Documente Academic
Documente Profesional
Documente Cultură
Ultracapacitor Applications
Ultracapacitor
Ultracapacitor Applications
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
Ultracapacitor
Applications
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)
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:
The Institution of Engineering and Technology
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.
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
Glossary 347
Index 351
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
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
Battery Capacitor
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)
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.
+ +
+ ++ +
+ + 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)
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).
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
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
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
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
3
30
0
A 2.5
20
0
2
10
–1 0
1.5
00
–3 0
1
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
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
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
1
Available in MathCAD software from MathSoft ver. 14 used here.
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
a b
c
–Imaginary Z
d
b
c e
d
a
e
Real Z
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
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).
Cdl
Rs
Rct Zw
W
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
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
Imaginary part of Zw
Imaginary part of Zw
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
10
W(i) W(i) 1 W(i)
1
Magnitude |Zw|
Magnitude |Zw|
Magnitude |Zw|
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
Angular frequency Angular frequency Angular frequency
Table 1.4 Representative Nyquist and Bode plots of a small 10 F ultracapacitor with Zw in Cdl branch
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)
Magnitude |Zw|
Magnitude |Zw|
Magnitude |Zw|
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
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.
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.
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 1.13 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,
E E
d d
f(x) E(x) n(x)
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
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.
Q2 ð8100Þ2
F¼ ¼
4per e0 d 2c 12:56 ð37:5Þð8:854 1012 Þð44:18 1018 Þ
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
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.
(+) Terminal
Jelly roll
Paper
Film
Film Foil
(–) Terminal
Re Ri Re
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
1.2 × 10–4
0
0
−50 −20 10 40 70 100
–40 a(i), x 70
Temperature (°C)
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.
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
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
the f ¼ 60 mm can diameter size and time constant t*0.65 s on average as shown
in Figure 1.17.
1 × 103
C(i)
C1
100
10
1 × 10–4 1 × 10–3 0.01
R(i), Rmx
ESR (Ohm)
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
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.
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
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
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
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
Rct Cj
Rs
Cd
R
e–
D
e–
C
(+) (–)
Cathode Anode
C = Charge
D = Discharge
= Ion
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.
6C þ Liþ þ e ! C Li ð1:53Þ
DC 6
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.
1
–
ELIC
0 t
tchg tdch – EDLC
tdch – LIC
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
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:
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.
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
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
Seminar on Double Layer Capacitors and Hybrid Energy Storage Devices,
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
1 1 1
¼ þ ; where Ca U ¼ Ck U ¼ C dlc (2.2)
C cell C a U C k U
1
Models are available for download from www.ansoft.com for users of Simplorer version 8.0 software.
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
Opposite electrode
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.
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].
+ # #
~ V C C1 C2 # Rsd
I1
VM1 XY
~ XY
~ XY
~
XY_LINT1 XY_LINT2 XY_LINT3
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.).
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
Rsa Cs
Rconn Lconn Rs
Uc
TLU: C(Uc) C0
Rp
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.
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)
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)
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)
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
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
Current (A)
12
10
8
6
4
2
0
0 5 10 15 20 25 30 35 40 45
Shell #
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
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.
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
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
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.
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.
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
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
–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
1 1
¼ (2.16)
tca ðRs þ Rsa ÞC s
0:455
Cs ¼ (2.17)
Rs þ Rsa
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
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.
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 )
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
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.
Intrinsic capacitance (F) Differential capacitance (F) Inflection point voltage (V)
Ca ¼ 2770 Cb ¼ 520 Ux ¼ 0.9
Capacitance
2500
2000
1500
1000
500
0
0 0.5 1 1.5 2 2.5 3
Voltage
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.
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
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)
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)
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).
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.
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
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).
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)
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.
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
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
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
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.
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 2.26, 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:
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.
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
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.
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.
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
Magnified figure
UR
ΔU3
U1
Voltage (V)
ΔU3
U2
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)
2W 0
C¼ (2.38)
ð0:9U R Þ ð0:7U R Þ2
2
0:25U 2R
Pdm ¼ (2.39)
ESRdc M
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.
Uc
TLU: C(Uc) Co
Rp
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)
CUL
Test C/C 643 Ω
t y V0 = 1.366
~
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
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
Ic, Uc
Rth1
Cth1
Rth2
Cth2
Rth_NM
Cth_NM
Ta
Gnd
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
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.
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)
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
2.5 2.5
VM1.V (V)
VM1.V (V)
2 2 2
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
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).
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
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.
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
Table 2.8 Energy storage system current magnitudes by component for battery
electric vehicle on Urban Dynamometer Drive Cycle
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)
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
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.
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)
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
– +
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
–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.
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)
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
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
Battery pack
Uc T Ud
Ultracapacitor Id
S1
pack L1 D1
Ic C2
C1 S2
D2
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
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%
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))
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))
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
● Battery heating reduced by 58%
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.
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
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)
ch002
IMC M/G
= ~
EMC
Id
Ud
ESS
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
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)
10,000
− a.c. 0
70
60
Power (W)
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
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.
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
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
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
25
Power
20 Voltage
15
10
0
1992 1996 2000 2004 2008 2012 2016
Year
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.
50
40
30
20
HEV
10
0
1000 10,000 100,000 1,000,000
Cycles (#)
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
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)
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.
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).
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.
For the energy density calculation, apply the stated cell diameter and length using
(3.3).
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.)
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
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
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
Stored Available
Ws(t)
WC Wa
Wdc Wdd
t
Dissipated Dissipated
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
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).
U 2mx
PML ¼ ð3:20Þ
4ESRdc
U 2mx
SPML ¼ ð3:21Þ
4ðESRdc ÞM
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
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)
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
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
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
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).
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.
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
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).
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
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).
MUL1.VAL
940.00
920.00
900.00
880.00
850.00
0 2.5 5.0 7.5 10.0 15.0
C0 2
3150 2
W sto ¼ U mx U 2cf ¼ 2:7 1:472 ¼ 8078 J ð3:48Þ
2 2
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
+Uc0
Uc(t) U0(t)
C0
ic(t) P0(t)
Figure 3.15 Ideal ultracapacitor equivalent circuit under constant power loading
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
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
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
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.
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
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,
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
Figure 3.16 Ragone plot of 3000 F ultracapacitor with discharge voltage level as
parameter
C 2
3C 2
W avail ¼ U mx U 2c ¼ U ¼ 8201 J ð3:63Þ
2 8 mx
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
Boost ð1 d 1 ÞU d 1
I
ð1 d 1 Þ 0
UdI0
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.
Table 3.5 Specifications of Nissan Leaf BEV and lithium-polymer battery pack
(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
$25, 000
$20, 000
Cost
$16, 667
$15, 000
$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%
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.
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
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.
Useable δSOC
energy with ultracapacitor
+ converter
30
Power at EOL
20
20
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
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.
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
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?
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.
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
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].
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
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
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.
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.
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
(a) (b)
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
Taking the conjugate of (4.9) and clearing the denominator results in the
expression for I s ,
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
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
Figure 4.5 Transmission line with fixed and switched load and fixed VAR
injection
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
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 924:7ffð24:4Þ ð4:18Þ
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
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
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.
¼ 850:2ffð8:37Þ ð4:21Þ
SCADA
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
80.00
40.00
20.00
–500.00
0 25.00 50.00 75.00 100.00 125.00 150.00
t
(a)
Source voltage and current
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
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
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
Ud
Ess
VSC inverter
magnitude, Bc. With these modifications, (4.22) can be represented as (4.23), which
includes the power law dependency on bus voltage, Vr.
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).
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.
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
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).
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
Q-V droop
Vs
V0
V
dV kQ
f
Vr ΔV
Ir
Q0 Q
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
30 min
10 h/year
UH No power reduction
1h
10% reduction
UHF
35 min 3 min
Power Normal No requirement for
Voltage
ULF
1h
10% reduction
UL
47.5 49 50.5 51 53
Frequency (Hz)
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
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.
0.18 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)
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
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
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]
Commercial applications 151
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
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
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]
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
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
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.
deteriorated in the late afternoon to partly cloudy. The power variability due to
cloud passing is shown by the nearly 8-MW power fluctuations.
700
600
Current (A)
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
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
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
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.
1
Mears, ‘Application of Energy Storage to Enhance Wind Generation,’ DOE Energy Storage Peer
Review Meeting, 10–11 November 2004
● 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
SCADA
(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
Source voltage and current
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)
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, . . . .
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
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
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.
2F
iFC ¼ n_ H ð5:2Þ
Nc 2
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.
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
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
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
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
600.00
500.00
400.00
Pdisp
300.00
200.00
100.00
0
0 10.00 20.00 30.00 40.00
t
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.
(a) (b)
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
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.
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
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.
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
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).
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.
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)
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.
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
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.
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.
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.
2.00
1.50
P0
1.00
500.00
–300.00
Figure 5.12 Representative power profile of the 495HF excavator (power in MW,
time in seconds)
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
495HF terminal power (kW)
8S × 16P × HTM module
Rconn Lconn 14.4 mΩ 21 F 2.80
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,
and energy output and dissipated are shown graphically in Figure 5.14. The effi-
ciency over the full operating profile is approximately 97.8%.
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)
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.
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.
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,
Rsa C1
8S × 9P × HTM module
Rconn Lconn 25.6 mΩ 11.8 F
Rs
2.98 mΩ 1.5 PH
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
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)
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
35
30
25
20
15
10
5
0
0 2000 4000 6000 8000 10,000 12,000 14,000 16,000
Time (s)
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
_ H2 LHVH2 4
W_ comb ¼ M !M_ H2 ¼ 23:96 ¼ 0:1995 g=s
120:1
(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.
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.
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:
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
● 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
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
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).
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
(b) Given that gxm ¼ 4.75:1, then cascading (6.6) and (6.7) down the driveline
results in
weng 230:38
wdl ¼ ¼ ¼ 48:5 rad=s ð6:9Þ
g xm 4:75
(c) Given that gfd ¼ 2.83:1, so cascade (6.8) and (6.9) to the driven wheels to get
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
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.
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
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
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
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
Table 6.3 Transit bus energy storage system types and ratings [5]
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.
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
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.
50
0
–50 0 5 10 15 20 25 30 35 40 45 50
–100
–150
–200
Time (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]
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.
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 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
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
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.
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
(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)
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
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 6.9 Standard modules for bus, wind turbine, and train applications
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
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)
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.
(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
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.
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)
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.
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.
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).
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 (#)
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
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,
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
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
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
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
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
50 50
45 45
40 40
Fraction of time (%)
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.
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)
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.
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
Ultra-capacitor
Controlled regene-
Stop & rative braking Wheels
Dry clutch
start
Engine
(gas or diesel) Transmission
Tank
Electric power
Drive power
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)
Ultra-capacitor
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,
Start-stop
Recuperation
Active boosting
Electric driving
Power consumer
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.
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
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,
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.
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 7.9 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.
V, P (V)
IMC ~ M/G
=
EMC
Id
Ud
ESS
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)
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)
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
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)
(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
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
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
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?
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.
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
Continuously variable
transmissions CVT
Electric
Mechanical Hydraulic
eCVT
IVT (Renault)
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
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
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.
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.
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
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.
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
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
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.
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:
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.
Wheels
Epicyclic
Inverter
274 V
MG2 R
ICE C FD
Battery
MG1 S
Ultra-
capacitor
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
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
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
(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:
The engine must therefore be started and run at 2235 rpm to avoid over-
speeding of the electric motor-generators.
(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
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
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.
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)
Pe + Pb
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
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)
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)
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)
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
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
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)
20
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)
200
Torque (N m)
100
0
–2000 –1500 –1000 –500 0
–100
–200
–300
Speed (rad/s)
50
0
–2500 –2000 –1500 –1000 –500 –50 0
–100
–150
–200
Speed (rad/s)
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.
● 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.
Pe + Pb
Pm motor
P e – Pg Generator Pg
Pe
Engine
direct to
wheels
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).
Wheels
Epicyclic
Inverter
274 V
MG2 R
ICE C FD
Battery
MG1 S
Ultra-
capacitor
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%
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
2. J.M. Miller, Propulsion Systems for Hybrid Vehicles, 2nd edn., The Institution
of Engineering and Technology (IET), Michael Faraday House, Stevenage,
United Kingdom, 2010
3. 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
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
Engine Engine
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
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)
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
(a)
Engine
Battery control unit
Planetary gearset A
Planetary gearset B
engaged friction
Two selectively
clutches
Dual-mode electronically
variable transmission
(b)
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.
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
● 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.
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
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
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.
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
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
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
1 1
w2 ¼ þ k1k2 1 wi ð9:13Þ
K K
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
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
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.
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
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
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.
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)
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.
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
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)
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.
53 kW
3-cylinder
1-L genset
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).
(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)
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
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:
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
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
(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
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.
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
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
2. J.M. Miller, Propulsion Systems for Hybrid Vehicles, 2nd edn., The Institution
of Engineering Technology (IET), Stevenage, Herts, United Kingdom, 2010
3. 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
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
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)
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)
385 7
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
FðtÞ ¼ 1 eðt=aÞ
b
ð10:1Þ
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.
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.
C(U) = C0 + kuU
{
t→∞
– + Vd.c.
Agilent Agilent
6031A + – 34970A
w/34901A
Rext
Jumper
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.
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.
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.
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.
Failures
Wear-out
dFðU Þ 1 2
¼ pffiffiffiffiffiffi eðU mÞ =2s
2
f ðU Þ ¼ 81<U <1 ð10:7Þ
dU 2ps
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)
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)
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
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
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)
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.
H2 (k) 0.06
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
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:
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).
normal distribution function. The metrics of this distribution are given in (10.16)
through (10.19).
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
The expectation value and variance of the log-normal function are given as
(10.22) and (10.23).
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
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
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.
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)
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 10.16 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.
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.
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).
Ci + 20%
Cμ
Ci
–t
Cn C0e τ
Capacitance (F)
a2t2 + a1t + a0
–25%*Ci
b1t + b0
C2y
C3y
Ceol
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
the exponential form of (10.31) results in the more common form of this life esti-
mation function as
n
Ux
MSLðU x ; T x Þ ¼ MSLðU 0 ; T 0 Þ eðEa =kÞ½ð1=T 0 Þð1=T x Þ ð10:33Þ
U0
n a b
Ea 1 1 Ux Ux Tx
LnðMSLx Þ ¼ LnðMSL0 Þ Ln D ð10:34Þ
k T0 Tx U0 U0 T0
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
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
● MSL(2.3 V, 40 C) ¼ 70,000 h?MSL(2.3 V, 55 C) ¼ 39,993 h, a 42.9% life
reduction
● MSL(2.7 V, 40 C) ¼ 47,128 h?MSL(2.7 V, 55 C) ¼ 19,489 h, a 58.6% life
reduction
0 t (s)
0 100 150 258 288
Loading acceleration Cruise deceleration
1400
1200
Cummulative time (h)
1000
800
600
400
200
0
0 5 10 15 20 25 30 35 40 45
Temperature bins (°C)
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
● MSL(2.6 V, 30 C) ¼ 84,542 h for 10.6% of time by year and 20% by voltage?
0.156
● MSL(2.1 V, 30 C) ¼ 101,722 h for 78% of time by year and 15% by voltage?
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
● MSL(2.6 V, 45 C) ¼ 43,608 h for 10.6% of time by year and 20% by voltage?
0.021
● MSL(2.1 V, 45 C) ¼ 54,176 h for 78% of time by year and 15% by voltage?
0.0156
● MSL(1.3 V, 45 C) ¼ 65,120 h for 10.4% of time by year and 65% by voltage?
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.
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
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.
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.
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
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
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.
(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)
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
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)
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)
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
BCAP0350
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
Table 11.1 Dielectric withstand potential, a.c. excitation 40–70 Hz, for 60 s
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
dV
I stray ¼ C s ¼ ð3:3 109 Þð757:6 V =msÞ ¼ 2:5 A ð11:6Þ
dt
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
Cs1
Cs2
Cs3
Csk Cs(k + 1)
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.
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.
(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)
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
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
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
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
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
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
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).
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
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.
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.
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.
I0
0 t
I (A)
–I0
Uc
Umx
Umx
2
0 t
I0
0 t
I (A)
–I0
Uc
Umx
Umx
2
0 t
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
Constant rms
I1
I2
I3
I (A)
0 t
–I3
–I2
–I1
I3
0 t
–I2
–I1
I (A)
U
U00
0 t
0 t
I (A)
Uc
Umx
Umx
2
0 t
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
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
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
References
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.
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.
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
Source
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
1 108
c ¼ pffiffiffiffiffiffiffiffiffi ¼ pffiffiffiffiffiffiffiffi ¼ 3 108 m=s ð12:8Þ
m0 e0 1=9
! !
Z 0 H r ¼ n Er ð12:9Þ
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
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.
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.
Power distribution
rod (kHz)
Magnetic shield
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.
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 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
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.
0 t
Uc
Umx
Umx
2
0
t
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
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
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
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
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
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
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
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
1. J.M. Miller, Propulsion Systems for Hybrid Vehicles, 2nd edn., The Institution
of Engineering Technology (IET), Michael Faraday House, Stevenage,
United Kingdom, 2010
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
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
Ultracapacitor Applications
Ultracapacitor
Ultracapacitor Applications
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