repeatedly and (2) d.c.
life under constant voltage and temperature stress. When we speak of power
cycling, we think of the current and voltage waveform shown in Figure 10.1, where
the current is a quasi-square wave of discharge and charge such that the ultra-
capacitor voltage is maintained within its voltage window. The ultracapacitor
nominal voltage window represents a voltage swing of 1.35 V for a device being
cycled between Umx ¼ 2.7 V and Umx/2 ¼ 1.35 V.
Figure 10.1 is generic in the sense that charges exchanged during the discharge
and charge portions of a cycle are matched by the cell voltage swing within its
working voltage window of 1.35 V. This fact is made clear by (10.1) and a quick
50 3.5
Charge (A) Discharge (A)
40
Voltage (V) 3
30
20 2.5
10
Current (A)
Vscap (V)
2
0
1.5
–10
–20 1
–30
0.5
–40
–50 0
0 50 100 150 200
Time (s)
Figure 10.1 Illustration of ultracapacitor cycling test, T ~14 s per pulse
CH010 7 June 2011; 14:23:41

exercise taking the current as 300 A and a 3000 F cell such that Q ¼ Ccell
dUcell ¼ 3000(1.35) ¼ 4050 C, from which T ¼ Q/I0 ¼ 4050 F/300 A ¼ 13.5 s.
When the ultracapacitor is cycled, the internal heating raises its core tem-
perature, and it is this temperature that influences life. Researchers [1,2] have
shown that under continuous cycling conditions, the cell core temperature rises
approximately 2 C above ambient under transient conditions and to within 1 C
during steady state. This means that for all intents, a measurement of cell terminal
temperature is representative of cell core temperature as illustrated by Figure 10.2.
The finding shown graphically as Figure 10.2 is very important because it
confirms that estimating cell service life as a function of its temperature and vol-
tage is adequately predicted by measuring the cell temperature and voltage at its
termination posts. Use will be made of these facts in this chapter, and the reader is
encouraged to make note of this fact.
50
45
Temperature
at front +
40
Temperature (°C)
35
Temperature inside
the supercapacitor
30
25
20
0 1 2 3 4
Time (h)
Figure 10.2 Cell temperature at positive (þ) post/terminal versus internal

temperature
Ultracapacitor cells or modules subjected to either method of life cycle estimation

noted above respond through a monotonic drift of their electrical parameters, ESRdc,
and capacitance, Ccell. This fact is verified by all investigations known to this author.
The result shown as Figure 10.3 best illustrates the parameter shifts to be expected. The
drift in cell capacitance is most pronounced and most easily quantified, whereas the
drift in ESRdc is less apparent, but it does tend to increase with stress and aging as
expected. For a 3000 F, 2.7 V ultracapacitor cell, a quick resistance survey shows that
termination resistance constitutes roughly 20% of its overall
ESRdc ¼ Rconn þ ESRelectronic þ ESRionic ¼ 70 mW þ 130 mW þ 160 mW ¼ 360 mW
ðe:g:; 70=360 ¼ 0:1944; or 20%Þ
CH010 7 June 2011; 14:23:41

Life cycle testing 281
385 7
End of life (ESR)

365 6
345 5
Resistance (mΩ)
Capacitance (F)
325 4
305 3
End of life (Capacitance)
285 2
265 1
0.001 0.01 0.1 1 10 100 1000 10000
Time (years)
Figure 10.3 Cell capacitance and ESRdc of a Maxwell D Cell, 350 F, 2.7 V,
showing end-of-life limits
The log-linear plot in Figure 10.3 shows that the D Cell ESRdc increases from
3.2 mW at BOL to approximately 3.5 mW at EOL, an increase of only 9% in this
case. Cell capacitance fade is most pronounced with life cycle testing and drops
from 350 F at BOL to its EOL value of 295 F in Figure 10.3 from Reference 2. The
chart of Figure 10.3 is representative of ultracapacitor parameter fade and not an
exact representation to be used for product service life calculations.
Before proceeding to sections 10.1 on more detailed d.c. life evaluation and
10.2 on self-discharge characterization, we take a short literature review on life
testing of capacitors. Sankaran et al. [3] experimentally evaluate aluminum elec-
trolytic capacitors used in power electronic inverter as d.c. link filters to bypass
switching ripple current. In this application the capacitor is exposed to long-term
voltage and temperature stress, where voltage stress is mainly uniform at rated d.c.
working voltage conditions but temperature is strongly dependent on automotive
temperature environment and core heating due to ESR and Irms. The electrolytic
capacitor wear-out mechanisms are proposed to be due to electrolyte loss via vapor
diffusion through the seals and deterioration of the electrolyte. It is worth con-
sidering these mechanisms for relevance to our case of the electrochemical double-
layer capacitor. Certainly, as with the aluminum electrolytic capacitors, the ultra-
capacitor pressurizes during use, and this internal pressure will slowly leak via the
seals. The electrolyte ions are being forced into and out of activated carbon cores
and oscillating within the paper separator so that it should be apparent that the
electrolyte will suffer some form of deterioration. Any form of cell deterioration
will appear as parameter fade and, therefore, a wear-out mechanism.
Sankaran et al. [4] explored the effects of power cycling on power electronic
module transistor stacks and modules consisting of semiconductor die soldered to
heat spreaders and wire bonds for interconnections. One of the more pronounced
CH010 7 June 2011; 14:23:41

effects of power cycling is to expose the wire to semiconductor metallization inter-

face to thermal and electromechanical stress that eventually result in fracturing of the
wire bond and subsequent failure. Similar situations exist in the ultracapacitor cell as
aluminum–aluminum welds as interference fits and termination welds.
Miller et al. [5] emphasize that it is the voltage and temperature that contribute to
loss of service life and not the cycling of the cell. In their experimental work, several
groups of ultracapacitor cells were subjected to different combinations of voltage and
temperature stress; half the groups within rated stress levels and half outside normal
stress limits. In their analysis, the time to failure, or service life as it will be known here,
is a defined failure based on cell parameter fade – nominally capacitance fade. Using
an Arrhenius–Eyring life relationship, where t ¼ characteristic life of a Weibull sta-
tistical distribution (10.1), and equating this to a modified Arrhenius relationship for
temperature and stress factors (10.2) result in the expression for characteristic life
(t, when 63.2% of parts have failed the prescribed EOL criteria).
FðtÞ ¼ 1 eðt=aÞ
b
ð10:1Þ
t ¼ A eB=t eDU ð10:2Þ
The Weibull shape factor, b, remains the same if wear-out has the same
underlying cause, and the characteristic life, a, is that point in time at which 63.2%
of the starting population have failed for this constant b. The characteristic life, t,
given by the Arrhenius-Eyring relationship (10.2), has three undetermined coeffi-
cients ‘A,’ ‘B,’ and ‘D’ plus the nominal life at the rating point (T0,U0). These two
expressions will be elaborated on in more detail in the following sections.
Briat et al. [6] target the determination of ultracapacitor reliability exposed to
pulsed current profiles having the same Irms and, therefore, with a premise that the
same core temperature rise should result in the same parameter aging effects. This,
however, is not what the authors found in experimental studies. Rather, the current
profiles of high duration but short dwell time resulted in faster parameter dete-
rioration, even though Irms was the same as for more symmetrical (softer, lower
peak-to-average ratio) current pulses. The rationale for this behavior was proposed
as due to the rate at which the cell core temperature changed during the course
of the pulse, with a high-amplitude, short-duration pulse resulting in faster aging.
Their conclusion was that additional experimental work is warranted to validate
this observation.
Kawaji and Okazaki [7] view ultracapacitor reliability in customer applications
as life due to electrical stress, mechanical endurance in the vehicle environment, and
safety. In their work, the ultracapacitor electrical service life is due to voltage
and temperature stress. Similarly, mechanical endurance is the ultracapacitor life in
a thermal and vibration environment, and lastly, safety of life is how well it holds
up in terms of electrical leakage and short circuit. Their concerns are prompted by
ultracapacitors put into service in transit bus applications where service life is on
the order of three times that of passenger cars and light trucks.
CH010 7 June 2011; 14:23:42

The authors in Reference 8 validate the ultracapacitor life metrics previously

discussed by showing that regenerative brake current waveforms in railway appli-
cations contribute to high internal power dissipation and the need for thermal
management to ensure adequate service life. These authors also point out the need
for better understanding the effects of ripple current on ultracapacitor life because
internal heating due to ripple currents may be higher than that for softer current
exposure.
One point to be clear on is that service life estimation is based on ultracapacitor
parameter fade, where the most consistent parameter is cell capacitance. The
question that comes up in regard to frequency response of ultracapacitors is capa-
citance, and also resistance, have frequency dependence. This is not the case as
these parameters are characteristics of materials and geometry. For example,
resistance is calculated as the product of a material property, conductivity, and the
ratio of area, ‘A,’ to length, ‘l,’ or electrode separation distance, ‘d.’ The same can
be said for capacitance and inductance.
U l
R¼ ¼ ; s½¼ðW mÞ1 ð10:3Þ
I sA
Q sA F Coul
C¼ ¼ ; e½¼ ½ ¼ ð10:4Þ
U d m Vm
l ml H Wb
L¼ ¼ ; m ½ ¼ ½¼ ð10:5Þ
I A m Am
In (10.4) and (10.5), the material properties for permittivity and permeability
are introduced and their definitions relative to the electromagnetic quantities of
electric flux and magnetic flux are shown. For example, permittivity is funda-
mentally electric charge (Coulomb) per Volt-meter, or Farad/meter, and for per-
meability it is fundamentally magnetic flux (Weber) per Ampere-meter, or Henry/
meter. The effect of frequency comes in through the complex form of these con-
stitutive relations of conductivity, permittivity, and permeability, but the steady
state relationships remain as given.
10.1 Leakage current influence

All electrochemical cells exhibit electrical leakage that causes a gradual loss of
stored charge. The fundamental cause of leakage is not well understood and is
perhaps an artifact of charge redistribution within deep pores of the electrode,
electrode pore pinch-off trapping charge, or some electrolyte shuttle type of
behavior that transfers charge between electrode double layers. Experimental evi-
dence quantifies leakage in two forms, self-discharge and leakage under constant
voltage. These two forms of leakage are captured in Figure 10.4 as manifestations
of some common underlying charge loss phenomena.
CH010 7 June 2011; 14:23:42

Internal charge loss
C(U) = C0 + kuU
Natural decay Voltage pinned leakage
Voltage decay Voltage is constant

Continuous charge loss Charge is replenished
Characteristic Characteristic
t
idecay = kuU + C0U + kuUU ileak (t) = Lim {UR e } + dQdt
τ
{
t→∞
Figure 10.4 Taxonomy of ultracapacitor leakage
Figure 10.4 quantifies charge loss as being natural decay or self-discharge or

voltage-pinned leakage. Self-discharge is the case when the cell is charged to rated
potential and then allowed to sit under open-circuit conditions. Leakage is when the
cell is charged to rated potential, then held under constant voltage and current
measured. Figure 10.5 is an illustration of the leakage current test showing the cell-
charging apparatus and monitoring equipment.
– + Vd.c.
Agilent Agilent
6031A + – 34970A
w/34901A
Rext
Jumper
Figure 10.5 Ultracapacitor leakage current test
For more discussion on leakage and natural decay testing, please refer back to
section 2.1. For our purposes here, it is the effects of voltage that are of interest. It is
well known that charge loss diminishes with time and that it is higher at higher
temperatures. As ultracapacitors age, their leakage reduces and eventually becomes
nearly constant with square root of time. The influence of temperature on leakage is
shown in Figure 10.6, where the higher the temperature (reciprocal scale), the
higher the leakage current. This is consistent with Arrhenius theory in that elec-
trochemical activation increases with increasing temperature.
CH010 7 June 2011; 14:23:42

Figure 10.6 shows that cell leakage current will be 10 mA at 2.7 V and 0 C
(3.66/K), dropping to less than 2 mA at 40 C (4.3/K), then growing to approxi-
mately 500 mA at þ60 C (3.0/K). The manufacturing criterion for a new cell of the
D Cell family is that leakage should not exceed 0.85 mA/F at room temperature.
Room temperature is 3.7 on the abscissa of the chart in Figure 10.6, for which
leakage at full voltage is well below 350 F 0.85 mA/F ¼ 300 mA.
10
2.5 V
2.6 V
1 2.7 V
Leakage current (mA)
2.8 V
2.9 V
3.0 V
0.1
0.01
0.001
2.8 3 3.2 3.4 3.6 3.8 4 4.2 4.4
1/Temperature (1000/K)
Figure 10.6 Leakage current versus temperature (60 C at left to 40 C at far
right) BCAP0350 cell
Endurance test criteria call for a service life of 1000 h at full voltage and
temperature stress. This means that the cell must lose not more than 20% of
capacitance nor gain more than 25% ESR for 1000-h exposure to 2.7 V and 65 C.
Life test under d.c. life conditions calls for a service life of 10 years (87,600 h) at
25 C. During d.c. life testing, the capacitance must not drop more than 20% and
ESR should not increase more than 100% of initial value.
The researchers in Reference 1 found that leakage current characterization
revealed two quite different activation energies, Ea, for the D Cell ultracapacitor
depending on temperature regime. For cold temperatures, 0–0 C, Ea ¼ 0.22 eV,
whereas for higher temperatures, 0–60 C, Ea ¼ 0.57 eV. This effect can be
appreciated by expanding (10.2) for A ¼ MSL0, and B ¼ Ea/k, where ‘k’
is Boltzman’s constant (8.62 105 eV/K) to (10.6), where MSL ¼ mean
service life.
MSLx ¼ MSL0 eðEa =kÞ½ð1=T 0 Þð1=T x Þ ð10:6Þ
CH010 7 June 2011; 14:23:42

Example 10.1: To illustrate the significance of (10.6) and the different activation
energies over the temperature range, calculate the mean service life (MSL) for both
levels of activation energy. Take the nominal temperature as T0 ¼ 21 C, and show
the data both in tabulated and in chart form.
Solution: A straightforward approach to this example is to solve (10.6) for both
cases of activation energy and to plot the result. Figure 10.7 illustrates the results in
a log-linear relationship showing the slope change when crossing 0 C. Note that the
cold temperature slope intersects the warm temperature curve at the nominal tem-
perature T0 as it should since the exponent is identical to unity at that point.
MSLx versus temperature

10.00 Tx Kx Ea MSLx
–40 233 0.22 9.7056
–30 243 0.22 6.1838
–20 253 0.22 4.0829
1.00
–10 263 0.22 2.7822
–60 –40 –20 0 20 40 60 80 0 273 0.22 1.9499
0 273 0.57 5.6413
0.10 10 283 0.57 2.3970
20 293 0.57 1.0798
30 303 0.57 0.5127
0.01 40 313 0.57 0.2553
Temperature (°C) 50 323 0.57 0.1327
60 333 0.57 0.0718
Figure 10.7 Service life approximation based on Arrhenius only for ultracapacitor
The result of Example 10.1 shows clearly that strictly on the basis of an Arrhenius
relationship for electrochemical activity, ultracapacitor service life would drop by two
orders of magnitude when operated at 60 C versus operating at 0 C because of the
higher activation energy in this temperature regime. At cold temperatures and taking
the cold temperature value of Ea out to 20 C shows that for the same temperature band
of 60 C, MSL will change by one order of magnitude, substantially less, so why life
expectancy in cold conditions is only modestly higher than it is at 0 C.
Inclusion of the voltage dependency term introduced in (10.2) is going to modify
this Arrhenius only relationship further and result in higher-temperature expected life
being longer than what would be estimated from Arrhenius only. It is found from long-
term d.c. life experiments that the activation energy of the acetonitrile solvent-based
ultracapacitor, Ea ¼ 0.5 eV, is very close to the literature value of Ea ¼ 0.57 eV cited
above. This fact, that in both cases experimental evidence shows that activation energy
is about the same, means that a single wear-out mechanism governs service life.
10.2 Reliability versus service life

It is very important to understand the distinction between reliability and service life
of a component or system. Reliability is a subset of quality and is defined as
follows [9]:
CH010 7 June 2011; 14:23:42

Reliability is defined as the probability that a product performs its inten-

ded function without failure under specified conditions for a specified
period of time. The definition contains three important elements: intended
function, specified period of time, and specified conditions.
When discussing reliability, most engineers think of an exponential distribu-
tion with constant failure rate, l. They also associate an exponential reliability
function, called the survival function, R(t), with the classical bathtub curve shown
in Figure 10.8 characteristic of this distribution. Service life means delivering an
intended function until a defined failure occurs. Parameter fade by a specified
percentage from initial value at EOL is a defined failure.
Failures
Early failures Time
Wear-out
Figure 10.8 Bathtub curve representative of exponential reliability function.

Early failure period: hazard rate l decreases over time; central
random failure period: hazard rate is constant over time; wear-out
failure period: hazard rate increases over time
The following explanatory material is explained in detail in References 9–13,

and in particular References 10–11 give insights into the Maxwell Technologies
ultracapacitor aging mechanisms through experimental investigation. References
12 and 13 are more in-depth treatments of accelerated life testing and case studies
of military systems. Turning our attention now to the topic of reliability, we define
the key metrics to be used in all studies of reliability:
● Probability density function, pdf or f(t)
● Cumulative distribution function, cdf or F(t)
● Reliability function, R(t), is called the survival function, the probability of
success
● Hazard function, h(t), is also called the failure rate for failures per unit time
* The most familiar illustration for h(t) is the reliability bathtub curve shown
as Figure 10.8
● Cumulative hazard function, H(t)
● Percentile, tp, is the time by which a specified fraction (p) of a population fails.
Percentile is the inverse of the cumulative distribution function, F(t). So,
tp ¼ F 1(p)
● Mean time to failure (MTTF) is the expected life, E(t), of a nonrepairable
product
● Variance, Var (T ) is a measure of the spread of a life distribution
CH010 7 June 2011; 14:23:42

Each of the reliability metrics will now be defined in terms of a reliability

function. As a case in point, we apply the normal distribution, N(m,s), to illustrate
the metrics and their definition. For example, any function having a mean, or
expected value m, and variance, s2, is a candidate. Application examples can be
voltage, rotations, distance, actuations, capacitance, and so on. As an example to
illustrate the normal distribution, we select the automobile electrical distribution
system, or PowerNet, voltage level as a case study.
Example 10.2: The automobile alternator regulator has a well-defined set point
that varies with temperature to maintain the vehicle battery at full float voltage
regardless of ambient conditions. This set point, crisp as it may be, however, has
manufacturing and component tolerances that result in a normal distribution of the
set point over a population of automobiles produced by an original equipment
manufacturer. The vehicle power train control module (PCM) regulates the Pow-
erNet to a potential, U, having mean value 14.2 V and various levels of standard
deviation to account for product tolerance. Figure 10.9 illustrates the normal dis-
tribution for this case where the pdf is defined as
dFðU Þ 1 2
¼ pffiffiffiffiffiffi eðU mÞ =2s
2
f ðU Þ ¼ 81<U <1 ð10:7Þ
dU 2ps
N (u, s) for u = 13.5 V and s = 0.5, 1.0, and 1.5

0.8
0.798
Voltage variability
0.6
f.1 (k)
f.2 (k)
0.4
f.3 (k)
0.2
0
0
10 12 14 16 18 20
10 U (k) 19.9
System voltage (U)
Figure 10.9 Normal, or Gaussian function, N(14.2, s) for various standard

deviations in PowerNet set point (s ¼ 0.5 curve has amplitude
0.8, s ¼ 1.0 curve amplitude is 0.4, and s ¼ 1.5 curve amplitude
is <0.3)
The cumulative distribution function, F(U), identified in (10.7) is the popula-

tion fraction relative to this regulation voltage set point (Figure 10.10). This is not
a failure, but tolerance about a design point where the variance must be kept
CH010 7 June 2011; 14:23:43

Cummulative distribution function of PowerNet voltage

1
1
0.8
f1 (k)
cdf (U)
0.6
f2 (k)
0.4
f3 (k)
0.2
0
0
10 12 14 16 18 20
10 U (k) 19.9
System voltage (U)
Figure 10.10 Cumulative distribution function for the PowerNet voltage,

U0 ¼ m ¼ 14.2 V (s ¼ 0.5 curve, F1, has steepest slope; s ¼ 1.0
curve, F2, has medium slope; and s ¼ 1.5 curve, F3, has shallow
slope)
very low to ensure proper function of vehicle ancillaries and accessories. It is clear
from Figure 10.9 that a standard deviation s < 0.5 is highly desirable in this
application.
ð1
1 U m
FðU Þ ¼ f ðxÞdx ¼ erf ð10:8Þ
1 2 s
Next, think of the impact that system voltage variation will have on the
vehicle headlamps where lamp power is a measure of illumination according to a
power law, Plamp ¼ P0(Ulamp/U0)a, where a 1.6. For this case, small changes in
lamp potential relative to the PowerNet set point will have dramatic impacts on
illumination and lamp life. The reliability function, R(U), defined by (10.9), is
plotted as Figure 10.11 and should be interpreted as the fraction of the population
of vehicles produced having PowerNet voltage set point on the high side. Vehicles
with PowerNet set point above the mean are more likely to experience early
failure of incandescent headlamps; hence, the population of vehicle lamps up to this
point are the survivors and those above are going to fail, in the context of this
example.

1 1 U m
RðUÞ ¼ 1 FðU Þ ¼ erf ð10:9Þ
2 2 s
The hazard function, h(U), for the normal distribution function illustrates how
the failures progress in time (10.10). In the context of this PowerNet example and
its connected headlamps, the hazard function must be interpreted as a measure of
CH010 7 June 2011; 14:23:43

Reliability function for the PowerNet voltage example

1
1
0.8
R.1 (k)
0.6
R2 (k)
R(t)
R.3 (k) 0.4

0.2
0
0
10 12 14 16 18 20
10 U(k) 19.9
System voltage (U)
Figure 10.11 Cumulative distribution function for the PowerNet voltage,

U0 ¼ m ¼ 14.2 V (s ¼ 0.5 curve, R1, has steepest (–) slope; s ¼ 1.0
curve, R2, has medium slope; and s ¼ 1.5 curve, R3, has shallow (–)
slope)
how fast the lamps will fail as a function of electrical distribution system voltage
level relative to the set point voltage, U0 (Figure 10.12).

Pðu < U u þ DujU>u Þ 1 dRðU Þ f ðUÞ
hðU Þ ¼ limDu!0 ¼ ¼
Du RðU Þ dU RðUÞ
f ðU Þ
¼ ð10:10Þ
ð1=2Þ ð1=2Þerf ½ðU mÞ=s
Hazard function for the N (u, s) pdf

5
5
4
h.1 (k)
3
h2 (k)
h(t)
h.3 (k) 2
1
0
0
10 12 14 16 18 20
10 U(k) 20
System voltage (U)
Figure 10.12 Hazard function for the PowerNet voltage, U0 ¼ m ¼ 14.2 V (s ¼ 0.5
curve, h1, is leftmost; s ¼ 1.0 curve, h2, middle; and s ¼ 1.5 curve,
h3, on far right)
CH010 7 June 2011; 14:23:43

The cumulative hazard function, H(U), indicates the rate of system voltage
change, in this case U, relative to the set point. That is, for the system voltage set-
point case having smallest standard deviation (s ¼ 0.5 in this example), the cumu-
lative hazard function is zero until it comes closest to the mean value. For wider
distributions, this range of zero H(U) shifts to correspondingly lower system voltages,
indicating that larger fractions of the vehicle (headlamp) population will be more
stressed. For example, if instead of a normal distribution, an exponential distribution
of system voltage had been selected, then its hazard rate, l, would be a constant and
its H(U) would therefore be an increasing function, similar to Figure 10.13, but linear.
Cummulative hazard function

0.1
0.1
H.1 (k) 0.08
H(t)
H2 (k) 0.06
H.3 (k) 0.04
0.02
0
0
10 12 14 16 18 20
10 U(k) 15
System voltage (U)
Figure 10.13 Hazard function for the PowerNet voltage, U0 ¼ m ¼ 14.2 V (s ¼ 0.5
curve, H1, is rightmost; s ¼ 1.0 curve, H2, middle; and s ¼ 1.5
curve, H3, on far left)
In general, the H(t) function indicates whether the failure rate, h(t), represents
an increasing failure rate (IFR), constant failure rate (CFR), or decreasing failure
rate (DFR). These metrics are very important in understanding system warranty,
but this topic is outside the scope of this book.
A summary of the more common probability distribution functions and their
reliability metrics are now presented to summarize this coverage of reliability. The
reliability functions covered are as follows:
● Exponential
● Normal
● Log-normal
● Weibull
10.2.1 Exponential reliability function

The key feature of this distribution is that its hazard rate is constant (CFR). This
means that the probability of a surviving product failing in the next small interval
of time is independent of time. This is the memory-less property. For an expo-
nential reliability function, random failures are caused by external shocks or
CH010 7 June 2011; 14:23:43

unexpected load changes. If each shock causes a failure, then the product life can
be approximated with an exponential distribution. Shocks of mechanical, thermal,
or electrical nature can generally be modeled as a Poisson process. If a product is
exposed to stress that exceeds some threshold, then it will fail if it adheres to an
exponential distribution. For example, semiconductor devices such as diodes,
transistors, and the like follow an exponential reliability function. Such devices
have no degradation or physical wear-out mechanisms, so the exponential relia-
bility function is appropriate. For devices that do fail due to some wear-out
mechanism such as mechanical components, a jet engine for example, or an elec-
trochemical capacitor or battery, the Weibull function is more appropriate.
The reliability metrics for the exponential reliability function are presented in
(10.11) through (10.15), where l ¼ failure rate or hazard function and q ¼ 1/l is the
mean time to fail (MTTF). The cumulative distribution function (10.11) for an
exponential reliability function is as follows:
FðtÞ ¼ 1 elt ð10:11Þ
The probability density function for the exponential reliability function is the
derivative of the cumulative function given as (10.12).
dFðtÞ
f ðtÞ ¼ ¼ lelt ð10:12Þ
dt
The survivor function, or reliability for the exponential case, is just the expo-
nential with hazard rate exponent.
ð1
RðtÞ ¼ 1 FðtÞ ¼ f ðxÞdx ¼ elt ð10:13Þ
t
The failure rate, or hazard function as noted above, is the ratio of the expo-
nential probability density function to its reliability function, which is constant.
f ðtÞ
hðtÞ ¼ ¼l ð10:14Þ
RðtÞ
Last, the cumulative hazard function for an exponential case is the product of
the constant failure rate, failures per unit time, multiplied by time according to
(10.15), the Taylor series expansion of H(t) showing that the cumulative hazard
function is an increasing function of time (IFR).
HðtÞ ¼ 1 RðtÞ lt ð10:15Þ
10.2.2 Normal reliability function

Most natural phenomena such as the height of individuals in a human population or
the tolerance of manufactured components and products follow the Gaussian or
CH010 7 June 2011; 14:23:44

normal distribution function. The metrics of this distribution are given in (10.16)
through (10.19).
Eff g ¼ m; expected value ð10:16Þ
varff g ¼ s2 ; variance of function; f ð10:17Þ
These definitions of expected value, or mean, and variance lead to the defini-
tion of the normal cumulative distribution function that was applied in Example
10.2. The cumulative distribution (10.18) yields the probability density function
after one differentiation operation (10.19).
ðt
1 2 2
FðtÞ ¼ pffiffiffiffiffiffi e½ðxuÞ=2s dx ð10:18Þ
1 2ps
1 2 2
f ðtÞ ¼ pffiffiffiffiffiffi e½ðtmÞ=2s ð10:19Þ
2ps
The reliability R(t) and hazard function h(t) for a normal distribution were
given in (10.9) and (10.10), respectively, and discussed in the context of vehicle
distribution system voltage. A new topic here is to consider the case of a group of
devices or products, each having a normal probability distribution function, where
each member of the group can be assumed to be independent of the others.
Therefore, this class of independent random variables acting together can be shown
to have a mean and variance of the population given by
XN XN
m¼ i¼1
mi and s¼ i¼1
si ð10:20Þ
10.2.3 Log-normal reliability function

The log-normal reliability function is of most interest to engineers engaged in
warranty analysis. For example, when the hazard function in the log-normal dis-
tribution begins to decrease (DFR), it is an indication that the warranty period is
exhausted. Once exhausted, products fail due to wear-out and normal degradation.
They have reached their end-of-life (EOL).
In this section the log-normal reliability function is defined, but not elaborated
on because our interest for ultracapacitors lies in the Weibull distribution. The
cumulative probability function for a log-normal reliability function is
ðt
1 f½LnðxÞm=2s2 g
2 lnðtÞ m
FðtÞ ¼ pffiffiffiffiffiffi e dx ¼ j ð10:21Þ
1 2psx s
CH010 7 June 2011; 14:23:44

The expectation value and variance of the log-normal function are given as
(10.22) and (10.23).
EfFg ¼ eðm þ s =2Þ

2
ð10:22Þ
VarfFg ¼ eð2m þ s Þ ½es 1

2 2
ð10:23Þ
10.2.4 Weibull reliability function

Weibull analysis provides a simple graphical solution of product life be it operating
cycles, operating time, or number of times a product is operated such as an auto-
mobile engine, a fuel cell, or a battery. Weibull paper abscissa is some measure of
life such as the characteristics noted above, and the ordinate is the probability of
occurrence of the event. When product failure data is plotted on the Weibull chart,
the straight line fit to the data intersects the 63.2% probability of occurrence line as
characteristic life, a, and its slope is the shape factor, b. One advantage of Weibull
analysis is that it provides useful data even for very small sample sizes. The Wei-
bull shape parameter is used to determine which member of the Weibull failure
distribution family best fits the data in question. For a given set of failure data, even
for 7 or 13 data points, an estimation of the shape parameter can be made. For
example, the following probability distribution functions result from the particular
value of b listed:
● b ¼ 0.5 represents infant mortality that may be attributed to inadequate burn-in,
green parts run, misassembly, or some other quality problem.
● b ¼ 1.0 represents random failures (i.e., the exponential distribution function)
where the failures are independent of time. Such failures are due to main-
tenance errors, electronic failures, or a mixture of problems.
● b ¼ 3.0 is early wear-out. For example, a low cycle fatigue failure.
● b ¼ 6 and higher is indicative of age and wear-out as would be expected from
long service life followed by rapid degradation.
There are two useful and unique shape parameters:
● b ¼ 2.0 is the Raleigh probability distribution function used to quantify wind
speed for wind turbine generation systems. Most wind turbine sites around the
world represent a Weibull wind speed distribution having 1.5 < b < 2.5, so a
value of 2.0 is representative and most often used.
● b ¼ 3.44 approximates the Gaussian or normal probability distribution
function.
Additionally, the Weibull distribution can be used to determine the onset of
failure. For example, shape factor, b1, represents expected life when 1% of the
population fails. The second parameter of interest a, the characteristic life or scale
factor, occurs at the 63.2 percentile of the distribution (dashed horizontal line on
Weibull paper). The point where the beta slope line crosses the 63.2 percentile of
population line is the characteristic life.
CH010 7 June 2011; 14:23:44

For historical perspective, Weibull [12] delivered his hallmark paper on the
subject in 1951. He claimed that his distribution applied to a wide range of pro-
blems. His examples ranged from the yield strength of steel to the size of adult
males born in the British Isles. Other examples of where one might use the Weibull
distribution include the following:
● The project engineer reports three failures of a component after being in ser-
vice for 6 weeks. The program manager asks: ‘How many failures are pre-
dicted for the next three months, six months and one year?’ The Weibull
reliability distribution function can answer this, even though so few data points
are available.
● An automobile company chief program engineer may inquire: ‘How many
spare parts of a particular item must be ordered knowing they have a two or
three year lead time. How may his engineer forecast the number of engine
modules that can be expected to be returned to a depot for three to five years
hence month by month?’
● An automobile company program manager may inquire: ‘If the new Engi-
neering Change Order eliminates an existing failure mode, how many units
must be tested for how many hours without any failures to demonstrate with
90% confidence that the old failure mode has either been eliminated or sig-
nificantly improved?’
For the Weibull reliability distribution function, (10.24) through (10.28)
describe the cumulative distribution function, cdf, probability density function, pdf,
reliability or survivor function, R(t), hazard or failure rate function, h(t), and
cumulative hazard function, H(t).
FðtÞ ¼ 1 eðt=aÞ
b
ð10:24Þ
b ðb1Þ ðt=aÞb
f ðtÞ ¼ t e ð10:25Þ
ab
RðtÞ ¼ eðt=aÞ
b
ð10:26Þ
b t ðb1Þ
hðtÞ ¼ ð10:27Þ
a a
t
HðtÞ ¼ b
ð10:28Þ
a
Example 10.3: Rewrite (10.23) for the case of a Rayleigh distribution function of
wind speed having scale factor ‘c’ and shape factor ‘k’ where the pdf is given in
terms of time at speed, h(v). That is, the fraction of time wind speed is between v
CH010 7 June 2011; 14:23:44

and v þ dv. The scale factor c is representative of the proportion of time wind speed
is high.
Solution: Replace the slope and characteristic life parameters in (10.23) with scale
and shape factors for the probability distribution function for wind speed as
k vðk1Þ ðv=cÞk
hðvÞ ¼ e ;0 v < 1 ð10:29Þ
c c
Then set k ¼ 2.0 in (10.29) and simplify
2 v ðv=cÞ2
hðvÞ ¼ e ð10:30Þ
c c
Here, (10.30) is the Rayleigh distribution of wind speed common to most wind
turbine sites.
A useful comparison is to illustrate the Weibull pdf for a fixed-scale factor
a ¼ 2200 h and a range of shape factors, b ¼ 0.5, 1.0, 3.0, and 6.0, as shown in
Figure 10.14.
Weibull pdf for b = 0.5, 1.0, 3.0, 6.0

1.5 × 10–3
1.018 × 10–3
Probability of failure
f1 (k) 1 × 10–3
f2 (k)
f3 (k) 5 × 10–4
f4 (k)
0
0
0 1 × 103 2 × 103 3 × 103 4 × 103 5 × 103
200 t(k) 4.16 × 103
Time (h)
Figure 10.14 Weibull pdf for a ¼ 2200 h and range of shape factors (b ¼ 0.5
lowest exponential, 1.0 linear exponential, 3.0 and 6.0 highest
amplitude)
A carbon–carbon ultracapacitor has a characteristic life a ¼ 1183 h and shape

factor b ¼ 15.7 as shown representatively in Figure 10.15 as a linear plot of its pdf
(10.25) versus life.
CH010 7 June 2011; 14:23:44

MC3000 cell Weibull pdf a = 1183, b = 15.7

5 × 10–3
4.9 × 10–3
4 × 10–3
Probability
3 × 10–3
f(k)
2 × 10–3
1 × 10–3
0
0
10 2
10 2
10 2
10 2
10 3
10 3
10 3
10 3
0
×
3
8
2.
4.
6.
1.
1.
1.
1.
50 t(k) 1.6 × 103
Time (h)
Figure 10.15 Weibull pdf for a representative ultracapacitor
Figure 10.16 shows the reliability function, R(t), for the ultracapacitor cell
having the characteristic life and shape factors listed above. In this case it is
apparent that all units survive until time t 900 h when the characteristic life is
1183 h, but at t ¼ 1350 h, virtually 100% of the cells will have failed.
MC3000 cell survivor function

1
1
Probability of surviving
0.8
0.6
R(k)
0.4
0.2
0
0
10 3
10 3
10 3
10 3
0
0
22
45
67
90
×
5
35
8
12
57
1.
1.
1.
1.
50 t(k) 1.646 × 103

Time (h)
Figure 10.16 Weibull survivor function for a representative ultracapacitor cell
The probability density function and reliability function are excellent repre-
sentations for ultracapacitor reliability. In the next section, our focus shifts back to
mean service life (MSL).
CH010 7 June 2011; 14:23:44

10.3 Mean service life

Mean service life (MSL) can be defined as a component or product delivering its
intended function over a specified period of time under specified conditions. At a
point in time when some parameter of the component or product first falls outside
its defined tolerance, the component or product is deemed to have failed in terms of
its inability to meet specified performance. For ultracapacitors, the defined failure
can be capacitance dropping below a threshold value or its ESRdc increasing
beyond a threshold value. In the case of ultracapacitors, the threshold values
defining a failure are as follows:
● Capacitance fade by 25% relative to its initial value. Initial value is the com-
ponent’s specified minimum capacitance, which is not the same as the pro-
duction mean of the population.
● Resistance, ESRdc, increasing to 100% of its initial value. For ESRdc this
means the initial value, or top of the production spread, increases by 100%.
In all derivations of mean service life, the ultracapacitor capacitance fade
will be the parameter of choice to make life estimates. The production spread of
capacitance is typically þ20%/0% of specified capacitance. This means the
Maxwell K2 3000 F cell, for example, will have an initial capacitance of 3000 F,
even though its mean value is 3300 F. It is very important to recognize that as
capacity fades the population distribution shown graphically as the bell curve in
Figure 10.17 essentially slides along the capacity fade curve with only modest
increase in its variance.
Ci + 20%
Cμ
Ci
–t
Cn C0e τ
Capacitance (F)
a2t2 + a1t + a0
–25%*Ci
b1t + b0
C2y
C3y
Ceol
0 1500 T2y T3y Teol3

t
Figure 10.17 Definition of ultracapacitor capacitance fade
CH010 7 June 2011; 14:23:45

Three facts are clear by inspection of Figure 10.17.

1. Initial capacity fade is rapid and decreases exponentially over the first 1500 h
of service.
2. During the initial portion of service life, in this case 2 years of continuous
duty, the capacitance fade is quadratic.
3. Over the final major stage of service life, capacity fade becomes very linear
until the EOL defined threshold is reached.
Given these observations we can say that MSL is strictly a defined failure
limited process that says nothing about material, process, or production defects and
their related failures. Such failures adhere to the Weibull statistics stated above, and
the reader is encouraged to consult References 10 and 11 for more insights. Rather,
MSL is a tool to predict life under conditions of voltage and temperature stress. It
was shown in earlier examples in this chapter that many life estimation methods
rely strongly on Arrhenius methods of life estimation under conditions of tem-
perature stress for components having specified electrochemical activation energy.
More refined life estimation methods for electrolytic capacitors modified the fun-
damental Arrhenius approach with the inclusion of voltage stress as done by
Arrhenius-Eyring methods. This approach, in fact, was implemented in the treat-
ment of ultracapacitor life prediction but found to err in overestimation when
voltage and temperature stress levels acting in concert were encountered. The error
was greater as voltage and temperature reached their rated values and even worse
for overstress conditions.
Ultracapacitor endurance specifications are a case in point. Endurance spe-
cifies the ultracapacitor life in hours that it can meet under full voltage and full
temperature stress. Endurance is typically 1500 h under such conditions, over
which time capacity fade does not exceed 20%. For real-world applications, this
defined failure due to capacity fade is modified to a 25% loss as shown in
Figure 10.17. Experimental methods require the long-term testing of ultracapacitor
groups, each composed of several cells, being subjected to prescribed voltage and
temperature stress. Half the groups are tested within rated conditions, and half
outside rated voltage and temperature. Actual data reduction methods and results
are manufacturer proprietary and not discussed in this book. Instead, we use the
results to build the MSL estimation method.
Nominal stress is defined as the voltage–temperature pair (U0,T0) ¼ (2.3 V,
40 C) in developments to follow. Let the service life at the nominal point be MSL0,
and at the test point, MSLx, then according to Arrhenius-Eyring theory and its
extension by Prokopowitz and Vaskas [14] we have

Ea 1 1 Ux n
lnðMSLx Þ ¼ lnðMSL0 Þ ln ð10:31Þ
k T0 Tx U0
According to (10.31), the logarithm of mean service life is the sum of MSL at
the nominal point minus Arrhenius and minus voltage stress contributions. Taking
CH010 7 June 2011; 14:23:45

the exponential form of (10.31) results in the more common form of this life esti-
mation function as
MSLx ¼ MSLðU x ; T x Þ ¼ MSLðU 0 ; T 0 ÞeðEa =kÞ½ð1=T 0 Þð1=T x Þ enLnðU x =U 0 Þ

ð10:32Þ
n
Ux
MSLðU x ; T x Þ ¼ MSLðU 0 ; T 0 Þ eðEa =kÞ½ð1=T 0 Þð1=T x Þ ð10:33Þ
U0
For Maxwell Technologies large ultracapacitor cells, the coefficients shown in

(10.33) are
● Ea ¼ 0.5 eV, the activation energy typical of electrochemical capacitors
● k ¼ 8.62 10–5 eV/K, Boltzmann’s constant
● n ¼ 0.6, the voltage power law exponent
It was found through long-term experimental work that (10.33) is not sufficient to
make accurate life estimates of the ultracapacitor. Less error contribution to MSL
was found if the Prokopowitz–Vaskas method was modified to include cross-cou-
pling of voltage and temperature stress. When this approach is taken, (10.31) is
augmented with an additional coefficient and the power law cross-coupling factors
of voltage and temperature stress level.
n a b
Ea 1 1 Ux Ux Tx
LnðMSLx Þ ¼ LnðMSL0 Þ Ln D ð10:34Þ
k T0 Tx U0 U0 T0
For the modification to include cross-coupled voltage and temperature stress,

(10.34) can be cast into an exponential form as
n
Ux
emðU x =U 0 Þ ðT x =T 0 Þb ðEa =kÞ½ð1=T 0 Þð1=T 0 Þ
a
MSLðU x ; T x Þ ¼ MSLðU 0 ; T 0 Þ e
U0
ð10:35Þ
As a result of the cross-coupling terms, the coefficients found for (10.33) no

longer hold and must be readjusted for the contribution of stress cross-coupling.
When this is done, the coefficients necessary for (10.35) to accurately estimate
large cell ultracapacitor life are
● Ea ¼ 0.3 eV, the activation energy typical of electrochemical capacitors
● k ¼ 8.62 105 eV/K, Boltzmann’s constant
● n ¼ 0.6, the voltage power law exponent
● m ¼ 0.023, the cross-coupling factor gain
● a ¼ 16, voltage factor power law coefficient
● b ¼ 17, temperature factor power law coefficient
CH010 7 June 2011; 14:23:45

Cell life estimation from d.c. life test data for MC3000 F
12
11.819
ML.25
11.143
ML.40
MC3000 estimated life (Ln (h))
ML.55 10.286
ML.65
ML.70 9.429
Data.40
8.571
Data.55
Data.70 7.714
Data.75
ML.75 6.857
6
6
1.8 1.9 2.0 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9
1.8 U2 2.9
Cell potential (V)
Figure 10.18 MSL chart for ultracapacitor cells exposed to voltage and
temperature stress
Figure 10.18 is the resulting mean service life estimate for the ultracapacitor
cells when voltage ranges from 1.9 to 2.9 V with temperature as a parameter. In this plot,
the nominal point (2.3 V, 40 C) is located on the prediction curve second from the top.
In Figure 10.18, the data shown modeled by (10.35) is d.c. life testing data
taken over a continuous testing period of 3 years, or close to 30,000 h under stress.
Notice that test data points at eight stress level conditions are approximated very
accurately by (10.35) except for the one group of cells that are at excessive tem-
perature of 75 C, well outside the normal operating temperature range, so even
with the cross-coupling of stress, this model is not capturing the fast capacity fade
at this overtemperature condition. The reason is that unmodeled phenomena are
presenting that accelerate capacity fade faster than prediction. An example will
help clarify the method and results presented in Figure 10.18.
Example 10.4: Given that MSL0 ¼ 70,000 h, or ln(MSL0) ¼ 11.156 on the chart,
compute the service life when the voltage is raised to 2.7 V at 40 C.
Solution: This is a direct application of (10.35), but with no temperature variation,
only the influence of voltage stress. For an increase in operating voltage from
U0 ¼ 2.3 V to Ux ¼ 2.7 V, the predicted service life becomes
n
Ux
emðU x =U 0 Þ ðT x =T 0 Þ eðEa =kÞ½ð1=T 0 Þð1=T x Þ
a b
MSLðU x ; T x Þ ¼ MSLðU 0 ; T 0 Þ
U0
16
¼ 70; 000ð1Þe0:023ð2:7=2:3Þ ¼ 47; 128 h
Or taking the natural logarithm of MSLx to yield ln{MSLx} ¼ 10.761.
CH010 7 June 2011; 14:23:45

Referring to Figure 10.18, it is apparent that this is very close to the data point
(2.7,40) for which MSL ¼ 10.741. The reduction is therefore a factor of 1 (0.908)
(0.7415) ¼ 1 0.6733 ¼ 0.327. The life reduction by increasing operating voltage
from 2.3 to 2.7 V at a temperature of 40 C is 32.7% less life.
The reader is referred to Exercise 10.5 for an illustration of what a corresponding
increase in temperature means to service life, all else being equal. It can be seen that
for a 17% increase in operating voltage, the life reduces by approximately 33%. For a
37% increase in operating temperature, the life reduction is found to be 43% at a
nominal voltage of 2.3 V. If the same temperature change is taken at a higher voltage
of 2.7 V, it is found that life reduces from the 47,128 h found in Example 10.4 to
19,489 h, or somewhat over 58%. This shows the need for cross-coupling and how
voltage and temperature stress interact at their extremes to cause higher life reduction
than what voltage and temperature alone would predict. Summarizing
● MSL(2.3 V, 40 C) ¼ 70,000 h?MSL(2.7 V, 40 C) ¼ 47,128 h, a 32.7% life
reduction
reduction
reduction
10.4 Comprehensive cycle life testing

Just as there was more to ultracapacitor service life than the stress factors of voltage
and temperature acting independently, the same can be said of estimated life
derived from d.c. life assessment as discussed in section 10.3 being modified for the
case of cycling. When the ultracapacitor is cycled, its voltage transitions between
rated and half rated according to the charge transfer conditions of applied current.
For cycling conditions the voltage is no longer stationary at one level but spends
time at different levels. The service life under such real-world operating conditions
is not so straightforward as the developments leading to (10.35) might indicate. The
stresses are no longer stationary processes but can be highly variable.
To illustrate the methodology of life estimation under real-world conditions, an
example is useful to gain necessary insights. For this example, a hybrid transit bus
that operates over a fixed route in a large metropolitan area is considered. The bus
operates 2 shifts per day, 6 days per week with down time of only 8 days per year
for maintenance, including a couple of holidays. The ambient temperature over a
year is given statistically, and the ultracapacitor modules are packaged in an area of
the bus so that ambient air provides some cooling. No excessive stress levels are
otherwise imposed. So, let’s proceed to this example.
Example 10.5: A hybrid transit bus in revenue service operates for T ¼ (2/3 day)
(24 h/day)(6 days/week)(51 weeks/year) ¼ 4896 h/year. Therefore, the ultracapacitor
modules are not energized for 3864 h/year, or 44% of the time. When the bus is in
service, it makes 200 stop–go maneuvers per day, or 200 cycles in 16 h for a
CH010 7 June 2011; 14:23:46

charge–discharge period on average of 240 s. The ultracapacitor charge–discharge

period can therefore be represented as a trapezoidal waveform that can further be
assumed to range from 0.6Umx to 0.95Umx, a voltage swing of 0.35 pu. Figure 10.19
illustrates the voltage swing and the yearly ambient temperature excursions. Vol-
tage excursions for the life estimation are compiled in Table 10.1, and temperature
excursions are tabulated in Table 10.2.
Solution: The solution will be carried out graphically to quantify voltage and
temperature ranges, then analytically.
To keep this analysis of composite MSL tractable, the percent time at voltage
has been compiled into four bins and annual temperature has been partitioned to
0.95 Umx 2.6 V/cell
0.775 Umx 2.1 V/cell
0.60 Umx 1.62 V/cell
0 t (s)
0 100 150 258 288
Loading acceleration Cruise deceleration
Phoenix AZ annual temperature (°C) 8736 h

1600
1400
1200
Cummulative time (h)
1000
800
600
400
200
0
0 5 10 15 20 25 30 35 40 45
Temperature bins (°C)
Figure 10.19 Ultracapacitor module voltage swing and temperature environment

of hybrid transit bus
CH010 7 June 2011; 14:23:46

Table 10.1 Voltage bin data for transit bus operating

in Phoenix, Arizona
Cell voltage Percent time Yearly percent

bin energized when ON time at voltage
2.6 34.7 20
2.1 28 15
1.6 37.3 21
0 44
Total 100 100
Table 10.2 Temperature histogram data for Phoenix, Arizona
Temperature T(t) estimated Percent time at Yearly percent at

bin ( C) hours temperature temperature
5?0 0 0 10.6 at 5 C
0?5 154 1.76
5?10 775 8.8
10?15 1547 17.7 78 at 25 C
15?20 1358 15.5
20?25 1246 14.2
25?30 1519 17.4
30?35 1246 14.2
35?40 789 9.0 10.4 at 40 C
40?45 120 1.37
> 45 6 0.068 1
Total hours 8760 100 100
three bins. The <1% at temperature greater than 45 C has been neglected. The
fact that ultracapacitor voltage resides in one of four bins and temperature in one of
three bins means that MSL must be calculated using (10.35) for 12 cases. Knowing
that MSL will change little for voltage stress below 1.9 V, the lowest two voltage
stress cases are lumped together as 65% of time at 1.3 V/cell. The temperatures
indicated in Table 10.2 are ambient temperatures, and the cells are being exercised
by the transit bus in service with air cooling. Therefore, as an approximation for
this example, the cell core temperature is taken as 5 C above ambient, making the
three temperature bins of interest: 10, 30, and 45 C. The cases and their solution for
large cell ultracapacitors such as the 3000 F cell are set up.
● MSL(2.6 V, 10 C) ¼ 200,007 h for 10.6% of time by year and 20% by
voltage?0.021
● MSL(2.1 V, 10 C) ¼ 228,661 h for 78% of time by year and 15% by voltage?
0.0159
● MSL(1.3 V, 10 C) ¼ 252,205 h for 10.4% of time by year and 65% by voltage?
0.069
CH010 7 June 2011; 14:23:46

0.156
0.117
● MSL(1.3 V, 30 C) ¼ 111,972 h for 10.4% of time by year and 65% by
voltage?0.507
0.021
0.0156
0.0676
It is far easier to view these composite results for time at temperature and time
at voltage by putting this data into a matrix so that the nine stress points can be
easily grasped. In the bullet list above, the final entry per row is the composite time,
or per unit time, at temperature and voltage for that MSL value.
Table 10.3 is the basic set of stress points and weight factor at each of the nine
stress points. A composite mean service life (MSL) for the hybrid transit bus
ultracapacitor pack is next condensed by taking the harmonic average of all the
stress points and weights from Table 10.3. The result is a single number, an MSL
for the ultracapacitor application for the given conditions and stress survey.
Table 10.3 Matrix of stress levels, time (h) at composite stress

level, and estimated life
Temperature Matrix of pu stress condition

45 C 0.0676 (65,120) 0.0156 (54,176) 0.021 (43,608)
30 C 0.507 (111,972) 0.117 (101,722) 0.156 (84,542)
10 C 0.069 (252,205) 0.0159 (228,661) 0.021 (200,007)
Voltage 1.3 V 2.1 V 2.6 V
1
MSLcomp ¼
ðW 1 =MSL1 Þ þ ðW 2 =MSL2 Þ þ ðW 3 =MSL3 Þþ
ð10:36Þ
ðW 4 =MSL4 Þ þ ðW 5 =MSL5 Þ þ ðW 6 =MSL6 Þþ
ðW 7 =MSL7 Þ þ ðW 8 =MSL8 Þ þ ðW 9 =MSL9 Þ
1
MSLcomp ¼
ð0:0676=65;120Þþð0:0156=54;176Þþð0:021=43;608Þþ
ð10:37Þ
ð0:507=111;972Þþð0:117=101;722Þþð0:156=84;542Þþ
ð0:069=252;205Þþð0:0159=228;661Þþð0:021=200;007Þ
CH010 7 June 2011; 14:23:47

106
MSLcomp ¼ 106
1:038 þ 0:2879 þ 0:4816 þ 4:528 þ 1:1502 þ ¼ ¼ 102; 262 h
9:7788
1:845 þ 0:2736 þ 0:0695 þ 0:105
ð10:38Þ
The final answer is (10.38), an expected life for the ultracapacitor pack
in a hybrid transit bus application under specified drive cycling conditions
and yearly ambient temperatures of 102,262 h, or 11.67 years. This is the service
life estimate before the ultracapacitor cells have lost 25% of their initial capaci-
tance in an environment where voltage and temperature stress are real-world
conditions.
Example 10.5 can of course be refined to even smaller voltage and tem-
perature bins as needed for accuracy, but the end result is the same: a composite
mean service life can be estimated by using time at voltage and time at tem-
perature as weight factors for each stress point estimated life. At the time of this
writing experimental work was under way to characterize the influence of power
converter ripple current on ultracapacitor parameter fade and long term service
life [15]. The need for such testing is apparent when one considers that modern
power converters and inverters rely on electrolytic and polymer film ripple filter
capacitors in their dc-link and that these filtering components bypass a good deal
of dc source ripple, but not all. The extent that power converter ripple current
passes through the battery, ultracapacitor or battery plus ultracapacitor combi-
nation energy storage system and causes electrochemical deterioration in
hybrid electric, plug-in hybrid electric and battery electric vehicles remains as a
research topic.
Exercises
10.1 Using (10.30) for the Rayleigh distribution function for wind speed, (A)
compute the average wind speed V0 and (B) h(v) for this case.
Ð1 Ð1 2
Ans: (A) For this case, V 0 ¼ 0 xhðxÞdx ¼ c22 0 x2 eðx=cÞ dx
pffiffiffi of wind speed scale factor, c, results in
Solving for V0 in terms
c ¼ p2ffiffipffi V 0 ; V 0 ¼ 2p c
(B) Substituting the value for c into (10.30) results in

p v 2 v 2
hðvÞ ¼ eðp=4Þ eðv=V 0 Þ ¼ 0:72 eðv=V 0 Þ
2 V0 v0
10.2 From Exercise 10.1 part (A), what is the relative scale of the Rayleigh dis-
tribution function scale factor and the computed average wind speed for this
case?
Ans: V0 ¼ 0.88c, or to a first order, the average wind speed in a Rayleigh
distribution function is the scale factor of the Rayleigh distribution function.
CH010 7 June 2011; 14:23:48

10.3 Using Figure 10.16 and (10.25), compute the reliability function, R90, at
which 10% of the units have failed given the characteristic time and slope
factors used in the derivation of Figure 10.16.
Ans: For this case, p ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
R90 ¼ 0:90 ¼ eðt=aÞ ;
b
t
a ¼ b
lnðR90 Þ; t ¼ 0:8665a ¼ 1025 h
10.4 A class of carbon–carbon ultracapacitors are found to adhere to Weibull
reliability statistics with a shape factor, b ¼ 5, and an R10 survivor level
when t ¼ 105,000 h. What is the characteristic life of these ultracapacitors
for an application that requires 10 year life before 63% have failed?
Ans: Follow the procedure outlined in Exercise 10.3 to obtain
t 105; 000
R10 ¼ 0:10; a¼p ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ p ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ 88;870 h; or 10 years
b
lnðR10 Þ 5
lnð0:1Þ
10.5 Repeat Example 10.4 for the case of a 33% increase in temperature from the
nominal point (2.3 V, 40 C) to (2.3 V, 55 C), and compare the resulting life
reduction to that found in Example 10.4 for a 33% increase in cell-operating
potential. In this exercise, MLS0 ¼ 70,000 h, and all coefficients are as given
in the example.
Ans:

U x n mðU x =U 0 Þa ðT x =T 0 Þb ðEa =kÞ½ð1=T 0 Þð1=T x Þ
MSLðU x ; T x Þ ¼ MSLðU 0 ; T 0 Þ e e
U0
and substituting coefficients and values
0:6
2:3 16 17
MSLð2:3; 55Þ ¼ MSL0 e0:023ð2:3=2:3Þ ð328=313Þ e3480½ð1=313Þð1=328Þ
2:3
¼ 70;000ð0:95Þð0:6014Þ ¼ 39;993 h
ln(MSLx) ¼ 10.596, and the data point for this test group is 10.586, a very
good fit. The life reduction for this 33% increase in operating temperature
results in a reduction of 42.9%, which is somewhat larger life loss than that
for a corresponding voltage increase.
10.6 Demonstrate that the same result given in (10.37) is obtained when the nine
MSLx values listed are given as natural logarithms.
Ans: For this case, (10.37) is written as
1
LnðMSLcomp Þ ¼
ð0:0676=11:084Þ þ ð0:0156=10:899Þ þ ð0:021=10:683Þþ
ð0:507=11:626Þ þ ð0:117=11:530Þ þ ð0:156=11:345Þþ
ð0:069=12:438Þ þ ð0:0159=12:339Þ þ ð0:021=12:206Þ
103
¼ ¼ 11:688
85:55
Therefore, taking exp(11.688) ¼ 119,216 h, reasonably close counting for
round-off errors.
CH010 7 June 2011; 14:23:48

References
1. R. Kotz, M. Hahn, R. Gallay, ‘Temperature behavior and impedance funda-
mentals of supercapacitors’, Journal of Power Sources, vol. 154, pp. 550–55,
2006
2. H. Gualous, H. Louahlia-Gualous, R. Gallay, A. Miraoui, ‘Supercapacitor
thermal modeling and characterization in transient state for industrial appli-
cations’, IEEE Transactions on Industry Applications, vol. 45, no. 3, pp.
1035–44, 2009
3. V. Anand Sankaran, F.L. Rees, C.S. Avant, ‘Electrolytic capacitor life testing
and prediction’, IEEE 32nd Industry Applications Society Annual Meeting,
Sheraton Hotel, New Orleans, LA, vol. 2, pp. 1058–65, 5–9 October 1997
4. V. Anand Sankaran, C. Chen, C.S. Avant, X. Xu, ‘Power cycling reliability of
IGBT modules’, IEEE 32nd Industry Applications Society Annual Meeting,
Sheraton Hotel, New Orleans, LA, vol. 2, pp. 1222–27, 5–9 October 1997
5. J.R. Miller, I. Goltser, S. Butler, ‘Electrochemical capacitor life predictions
using accelerated test methods’, Proceedings of the 42nd Power Sources
Conference, Philadelphia, PA, pp. 581–4, 12–14 June 2006
6. O. Briat, W. Lajnef, J-M. Vinassa, E. Woirgard, ‘Power cycling tests for
accelerated ageing of ultracapacitors’, Microelectronics Reliability vol. 46,
pp. 1445–50, 2006
7. T. Kawaji, A. Okazaki, ‘Reliability evaluation for electric double layer
capacitor’, The 23rd International Battery, Hybrid and Fuel Cell Electric
Vehicle Symposium, EVS23, Anaheim, CA, 2–5 December 2007
8. J-Y. Kim, S-J. Jang, B-K. Lee, C-Y. Won, C-M. Lee, ‘A calculation of pre-
dicting the expected life of super-capacitor following current pattern of rail-
way vehicles’, The IEEE 7th International Conference on Power Electronics,
ICPE2007, Daegu Conference Center, Daegu, Korea, pp. 978–83, 22–26
October 2007
9. G. Yang, Life Cycle Reliability Engineering, John Wiley & Sons, Inc.,
Hoboken, NJ, 2007
10. J.R. Miller, ‘Reliability assessment and engineering of electrochemical
capacitors’, The 18th International Seminar on Double Layer Capacitors and
Hybrid Energy Storage Devices, Embassy Suites Deerfield Beach, Deerfield
Beach, FL, 8–10 December 2008, invited tutorial
11. J.R. Miller, A.D. Klementov, S. Butler, ‘Reliability investigation of 3000F
Maxwell electrochemical capacitor cells’, The 18th International Seminar on
Double Layer Capacitors and Hybrid Energy Storage Devices, Embassy
Suites Deerfield Beach, Deerfield Beach, FL, 8–10 December 2008
12. R.B. Abernethy, J.E. Breneman, C.H. Medlin, G.L. Reinman, Weibull Analysis
Handbook, Pratt and Whitney, West Palm Beach, FL. Government Products.
Report supplied by Storming Media, now out of print. November 1983
13. W. Nelson, Accelerated Testing: Statistical Models, Test Plans, and Data
Analysis, Wiley-Interscience Publication, New York, 1990
CH010 7 June 2011; 14:23:49

14. T.I. Prokopowitz, A.R. Vaskas, Research and Development Intrinsic Relia-
bility, Subminiature Ceramic Capacitors, Final report, ECOM-9075-F, NTIS
A0-864068, 1969
15. J.M. Miller, P.J. McCleer, ‘Electrical and thermal investigation of power
electronic converter ripple current on the ultracapacitor’, The 20th Interna-
tional Seminar on Double Layer Capacitors and Hybrid Energy Storage
Devices, Embassy Suites Deerfield Beach Hotel, Deerfield Beach, FL, 6–8
December 2010
CH010 7 June 2011; 14:23:49

CH010 7 June 2011; 14:23:49
Chapter 11
Abuse tolerance
The chapter focuses on ultracapacitor abuse. But what exactly constitutes abuse
conditions? The answer is one or a combination of the following factors: voltage,
temperature, vibration, shock, extreme current, such as short circuit, and reverse
polarity. Clinton Winchester [1] of Naval Surface Warfare Center (NSWC)
Caderock division, Maryland, presented results of ultracapacitor cell overvoltage
and overtemperature abuse testing performed for the application of these cells in
Navy equipment. This reference on abuse testing is an excellent introduction to the
topic because it illustrates the response these cells have to abuse.
The abuse testing presented in Reference 1 was prefaced by proper preparation
of the test cells, in this case Maxwell Technologies 3000 F ultracapacitors.
Figure 11.1 illustrates the cell preparation method. First, the cell is wrapped with
thermal tape and thermocouples are installed on center of cell and at terminal posts.
Second, the cell is covered with insulation to obtain near adiabatic conditions, and
lastly, the cell is connected to laboratory instrumentation. After this, the testing
begins and electrical and thermal responses are measured and logged using data
acquisition equipment.
11.1 Need for abuse testing

All the applications covered in this book assume that the ultracapacitors design-in
process had been completed correctly and that its service life is in accordance with
the prescribed conditions. But sometimes, either a part malfunctions causing an
overstress on adjacent parts or a system fault occurs causing excessive voltage or
current to be present in the cell. Abuse testing is necessary to determine how an
ultracapacitor responds to such abuse and for how long it can tolerate abuse of a
given magnitude. DeJarnette et al. [1] decided to subject the ultracapacitor to
overtemperature conditions by externally heating the cell with the thermal wrap as
shown in Figure 11.1b. This is an example of thermal ramp and its results are
shown in Figure 11.2.
The thermal tape when electrically powered was capable of raising the
temperature of the ultracapacitor cell mass to 700 C, well beyond the temperature
at which internal pressure built up sufficiently (*15 bar) to vent. A vent in fact
occurs when the cell temperature reaches 200 C. This condition took
ch011 6 June 2011; 18:37:54

(a)
(b)
(c)
Figure 11.1 Preparation of cell for abuse testing: (a) selected cell, (b) wrapped
with thermal tape, and (c) insulation covering applied
Thermal ramp
800
700
600
500
T (°C)
400
300
200 Vents
100
0
0 10 20 30 40 50 60 70 80
t (min)
Figure 11.2 Ultracapacitor cell (3000 F) subjected to thermal ramp
ch011 6 June 2011; 18:37:55

Abuse tolerance 313
approximately 42 min. Since this process is nearly adiabatic, the thermal power
input to the cell can be computed as (11.1).
ð tf
1 Pd t f
T ðtÞ ¼ Pd dt ¼ ð11:1Þ
C th 0 C th
For this externally applied heat source, the power that is transmitted into the
cell mass under adiabatic conditions can be calculated using (11.1), knowing
tf ¼ 2520 s, T(tf) ¼ 200 C, and estimating Cth for the cell as 600 J/K. Using these
quantities, the injected power and energy would need to be
T ðtf ÞC th 200 Cð600 J=CÞ

Pd ¼ ¼ ¼ 47:6 W ð11:2Þ
tf 2520 s
W d ¼ Pd tf ¼ 47:6ð2520Þ ¼ 120 kJ ð11:3Þ

When this amount of energy is being injected into the cell, its pressure rises
until the cell fuse activates and the electrolyte is expelled. DeJarnette et al. [1]
investigated the nature of the vented liquids and gases. Suffice to say that when an
ultracapacitor is subjected to such extreme abuse, a large fraction of its electrolyte
will be expelled. However, this topic is outside the scope of this book.
11.2 Overvoltage and overcurrent abuse

This section considers two particular abuse conditions: first, an application of
voltage beyond the cells rated working voltage, and second, overcurrent abuse. The
previous section discussed abuse tolerance to external thermal source,
but DeJarnette et al. [1] also executed an overvoltage test on the same cells.
Figure 11.3 summarizes the overvoltage abuse test.
In this test, the following sequence of events are captured (Figure 11.3):
● Charging current of 3 A is applied until the cell enters an overvoltage state of
3.1 V, then the current is relaxed in 60- to 80-min interval as shown in the figure.
● Cell voltage holds at approximately 3.1 V for 70 s, then at approximately
80 min the voltage had to be increased to 6 V in order to maintain a constant
3 A current. This interval lasts another 70 s until the current is increased.
● At approximately 150 min when the current is increased to 6 A, the cell voltage
rises slowly, then faster as the current is increased a final time to 9 A.
● As the cell voltage rises past 5 V, the cell temperature approaches 80 C and its
vent opens expelling electrolyte. The test is then terminated.
For this overvoltage abuse test, the 3000 F cell tolerated substantial voltage abuse
for nearly 220 min (13,200 s) before venting. The power dissipated internally is now
substantially lower than was injected by the external ramp test. In fact, with a cell
potential of 3.1 V and 3 A of leakage current, the power dissipated internally is only
9.3 W. The temperature response of the cell can be found using (11.1). In this case,
ch011 6 June 2011; 18:37:55

Overvoltage
U (V)
I (A)
110
100 10
Vents
80 8
T (°C)
V
60 6
T U
I
40 I 4
20 2
0
0 50 100 150 200 250 300
t (min)
Figure 11.3 Overvoltage abuse test results on 3000 F cell
T(t) ¼ (Pd/Cth)t ¼ 0.0155t. If this power were all dissipated internally, which is not for
such a slow experiment, the cell temperature would reach T ¼ 0.0155(13,200 s) ¼
204.6 C. This is surprisingly close to the temperature found in Figure 11.3, but is not
correct. Measurements made indicate that the cell temperature reached approximately
80 C. Therefore, a considerable amount of the input power is being conducted out,
possibly by the leads connected to the cell terminal posts.
To circumvent this error in power assessment by heat conduction and con-
vection from the cell during the test, a much faster test can be performed. In this
case, suppose the same cell is subjected to an externally applied overcurrent of 100 A.
This would be very close to an adiabatic condition for the cell because the time
for internal pressure to reach fuse opening would be much shorter. Figure 11.4 is
an illustration of what to expect for sustained high overcurrent.
Constant current
T(t) 100
I0 = 100 A
80
T (°C)
Pd Cth 60 T(t)
40
P(t)
20
Ta 0
0 1 2 3 4 5
t (min)
Figure 11.4 Ultracapacitor overcurrent abuse and near adiabatic response
ch011 6 June 2011; 18:37:55

Abuse tolerance 315
As the schematic in Figure 11.4 indicates, the short test time can be modeled as
a power dissipation source internal to the cell and the cell mass thermal capaci-
tance. As power flows into the thermal capacitance as shown in Figure 11.4, its
temperature increases linearly with time as per (11.1), also shown in Figure 11.4. In
this example, it takes a 3000 F cell with t ¼ 81 s or just over a minute to fully charge
when subjected to I0 ¼ 100 A of constant current. The shallow initial slope of the
T(t) response depicted in Figure 11.3 is due to cell charging, then entering an
overvoltage condition by 2 min, and finally, the linear rise of temperature due to the
heating effect of 100 A of leakage current in the cell.
The cell undergoing overcurrent abuse test in Figure 11.4 has an approximate
internal power dissipation and heat energy accumulation due to extreme leakage
current of
C th T 600ð70Þ
Pd ¼ ¼ ¼ 233 W ð11:4Þ
tf 180
W d ¼ Pd tf ¼ 233ð180Þ ¼ 41:9 kJ ð11:5Þ
The amount of fast internal energy dissipation given by (11.5) is sufficient to

build internal pressure to activate the cell fuse. Figure 11.5 is an example of abuse
testing of a BCAP0350 cell to failure. In these tests, both overvoltage at 4 V and
overcurrent to 20 A were sufficient to fail the pressure fuse in just a few minutes.
Cell internal temperature builds to approximately 100 C, well beyond the 83 C at
which the electrolyte solvent vaporizes.
BCAP0350
Figure 11.5 Illustration of cells abuse tested to failure (350 F ultracapacitors,

D cell package)
Cyrus Ashtiani [2] presented results on automotive expectations of ultra-

capacitors in mobile systems from the perspective of overvoltage, overcurrent and
overtemperature abuse. One outcome of that work was to place ultracapacitors in
the hazard level context of batteries as outlined by the European Center for
ch011 6 June 2011; 18:37:55

Automotive Research (EUCAR). Figure 11.6 summarizes the hazard levels of the
EUCAR scale and what this means to ultracapacitor installations.
Hazard level
0 1 2 3 4 5 6 7
Loss of functionality
Defect
Leakage
Venting
Fire or flame
Rupture
Damage
Repair of protection device needed
Irreversible damage, repair needed
Reaction or thermal runaway
Electrolyte loss
Electrolyte loss > 50%
Flying parts
Ejection of active mass
Explosion
Disintegration
Accept Reject
Figure 11.6 EUCAR Hazard Level scale
Ultracapacitors tested to abuse by overvoltage, overtemperature and sudden

polarity reversal exhibited the following types of failure:
● No: Hazard Level 7 (disintegration or explosion).
● No: Hazard Level 6 (case rupture with ejection of active mass and flying parts).
● No: Hazard Level 5 (flames and fire observed).
● Yes: Hazard Level 4 (electrolyte loss of > 50% weight, but no reaction or
thermal runaway).
● Yes: Hazard Level 3 (loss of electrolyte and loss of functionality occurred in
several tests).
The bottom line on all the abuse examples given is that ultracapacitors can
tolerate abuse for a considerable amount of time. What was not presented is the fact
that if the abuse condition is removed before cell temperatures approach 100 C
then the cell can recover and resume working, albeit with some irreversible loss of
capacitance.
ch011 6 June 2011; 18:37:55

Abuse tolerance 317
11.3 Insulation resistance and high potential

International standards exist so that component and product manufacturers
can certify their products for compliance with internationally acceptable perfor-
mance levels. Energy storage modules for application to passenger vehicle,
transit bus, and metro rail, for example, must pass test standards for safety and
performance. Insulation resistance (IR) or dielectric withstand potential is one
such measure designed to ensure that products designed to operate in high-
voltage customer equipment do not pose shock hazard to users and mainten-
ance personnel. Capacitors are governed by U.S. Underwriters Laboratory
UL810 and electrochemical capacitors by UL810A, and the International Elec-
tro-technical Commission (IEC), such as IEC60077. Table 11.1 is an excerpt
from IEC60077 for dielectric withstand potential of ultracapacitor modules by
voltage class.
Table 11.1 Dielectric withstand potential, a.c. excitation 40–70 Hz, for 60 s
IEC voltage <36 V 36–60 V 60– 300– 660–1200 V 1200 V to

range 300 V 660 V 10 kV
Withstand 750 V 1000 V 2Ud þ 1 kV 2 Ud þ 1.5 kV 2Ud þ 2.5 kV
voltage
Typical 16 V 48 V HEV HEV, Transit bus, Metro rail,
application module module PHEV, light rail high
@ Ud BEV speed rail
maximum
For example, Maxwell Technologies Heavy Transportation Module, HTM125,

is typically used in transit bus and light rail at 700 and 1100 Vdc so that the module
is HiPot tested per UL810A to 4 kVac for 1 min. Figure 11.7 shows the equipment
setup for a HiPot test during module manufacturing. In a HiPot test, the internal
ultracapacitor cell block terminals are tied together at the module posts and any
internal electronics are grounded and tied to the module case. HiPot equipment
ground terminal is then connected to the module case and its hot terminal is con-
nected to the ultracapacitor pack terminals. The intent of the test is to validate the
integrity of the ultracapacitor pack cell block insulation system to the aluminum
heat spreaders and chassis.
For the ultracapacitor module, the correct procedure of testing is to first make
the connections then to dial-up the voltage on the test unit from 0 to 2.5 kVdc (or
1785 Vac) no faster than a 10-s ramp charge and 10-s ramp discharge. The reason
for this is that just touching 2.5 kV to the terminals would cause a large inrush
current to charge the module stray capacitance (Table 11.2), resulting in an inac-
curate current reading.
ch011 6 June 2011; 18:37:55

Figure 11.7 HiPot test configuration performed at a manufacturing facility.

BMOD0165-P048 module certified for Ud ¼ 750 Vdc and IR of
2.5 kVdc
Table 11.2 Module parasitic capacitance and

inductance for 3000 F cell modules
Module Number Cs Ls
rated cells (nF) (nH)
voltage (V)
16 6 2.2 372
48 18 3.3 1116
125 48 4.5 2976
Example 11.1: If the 48 V ultracapacitor module shown in Figure 11.7 is HiPot

tested at 2.5 kVdc and the operator just touches the hot lead to the module terminal,
then would the equipment register a fault if the certification level is current not to
exceed 2 mA?
Solution: Given that the stray capacitance of the BMOD0165-P048 module is
Cs ¼ 3.3 nF, and assume that the operator touching the terminal makes contact in
20 ns. Given these conditions, the dV/dt ¼ 0.7576 V/ns, or 757.6 V/ms.
dV
I stray ¼ C s ¼ ð3:3 109 Þð757:6 V =msÞ ¼ 2:5 A ð11:6Þ
dt
ch011 6 June 2011; 18:37:56

Abuse tolerance 319
This exceeds the maximum threshold by a factor of over 1000. Therefore, the best
practice is to slowly ramp the HiPot tester voltage, slowly but continuously over
the 10 s.
The previous example brings out an important point and that is the stray
capacitance of modules. Experimental testing verifies that the common modules
applied in customer applications have the following parasitic capacitance, Cs. The
parasitic inductance is an approximate value determined from the modeled induc-
tance for interconnect straps. This inductance for a 3000 F cell strap is approxi-
mately 22 nH and that of the cells themselves is approximately 40 nH.
A convenient way to view the locations of these parasitic elements is to con-
sider an NsS MP Ccell module as shown in Figure 11.8, where stray capaci-
tance, Cs, is from cell interconnects and cell sides to metal case components;
inductance, Ls, is in series with the internal cell capacitance; and insulation resis-
tance is from the cell casing to module chassis metal.
Ic, Uc
IR1
Cs1
IR2
Cs2
IR_N M
Cs_N M
Ta
Gnd
Figure 11.8 Module parasitic elements
Each of these parasitic elements will now be discussed in detail. Parasitic, or

stray capacitance, Cs, consists of two main elements: (1) stray capacitance of the
interconnect straps to heat spreader metal of the chassis, Csh, and (2) stray capa-
citance of the can sides to metal chassis, Cs. These stray capacitances are illustrated
in Figure 11.9 as capacitance from interconnect straps via thermal insulation
ch011 6 June 2011; 18:37:56

Csh1 Csh2 Csh3
Cs1
Cs2
Cs3
Csk Cs(k + 1)
Figure 11.9 Stray capacitance of ultracapacitor module
material to the metal heat spreader and capacitances from the metal case to the
cylindrical cans.
Stray capacitance of the interconnect straps, shown in Figure 11.9, to an alu-
minum heat spreader can be calculated using the classical formula for a parallel
plate capacitor where the dielectric is taken as a 0.75-mm thick elastomer thermal
conducting insulator having relative permittivity er * 5. For this case the stray
capacitance to heat spreader, Csh, is
er e0 Astrap
C sh ¼ ð11:7Þ
telast
Consider the case when the strap has dimensions L ¼ 90 mm and W ¼ 20 mm,
so Astrap ¼ 1800 mm2.
er e0 Astrap 5ð109 Þð1:8 103 Þ

C sh ¼ ¼ ¼ 0:106 nF ð11:8Þ
telast ð0:75 103 Þ36p
The value of Csh given by (11.8) is per interconnection strap for the dimensions
provided. Next, the stray capacitance from ultracapacitor cylindrical can sidewalls
to the metal chassis aluminum is estimated. For this case the geometry given in
Figure 11.10 is analyzed.
ch011 6 June 2011; 18:37:56

Abuse tolerance 321
(a)
s Aluminium can
rc
dq
s
d
d s
d d1 Aluminium
chassis wall
p ds ds ds −p
q=
2 2
s = rc dq
ds = s cosq = rc cosq dq
d = d1 + rc (1 − cosq )
(b)
Figure 11.10 Geometry of ultracapacitor can to chassis stray capacitance:

(a) portion of an ultracapacitor module and (b) the geometry of can
sidewall to chassis to be solved
Given an ultracapacitor can sidewall length, Lc, and the definition of the
differential linear area, ds, shown in Figure 11.10b, an expression can be given for
differential capacitance of each differential strip along the chassis wall to obtain
total stray capacitance, Cs, as
ð p=2 s ð p=2
rc cos q
C s ¼ er e0 Lc d ¼ er e0 Lc dq ð11:9Þ
p=2 d d
p=2 1 þ r c ð1 cos qÞ
The solution to the differential capacitance along the ultracapacitor module

sidewall for this illustration neglects the mixed permittivity of air and dielectric
ch011 6 June 2011; 18:37:56

material used and assumes that only the can polymer dielectric, or a
Nomex material such as Dupont Type 994 Nomex paper, is used. Equation (11.9)
can be rewritten in a simpler way by considering only the integral part and solving
it first.
ð p=2 ð p=2
rc cos q cos q
dq ¼ dq ð11:10Þ
p=2 d 1 þ rc ð1 cos qÞ p=2 ð1 þ d 1 =rc Þ cos q
In (11.10), the definite integral limits can be changed by invoking symmetry

and the constant in the denominator written as A ¼ (1þd1/rc), note that A > 0 in all
cases. The reader is challenged to derive the answer to the definite integral analy-
tically, which is nontrivial, or to use a comprehensive table of integrals and to
check this with a numerical solver. For the illustration geometry shown in
Figure 11.10b, the can-to-wall spacing d1 ¼ 1 mm, and for a standard 60 mm can,
rc ¼ 30 mm. This means that A ¼ 1.0333 > 0 as given. MathCAD Maple solver
gives the definite integral over 0–p as 9.334 so that value will be used as a check on
the analytical approach.
ð p=2 ðp
cos q cos q
dq ¼ dq
p=2 ð1 þ d 1 =rc Þ cos q 0 ð1 þ d 1 =rc Þ cos q
ðp ð11:11Þ
cos q
¼ dq
0 A cos q
Using integral 3.613 from Reference 3, it is shown that a definite integral of

the form in (11.12) can be modified to fit the expression in (11.11) as follows:
ðp
cos nx p
dx ¼ 2 ; for a2 > 1 ð11:12Þ
0 1 2a cos x þ a 2 ða 1Þan

d1 1 þ a2
A¼ 1þ ¼ ð11:13Þ
rc 2a
Solving (11.13) for ‘a’ in terms of the dimensions of Figure 11.10b geometry
results in
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
!
2 u
u d 1 d 1 2
d 1 d 1 d 1 t 2 þ
ða2 1Þ ¼ 4 þ 2 þ2 1þ ð11:14Þ
rc rc rc rc rc
ch011 6 June 2011; 18:37:57

Abuse tolerance 323
Substitute (11.14) into (11.12) for the case n ¼ 1 and finally substitute the
resulting expression into (11.9) as the final solution.
2p
C s ¼ er e0 Lc
ða2 1Þ
2per e0 Lc
¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð11:15Þ
4ðd 1 =rc Þ þ 2ðd 1 =rc Þ2 þ 2ð1 þ d 1 =rc Þ ð2ðd 1 =rc Þ þ ðd 1 =rc Þ2 Þ
Solving (11.15) for the 3000 F cell geometry yields

9
2p 10 2p
C 0s ¼ er e0 Lc 2 ¼5 ð0:138Þ ¼ 0:0569 nF ð11:16Þ
ða 1Þ 36p 0:6735
The results are summarized in Table 11.3.
Table 11.3 Module parasitic capacitance calculations for

3000 F cell modules
Module rated Number C sh þ C 0s ðnFÞ Cs

voltage (V) cells, Nc (nF)
16 6 ¼ N c C sh þ ðN c þ 2ÞC 0s 1.091
48 18 ¼ N c C sh þ ðN c þ 4ÞC 0s 3.160
125 48 ¼ N c C sh þ ðN c þ 4ÞC 0s 8.047
Comparing the results of calculated stray capacitance given in Table 11.3 with
measured stray capacitance given in Table 11.2 shows reasonable estimates for the
gross approximations made in dielectric constant for the calculated cases.
Example 11.2: Compute the displacement current that would flow into each of the
modules listed in Table 11.2 when the HiPot test voltage is 500 V for the 16 V units,
2.5 kV for the 48 V units, and 4 kV for the 125 V units.
Solution: For each case, the displacement current would be I q ¼ C s ðDV =DtÞ so
that the calculations come out to
DV 500
16 V : I q ¼ C s ¼ 2:2 109 ¼ 110 nA
Dt 10
DV 2500
48 V : I q ¼ C s ¼ 3:3 109 ¼ 825 nA
Dt 10
DV 4000
125 V : I q ¼ C s ¼ 4:5 109 ¼ 1:8 mA
Dt 10
ch011 6 June 2011; 18:37:57

Lastly, the insulation resistance is generally specified in terms of MW to

describe the amount of leakage current that would be present when full voltage is
applied to the module in steady state. The values of IR are nominally in
ohms 1012. There is generally no issue with IR, but in some cases and depending
on insulating materials used, the leakage current could be an issue or shock hazard.
Nomex material is a cellulose paper that will absorb moisture when placed at
various levels of relative humidity (RH). The Dupont Type 994 Nomex will absorb
approximately 4% moisture content when exposed to 50% RH for 150 h. In a 50%
RH environment, the volume resistivity of Nomex 994 paper is r ¼ 21016 W-cm.
Exercise 11.5 discusses this particular case.
11.4 Vibration requirements
Customers that use ultracapacitor modules in their equipment will specify the
vibration specifications that the module will be subjected to. Typically, specifica-
tions such as SAE J2380 quantify the vibration levels the module should be sub-
jected to without any internal faults or loss of performance.
Figure 11.11 is a typical vibration specification that an ultracapacitor manu-
facturer would need to verify compliance to. In this figure, peak vibration levels of
0.8 g would be applied for 4 million occurrences but only 5000 events when the
peak vibration level is 5 g, here g ¼ gravity ¼ 9.802 m/s2. Battery manufacturers
must comply with vibration standards that typically call out 0.8grms for 16 h when
the battery is at 60% state-of-charge (SOC).
Cumulative exposure of vibration

5
4 Vertical axis
Peak acceleration (G’s)
Longitudinal axis
3
0
100 1000 10,000 100,000 1000,000 10,000,000
Cumulative occurrences per axis
Figure 11.11 Vibration specification
Example 11.3: How much force will the ultracapacitor pack mass, Mpak ¼ 11 kg, of
a BMOD0165-P048 module experience when subjected to 5 g along its z-axis? The
z-axis is taken as an axis along the axial direction of cells.
ch011 6 June 2011; 18:37:58

Abuse tolerance 325
Note: Apply Newton’s law to calculate the expected force and assume that the only
restraint is the thermal elastomer between the cell interconnects and the module
heat spreader top and bottom.
Solution: For this case, Fpk ¼ Mpaka ¼ Mpak(5 g) ¼ 11 kg(5)(9.802 m/s2) ¼ 539 N of
force along the pack z-axis.
The concern will therefore be for how long and how many such occurrences
can any internal component tolerate before it encounters mechanical fatigue.
Module designers must ensure that the internal connections are sufficiently robust
to tolerate this vibration level without fatigue.
11.5 Ultracapacitor cycling exposure
The types of cycling ultracapacitors that are being exposed to in-laboratory eva-
luations and in customer applications are summarized in this section for com-
pleteness. The waveforms discussed are not abuse conditions, but do represent
challenging current profiles, especially when repetitive over many thousands of
cycles. Figure 11.12 is the square wave current profile that has been used in thermal
evaluations of the ultracapacitor. This waveform gives the ultracapacitor electrolyte
no time to settle between charging and discharging and may represent a stressful
condition in terms of electrolyte aging.
Symmetric square wave
I0
0 t
I (A)
–I0
Uc
Umx
Umx
2
0 t
Figure 11.12 Square wave testing waveform
ch011 6 June 2011; 18:37:58

Symmetric quasi-square wave
I0
0 t
I (A)
–I0
Uc
Umx
Umx
2
0 t
Figure 11.13 Quasi-square wave profile
Figure 11.13 illustrates the quasi-square wave testing profile that is used by man-
ufacturers to characterize electrical and thermal parameters for cell and module and
long-term cycling performance. This waveform provides the ultracapacitor with
some amount of dwell time between charging and discharging for ionic settling. It
is important to understand what differences in service life exist between exposure
to square wave and quasi-square wave, if any.
A profile that is used extensively in ultracapacitor characterization is the
variable peak to average waveform having constant rms value. Figure 11.14 depicts
the constant rms waveforms. Experience reveals that current waveforms having
different crest-to-average values of current, such as this profile, results in differ-
ences in ultracapacitor life because of the different rates at which charge is
removed and restored, even though this is done symmetrically.
The final two profiles to be described represent automotive current profiles for
engine cranking followed by charge replenishment via the engine-driven alternator
and the second is an energy recuperator to store energy from opportunity charging via
the engine alternator. The engine starting current profile, shown in Figure 11.15, is
unique in that it is highly asymmetric having very high discharge pulse of several
hundred amperes followed by relatively low recharging at 50–100 A. The implica-
tions on ultracapacitor life for this profile are not clearly defined at present.
The last current profile to be considered for ultracapacitor exposure is the
energy recuperator type asymmetric waveform. In this waveform, the ultracapacitor
is exposed to a high charge pulse lasting 1–5 s of maximum output current of a
boosted alternator followed by a long shallow discharge pulse. The significance of
ch011 6 June 2011; 18:37:58

Abuse tolerance 327
Constant rms
I1
I2
I3
I (A)
0 t
–I3
–I2
–I1
Figure 11.14 Constant rms profile
I3
0 t
–I2
–I1
I (A)
U
U00
0 t
Figure 11.15 Engine starting profile
this profile is that of an energy recuperator or microhybrid system in which the

alternator output is boosted during vehicle deceleration followed by relatively low
current draw representing the ultracapacitor powering vehicle loads during engine-
OFF intervals (Figure 11.16).
It is important that adequate understanding is obtained on the service life
implications of these operating profiles on the ultracapacitor. At the present state of
ultracapacitor technology, and battery for that matter, it is too early to have an
acceptable picture of how such very different current waveforms, applied over long
periods of time, will impact ultracapacitor parameters and service life.
ch011 6 June 2011; 18:37:58

Asymmetric charge balanced
0 t
I (A)
Uc
Umx
Umx
2
0 t
Figure 11.16 Energy recuperator profile
Exercises
11.1 Failure analysis on an ultracapacitor cell indicated that the fuse had been
activated. Quality control determined that the cell was part of a standard
module that had a history of being exposed to frequent controlled discharges
in its application so that the full pack voltage of 700 Vdc was drained to near
zero voltage during out-of-service periods. This particular cell had an initial
capacitance that was lower than other cells in the series string so that full
discharge caused it to experience repetitive reverse polarity. When the
module was subsequently charged to full potential, this cell was subjected to
continuous overvoltage resulting in internal pressure build to the inevitable
activation of its fuse. What EUCAR Hazard Level best characterizes this
failure?
Ans: Hazard Level 4 since some venting had occurred by no reaction or
thermal runaway was detected.
11.2 What if the fault noted in Exercise 11.1 occurred in the presence of an
ignition source, for example, an energized frayed conductor in or near the
pack that caused intermittent electrical parallel arcing. For clarification, a
series arcing condition occurs when a contact or connection intended to
conduct full load current is intermittent such that arcing is limited by the
load. A parallel arc is a shunt path around the load or part of the load, so it is
more energetic. Ultracapacitor electrolyte solvents are flammable, and if the
ch011 6 June 2011; 18:37:58

Abuse tolerance 329
venting liquid and vapor is contacted by this arcing source a flame can
occur. What Hazard Level would this constitute?
Ans: Hazard Level 5 since there was an ignition source present.
11.3 An ultracapacitor module intended for engine cranking is subjected to very
high, but manageable, pulse current. For example, the current through the
cell may be 750 mA/F of carbon loading, which is extreme but tolerable for
carbon–carbon ultracapacitors. However, in this particular case and for a
history of use such as depicted in Exercise 11.1, the cell fuse opens, elec-
trolyte vents but at the same time internal welds fracture and the internal
arcing ignites the venting electrolyte causing the cell to eject electrode and
case fragments. What EUCAR Hazard Level does this case correlate to?
Ans: Hazard Level 6 since not only cell venting but also flame is detected
and the cell ruptured causing ejection of active mass and flying parts.
11.4 Using relative permeability in the stray capacitance calculations of (11.16)
and if a split dielectric were assumed, for example, 50% split between
er1 ¼ 2 and er2 ¼ 5, calculate by what percent the stray capacitance calcu-
lated would decrease.
Ans: Capacitance of a split dielectric capacitor is given by
C ¼ ðx1 =er1 Þþðx
e0 A
2 =er2 Þ
; so; e0r ¼ e2 xe11þe
e2
1 x2
where x1 ¼ 0.5d ¼ x2, so the relative permittivity decreases by

e0r 2ð2Þð5Þ=ð2þ5Þ
er ¼ 5 ¼ 20=7
5 ¼ 5
2:857
So, the effective stray capacitance for a split dielectric would decrease by 43%.
11.5 For the cited volume resistivity of Nomex 994 cellulose paper used as an
insulating material in a 48 V ultracapacitor module, assume that this mate-
rial completely envelops the sidewall perimeter of the module where the
dimensions are H ¼ 157 mm, W ¼ 91 mm, and L ¼ 418 mm. (A) What is the
expected insulation resistance? (B) What would the leakage current be if
operated at maximum system voltage of 750 V with an ungrounded chassis?
(C) Does this pose a shock hazard if the maximum permitted human contact
current is <2 mA?
Ans:
(A) The area is Aside ¼ H W L ¼ 5.972 103 cm2.
3
ð0:7610
16
Þ
Therefore, IR ¼ rlA ¼ 2105:97210 3 ¼ 2:54 GW
(B) The leakage current would be
I leak ¼ UIR
dmx
¼ 2:5410
750
9 ¼ 0:295 mA
(C) No
11.6 Calculate the undamped resonant frequency for the ultracapacitor modules
listed in Table 11.2. The undamped resonance is what to expect for an
ch011 6 June 2011; 18:37:58

impulse applied at the terminals of the module during operation such as

coming from a power electronic converter.
Ans: The calculations are put into tabular form.
Module rated Number cells Cs (nF) Ls (nH) f0 (Hz)

voltage (V)
16 6 2.2 372 5.56Eþ06
48 18 3.3 1116 2.62Eþ06
125 48 4.5 2976 1.37Eþ06
References
1. H. DeJarnette, C. Winchester, T. Tran, C. Govar, J. Banner, ‘Preliminary

abuse tolerance assessment of acetonitrile based super-capacitors for navy
power applications’, Presentation to the 42nd Power Sources Conference,
Philadelphia, PA, 12–15 June 2006
2. C. Ashtiani, ‘New perspectives on ultracapacitors’, The 6th International
Advanced Automotive Battery Conference, AABC, Ultracapacitor Session,
UCAP, Renaissance Harbor Place Hotel, Baltimore, MD, 15–19 May 2006
3. I.S. Gradshteyn, I.M. Ryzhik, Table of Integrals, Series, and Products,
translated from Russian by Scripta Technica, Inc., Academic Press, New
York, 1980
ch011 6 June 2011; 18:37:59

Chapter 12
Future transportation systems
The future of transportation systems appears to be one of a transition from fossil-

fueled vehicles via various levels of electric vehicles having onboard electric
energy storage to a system characterized by external electrification. Hybrid electric
vehicles of all types are the vanguard of this transition defined by the gradual
downsizing of the internal combustion engine and rise in dominance of electric
propulsion. In the battery electric vehicle, the heat engine is completely eliminated,
but all electric range is compromised because of the low-energy density of electric
storage mediums relative to that of hydrocarbon fuels. The range-extended vehicle
(REV), discussed in Chapter 9, uses an onboard gasoline engine to augment the
electric propulsion so that vehicle range targets of 300 mi are satisfied. It is unlikely
that purely electric energy storage will reach the 12.4 kWh/kg specific energy of
gasoline. The holy grail of electric energy storage, the lithium-air battery at a the-
oretical specific energy of 11 kWh/kg, excluding the oxygen source, is today at
laboratory levels of approximately 2 kWh/kg according to researchers at IBM’s
Almaden Institute. When the oxygen-positive electrode is included, the theoretical
specific energy is 5.2 kWh/kg. A more direct approach to fully electric transpor-
tation without the burden of massive onboard electric energy storage is to transmit
electrical power wirelessly to the vehicle, thereby mitigating the need for exces-
sively large onboard energy storage.
This chapter highlights some of the more promising methods of transmitting
electric power wirelessly to a vehicle while stationary or moving. Stationary sys-
tems are today being heavily researched, and several companies are beginning to
market such systems. Techniques to transmit large amounts of power to a moving
vehicle are perhaps further in the future, but systems already exist that are capable
of such feats. Some of the existing stationary wireless power transfer systems
available commercially are as follows:
● HaloIPT (www.haloipt.com) is an inductive power transfer system for vehicle
battery charging.
● WREL (www.intel.com) is a wireless resonant energy link developed for
consumer electronics.
● Witricity is a high-frequency resonant energy transfer pioneered by MIT
Physics Research Department.
● Momentum Dynamics Corp., Malvern, PA, is developing wireless power
transfer for PHEV and battery EV charging.
ch012 6 June 2011; 18:45:47

At the time of writing this book, many investigators in Japan, Korea, and China
are researching on wireless power transfer technology capable of transmitting
kilowatt levels of electrical power to a vehicle. Early demonstrations, like the
Momentum Dynamics system, will be on plug-in hybrid cars; the real value may be
in wireless charging of electric transit bus used in mass transportation systems.
12.1 Future mobility systems

There are many concepts in existence or are proposed for future mobility systems.
For example, Miller [1] discusses many example systems used for personal rapid
transit (PRT) to automated highway systems (AHS). More relevant to our discus-
sions here is the work of Stephan et al. [2–4] that describes a hybrid electric vehicle
having means of being externally powered when on a dedicated guideway.
Figure 12.1 illustrates this PRogram for Individual Sustainable Mobility (PRISM)
as a dual-mode vehicle certified for highway and guideway use. In normal driving,
the PRISM vehicle is operated as a conventional hybrid or battery electric car.
Once on a dedicated guideway, the vehicle receives its propulsion power, and
power to replenish its onboard energy storage, from an energized rail, or cable
mounted on the guideway wall.
Figure 12.1 Guideway concept and dual-mode vehicle
To quantify the electric power transfer required for the PRISM vehicle shown
in Figure 12.1, consider the following representative vehicle parameters for such
tandem seating, narrow lane vehicles in general. A narrow lane vehicle (NLV) is a
class of vehicles designed such that one conventional vehicle highway lane that is
specified to be 12 ft (3.66 m) will be capable of accommodating two NLVs side by
side in lieu of a single conventional vehicle that is 6 ft wide (1.83 m). By definition,
an NLV is restricted to vehicle’s width of less than 44 in. (1.12 m).
Example 12.1: Compute the power that must be transmitted via a noncontacting, or
wireless, power link to the PRISM vehicle shown in Figure 12.1 and having the
vehicle parameters listed in Table 12.1. The vehicle is assumed to travel on a
guideway section having the nominal velocity, grade, and headwind specified as
worst-case condition.
ch012 6 June 2011; 18:45:47

Future transportation systems 333
Table 12.1 PRISM vehicle parameters

Vehicle mass, Mv kg 1000 Rated vehicle speed, V mph 150
Aerodynamic drag coefficient, Cd – 0.22 Nominal headwind, Vw mph 25
Roll resistance coefficient, Crr – 0.007 Nominal grade, gr % 6
Frontal area, Af m2 1.6 Nominal air density, r kg/m3 1.22
Solution: Since the vehicle is specified to be in steady state velocity, the propulsion
power is given by (12.1) and represents the power that must be transmitted to the
vehicle to sustain its velocity.
n h gr io
PðV Þ ¼ gC rr M v þ gM v sin tan1 V þ 0:5rC d Af ðV þ V W Þ3 ð12:1Þ
100
Pkinetic ¼ M v V V_ ¼ 0; Proll ¼ gC rr M v ¼ 4:6 kW ð12:2Þ

h gr i
Pgrade ¼ gM v sin tan1 ¼ 39:33 kW; Paero ¼ 0:5rC d Af ðV þ V W Þ3
100
¼ 102:7 kW ð12:3Þ
PðV Þ ¼ Proll þ Pgrade þ Paero ¼ 146:73 kW ð12:4Þ
Therefore, the noncontacting or wireless power transfer means must be sized to

efficiently transmit approximately 150 kW of power. ‘Efficient for vehicle appli-
cations’ means of the order of 97% transmitter to rectified d.c.-link power, not
counting transmitter electronics and antenna nor energy storage system efficiency.
In this case, the vehicle receiver and conversion equipment would be expected to
receive (1/0.97)(146.73) ¼ 151.27 kW, of which 4.54 kW will be dissipated as heat.
Exercise 12.2 examines the carbon emissions from a coal plant in grams CO2
per mile for the NLV described in Example 12.1. Exercise 12.3 compares the
yearly CO2 emissions for the same vehicle based on average time spent traveling
for passenger vehicles in the United States. The metric has to be changed since the
tendency is for people to commute further in the same allotted time in commute.
Interestingly, this has changed little over time from an average of 1.5 h/day whether
walking, horse and buggy, automobile, or train.
12.2 Wireless power transfer
A short brief on electromagnetic power transmission is given before proceeding

with this chapter. In Figure 12.2, an electromagnetic wave of wavelength, l, is
shown leaving a source antenna. Four field zones identified in antenna transmission
theory are defined in this figure:
ch012 6 June 2011; 18:45:47

Source
l/2p l Mixed polarization S=E×H

Near Near Transition zone Far field
field field 2l 2l → ∞
reactive radiative
Figure 12.2 Definition of antenna transmission zones

● Near field reactive zone: x (l/2p), a zone in which the near field wave
interacts strongly with the antenna so that some energy becomes radiative and
the remainder returns to the antenna. In optics, the near field reactive zone
wave is termed an evanescent wave.
● Near field radiated zone: x < l, a zone where electromagnetic field parameters
E and H are interacting and one or the other still dominates.
● Transition zone: x < 2l, the zone where the wave polarization is becoming
differentiated and E and H are assuming their free space values.
● Far field zone: x > 2l, the purely radiated wave zone in which electromagnetic
radiation continues on to infinity.
The Witricity demonstrator of wireless power transfer, for example, operates at
9.9 MHz and uses a transmitting antenna coil that is resonated with a capacitor. The
capacitor is inserted into the antenna circuit to absorb a high electric field, E, and
thereby relies on magnetic field, H, for transmission. At this frequency, the wave-
length is 30 m and power transmission is demonstrated out to be 2 m, well within
the near field reactive zone.
The constants of electromagnetic theory are free space (i.e., vacuum) permit-
tivity, e0, and permeability, m0, defined as
m0 ¼ 4p107 ðH=mÞ ð12:5Þ
109
e0 ¼ ðF=mÞ ð12:6Þ
36p
Calculating the surge impedance of free space using (12.5) and (12.6) yields
the well-known value, Z0, for vacuum.
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
rffiffiffiffiffi
m0 4pð36pÞð107 Þ
Z0 ¼ ¼ ¼ 120p ðWÞ ð12:7Þ
e0 109
ch012 6 June 2011; 18:45:48

Equation (12.7) is used extensively in antenna theory and is recognized as

377 W of free space impedance. Also, the velocity of electromagnetic wave pro-
pagation, c, is derived from (12.5) and (12.6) as
1 108
c ¼ pffiffiffiffiffiffiffiffiffi ¼ pffiffiffiffiffiffiffiffi ¼ 3 108 m=s ð12:8Þ
m0 e0 1=9
In the far field, when electromagnetic power ! ! !

S ¼ E H has well-
differentiated E and H fields, the receiver magnetic field, Hr (A/m), can be calcu-
lated using the impedance of free space and the measured electric field, Er (V/m), at
the receiver, where n ¼ unit normal vector, as
! !
Z 0 H r ¼ n Er ð12:9Þ
A useful example to link these concepts together is found in the design of

a directed dipole antenna used commercially in broadcast television and radio.
Figure 12.3 shows such antenna construction schematically where the director lies
in the near field and reacts with the driven antenna. Acting in concert, the reflector
and director modify the dipole radiation pattern to a more focused beam to achieve
a measure of antenna gain. In this sort of antenna, gains up to 7 dB are practical.
ld la lr
Radiation
pattern
O ⇒ s
lr
ld la lr ≅ 0.9
l/4
d1 d2 d2
≅ 0.8
la = ld = d1= l/4 l/4
Figure 12.3 Illustration of directed dipole antenna
A challenge of wireless power transfer is obtaining a highly directed beam so

that highest possible transmission efficiency is realized. In some inductive power
transfer systems in use today, efficiencies more than 90% from transmitter antenna
to d.c. link power are being achieved. The Halo IPT, for example, operates with
resonant antenna coils at 20 kHz over transmission distances of 10 in. and power
level of 2 kW. For higher power levels consistent with SAE J1772 Level 1 and
Level 2 means that not only will power electronic and antenna efficiencies need to
be increased but highly directed beams must also be realized. A reason for this is
that power transmission into the bottom of a parked vehicle will require minimal
leakage fields adjacent to the transmitting antenna. Today’s levels of magnetic field
in the fringe pattern are Hfringe < 60 mG. This may be difficult to realize since the
E and H fields in the near field are very complex and variable.
ch012 6 June 2011; 18:45:48

At the time of this writing the Society of Automotive Engineers (SAE) formed
a new committee to pursue standardization of wireless power transmission (WPT),
also known as inductive power transfer (IPT). The SAE J2954 Wireless Charging
Task Force has the charter to standardize WPT for light duty vehicles by 2015.
During early 2011 the International Electrotechnical Commission (IEC) convened a
similar working group.
12.2.1 Coaxial winding transformer

One noncontacting power transmission system that does not suffer from fringe
fields of any consequence is the coaxial winding transformer (CWT) [5] modified
for use in a moving vehicle [2]. In the CWT, power is inductively coupled via a
single turn transformer having a gapped core that glides along a power cable. This
power cable is the energized rail on a guideway Modern power electronics now
makes it feasible to consider transmitting high levels of electric power at kilohertz
frequencies. The coupling efficiency of the CWT requires very tight gap tolerances
of 1 mm so as to provide some means of guidance and positioning. For example, the
read/write head is positioned with very high precision over a high-speed magnetic
pattern in a computer hard drive. Similar technology would be employed in posi-
tioning the CWT in an NLV traveling at guideway speeds of 95–150 mph and
higher in the future. It is also practical to consider electrodynamic positioning at
these speeds using a Halbach array, for example. Regardless of the method used,
position accuracy is extremely important because any gap mismatch on the CWT at
speed would result in near instantaneous destruction of the magnetic core at its gap.
To better appreciate the challenge of using a CWT, consider that this device is
a single-turn current transformer that must transmit high power levels at reasonable
primary currents supplied by utility converters at intervals along the guideway.
Suppose further that the magnetic flux density in the highly permeable core is
limited to 1 T. If a physical gap lg ¼ 1 mm is the best a positioning system can hold
the tolerance at high speed, then what level of primary current is necessary to
magnetize the core of this device?
m0 N I p
Bg ¼ ¼1T ð12:10Þ
lg
Bg lg 1ð1 103 Þ 1
Ip ¼ ¼ 7
¼ 104 A ¼ 795:8 A ð12:11Þ
m0 4p 10 4p
High primary current levels in a CWT are necessity since N ¼ 1. This however
complicates the utility side converters that now must be sized to handle large
excitation current. There is also a very challenging issue of the high power
requirements of high-speed NLVs on a guideway. Depending on the vehicle
headway, the power burden on the guideway power supplies could be prohibitive
and force platooning of vehicles.
ch012 6 June 2011; 18:45:48

Power distribution
rod (kHz)
Magnetic shield
Conductive inner layer
Figure 12.4 Coaxial winding transformer cross section
Figure 12.4 shows the cross section of a CWT having single-turn primary and
secondary. The primary is the utility-driven power rail, and the secondary is the
single-turn copper inner conductor that supplies its power to the vehicle ESS and
traction drive system. The primary rail, power distribution rod in the figure, is a
conducting solid tube or rod supported on a specially designed arm that is both
magnetically conducting in circumferential direction and electrically insulating
along its axis.
The CWT core is magnetic iron, powdered iron core, or other very low-loss
magnetic material suitable for kilohertz power transmission. The conductive inner
layer in Figure 12.4 is a copper sheath of sufficient thickness to conduct the high
secondary current to the moving vehicle. The main design challenges of the CWT are:
● Very high permeability but low electrical conductivity magnetic material for
low loss at high operating frequency and wide flux excursions.
● Tight, and very high tolerance, physical gap of CWT to the power rail support arm.
● High positioning accuracy actuator to hold the CWT near concentric with the
power rail all the while holding a tight air gap between the CWT cut core and
magnetic portion of the support arm.
● Actuator positioning at low vehicle speed transition to electrodynamic posi-
tioning at high vehicle speeds.
The authors in Reference 6 describe a contactless energy transfer system
(CETS) based on an E-core transformer operating at 300 kHz. This system is
similar to a CWT but without the need for concentric magnetic gap. The CETS
achieves a magnetic coupling coefficient, k ¼ 0.6, over a gap d ¼ 10 mm and power
transfer of 60 W at 88% efficiency.
12.2.2 Inductive power transfer

Inductive power transfer systems are being investigated now for application to
stationary charging of electric vehicles [7–9]. Inductive power transfer relies on
resonant tuned transmitter and receiver coils, typically at 20–760 kHz, with some
ch012 6 June 2011; 18:45:48

concepts at 9.9–38 MHz. Imura et al. [10] describe electromagnetic resonance

coupling for wireless power transfer that exploits the fact that weak coupling at
resonance can transfer energy with high efficiency. In their work, transmit and
receive coils of r0 ¼ 150 mm consisting of dw ¼ 2 mm wire size with a coil pitch,
p ¼ 3 mm, are shown to transmit PIPT ¼ 100 W over significant air gaps. Inductive
power transfer, PIPT, is done at high efficiency, h ¼ 97% for gaps, g ¼ 100–200 mm
and h ¼ 80% at g ¼ 250 mm. This is a very good efficiency for large air gaps and is
achieved at a resonant frequency of f ¼ 16 MHz and coupling coefficient, k ¼ 0.057.
Budhia et al. [11] describe a three-phase approach to inductive power transfer
to a moving vehicle, in this case for PHEV or BEV at power levels of 20–30 kW. A
demonstrator system was built having gap, g ¼ 60 mm, operating at frequency,
f ¼ 38.4 kHz, and with primary current of 22.5 Arms that was able to tolerate lateral
offsets of 120 mm. The power output of IPT, an electromagnetic resonance cou-
pling technology, depends strongly on the quality factor, Q, of the resonant coils.
Typical range of quality, 4 < Q < 6, provides reasonable range of primary inverter
(kVA) to output power (kW) rating. These authors show that output power at the
receiver is given as the product of Q with resonant system open-circuit voltage and
short-circuit current as

MI p M2
P0 ¼ U oc I sc Q ¼ ðwMI p Þ Q ¼ wI 2p Q ð12:12Þ
Ls Ls
In (12.12) the mutual inductance, primary (transmit) to secondary (receive)

coils, is M, primary current, Ip, secondary inductance, Ls, and output power P0. At
such high frequency, the primary and secondary coils are Litz wire to minimize
eddy current losses in the conductors.
The topic of IPT has been introduced in this chapter for the purpose of increasing
awareness of this growing trend and for highlighting the merits of wireless power
transfer to a stationary or moving vehicle. Significant benefits of IPT include:
● Immunity to leakage currents and ground faults between the vehicle under
charge and the utility mains.
● All weather charging of high-voltage batteries and ultracapacitors without need
of high-current connectors.
● User-friendly means of vehicle charging that are tolerant of significant lateral
mismatch between transmit and receive coils.
● Hassle-free vehicle charging.
12.3 Ultracapacitor applications in IPT

Because of ease of use and high tolerance of vehicle positioning over the trans-
mitting coil in an IPT system, this may well be the preferred system of vehicle
charging in the near future. Inductive power transfer means that regardless of
weather conditions or parking accuracy, a vehicle electric energy storage system
can be replenished autonomously. A proximity sensor determines that a vehicle
ch012 6 June 2011; 18:45:49

receiver coil is positioned in proximity to its IPT transmitter coil and, after proper
communication to identify the type ESS and the utility rate structure the power
flow is managed for demand and price.
Example 12.2: Consider the electric transit bus having an ultracapacitor-only ESS
that operates over a fixed route with a maximum travel between stops of 1 km. The
bus has an energy demand, including cabin air conditioning of 1200 Wh/km. If the
bus is to recharge in no longer than 90 s, then what transfer power level must an
inductive power transfer system operate at? Assume that the ultracapacitor ESS
pack must be recharged at 60% of full capacity.
Solution: For this example, the ultracapacitor pack requires energy replenishment
of 1200 Wh in 60 s. It is also apparent that this bus has a rated pack of (1/0.6)
(1200) ¼ 2000 Wh of storage. This example is not intended to comment on what
technology the ultracapacitor must be, perhaps a hybrid capacitor, given the high
energy level in question.
The transmission power level at the transit bus receiver side converter is
therefore
gW d leg ð1:2 kWh=kmÞð1 kmÞ
Prec ¼ ¼ ¼ 48 kW ð12:13Þ
tchg 90 s
This amount of power is entirely feasible for a Level 3 charger.
The research challenge is therefore in the design of an IPT capable of such
high power level and how to implement this along the route bus stops, each of
which requires a utility side converter. Figure 12.5 illustrates how ultracapacitors
may be employed with advantage in a transit bus that operates over a fixed route as
discussed in Example 12.2. For example, an ultracapacitor-only ESS hybrid transit
bus will operate with Ud * 600 V, and Wuc * 2.5 kWh of deliverable energy for
3–5 km of driving.
Ultra-
capacitor
P0 ESS
10 pu Converter
Electric
drive
+ RCVR +
XMTR
Concrete or pavement
Ud
Converter Charger
10 pu 1 pu Utility
Ultra-
capacitor
ESS
Figure 12.5 Wireless charging of ultracapacitor-only transit bus with converters

rated for 15-min headway
ch012 6 June 2011; 18:45:49

Figure 12.5 highlights the cycling demands for ultracapacitor energy storage at
the bus stop converter shown as the utility interface charger and for the vehicle.
These two ultracapacitor energy storage systems have nearly reciprocal cycling
requirements of the type introduced in the last section of Chapter 11. The utility
charger operates at 1 per unit (pu) to replenish the utility side ESS during the time
the bus is on its route. Then, when a bus stops over the IPT transmitter (XMTR in
Figure 12.5), the system is energized and the utility side ESS discharges at 10 pu
through the high-power IPT converter into the bus receiver (RCVR in Figure 12.5).
The transit bus onboard ultracapacitor ESS is then fast charged by the 10 pu
onboard charger in 90 s. For example, if the bus fully depletes the 2.5 kWh of
deliverable energy, then it must be replenished at a 100 kW rate at the bus receiver
converter output. Industrial IPT systems [11] are designed to operate at 50–200 kW,
so this power level is consistent with present industrial practice. Figure 12.6 illus-
trates the utility side converter current and voltage waveforms to be expected at the
ultracapacitor ESS. This figure is drawn in the same way as that for ultracapacitor-
assisted engine starting in a microhybrid application.
I3
0 t
–I2
–I1
U
Uco
0
t
Figure 12.6 Utility side ultracapacitor ESS current waveform
In the ultracapacitor-only transit bus example, the utility side converter ESS
experiences a very high discharge current necessary to support the 100-kW power
transfer discussed earlier. The current can be stepped as shown to relax thermal
stress on the transmit and receive coils, but in general it will have this character.
After the fast discharge during the bus charging period, the utility side converter is
slowly recharged in much the same manner as an ultracapacitor would be in an
energy recuperator or microhybrid on a vehicle.
Figure 12.7 depicts the vehicle side ultracapacitor fast recharge current
waveform and a stylized representation of charge depletion during bus service
along its route. The key point is that ultracapacitor recharging can be very fast,
regardless of environmental conditions, and that the bus can be put into service
immediately following the fast recharge.
ch012 6 June 2011; 18:45:49

Asymmetric charge balanced

I (A)
0 t
Uc
Umx
Umx
2
0
t
Figure 12.7 Vehicle side ultracapacitor ESS current waveform
According to Figures 12.6 and 12.7, the ultracapacitor cycling is one of fast
discharge (utility ESS) and fast charge (vehicle ESS) followed by shallow charge
and discharge, respectively. The purpose of this is to highlight the need for more in-
depth understanding of ultracapacitor (or battery for that matter) service life under
high-stress current charge and discharge. Miller and McCleer [12] investigated the
thermal response of the Maxwell BMOD0165-P048 ultracapacitor module under
clean and ripple current cycling conditions (Figure 12.8).
Ripple current
105.00
AM1. I (A)
50.00
0.00
Irms
–50.00
–105.00
0
00
.0
.0
.0
.0
5.
20
40
60
80
11
Figure 12.8 Ultracapacitor module under cycling having superimposed power

converter ripple current (I0 ¼ 90 A and dI ¼ 30 App)
The module under clean cycling having exposure to a quasi-square wave of

current of magnitude, I0, and dwell time, dT, is shown graphically in Figure 12.9
for which its rms value is
ð dT pffiffiffi
1 dT
I 2rms ¼ I 20 dt ¼ I 20 ; I rms ¼ dI 0 ð12:14Þ
T 0 T
ch012 6 June 2011; 18:45:49

Ripple current
90.00
50.00
25.00
AM1.I (A)
0
–25.00
–50.00
–90.00
0 20.00 40.00 60.00 80.00 100.00 115.00
t
Figure 12.9 Clean current cycling control case, I0 ¼ 90 A
The surprising results are summarized in Figure 12.10 for the thermal response
of the module positive terminal temperature response when one module is sub-
jected to the clean quasi-square wave cycling and a second, identical module, is
subjected to the same quasi-square wave of current of the same magnitude, I0, but
with superimposed ripple current 2dI ¼ 30 App. It is surprising that the total rms
current difference between the two cases, when I0 ¼ 80 A, is very small as eval-
uated in Exercise 12.6.
BMOD0165-P048 meas temp pos versus modeled BMOD0165-P048 meas temp pos versus modeled
36 45
35,504 41,893
34 41
Temperatue (°C)
Temperatue (°C)
32 37
Tpos Tpos
30 Testpos 33 Testpos
28 29
26,151 28,429
26 25
3
0 4 × 10 8 × 103 1.2 × 104 1.6 × 104 2.0 × 104 0 4 × 103 8 × 103 1.2 × 104 1.6 × 104 2.0 × 104
0 Time (s) 20,000 0 Time (s) 20,000
Figure 12.10 Measured versus modeled temperature of 48-V ultracapacitor

module at its positive terminal (left: control case clean quasi-square
wave current, Rth case ¼ 0:173 K=W ; C th case ¼ 25; 000 J =K;
tcase ¼ 3870 s; right: superimposed ripple current case
Rth case ¼ 0:158 K=W ; C th case ¼ 22; 000 J =K; tcase ¼ 3410 s)
The calculated thermal resistance, Rth, thermal capacitance, Cth, and thermal
time constant are not the important findings in this experimental work. The goal
is not to quantify the thermal parameters so much as to evaluate the difference
in steady state temperature rise of the module with superimposed high-frequency
ripple current 12 C versus the module with clean quasi-square wave current 9 C.
In this particular case and as demonstrated in Exercise 12.6, the ripple case rms
current is only 0.6% higher than the clean current case, yet its temperature is
ch012 6 June 2011; 18:45:49

higher by 33% (12 C for ripple vs. 9 C for clean). This evidence, limited though it
is, indicates that something is going on to cause the module temperature to rise higher
when ripple current is present. One plausible answer to this finding comes from Briat
et al. [13]:
The difference observed for these results may reveal an influence of the
profile discontinuity on aging. In fact, even if a common RMS current value
is supposed to lead to the same heating, the temperature shapes during a
cycle are different and can induce differences in aging mechanisms. How-
ever, this effect has to be verified by additional power cycling tests.
Referring again to the waveforms shown in Figures 12.6 and 12.7, it is
important to gain a deeper understanding of the stress imposed by high rate dis-
charging (utility ESS) and charging (vehicle ESS) for an ultracapacitor-only transit
bus, or other vehicle. The current stress and attendant influence not only the ageing
mechanisms but also the overall service life in such applications. This is a matter of
considerable interest and it points to the need for a program to characterize ultra-
capacitors under asymmetric current loading having vastly different charge–dis-
charge characteristics.
Exercises
12.1 Compute the vehicle electrical power for Example 11.1 if the vehicle speed
were specified as 300 mph.
Ans: P(V) for rolling and grade double but aerodynamic power is higher by
(145.2778.22)3 ¼ 6.406 times. Therefore, the total power for twice the speed
is PðV Þ ¼ 2ð4:6 þ 39:33Þ þ 658:51 kW ¼ 746:37 kW.
12.2 If the emissions from a coal-fired electrical generating plant are given as
583 g CO2/kWh, then what is the generating station emissions burden to
supply the narrow lane vehicle of Example 11.1 versus the same vehicle at
higher speed in Exercise 11.1? For this exercise assume that the utility to
guideway transmission efficiency, including the power converter to energize
the guideway rail, is composed of transmission efficiency of 92%, utility
side converter efficiency of 96%, and power transmitter efficiency of 95%.
Then h ¼ (0.92)(0.96)(0.95) ¼ 0.839 is the efficiency to deliver utility plant
generated electricity to the vehicle receiver. Vehicle receiver efficiency is
97%, so its overall generating plant to NLV d.c. link is htot ¼ 0.839(0.97) ¼
0.814.
Ans: The NLV at V ¼ 150 mph has P(V) ¼ 146.73 kWh and covers a distance
of 1 mi in t ¼ 24 s, yielding a vehicle energy consumption of
W150 ¼ 1.202 kWh/mi reflected back to the generating plant. At a vehicle
speed V ¼ 300 mph, the vehicle energy consumption is W300 ¼ 3.06 kWh/mi,
and the vehicle travels 1 mi in 12 s. The carbon emissions for both cases are
therefore
ch012 6 June 2011; 18:45:49

CO2 (@150 mph) ¼ W150(583) ¼ 700.77 g/mi,

CO2 (@300 mph) ¼ W300(583) ¼ 1784 g/mi, a ratio of 2.54 times higher.
12.3 Suppose the average commuter in the future, traveling by each of the NLV
modes discussed in Exercise 12.2, allots a total daily commute time of 1.5 h.
Then what will be the yearly CO2 emissions of the generating plant if
electricity is generated by coal?
Ans: Today’s commuter averages 15,000 mi/year or approximately 330 h
driving. Using this as the basis to answer this question means a total of
(330 h/year)/1.5 h/day ¼ 220 days commuting per year.
NLV speed Average miles/year Average CO2/mi Yearly CO2

150 150 mph (330 h) ¼ 49,500 @700.77 g/mi 34,688 ton
300 300 mph (330 h) ¼ 99,000 @1,784 g/mi 176,616 ton
12.4 Quantify the term 7-dB antenna gain in terms of received voltage relative to
1 mV. The antenna gains are measured relative to 1 mV in signal transmission
and 1 mW in terms of power transmission.
U
Ans : 7 dBmV ¼ 10 log10 ; U ¼ 1 mV 107=10 ¼ 5 mV
1 mV
12.5 Extend the topic of rms current given by (12.14) for the quasi-square wave
of current shown in Figure 12.8 to the case of the same quasi-square wave
current but with superimposed higher-frequency triangular wave of current.
Tek Stop M pos: 300.0 µs CH3
Coupling
DC
Ripple current BW Limit

90 Off
100 MHz
50 Volts/Div
AM1.I (A)
20 4+ Coarse
0 Probe
–20 10 ×
voltage
–50 Invert
1+
–90 Off
Plus
0 20 40 60 80 115 CH1 5.00 V M 1.00 ms CH1\ 2.80 V
t CH4 10.0 A < 10 Hz
Ans:
pffiffiffi
Hint: First compute the rms value of a triangular wave as ð1= 3ÞdI, then
combine by superposition for the case of the quasi-square wave shown here
for reference.
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2
dI
I rms ¼ dI 20 þ pffiffiffi
3
ch012 6 June 2011; 18:45:50

12.6 Determine the ratio of clean quasi-square wave rms value as given by
(12.14) to that found in Exercise 12.5, and compute its value when I0 ¼ 80 A
and ripple dI ¼ 15 A.
Ans: The ratio of clean quasi-square current wave to that with superimposed
ripple of magnitude dI is

I clean 1 1
rms
¼ q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi ¼ ¼ 0:9942
I ripple 2
1:005836
rms ð1 þ ð1=3ÞðdI=I 0 Þ Þ I 0 ¼80 A; dI¼15 A
References
of Engineering Technology (IET), Michael Faraday House, Stevenage,
2. C.H. Stephan, J.M. Miller, J. Pacheco, L.C. Davis, ‘A program for individual
sustainable mobility’, Global Powertrain Congress, GPC2003, Ann Arbor,
MI, 23–25 September 2003
3. C.H. Stephan, J.M. Miller, L.C. Davis, ‘A program for individual sustainable
mobility’, International Journal of Vehicle Autonomous Systems, vol. 2,
pp. 255–77, 2004
4. C. Stefan, J.M. Miller, A Program for Individual Sustainable Mobility,
Frontiers in Transportation: Social Interactions, Hotel Rembrandt
Amsterdam Academic Club, Oudezijds Achterburgwal 235, Amsterdam,
The Netherlands, 14–16 October 2007
5. K.W. Klontz, D.M. Divan, D.W. Novotny, R.D. Lorenz, ‘Contactless power
delivery for mining applications’, IEEE Industry Applications Society Annual
Meeting, Ritz Carlton, Dearborn, MI, 28 September–4 October 1991
6. W. Zhang, Q. Chen, S.C. Wong, C.K. Tse, X. Ruan, ‘A novel transformer for
contactless energy transmission systems’, IEEE 1st Energy Conversion
Congress and Exposition, ECCE2009, Double Tree Hotel, San Jose, CA,
20–24 September 2009
7. A. Kurs, A. Karalis, R.M. Moffatt, J.D. Joannopoulos, P. Fisher, M. Soljacic,
Wireless Power Transfer via Strongly Coupled Magnetic Resonances,
American Association for the Advancement of Science (AAAS), vol. 317,
p. 83, 2007 [doi: 10.1126/Science.1143254]
8. N. Keeling, G.A. Covic, F. Hao, L. George, J.T. Boys, ‘Variable tuning in
LCL compensated contactless power transfer pickups’, IEEE 1st Energy
Conversion Congress and Exposition, ECCE2009, Double Tree Hotel,
San Jose, CA, 20–24 September 2009
9. C. Yu, R. Lu, Y. Mao, L. Ren, C. Zhu, ‘Research on the model of magnetic-
resonance based wireless energy transfer system’, IEEE 5th Vehicle Power
and Propulsion Conference, VPPC2009, Fairlane Technical Center,
Dearborn, MI, 7–10 September 2009
ch012 6 June 2011; 18:45:50

10. T. Imura, H. Okabe, Y. Hori, ‘Basic experimental study on helical antennas of

wireless power transfer for electric vehicles by using magnetic resonant
couplings’, IEEE 5th Vehicle Power and Propulsion Conference, VPPC2009,
Fairlane Technical Center, Dearborn, MI, 7–10 September 2009
11. M. Budhia, G. Covic, J. Boys, ‘Magnetic design of a three-phase inductive
power transfer system for roadway powered electric vehicles’, IEEE 6th
Vehicle Power and Propulsion Conference, VPPC2010, Lille Grand Palis
Conference Center, University of Lille, Lille, FR, 1–3 September 2010
12. J.M. Miller, P.J. McCleer, ‘Electrical and thermal investigation of power
electronic converter ripple current on the ultracapacitor’, The 20th Interna-
tional Seminar on Double Layer Capacitors and Hybrid Energy Storage
Devices, Embassy Suites Deerfield Beach Hotel, Deerfield Beach, FL, 6–8
December 2010
13. O. Briat, W. Lajnef, J.-M. Vinassa, E. Woirgard, ‘Power cycling tests for
accelerated ageing of ultracapacitors’, Elsevier, Microelectronics Reliability,
vol. 46, no. 2006, pp. 1445–50, 2006
ch012 6 June 2011; 18:45:50

Glossary
Adsorption pseudo-capacitance The phenomena of reduction–oxidation inter-

mediates formed on a conducting surface by charge transfer. A pseudocapacitance
arises from the fact that an increment in applied potential results in current flow
until a monolayer of charge is accumulated on the surface such that the surface can
be considered homogeneous and there is no lateral interaction of species.
Chemical bonds Chemical bonds are of three types: (1) ionic, (2) metallic, and
(3) covalent. Ionic bonds are generally molecules made up of a metal and a non-
metal. Metallic bonds are formed where atoms of a particular metal combine.
Covalent bonds form when two nonmetals combine. Bonds form when atoms or
molecules gain or lose electrons according to their valency.
Electrolyte Typically a liquid containing free ions that contribute to a certain
level of electrical conductivity of the solution. Most typical electrolytes are ionic
solutions, molten salts, and ionic liquids.
Faradaic A process involving chemical oxidation or reduction of a chemical
substance at an electrode. Faradiac current is generated by reduction–oxidation
reactions taking place.
Insertion chemistry During insertion an ion moves into a porous electrode, a
process known as intercalation. During the reverse process, or ion extraction from
the electrode, the ion moves out, a process known as deintercalation.
Intercalation pseudo-capacitance The phenomena of reduction–oxidation
intermediates
Electrolyte type varies for each type of electrochemical capacitor (EC)

BOL (beginning of life): Defines the cell (module) parameters at the time of
manufacture.
Charge: Ionic molecules in solution, electrons in conducting medium.
Current collectors: Metal foils used in each electrode to which the carbon
electrode films are laminated. Typically aluminum foil.
EOL (end of life): Defines cell (module) parameters at wear out. EOL for the
symmetric EC is defined as 20% irreversible loss of capacity and 100%
increase in ESR.
Internal resistance: Also known as equivalent series resistance (ESR) defines
the sum total of electronic (electron conducting) and ionic (ion conducting)
glossary 6 June 2011; 18:36:41

contributions inhibiting current flow. ESR increases by approximately 1.4

from 0 to –30 C in the symmetric EC and by 3 for asymmetric ECs.
Leakage current: Defines the bleeding off of accumulated charge at each
electrode’s double layer. Considered to result from impurities in the carbon that
disrupt the solvated ion dielectric sheath at the compact layer. Effect accel-
erates with increasing temperature. For example, in a symmetric EC, cell
potential loss varies with sqrt(time). Unlike the diffusion-limited leakage
phenomena in the symmetric EC, for asymmetric ECs the leakage is more
redox controlled and results in cell potential dropping with log(time).
Nominal voltage: For aqueous electrolytes it is typically 1.3 V, whereas for
organic electrolytes it can be 3 V. MXWL cells are labeled with maximum
operating voltage.
Overvoltage: Short-term overvoltage that the cell can tolerate for some seconds
but must not persiltage: A symmetric EC may be subjected to abuse conditions
such as 4–6 V at the terminals. The overvoltage results in charging current
becoming total leakage current in the device.
Polarity: Asymmetric ECs have polarization due to presence of a battery
electrode. Symmetric ECs technically do not have polarization, but presence of
impurities gives rise to redox reactions and some small preferential polarity,
hence the (+), (–) markings on MXWL cells.
Randles equivalent: An electronic RC network equivalent of electrode
dynamics that consists of a series resistance, Rs, and the parallel combination of
a capacitance, C, and parallel resistance, Rp. The Randles equivalent is a good
approximation to ultracapacitor voltage and current behavior.
Recycling: Defines the proper disposal procedures for discarding ECs. The pre-
sence of some hazardous materials, for example, electrolyte solvents and aqu-
eous alkali or acid solvents in asymmetric devices require appropriate handling.
Separator: Porous paper, polymer, or ceramic that prevents EC electrodes from
shorting together. Must be ion conducting (porous) and electron blocking.
SOC (state-of-charge): A measure of the available, or useable, energy content
of the EC (or battery). Unlike most batteries that exhibit ‘memory’ effects, the
EC is a nearly linear device in regard to SOC.
Storage life (or shelf life): Storage life is specified at temperature and humidity
conditions such as cool, dry area. Symmetric ECs should be stored in a dis-
charged state (generally with shorted terminals) to minimize any natural decay
phenomena. Asymmetric ECs are stored in a charged state and must be peri-
odically recharged to hold near 100% state-of-charge (SOC). The storage
location temperature should not drop below –60 C.
Surge: st (2.85 V for a 2.7 V cell).
Temperature effects: Capacitance is generally insensitive to temperature.
However, for redox components, the capacity will diminish at reduced
temperature.

Glossary 349
Warburg impedance: A special case of the constant phase element, Zcpe ¼

A/sa, where a [0,1], with a ¼ 0 being pure resistance and a ¼ 1 a pure
capacitance. The midway point in this domain is referred to as the Warburg
impedance where frequency response is proportional to square root of fre-
quency and phase on a Bode plot is 45 .
ði jÞ
Z ¼ pffiffiffiffiffiffiffiffiffiffi ðG:1Þ
sðwÞ
Generic types of electrochemical capacitors (ECs)
Asymmetric design: Electrodes are different materials, one activated carbon

(DLC electrode) and the opposing electrode is a battery type that stores charge
via chemical reactions, reduction–oxidation (redox).
Symmetric design: Same carbon material is used in both electrodes. Testing
generally imparts a (+) positive or (–) negative polarization.
Rugosity More aptly here, surface rugosit explains how the morphology of a
carbon particle surface contributes to specific capacitance of the ultracapacitor.

Index
Page numbers followed by f and t indicate figures and tables, respectively.
abuse testing automotive manufacturers

see also ultracapacitor abuse generation technologies 259
acetonitrile (AN) 17 Axion Power International 33, 73
active parallel architecture 74, 81–9
active parallel ESS barium titanate ceramics 5
and converter power 88t BAS: see belt alternator starter
down-convert case of 84–5, 85f, 86f (BAS)
up-convert case of 84, 84f batteries and ultracapacitors 94,
vehicle simulation for 87f 114–123
adsorption 10–11 battery
AFS Trinity 115 average current 88
AHS: see automated highway systems electric car 332
(AHS) heating 75–76, 96
aluminum 17 manufacturers 324
cathode foils 4 power 248
electrolytic capacitors 281 state-of-charge (SOC) 324
hybrid 25 battery and ultracapacitor, combination
aluminum–aluminum welds 282 of
anode electronic resistance, architectures for 77t
components of 37–8 battery electric bus 199–200
Ansoft Q3D Extractor 42 characteristics of 199t
Ansoft Simplorer 37, 40, 58, 79 battery electric vehicle (BEV) 4, 75,
Arrhenius–Eyring life relationship 282 82, 94, 272
Arrhenius-Eyring methods 299 cycling capability of batteries and
Arrhenius-Eyring theory 299 ultracapacitors 94
Arrhenius function 22 performance and economy obstacle
Arrhenius relationship 282, 286 for 96
Arrhenius theory 284 B0 Blue Car 115, 117f
Asahi Kasei 2 belt alternator starter (BAS) 213
asymmetric (hybrid) electrochemical belt integrated starter generator
capacitor products 25t (B-ISG) 213
asymmetric ultracapacitor 24–30 benchmark values on inductance 46
A&T Batt Co. 2 Betz coefficient 147
automated highway systems (AHS) BEV: see battery electric vehicle (BEV)
332 BHR (battery heat reduction) 76
index 7 June 2011; 15:27:27

B-ISG: see belt integrated starter cell model constituents 38f

generator (B-ISG) cell preparation method 311
BMW, Efficient Dynamics 227, 227f for abuse testing 312
Bode plot 14f, 15f, 29, 47, 49, 49f overvoltage test 313
Boltzman’s constant 285 cells abuse test, illustration of 315
brake specific fuel consumption cell voltage swing 279
(BSFC) 243 CETS: see contactless energy transfer
BSFC: see brake specific fuel system (CETS)
consumption (BSFC) charge–discharge energy 99f
Bucyrus shovel 177 Chevy Volt 272
buses Chevy Volt REV chassis, illustration
battery electric 199–200 of 272
conventional architecture 190 CIDI engines: see compression ignited
shuttle 206–8 direct injected (CIDI) engines
see also heavy transportation coaxial winding transformer (CWT)
vehicles 336, 337
magnetic iron, powdered iron core
capacitance 4, 5 337
capacitor–battery combinations 73, cold temperature, electronic resistance
76, 81 at 94
active parallel architecture 74, combination technologies 114, 115
81–9 Commonwealth Scientific and
passive parallel architecture 74, Industrial Research
78–81 Organization (CSIRO) 73
capacity fade, thermal effect 301 component rating 41
CapBat 73 compression ignited direct injected
carbide-derived carbon (CDC) 18 (CIDI) engines 172
carbon–carbon symmetric constant current (CC) conditions
ultracapacitor 221 charge–discharge behavior 98f, 99f
carbon–carbon ultracapacitor 17, 18f, discharge efficiency under 99–100
30, 37, 42, 296 energy efficiency under 97–8, 97f,
equivalent circuit model of 42f 99
cell round trip efficiency in 100
capacitance 20, 25, 32, 60, 283 constant failure rate (CFR) 291
fuse 315 constant power (CP) operation 198
inductance 44, 44f constant power–speed ratio (CPSR)
mass under adiabatic conditions 313 machines
overvoltage test 313, 314 internal permanent magnet (IPM)
cell characterization methods and 243, 244
standards 58 constant torque (CT) operation 198
EUCAR method 61–6, 62f contactless energy transfer system
cell interconnect 40, 42 (CETS) 337
parasitic resistance and inductance continuously variable transmission
42, 43f (CVT) 215, 259
resistance and inductance 43f types of 236
index 7 June 2011; 15:27:27

Index 353
conventional automatic transmission down-convert case of active parallel

Allison 6-speed 258 ESS 84–5, 85f, 86f
converter input–output parameters Dupont Type 994 Nomex paper 322,
under power invariance 117t 324
converter locations, in ultracapacitor
plus battery combination Earth moving equipment: see
storage system 115, 116t excavator
copper anode foil 4 East Penn Manufacturing Company 73
CP operation: see constant power (CP) Eaton UT3220 220-kVA three-phase
operation UPS
CT operation: see constant torque (CT) parameters of 130t
operation eclectic principle 96
current, leakage 315 Ecotec engine 272
CVT: see continuously variable eCVT: see electronic continuously
transmission (CVT) variable transmissions (eCVT)
CWT: see coaxial winding transformer Efficient Dynamics concept vehicle
(CWT) 227, 227f
cyclable energy storage 101 8S9PHTM model, in 495HF
cylindrical shell approximation, of excavator 182–5
spiral winding 45, 45f electrical distribution system
288, 290
DAQ bits, number of electrically variable transmissions
calculation of 64 (EVTs) 235, 257
d.c.-link potential 77, 79, 81, 83, 86 advantage of 268
d.c.–d.c. converter 75 analysis of 264
and ESS architecture 83f E1 and E2 base ratios 268
possible locations for 76 electric variator function 270
d.c.–d.c. converter duty ratio control full schematic of 270
parameters 115, 117t functional diagram
d.c.–d.c. converter location, in energy storage 263
combination technologies planetary and clutch designs 262
114 hydraulic controls 259
Debye length 17, 19, 20, 23 MG1 input power, trajectory of
decreasing failure rate (DFR) 291 267, 268
DFIG: see doubly fed induction MG1/MG2 rotational speeds 271
generator (DFIG) MG1/MG2 torque 266
dielectrics, properties of 6t in mode-1, low-speed input split
dielectric withstand potential 317 mode 263
discharge efficiency in mode-2 265
under constant current conditions motor-generator efficiencies 264
99–100 operational modes of 263
double-layer capacitance 50, 52, 53 shaft power 264
functional fit to 53f single-mode vs. 2-mode architecture
doubly fed induction generator 257
(DFIG) 146, 152–5 vehicle speed 271
index 7 June 2011; 15:27:27

electrical–thermal interdependent Toyota’s hybrid synergy drive

model 70f (HSD) 235
electric energy storage 331 6-speed automatic transmission
electric energy storage devices 2–3, 4 258, 260
Ragone chart of 3f Toyota Motor Lexus 600 h 259
electric material handling trucks ultracapacitors in 242
165–8 electronic double layer capacitors 1,
electric motor: see sun gear 20
electric motor-generators 271 electronic equivalent circuit model
electric power assist steering 42–58
application (EPAS) 41 parameter values 58t
electric vehicles electronic resistance, at cold
ultracapacitors, application of 238 temperature 94
see also hybrid electric vehicles ELINT 25–6
(HEV) energy conversion capacitance
electrochemical capacitors (ECs) 1, 2, method 64
10–16 energy density (ED) 3, 96–7
in automotive systems 93–4 energy dissipation 315
electrochemical energy storage, energy efficiency 65, 97, 98
taxonomy of 2f under constant current conditions
electrochemical impedance 97, 97f, 99
spectroscopy (EIS) 28, 40 energy-efficient homes 155
electronic continuously variable energy management strategy (EMS)
transmissions (eCVT) 197 74, 75, 82, 84, 114, 115
Allison 6-speed, characteristics of energy recuperator/microhybrid
260 system 327
BEV mode 244 energy storage 17, 94
d.c. link of 240 energy storage system (ESS) 143
d.c.–d.c. converter interface 247 in hybrid transit bus 195, 205–6
electric variator 239 mechanical power 238
with energy storage system 243 in shuttle bus 207t
GM 2-mode 259 state-of-charge (SOC) 272
hardware illustration of 246 ultracapacitor cycling 341
in hybrid electric vehicles (HEV) ultracapacitor-only transit bus 340
241, 242 utility/vehicle 343
IPM, high-torque electric machine vehicle
244 battery-active parallel
2-mode system 260–3 combination 270
power-circulating function 240 ultracapacitor 270
power flows in 252, 253 energy versus voltage of capacitor 5f
single-mode, comparison of 259 engine
single-mode power split 258 BEV mode 243
Camry-size HEV 248, 249 crankshaft 247
core elements of 239 four-cylinder engine 247
Ford hybrid system (FHS) 235 MG1 and MG2 operating maps 244
index 7 June 2011; 15:27:28

Index 355
ON–OFF, busyness of 250 fluctuation suppression (FS) 158

power flow 248 flywheel energy storage (FES) 131
torque clusters 249 vs. ultracapacitor, in UPS 133, 135t
engine cranking forcast hedge (FH) 158
torque conditions 238 Ford hybrid system (FHS) 235
engine engagement clutch 262 forklift trucks 165
engine mechanical power fuel cell powered 165f
mechanical fraction of 240 properties of propulsion current
engine torque 250 165t
histogram plot of 250 ultracapacitor ESS in 166–8, 168f
MG1 and MG2 torque–speed 250, see also electric material handling
251 trucks
for N/V strategy 249 Formula SAE (FSAE) 216
speed trajectory 243, 244 495HF excavator 181, 182f
entropy change 41 8S9PHTM model in 182–5
epicyclic gear set: see planetary gear Franklin, Ben 1
sets frequency response, ultracapacitor
equivalent circuit simulation model electrode pore shape on 10f
58, 59f front wheel drive (FWD) 246
equivalent series resistance (ESR) 21, FS: see fluctuation suppression (FS)
66 FSAE: see Formula SAE (FSAE)
E450 shuttle van 206 fuel cell (FC) 163–4, 163f
specifications for 207t forklift trucks 165f
ESS: see energy storage system (ESS) fuel cell vehicle applications
ethylene carbonate (EC) 32 d.c.-d.c. converter for 242
EUCAR method 61–66, 62f Fuji Heavy Industries (FHI) 31
European Center for Automotive Furukawa Battery Co. 73
Research (EUCAR) FWD: see front wheel drive (FWD)
hazard level scale 316
Evans Capacitor Co. 24 Gaussian distribution function 293
EVTs: see electrically variable gear
transmissions (EVTs) ring gear design 251
excavator 176–85 GFX: see grid frequency support
Bucyrus shovel 177 (GFX)
495HF 181, 182f GM Volt 227–8, 228f
Komatsu PC200-8 176–7 specifications of 229t
type/scale of 176f GM Volt plug-in hybrid electric
vehicle 272
Faraday’s constant 12 Goodenough, John 2
fast Fourier transform (FFT) 86 grid frequency support (GFX) 158
FC: see fuel cell (FC) grid side converter (FEC) 152
FEC: see grid side converter (FEC) grid voltage regulation 140
FES: see flywheel energy storage implementation, using VAR source
(FES) 145
FH: see forcast hedge (FH) without VAR compensation 142
index 7 June 2011; 15:27:28

Harbor cranes 168–76 hybrid transit bus 189–92, 201–6

heat conduction 314 acceleration characteristics 203f
heavy transportation vehicles 189–208 components 192t
battery electric bus 199–200 energy storage system types/ratings
hybrid electric transit bus: see hybrid 196t
transit bus ESS in 195, 205–6
hybridization of 189–90 lithium-ion battery in 195
propulsion system architectures of parameters/electric drive system
190, 190f attributes 191t
shuttle bus 206–8 propulsion architecture 194f
see also buses propulsion system route adaptation
Helmholtz layer 19, 23 198t
HEV: see hybrid electric vehicles with series-parallel propulsion
(HEV) architecture 197f
HiPot equipment 317 vs. hybrid automobile 208t
HiPot test 317 hybrid vehicles
configuration 318 power management in
HiPot tester voltage 319 ultracapacitor and battery 242
history, of ultracapacitors 1–2 Hyundai Enercell Co. Ltd. 216
Honda Civic 215
HSD: see hybrid synergy drive (HSD) ICE: see internal combustion engine
HTM125, transportation modules 131, (ICE)
134f IEC62391, application of 58–62, 61f
Bucyrus 495HF rope shovel and IEC62576 62, 64, 65, 65f
177–80 IEC characterization method for
hybrid capacitors 25, 30–3 EDLCs 60f
hybrid electric powertrain increasing failure rate (IFR) 291
GM 2-mode 261 increasing function of time 292
hybrid electric vehicles (HEV) 4, 10, inductance calculation by method of
213–33, 241, 242 shells 46t
benefits of 220, 220f industrial application, of energy
dynamic simulation of 250 storage 163–85
eCVT propulsion system earth moving equipment 176–85
architecture of 242 electric material handling trucks
functions of 223–5 165–8
low-end 221f harbor cranes 168–76
performance and economy obstacle rubber tired gantry cranes 168–76
for 96 insulation resistance (IR) 317
plug-in 227–32 integrated Starter-Alternator
power assist 225–7 Reversible System (iStARS)
types of 219–23 213
hybridized energy storage systems belt-driven microhybrid concept
114–15, 122 213, 214f
converter locations 116t internal combustion engine (ICE)
hybrid synergy drive (HSD) 235 239
index 7 June 2011; 15:27:28

Index 357
internal components of ultracapacitor ultracapacitor cycling test,

cell 38f illustration of 279
internal permanent magnet (IPM) 243 of ultracapacitors 279
internal rolled electrode structure, in lift actuator 166
ultracapacitor 39f line current 136
International Electrotechnical lithium-air battery 331
Commission (IEC) 317, 336 lithium-capacitor hybrid cell, operating
inverters: see motor-generator power principle of 32f
electronics lithium-ion battery 1, 30, 31f
ionic resistance 22, 39, 94 in hybrid transit bus 195
IPM: see internal permanent magnet lithium ion capacitor (LIC) 31–2
(IPM) lithium-ion cells 3
ISE Corporation 195, 196, 201 Ragone relation for 4f
iStARS: see Integrated Starter- lithium-ion insertion chemistry
Alternator Reversible System batteries 119–20, 120t
(iStARS) comparisons 121t
lithium ion (Li-ion) 73
ladder network ultracapacitor model theoretical and practical properties
40f of 121t
lead-acid (Pb-acid) battery 3, 73, 78, lithium–iron–phosphate (LFP) cells
79, 79f 120t, and ultracapacitor
theoretical and practical properties and ultracapacitor
of 121t power performance versus
leakage current 51, 51f temperature 95–6, 95f
characteristic of, with temperature as representative ESR (T) for
parameter 52f 94–5, 95f
least squares internal resistance lithium polymer
method 64 theoretical and practical properties
life cycle testing of 121t
Arrhenius–Eyring life relationship lithium–polymer battery 272
282 lithium titanate 25–6
cell capacitance and ESRdc 281 LMnO (spinel) 120t
cell temperature 280 LNCA 120t
leakage current influence 283–6 low-end hybridization 221f
mean service life (MSL) 298–302 low-voltage ride through (LVRT) 135
methodology of 302 LTO (lithium titanate) 25, 26, 120t
reliability vs. service life 286–97 LVRT: see low-voltage ride through
exponential reliability function (LVRT)
291–2
log-normal reliability function Maple 9
293–4 matched impedance power density 65
normal reliability function 292–3 MathCAD Maple solver
Weibull reliability function definite integral 322–3
294–7 Maxwell BMOD0165-P048
resistance, calculation of 283 ultracapacitor module 341
index 7 June 2011; 15:27:28

Maxwell’s laws of electromagnetics 5 Naval Surface Warfare Center

Maxwell Technologies 216 (NSWC) 311
K2 cell line 204, 204t NEC Tokin 1
Maxwell Technologies Heavy NEDC: see New European Driving
Transportation Module 317 Cycle (NEDC)
Maxwell Technologies Inc. 37, 66 New European Driving Cycle
Mazda Miata BEV 76, 84 (NEDC) 213
mean service life (MSL) 297, 305, 306 New York City Cycle (NYCC) 229,
for hybrid transit bus 305, 306 230f, 231t, 232
for ultracapacitor cells 301 nickel-based chemistries 3
voltage stress 304 nickel-cadmium (NiCd) 3, 25
microcycles 198, 198f, 204 theoretical and practical properties
Mitsubishi Fuso Truck and Bus of 121t
Corporation 195 nickel-metal-hydride (NiMH) 3, 73,
models of ultracapacitor 37 121t
capacitor–battery combinations 73 theoretical and practical properties
active parallel architecture 74, of 121t
81–9 nickel-zinc (NiZn) 3
passive parallel architecture 74, NiMH battery 253
78–81 Nissan Diesel Corporation 195
cell characterization methods and Nomex material 324
standards 58 relative humidity (RH) 324
EUCAR method 61–6, 62f nonlinear capacitance 8f, 9f
electronic equivalent circuit model NREL: see National Renewable
42–58 Energy Laboratory (NREL)
parameter values 58t NYCC: see New York City Cycle
simulation model validation 66–73 (NYCC)
modified Randles equivalent 11, 11f, Nyquist plot 11, 14t, 15t, 47, 49f
12, 12f
module electrical and thermal design overtemperature abuse testing 311
scenario 69, 69f oxygen-positive electrode 331
module parasitic elements 319
molar mass 18 parallel plate capacitor 5
motor-generator (MG2) 263 parallel resistance 50
angular speed of 239 parameter values, of ultracapacitor
mechanical shaft of 239, 240 products 41, 48t, 58
power electronics 262 passive parallel architecture 74,
motor-generator power electronics 78–81
261, 262 PC200-8 excavator, Komatsu 176–7
MSL: see mean service life (MSL) peak power test 101
PEM fuel cell: see proton exchange
nanophosphate 120t membrane (PEM) fuel cell
National Renewable Energy personal rapid transit (PRT) 332
Laboratory (NREL) 215 PES: see Power and Energy Society
natural decay behavior 51, 52f (PES)
index 7 June 2011; 15:27:29

Index 359
PHEVs: see plug-in hybrid electric Power transmission 136f

vehicles (PHEVs) PRogram for Individual Sustainable
photovoltaic (PV) systems 155–8 Mobility (PRISM) 332
solar array 155, 156f Prokopowitz–Vaskas method 300
planar spiral, self-inductance of 47 propylene carbonate (PC) 17, 32
planetary gear sets 235, 236, 237, 262, Proton exchange membrane (PEM)
263 fuel cell 163–4
angular speed relation 236, 237 properties of 164t
power flow in 237 PRT: see personal rapid transit (PRT)
power-splitting device 237 PSA Peugeot Citroen system 213
transistor-like block diagram 236 pseudo-capacitance effect 8, 29
planetary ring gear PTC: see positive temperature
torque conditions 238 coefficient (PTC)
plug-in hybrid electric vehicle PV systems: see photovoltaic (PV)
(PHEV) 4, 74, 135, 227–32 systems
cycling capability of batteries and
ultracapacitors 94 quaternary tetraethylammonium 17
performance and economy obstacle
for 96 Ragone chart 3
Volt 227–8, 228f, 229t of electric energy storage devices 3f
Poisson–Nernst–Planck theory 16 for lithium-ion cells 4f
positive temperature coefficient Ragone relationships 112–14
(PTC) 222 Randles circuit 41
power and energy 93 Randles equivalent 11, 12, 12f
Ragone relationships 112–14 range extended vehicle (REV) 272
specific energy and energy density engine and functional schematic
96–100 273
specific power and power density reactive power 139
101–12 REC: see rotor side converter (REC)
ultracapacitors and batteries regenerative energy storage systems
114–23 (RESS) 101
Power and Energy Society (PES) 155 reliability 287
power assist hybrid electric vehicles renewable energy resource (RER)
225–7 157
performance metrics for 225f renewable energy sources (RES) 135
power buffer RER: see renewable energy resource
ultracapacitor, application of 250 (RER)
power converter 306 RES: see renewable energy sources
power electronic inverters 238, 243 (RES)
PowerNet set point 289 REV: see range extended vehicle
PowerNet voltage, hazard function (REV)
290, 291 ring gear
power performance 66, 95, 101 ICE power and MG2 power 240
power splitting devices 262 rotor side converter (REC) 152
power-split transmissions 235, 239 round trip efficiency 100
index 7 June 2011; 15:27:29

RTG cranes: see rubber tired gantry using CSC power inverter 141f
(RTG) cranes using VSC power inverter 142f
rubber tired gantry (RTG) cranes state-of-charge (SOC) 17, 165
168–76 static synchronous series compensator
deceleration rates 175t (SSSC) 140
regeneration power level 175 static VAR compensator (SVC) 136,
140
self-discharge data versus log(time) aluminum electrolytic capacitors
52, 53f and 143
self-inductance of planar spiral 47 and STATCOM 145
SGIC: see smart grid information statistical process control (SPC) 56
clearing house (SGIC) straddle crane: see rubber tired gantry
shuttle bus 206–8 (RTG) cranes
ESS pack parameters 207t sulfuric acid electrolyte 26, 33
simulation model validation 66–73 sun gear
single-mode power split definition of 238
drive cycle evaluations 246–53 MG1, angular speed of 240
electronic continuously variable shaft torque 240
transmission (eCVT) 239–42 supercapacitor 1, 4
ultracapacitor application in surface electrolyte interphase (SEI)
242–5 30, 33
650 F ultracapacitor cell 66 surface redox 8
smart grid information clearing house SUVs: see sport utility vehicles (SUVs)
(SGIC) 155 SVC: see static VAR compensator
SOC: see state-of-charge (SOC) (SVC)
Society of Automotive Engineers symmetric solvent and salts, properties
(SAE) 336 of 17t
sodium-metal chloride symmetric ultracapacitor 17–24, 30,
theoretical and practical properties 33
of 121t
sodium-sulfur Tafel relation 8, 28
theoretical and practical properties tantalum pentoxide 24, 25
of 121t TC: see transmission curtailment (TC)
SOHIO (Standard Oil Company of temperature histogram data
Ohio) 1 for Phoenix, Arizona 304
Sony 2 tetraethylammonium tetraflouroborate
specific energy (SE) 3, 96, 97 (TEATFB) 17
specific power and power density tetraflouroborate 17
101–12 THD: see total harmonic distortion
sport utility vehicles (SUVs) 235, 258 (THD)
Allison hybrid system (AHS) 271 thermal capacitance 69
SSSC: see static synchronous series thermal characterization 66
compensator (SSSC) thermal insulation 319
STATCOM 140–1 thermal management 69
SVC and 145 thermal tape 311
index 7 June 2011; 15:27:29

Index 361
3000 F cell modules inductive power transfer systems

module parasitic capacitance 318, 337–8
323 Witricity demonstrator
overvoltage abuse test 314 of wireless power transfer 334
3000 F ultracapacitor 39, 47 TS: see time shifting (TS)
functional fit to measured data of turbine power coefficient 147
54f 2-mode power split
moment matched equivalent circuit electrically variable transmissions
model of 40f (EVTs) 257
Nyquist and Bode plots for 49f operational modes of 263–9
Ragone plot of 113f 6-speed automatic transmission
validation of 66, 68f 258
time shifting (TS) 157 electronic continuously variable
torque–speed scatter transmissions
of MG1/MG2 251, 252 essentials of 260–3
total harmonic distortion (THD) 130 ultracapacitor application in
Toyota hybrid system (THS) 235 269–71
transmission curtailment (TC) 157 plug-in hybrid electric vehicle
transmission line voltage drop 144 Volt 272–5
transportation systems, future of 331 types of, ultracapacitors 1
antenna theory 335 asymmetric ultracapacitor 24–30
antenna transmission zones, electrochemical capacitors
definition of 334 10–16
dipole antenna, illustration of 335 hybrid capacitors 30–3
dual-mode vehicle 332 symmetric ultracapacitor 17–24
electric transportation 331
electromagnetic power UDDS: see Urban Dynamometer Drive
transmission 333 Cycle (UDDS)
inductive power transfer (IPT) UltraBattery 73
systems 336 ultracapacitor
IPT system, ultracapacitor capacitance fade
applications 338–43 definition of 298
mobility systems 332–3 geometry of 321
narrow lane vehicle (NLV) 332 vs. flywheel energy storage, in UPS
plug-in hybrid cars 332 133, 135t
PRISM vehicle 332 ultracapacitor abuse
stationary wireless power transfer abuse testing, need for 311–13
systems 331 cycling exposure 325–8
ultracapacitor cycling 341 dielectric withstand potential 317
ultracapacitor-only transit bus, factors 311
wireless charging of 339, 340 insulation resistance 317–24
wireless power transmission (WPT) overcurrent 313–16
333–8, 336 overvoltage 313–16
coaxial winding transformer vibration requirements 324–5
336–7 vibration specification 324
index 7 June 2011; 15:27:30

ultracapacitor–battery combinations HEV strategy response 249

74 to US06 drive cycle 249, 250
ultracapacitor cells 280, 282, 306, 312 US06 Drive Cycle 249
overvoltage 311
ultracapacitor leakage valve regulated lead acid (VRLA)
current test 284 battery 216
taxonomy of 284 van der Waals force 27t
ultracapacitor manufacturer, variable structure controlled circuit
compliance 324 (VSC) 83
ultracapacitor module 317 vehicle-2-grid (V2G) 135
measured vs. modeled temperature vehicle system controller (VSC) 197
342 Vestas 4.5-MW wind turbine 146
stray capacitance of 320 V2G: see vehicle-2-grid (V2G)
voltage 303 volt, GM 227–8, 228f
ultracapacitor-only transit bus 343 specifications of 229t
ultracapacitor overcurrent abuse voltage bin data
adiabatic response 314 for transit bus operating
ultracapacitor products, parameter in Phoenix, Arizona 304
values of 48t voltage stabilizer 135–45
ultracapacitors, automotive voltage transitions 302
expectations of 315 voltage versus time of capacitor 5f
ultracapacitors cycling voltage waveform 279
constant rms profile 327 Volt range-extended vehicle 272
energy recuperator profile 328 VRLA battery 79, 80: see valve
engine starting profile 327 regulated lead acid (VRLA)
types of 325 battery
waveforms 325 VSC: see vehicle system controller
quasi-square wave profile 326 (VSC)
square wave testing 325 Vue Greenline, mild-hybrid vehicle
uninterruptible power supply (UPS) 214, 215f, 216
129–35 dynamometer testing of 216
Eaton UT3220 220-kVA three-
phase, parameters of 130t Warburg impedance 11, 11f, 13, 41
efficiency of 129 Warburg modified Randles equivalent
installation 131–2, 131f, 132f 11f, 12
run time 130 WEC: see wind energy converter (WEC)
ultracapacitor vs. flywheel energy Weibull distribution 293, 294
storage in 133, 135t Weibull function 292
up-convert case of active parallel ESS Weibull paper abscissa 294
84, 84f Weibull probability density functions
UPS: see uninterruptible power supply 147–8, 148f
(UPS) Weibull reliability distribution
Urban Dynamometer Drive Cycle function 295
(UDDS) 75, 75f, 76, 81f, 216, Weibull survivor function
229, 230f, 231t, 232, 249 for ultracapacitor cell 297
index 7 June 2011; 15:27:30

Index 363
wide-open-throttle (WOT) speed 225 WOT speed: see wide-open-throttle

wind energy converter (WEC) 146 (WOT) speed
mechanical power output from 147
wind turbine systems 146–55 XH-150 plug-in hybrid 115
in phantom view 149, 150f
sizes 146 zero emission vehicle (ZEV) 189
ultracapacitor module 149, 151, 151f ZEV: see zero emission vehicle
use of energy storage in 152 (ZEV)
Vestas 4.5-MW 146
index 7 June 2011; 15:27:30

Power and Energy Series 59
Ultracapacitor
Energy storage and in particular electrical storage of energy
has become a very talked about topic in circles ranging
from lay persons, in regard to hybrid and battery electric
Dr. John M. Miller was vice president of systems and
applications at Maxwell Technologies when this book was
written. He is currently with the U.S. Department of Energy, Oak
Applications
vehicles, to professionals, and certainly by legislators Ridge National Laboratory, National Transportation Research
Center. He is also founder and principal engineer of J-N-J Miller
and energy policy makers in government. But even
Design Services, P.L.C. Dr. Miller worked for 20 years in the
to professionals the distinctions between physical and automotive industry, leading several hybrid vehicle technology
chemical forms of electric energy storage are unclear programs including 42V Integrated Starter Alternator, ISG, for
and at times poorly understood, if at all. This book takes application into a SUV. He has been active in collaborations at
a critical look at physical storage of electricity in the the industry and government levels including the NSF funded
devices known collectively as electrochemical capacitors systems center for Future Renewable Electric Energy Delivery
and particularly as ultracapacitors. In its 12 chapters, this and Management, FREEDM. He was actively engaged in MIT’s
text covers ultracapacitor and advanced battery topics Consortium on Advanced Automotive Electrical and Electronic
Components and Systems and has served as Adjunct Professor
John M. Miller
with emphasis on clear understanding of fundamental
of Electrical Engineering at Michigan State University and at
principles, models and applications. The reader will Texas A&M University. Dr. Miller has authored over 170 technical
appreciate the case studies ranging from commercial publications, holds 53 U.S. patents, and authored or co-authored
to industrial to automotive applications of not only five books. He is a Fellow of the IEEE, Member of SAE, 2009
ultracapacitors, but of these power dense components in recipient of the IEEE Kliman Innovator award, and 2010 recipient
combination with energy dense battery technologies. of the IEEE Power Electronics Society distinguished service award.
Miller

www.theiet.org
978-1-84919-071-8
PBPO0590.indd 1 27/05/2011 16:35:40

Ultracapacitor Applications

Încărcat de

Informații document

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Ultracapacitor Applications

Încărcat de

Drepturi de autor:

Formate disponibile

Power and Energy Series 59

The Institution of Engineering and Technology

PBPO0590.indd 1 27/05/2011 16:35:40

PRELIMS 7 June 2011; 18:2:38

PRELIMS 7 June 2011; 18:2:38

The Institution of Engineering and Technology

PRELIMS 7 June 2011; 18:2:38

British Library Cataloguing in Publication Data

ISBN 978-1-84919-071-8 (paperback)

Typeset in India by MPS Ltd, a Macmillan Company

PRELIMS 7 June 2011; 18:2:39

PRELIMS 7 June 2011; 18:2:39

3 Power and energy 93

4 Commercial applications 129

PRELIMS 7 June 2011; 18:2:39

5 Industrial application 163

6 Heavy transportation application 189

7 Hybrid electric vehicles 213

8 Single-mode power split 235

9 2-mode power split 257

10 Life cycle testing 279

PRELIMS 7 June 2011; 18:2:39

10.3 Mean service life 298

11 Abuse tolerance 311

12 Future transportation systems 331

PRELIMS 7 June 2011; 18:2:39

PRELIMS 7 June 2011; 18:2:39

PRELIMS 7 June 2011; 18:2:39

Ultracapacitors, or to be technically correct, electronic double layer capacitors,

ch001 7 June 2011; 13:4:39

Electrochemical energy storage

Primary Secondary Electrostatic Electrolytic Electrochemical

Most popular today Active research

Figure 1.1 Taxonomy of electrochemical energy storage

● Insertion chemistry and intercalating cathode of LiCoO2 patented by John

ch001 7 June 2011; 13:4:39

Electrochemical capacitors would be located in the area of approximately 10

ch001 7 June 2011; 13:4:39

Lithium-ion power versus energy including hybrid-cap Very high power

Figure 1.3 Ragone relation for lithium-ion cells

ch001 7 June 2011; 13:4:39

ch001 7 June 2011; 13:4:39

Table 1.1 Properties of dielectrics

Material K-factor Ubd (MV/m)

ch001 7 June 2011; 13:4:39

Equation (1.6) is the familiar expression for capacitor energy as shown in

The integral in (1.9) is easily solved by referring to tables of integrals for

ch001 7 June 2011; 13:4:40

Onset of nonideal 2600

Time (at constant current i) Current

ch001 7 June 2011; 13:4:40

Figure 1.7 Data and fit of nonlinear capacitance C(u)

The solution to (1.18) is analytically difficult to obtain in closed form so a

ch001 7 June 2011; 13:4:40

Table 1.2 Solution to the nonlinear charge versus potential of the

Component of (1.18) to integrate Result of numerical integration

W fb ¼ C b 0 mx x tanh Uxx U x dx Wfb ¼ 1.385  103

1.1 Electrochemical capacitors

Figure 1.8 Influence of electrode pore shape on ultracapacitor frequency

ch001 7 June 2011; 13:4:41

ch001 7 June 2011; 13:4:41

Z w ðwÞ ¼ Z 0 ðwÞ jZ 00 ðwÞ ð1:21Þ

Figure 1.10 The Warburg-modified Randles equivalent circuit for an EC

W fb ¼ C b 0 mx x tanh Uxx U x dx Wfb ¼ 1.385 103

Diffusion Dc ¼ 5.6 10–8 Dc ¼ 9 10–7 Dc ¼ 8.8 10–5

Diffusion coefficient Dc ¼ 5.6 10–8 Dc ¼ 9 10–7 Dc ¼ 8.8 10–5

¼ 3:56 1026 N ð1:31Þ

The electrostatic pressure given by (1.32) is 3.09 TGPa (terra (1012)

Van der Waals force ( 10–11 N) Approach distance, rw