Journal of Energy Storage 12 (2017) 297–304

Contents lists available at ScienceDirect

Journal of Energy Storage

journal homepage: www.elsevier.com/locate/est

Analytical sizing methods for behind-the-meter battery storage$

Di Wua,* , Michael Kintner-Meyera , Tao Yangb , Patrick Balduccia
a
Pacific Northwest National Laboratory, Richland, WA 99354, USA
b
University of North Texas, Denton, TX 76203, USA

A R T I C L E I N F O A B S T R A C T

Article history:
Received 12 January 2017 In behind-the-meter application, battery storage system (BSS) is used to reduce a commercial or
Received in revised form 21 April 2017 industrial customer's payment for electricity use, including energy and demand charges. The potential
Accepted 22 April 2017 value of BSS in payment reduction and the optimal size can be determined by formulating and solving
Available online 4 June 2017 standard mathematical programming problems. In such mathematical programming methods, users
input system information such as load profiles, energy/demand charge rates, and battery characteristics
Keywords: to construct a standard programming problem, which typically involves a large number of constraints
Battery sizing and decision variables. The problems are then solved by optimization solvers to obtain numerical
Behind-the-meter
solutions. Such kind of methods cannot directly link the obtained optimal battery sizes to input
Energy storage system
parameters and requires case-by-case analysis. In this paper, we present an objective quantitative
Optimization
analysis of costs and benefits for customer-side BSS, and thereby identify key factors that affect optimal
sizing. We then develop simple but effective guidelines for determining the most cost-effective battery
size. The proposed analytical sizing methods are innovative, and provide engineering insights on how the
optimal battery size varies with system characteristics. We illustrate the proposed methods using
practical building load profile and utility rate. The obtained results are compared with the ones using
mathematical programming based methods for validation.
© 2017 Published by Elsevier Ltd.

1. Introduction play an even more significant role in increasing operation

The operation of electric power sector requires flexibility to
realize instantaneous balance between generation and constantly
changing demand. Wind and solar power have been developed
rapidly in recent years world-wide. With their natural variability
and uncertainty, these renewable energy resources introduce
additional variation to power systems and present difficulties and
challenges to system operators. Energy storage system (ESS) has
been a candidate for meeting flexibility and reserve requirement
from power grid for years. Recent developments and advances in
energy storage and power electronics technologies are making the
application of energy storage a viable solution for increasing
flexibility and improving reliability of power systems. When
delivered via behind-the-meter application, energy storage can
flexibility and reducing energy cost.
Many studies have been devoted to various ESS technologies for
grid applications. The technical and economic characteristics of an
array of ESS technologies are reviewed and compared in [1–3],
including pumped hydro, battery storage, flywheel, compressed
air, superconducting magnetic energy storage, and advanced
capacitors. Studies [4,5] are dedicated to various battery technolo-
gies and methods of assessing their economic viability and impacts
on power systems. The identified grid applications include energy
arbitrage/load leveling, frequency regulation, load following,
spinning and non-spinning reserves, T&D deferral etc. A review
of several ESS technologies for wind power applications is provided
in [6]. There are also many studies focusing on the optimal
operation and sizing for BSS, and can be categorized into
transmission, distribution and customer-side applications.

Transmission: In [7], the authors evaluate the economic

$
This material is based upon work supported by the U.S. Department of Energy performance of NaS batteries for energy arbitrage and flywheels
(DOE), Office of Electricity Delivery and Energy Reliability through its Energy for regulation services based on fixed utilization factors for
Storage program, and Washington State Department of Commerce Clean Energy
NYISO and PJM systems. Ref. [8] incorporates realistic CAISO
Fund. Pacific Northwest National Laboratory is operated for the DOE by Battelle
Memorial Institute under Contract DE-AC05-76RL01830. regulation signals and battery responses to yield more granular
* Corresponding author. results. In [9], the authors investigate the application of battery
E-mail address: di.wu@pnnl.gov (D. Wu).

http://dx.doi.org/10.1016/j.est.2017.04.009
2352-152X/© 2017 Published by Elsevier Ltd.
298 D. Wu et al. / Journal of Energy Storage 12 (2017) 297–304
Nomenclature
a battery equivalent capital cost with respect to energy
size ($/kWh)
b battery equivalent capital cost with respect to power
size ($/kW)
Ecycle the maximum amount of load that can be shifted
from peak hours to off-peak hours considering
battery storage power capacity
Eoffpeak the amount of energy charged during off-peak hours
Emax battery storage energy capacity
Epeak the amount of energy discharged during peak hours
ND number of days in a year
NM number of months in a year
pmax battery storage power capacity
roffpeak off-peak rate ($/kWh)
rpeak peak rate ($/kWh)
Toffpeak duration (in h) of off-peak period
Tpeak duration (in h) of peak period
a capital recovery factor (levelized annual fixed-
charge rate)
b monthly demand charge rate ($/kW/month)
Dr saving in energy charge by cycling per unit energy
from off-peak to peak time period
hround battery round-trip efficiency
ropt optimal ratio of energy size to power size for a
battery storage system
storage system (BSS) for relief of transmission line thermal
constraints, and therefore increase the transfer capability. Based
on a case study, an economic analysis of benefits and costs is
provided. Ref. [10] presents control algorithms and sizing
strategies for using BSS to manage energy imbalance for variable
generation resources. The authors in [11] develop an optimal
control strategy of BSS for smoothing out the intermittent power
from the wind farm. In [12], an evaluation framework and co-
optimization are proposed to assess BSS economic performance
considering multiple grid applications simultaneously, including
energy arbitrage, balancing services, capacity value, distribution
upgrade deferral, and outage mitigation. In [13], a method is
proposed for identifying the sites where BSS should be located to
perform spatial-temporal energy arbitrage most effectively and
the optimal size of these systems.

Distribution and microgrid: The authors of [14] present an
optimal sizing method for BSS in a microgrid, where the unit
commitment problem with spinning reserve for microgrid is
formulated as a mixed linear integer problem. Ref. [15]
investigates the potential of using BSS in the public low-voltage
distribution grid, to defer upgrades needed to increase the
penetration of photovoltaics, where a multi-objective optimiza-
tion method is proposed to visualize the trade-offs between
three objective functions: voltage regulation, peak power
reduction, and annual cost. An optimal scheduling and cost-
benefit analysis for microgrid applications is developed in [16].
Dynamic programming is used to solve the optimal scheduling
problem and determine the optimal BSS power and energy
ratings for both isolated and grid-connected microgrids. The
authors in [17] formulate an optimization problem to minimize
microgrid energy cost using BSS subject to operational con-
straints. A recent optimization technique called grey wolf
optimization is then applied to solve the problem and determine
the optimal BSS size. In [18], the authors develop a primary
frequency control scheme for islanded microgrid using BSS. Then
based on maximum mismatch power and permissible durations
in the control scheme, the optimal battery sizing procedure is
proposed. In [19], the authors develop a frequency-based
approach to size a battery-supercapacitor energy storage system
for maintaining power balance of an isolated system with high
penetration of wind generation, thus to maintain the grid
frequency stability with the stochastic wind power fluctuations.

Customer-side: Customer-side BSS is gaining popularity among
commercial and industrial businesses as a cost-effective solution
to reduce energy and demand charges. In such an application, the
battery is utilized by customers for saving energy cost, rather
than by system operators for improving distribution system
operation. This is also referred to as behind-the-meter BSS
because the battery is literally “behind-the-meter”, on the
owner's property, not on the side of the electric utility. This
market is predicted to grow rapidly. In California Public Utilities
Commission's mandatory energy storage procurement targets,
its IOUs are required to collectively procure 200 MW energy
storage in the customer-side domain by 2020 [20]. According to
GTM Research, over 700 MW of distributed energy storage will
be deployed in the U.S. between 2014 and 2020 [21]. However,
only few studies have been carried out for economic analysis and
optimal sizing of battery storage in behind-the-meter applica-
tion. In [22], the authors develop a peak-shaving control
algorithm to determine battery charging and discharging
operation, and then calculate the economic benefits in demand
charge reduction for a few given battery sizes. Ref. [23] considers
an application bundle including both energy and demand charge
reduction. One application is treated as primary and determined
first, and then the other application is evaluated. Ref. [24]
explicitly formulates a co-optimization problem considering
both the energy and demand charges. Dynamic programming
technique is used to solve the problem to obtain the optimal
charging schedule. The charging and discharging efficiencies
have not been modeled. These papers provide different methods
in evaluating economic benefits for a given size of battery
storage, but have not addressed the optimal sizing problem
considering the trade-off between economic benefit and cost.
Therefore, the authors [25] study optimal sizing of a solar-plus-
storage system for utility bill savings and resiliency benefits.
Rather than formulating explicit optimal sizing problem, a
number of simulations are carried out to obtain the economic
and resiliency benefits for different solar-plus-storage sizes, and
then select the optimal one. In [26], the authors formulate linear
optimization problems for economic analysis and optimal sizing
of behind-the-meter BSS. This method explicitly models the
charging and discharging efficiencies and co-optimizes the
battery operation for both energy and demand charge reduction.
No analytical expression is given for optimal battery sizes as
functions of key factors such as load magnitude, energy/demand
charge rate, peak/off-peak time periods, and therefore cannot
provide much engineering insights on how the optimal battery
size varies with those factors. This motivates the work presented
in this paper.
This paper presents analytical sizing methods for behind-the-
meter battery storage. The main contributions of this paper are
summarized as follows.

Developing analytical sizing method for battery storage is a
challenging problem. Existing methods rely on standard
mathematical programming for battery operation and sizing
analysis, which cannot explicitly directly link the output to input
parameters and requires case-by-case analysis. This paper
proposes an innovative objective quantitative analysis of costs
 D. Wu et al. / Journal of Energy Storage 12 (2017) 297–304 299

and benefits for customer-side BSS with respect a few key charged during off-peak hours should be
factors.
Epeak

Practitioners need effective rule and guidelines for investment Eoffpeak ¼ ; ð2Þ
decision making on battery storage or utility rates design to
hround
stimulate the development of battery storage. This paper where hround is the BSS round-trip efficiency. Based on (1) and (2),
provides analytical expression of optimal battery sizes and the energy arbitrage is only profitable when the rate difference
corresponding benefits, and/or develop simple but effective between peak and off-peak hours is significant enough compared
guidelines and algorithms to meet such a need. with round-trip efficiency, i.e.,
roffpeak
The rest of this paper is organized as follows: Section 2 presents < hround : ð3Þ
rpeak
the benefit and cost analysis and develops rules and guidelines for
optimal sizing. In Section 3, practical utility rate and commercial Given (3) is satisfied, according to the daily saving expressed
building load profile are used to illustrate the proposed analytical in (1), annual benefit in energy charge reduction can be expressed
sizing methods. The optimal battery storage sizes are first as
determined using the proposed analytical methods, and then

roffpeak
compared with the results obtained using linear programming S ¼ ND Epeak rpeak  ; ð4Þ
methods for validation. Finally, Section 4 concludes the paper. hround
where ND is number of days in a year. On the other hand, the
2. Benefit-cost analysis and sizing guideline development levelized annual battery cost can be expressed as

In customer-side domain, the two main applications of BSS are C ¼ aðaEmax þ bpmax Þ; ð5Þ
energy charge and demand charge reduction. Energy charge is where a is the capital recovery factor (levelized annual fixed-
based only on the amount and time when energy is consumed. It charge rate), a is BSS capital cost with respective to energy size
reflects the operational cost in electricity generation and delivery. ($/kWh), Emax is BSS energy capacity, b is BSS capital cost with
Demand charge is based on the highest power consumption during respective to power size ($/kW), and pmax is BSS power capacity.
a billing period (typically a month). It is mainly designed to recover Therefore, levelized annual benefit can be calculated as
the investment in electricity generation and transportation
infrastructure. Separate charges for energy consumption and B ¼SC

roffpeak ð6Þ
demand more fairly distribute power system's operational and ¼ ND Epeak rpeak   aðaEmax þ bpmax Þ:
hround
investment cost to customers. This section presents benefit and
cost analysis for both energy and demand charge and derives rules Given that there exist some Emax and pmax such that B > 0, what
and guidelines for optimal sizing. In particular, Section 2.1 analyzes is the optimal ratio between battery energy size and power size so
energy charge reduction, while Section 2.2 analyzes demand that B is maximized? In order to maximize the annual benefit B, we
charge reduction. The two applications are studied jointly in need to maximize the daily energy discharged during peak hours
Section 2.3. Epeak, which is limited by BSS energy and power capacity, as
expressed in (7).
2.1. Energy charge reduction
Epeak  min fT peak pmax ; hround T offpeak pmax ; Emax g ð7Þ
In time-of-use (TOU) pricing program, instead of a single flat where Tpeak and Toffpeak are duration of peak and off-peak period,
rate for energy use, different rates are set for specified time periods respectively. Let us first fix pmax, and let Ecycle = min{Tpeakpmax,
on an advance or forward basis. TOU programs allow consumers to hroundToffpeakpmax}. It can be shown that
respond to the rates and manage their energy cost by shifting usage

roffpeak
to a lower cost period, which can not only help consumers save ND rpeak   aa > 0 ð8Þ
money but also reduce strain on the electric grid. Most of utilities hround
in the U.S. have launched TOU rate programs especially for is a necessary condition such that (6) is positive. Given (8) is
commercial customers. In California, a plan has been set forth to satisfied, the annual benefit B in (6) monotonically increases as
make TOU rate programs the commercial standard for all of the Epeak increases from 0 to Ecycle. Because of (7), increasing energy
state's major electric utilities. This paper considers TOU plans for size Emax beyond Ecycle only increases battery cost but cannot
energy charge. increase Epeak and therefore cannot increase saving in energy
The customer can charge the battery with cheap energy during arbitrage. Hence, the optimal energy size is
off-peak hours, and discharge the battery during peak hours when
energy price is high. This is referred to as energy arbitrage or peak Emax ¼ Ecycle ; ð9Þ
shaving. It should be noted that in energy charge reduction
application, the optimal battery operation and corresponding and optimal energy to power ratio is
benefits do not vary with load profile. In order to obtain positive ropt ¼ min fT peak ; hround T offpeak g: ð10Þ
benefits from energy arbitrage, the saving in electricity usage
during peak hours should be more than the cost of additional Replacing Epeak and Emax in (6) by roptpmax, after manipulation,
electricity consumption used to charge the battery during off-peak yields
hours, i.e., "
 !#
roffpeak aa ab
rpeak Epeak  roffpeak Eoffpeak > 0; ð1Þ B ¼ ND ropt pmax rpeak   þ : ð11Þ
hround ND ND ropt
where rpeak and roffpeak are the peak and off-peak rate, respectively, The physical meaning of important terms in (11) are explained as
Epeak and Eoffpeak is the amount of energy discharged during peak follows:
hours and charged during off-peak hours, respectively. In order to
maintain energy-neutral status of battery, amount of energy
roptpmax : Optimal energy size.
300 D. Wu et al. / Journal of Energy Storage 12 (2017) 297–304

rpeak  roffpeak : Daily energy charge savings per unit battery energy
h round

size.

NaDa : Levelized daily battery energy investment cost per
unit energy size.

ab : Levelized daily battery power investment cost per
ND ropt
unit energy size under optimal energy to power
ratio.

The derived analytical expression can be used for benefit

analysis and investment guidance, as illustrated with the following
example systems, whose parameters are listed in Table 1. These
parameters are developed based on the data in Refs. [22,26].
According to (4) and (11), benefit analysis and investment
guideline are explored in energy rate space and battery cost space
in Fig. 1(a) and (b), respectively, where the optimal energy to
power ratio is determined using (10).
The rate space in Fig. 1(a) is divided into three regions.

In the yellow region, condition (3) is not satisfied, therefore

resulting in negative saving from energy arbitrage.

In the green region, condition (3) is satisfied, but the daily energy
charge saving per unit battery energy size is not big enough to
cover the levelized daily battery energy investment cost per unit
energy size (under optimal energy to power ratio). Therefore,
investment in battery for energy cost saving is not cost-effective
either. The slope of the edge between yellow and green regions is
1/hround.

In the white region, (11) is positive. The expected saving in
energy charge is bigger than battery cost. It is interesting to note
that the edge between intersection white and green regions has
an intersection point with y-axis. Its value (about $0.16/kWh this
case) is the levelized daily battery energy investment cost per
unit energy size (under optimal energy to power ratio).

Therefore, in order to motivate a battery storage to participate Fig. 1. Operational and sizing guideline example. (For interpretation of the
energy shifting services, the energy charging rates offered by references to color in text near the figure citation, the reader is referred to the web
version of this article.)
utilities need to at least stay in the green region. Utility rates in the
white region provide even bigger difference between peak and off-
peak prices and therefore can help customers to financially justify load monotonically increases from minimum to maximum and
their investment in battery storage. It should be noted that in white monotonically decreases from maximum to minimum. Fig. 2
region, if the battery's power size is limited, its optimal energy size illustrates the idea of demand charge reduction. Without
can be determined using the optimal ratio given in (10), and vice discharging BSS during peak hours, the demand at the meter is
versa. On the other hand, without limiting power or energy size, D0. By discharging BSS during peak hours, the net demand at the
the profit optimization problem is unbounded. meter is reduced to DT, where T denotes the time duration of
The battery cost space in Fig. 1(b) is divided into two regions. original load that is higher than or equal to DT. The height of yellow
Per unit battery investment cost becomes more expensive as it region indicates the demand reduced at the meter, which is the
goes to upper and right, and therefore resulting in more same for all months in this case. As we continue lowering demand
investment cost than savings at some point. The boundary level, the annual incremental saving in demand charge can be
between “Investment” and “No investment” regions is quite expressed as:
vertical, which means the investment decision is more sensitive DS ¼ NM bDd; ð12Þ
to energy capacity cost than power capacity cost. This property can
be expected in general for practical Tpeak and Toffpeak. where NM is the number of months in a year, b is the demand
charge rate ($/kW/month), and Dd is the incremental change in
2.2. Demand charge reduction demand reduction. The effectively reduced demand by Dd, the
minimum of required increment in energy capacity is TDd.
Demand charge is a billing mechanism used to recover the cost
of providing transmission and distribution service to customers. It
is calculated based on the highest load during a billing period Table 1
Parameters of example system.
(typically a month). The monthly peak demand can be reduced by
discharging the battery during peak hours, and therefore reducing Param. Value Param. Value
demand charge. Both case (a) & (b) a 0.1168a hround 0.77
ropt 9h ND 365
2.2.1. Similar monthly peak load shape Case (a) a 500$/kWh b 125$/kWh
Case (b) rpeak 0.145$/kWh roffpeak 0.086$/kWh
For simplicity, let us first assume peak day load shape in each
month are approximately the same. In addition, it is assumed that a
15 years lifetime with 8% discount rate.
 D. Wu et al. / Journal of Energy Storage 12 (2017) 297–304 301

Fig. 3. Optimal load duration at new demand level (Topt) as a function of battery
cost, where demand charge rate (b in $/kW/month) is a parameter. The horizontal
plane is the maximal Topt limited by energy-neutral constraint.

Fig. 2. Illustration of demand charge reduction. (For interpretation of the

the peak is cut perfectly to fill the valley. The minimum demand is
references to color in text near the figure citation, the reader is referred to the achieved and Topt depends on load shape.
web version of this article.) Once Topt is determined, one can immediately find DT opt . The
corresponding optimal battery power size can be calculated as
According to (5), the corresponding minimum increment in
pmax ¼ D0  DT opt : ð18Þ
battery cost is

DC ¼ aðaT Dd þ bDdÞ: ð13Þ The optimal energy size can be calculated as

Z t0 þT opt
The incremental levelized net benefit can be calculated as Emax ¼ ½DðtÞ  DT opt dt; ð19Þ
t0
DB ¼ DS  DC ð14Þ
¼ DdðNM b  ab  aaTÞ: where D(t) is the load as a function of time, and t0 is the starting
time when load level reaches DT opt (please refer to Fig. 2).
It can be shown that

NM b  ab > 0 ð15Þ 2.2.2. Distinct monthly peak load shape

is a necessary and sufficient condition to receive positive net
benefit in demand charge reduction. Given (15) is satisfied, the
incremental benefit decreases as T increases, but the total benefit
increases with T as long as DB  0. The maximum benefit is
obtained when DB = 0, i.e.,
NM b b
T opt ¼  ;
aa a
which is independent of load shape. The demand charge rate
offered by utility need to satisfy (17) for battery storage to
participate peak demand reduction.
b > ab=NM :
Fig. 3 plots Topt as a function of a and b (BSS investment cost per
kWh and per kW), for a set of b (demand charge rate). As can be
seen, larger demand charge rate generates more incentives for
demand reduction, and therefore increment in Topt. In addition,
higher battery cost for per unit energy and power (a and b) reduced
the cost-effectiveness of demand charge reduction, and therefore
reducing Topt. Moreover, Topt is not sensitive to b, because the
second term on RHS of (16) is typically small compared with the
first term. As T and the energy discharged during peak hours
increase, at some point, there is not enough time to recharge the
battery during off-peak hours to maintain BSS energy-neutral
status over long term. This creates another constraint that limits T.
Given practical demand charge rate and battery cost, energy-
neutral constraint is inactive, and Topt is directly determined by
(16). However, in extreme cases where battery is very cheap
compared with demand charge rate, energy-neutral constraint
determines Topt. In that case, the net-load at the meter is flat, i.e.,
In practice, the peak day load shape in each month could be
quite different. In this case, an incremental reduction in demand
Dd in each month could be associated with different T and
therefore requires different amount of increment in BSS energy
capacity. Hence, the proposed method in Section 2.2.1 cannot be
used to determine the optimal sizes of BSS in this case.
Given (15) is satisfied, there always exist BSS energy and power
ð16Þ sizes such the net-benefit is positive. For month i, let Di0 denote the
original monthly load level, DiT i denote the corresponding reduced
monthly load level, and Ti denote the time duration of original load
that is higher than or equal to DiT i . It is not difficult to see that the
BSS power capacity only needs to be
ð17Þ
pmax ¼ max ðDi0  DiT i Þ; ð20Þ
i
because bigger power capacity increases BSS cost and smaller
power capacity cannot achieve the reduced load level DiT i for all the
months. Let imax denote the month index when the maximum is
achieved for (20).
Now, assuming the BSS energy size is increased by DE, the
monthly demand reduction is DT Ei and the annual saving in demand
charge is
X
NM
DE
DS ¼ b :
i¼1
Ti
It should be noted that effective demand reduction requires
additional power capacity, which is set by month imax, and the
required increment in power capacity is TD
i
E
. Therefore, the annual
max
302 D. Wu et al. / Journal of Energy Storage 12 (2017) 297–304

levelized incremental BSS cost becomes difference between (23) and (14) is NDDr T, which is the

 incremental saving from energy charge reduction. In practice,
DE
D C ¼ a aD E þ b NMb  ab is always positive, and NDDr  aa is typically negative.
T imax
Therefore, the incremental benefit decreases as T increases, but the
The incremental levelized annual benefit can be calculated as cumulative benefit increases with T as long as DB  0. The
maximum benefit is obtained when DB = 0, i.e.,
DB ¼ DS  DC !
XNM
1 b NM b  ab
¼ DE b  aða þ Þ T opt ¼ : ð24Þ
T T imax ð21Þ aa  ND Dr
i¼1 i
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
L
Compared with demand charge only case, saving from energy
charge helps to increase Topt and therefore the battery size. Once
The benefit increases as long as DB  0. Based on this analysis, Topt is obtained, one can calculate the battery size using (18) and
Algorithm 1 is developed to find the optimal BSS sizes. The sizing (19). In extreme cases when NDDr > aa, (23) is always positive and
analysis is based on the calculation of incremental benefit as a the problem is theoretically unbounded.
function of DE. Therefore, DE should be small. The proposed
algorithm uses binary search method on DE, and therefore ensures 2.3.2. Distinct monthly peak load shape
the accuracy for any initial value. Nevertheless, too small or too When the peak day load shape varies significantly from month
large initial value of DE could increase the iteration numbers. In to month, the analysis in Section 2.2.2 can be modified to include
this study, the initial value of DE is chosen as one percent of the energy charge component. In this case, the calculation in (21) is
peak hour load. In Algorithm 1, the initial Emax should be small modified to include saving in energy charge, and the increment in
enough to ensure that we start with a positive marginal benefit. We levelized annual benefit becomes
then gradually adjust Emax until it results in a zero marginal benefit. !
X
NM
1 b
The initial value of Emax also needs to be bigger than DE. In this DB ¼ DE N D Dr þ b  aða þ Þ ð25Þ
T T imax
study, we set the initial Emax as four times of initial DE. The i¼1 i
corresponding levelized annual benefit can be calculated as |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
L0
X
NM
Emax
B¼b  aðaEmax þ bpmax Þ: ð22Þ Hence, the optimal BSS scale can be determined using Algorithm 1
i¼1
Ti by replacing L in (21) at line 5 with L0 in (25).

3. Case studies
Algorithm 1. Battery sizing for demand charge reduction with
distinct monthly peak load shape. In this section, practical utility rate and commercial building
load profile are used to illustrate the proposed analytical sizing
1: Initialize Emax that is small enough to ensure positive benefit and DE
methods. Commercial buildings can be categorized based on their
2: repeat
3: principal building activities, including office, retail, warehouse,
Based on current Emax, calculate Ti and DiT i for each month i.
4: education, hotel, healthcare, grocery, restaurant, bank, and others.
For obtained DiT i , determine pmax, imax, and T imax using (20).
5:
The potential benefits and optimal sizes of BSS depend on building
Calculate L in (21).
6: if L  0 then electricity use pattern, which varies with building type and
7: Emax Emax + DE geographical location. The study in [27] develops a library of
8: else building load profiles, consisting of 68 building load data sets for
9: Emax Emax  DE six building types (office, retail, school, hotel, hospital and
10: DE DE/2
11: end if
warehouse) and four locations in different climate zones (San
12: until DE is small enough Francisco, Chicago, Houston, and New York City) from three
13: return Emax, pmax research institutes. Office buildings represent the most important
building type, accounting for about one fifth of total floor space and
building in the U.S. [28], and thereby we select office building load
profile in Houston to illustrate and validate the proposed analytical
2.3. Energy and demand charge reduction sizing methods in this paper.
The optimal battery storage sizes are first determined using the
In the case where both energy and demand charge reduction are proposed analytical methods, and then compared with the results
considered, BSS sizes and charging/discharging operation should obtained using linear programming methods presented in [26] for
be analyzed considering net-benefits from both applications. In validation. In particular, Section 3.1 and Section 3.2 study energy
practice, it is safe to assume that demand reduction occurs during charge reduction and demand charge reduction, respectively. The
peak-hours, and therefore discharging BSS during peak hours can two applications are studied jointly in Section 3.3.
help to reduce both energy and demand charges.
3.1. Energy charge reduction
2.3.1. Similar monthly peak load shape
In this case, calculation in (14) is modified to include energy In this subsection, optimal sizing and potential benefits of BSS
charge saving, and the increment in levelized benefit becomes are studied considering energy charge reduction only. The
assumed capital recover factor (a), round-trip efficiency (hround),
DB ¼ ðND T DdDr þ NM bDdÞ  aðaT Dd þ bDdÞ ; ð23Þ and battery capital cost (a and b) are the same as Case (b) listed in
¼ Dd½NM b  ab þ ðND Dr  aaÞT Table 1. The assumed charging and discharging efficiencies
(including both battery and inverter) are 0.887 and 0.868,
where Dr ¼ rpeak  hoffpeak , which is the reduced energy cost by
r
round respectively, which are needed in the linear programming
cycling per unit energy from off-peak to peak hours. The only methods. The peak hours are from 7am to 4pm, with a duration
 D. Wu et al. / Journal of Energy Storage 12 (2017) 297–304
Fig. 4. Levelized annual net-benefit vs BSS energy capacity.
of 9 h. The other hours are off-peak hours. Positive net-benefit from
energy charge reduction requires that off-peak and peak rates fall
in the write region in Fig. 1(a), Herein, it is assumed off-peak and
peak energy charge rates (roffpeak and rpeak) are 0.07$/kWh and 0.26
$/kWh, respectively.
With these parameters, one can verify that the necessary
condition of positive benefit in (8) is satisfied. The optimal energy
to power ratio can be determined by (10), which is 9 h. As analyzed
in Section 2.1, in white region, without limiting power or energy
size, the profit optimization problem is unbounded. Therefore, the
BSS maximum power is assumed to be 50 kW in this case, and the
corresponding optimal energy size is 450 kWh. The levelized
annual net-benefit can be calculated using (11), which is $754.7.
Next, optimal sizing is studied using the linear programming
methods presented in [26]. The benefits for different BSS energy
capacity are explored, as plotted in Fig. 4. The result is consistent
with the analysis in Section 2.1. Because the necessary condition in
(8) is satisfied in this example, the net-benefit monotonically
increases as Emax increases from 0 to Ecycle = min{Tpeakpmax,
hroundToffpeakpmax}, which is 450 kWh in this example. The net-
benefit reaches the maximum when BSS energy capacity equal
450 kWh, i.e., the optimal energy to power ratio is 9 h, and then
decreases as Emax increases. The corresponding maximum annual
benefit is the same as calculated using the proposed analytical
method.
3.2. Demand charge reduction
In this subsection, optimal sizing and potential benefits are
studied considering demand charge reduction only. The same
parameters in previous subsection are used here. In addition, the
demand charge rate is assumed to be $30/kW/month. For the office
Fig. 5. Net-benefit with different battery sizes.
303
Fig. 6. Annual saving vs. levelized annual cost with optimal BSS sizes.
building load profile in Houston used in this paper, the peak day
load shape in different months are quite different. Therefore,
Algorithm 1 is used to find the optimal energy and power size,
which are 43.6 kWh and 18.2 kW, respectively. The corresponding
levelized annual net-benefit can be calculated by (22), which is
$2003. This optimal size and the corresponding benefit are the
same as those obtained using the linear programming methods
presented in [26].
The annual benefits are explored in the energy and power
capacity space, as plotted in Fig. 5. It can be observed that benefits
do not vary significantly with energy or power size around the
optimum. In other words, a few different sizes of BSS generate
approximately the same performance on economic benefits. In
such a case, rather than simply taking the optimum, other factors
such as available models from BSS manufactures, budget and space
constraints should be considered to select the most appropriate
BSS size among a few good candidates.
3.3. Energy and demand charge reduction
In this subsection, optimal sizing is studied considering both
energy and demand charge reduction. All the parameters are the
same as Section 3.2, and the off-peak and peak energy charge rates
are assumed to be 0.86$/kWh and 0.145$/kWh, respectively. Using
the method proposed in Section 2.3.2, the optimal energy and
power sizes are found to be 76.4 kWh and 23.7 kW, respectively.
The corresponding levelized annual net-benefit is $2715.1.
Compared with the demand charge only case, additional saving
from energy charge helps increase BSS sizes and the corresponding
net-benefit. Fig. 6 plots the annual saving from energy and demand
charge reduction, as well as levelized annual cost for the optimal
BSS sizes. As can be seen, the demand charge reduction dominates
the total saving and contributes much more to net-benefit. This
result is quite typical for practical utility rates and existing BSS
technologies.
4. Conclusions
This paper presents innovative analytical sizing methods for
behind-the-meter BSS to reduce the electricity bill in energy and
demand charges, considering the trade-off between benefits and
costs. The proposed methods are based on objective quantitative
analysis and result in simple but effective rules and guidelines that
can be used to determine the most cost-effective battery size or
guide utility rate design for stimulating energy storage develop-
ment. With the proposed methods, optimal sizing and analysis
have been performed using typical commercial building load
profile and utility rates in a few cases. It is found that with much
simplified procedures, the proposed analytical sizing methods can
not only effectively identify the most cost-effective BSS sizes as
304 D. Wu et al. / Journal of Energy Storage 12 (2017) 297–304

other existing methods for a given system, but also provide [13] R. Fernandez-Blanco, Y. Dvorkin, B. Xu, Y. Wang, D.S. Kirschen, Optimal energy
important engineering insights on how different factors affect the storage siting and sizing: a WECC case study, IEEE Trans. Sustain. Energy (99)
(2016) 1, doi:http://dx.doi.org/10.1109/TSTE.2016.2616444.
optimal sizing and the net-benefits. [14] S.X. Chen, H.B. Gooi, M.Q. Wang, Sizing of energy storage for microgrids, IEEE
Trans. Smart Grid 3 (1) (2012) 142–151, doi:http://dx.doi.org/10.1109/
References TSG.2011.2160745.
[15] J. Tant, F. Geth, D. Six, P. Tant, J. Driesen, Multiobjective battery storage to
improve PV integration in residential distribution grids, IEEE Trans. Sustain.
[1] P.F. Ribeiro, B.L. Johnson, M. Crow, A. Arsoy, Y. Liu, Energy storage systems for
Energy 4 (1) (2013) 182–191, doi:http://dx.doi.org/10.1109/TSTE.2012.2211387.
advanced power applications, Proc. IEEE 89 (12) (2001) 1744–1756, doi:http://
[16] T.A. Nguyen, M.L. Crow, A.C. Elmore, Optimal sizing of a Vanadium Redox
dx.doi.org/10.1109/5.975900.
battery system for microgrid systems, IEEE Trans. Sustain. Energy 6 (3) (2015)
[2] I. Hadjipaschalis, A. Poullikkas, V. Efthimiou, Overview of current and future
729–737, doi:http://dx.doi.org/10.1109/TSTE.2015.2404780.
energy storage technologies for electric power applications, Renew. Sustain.
[17] S. Sharma, S. Bhattacharjee, A. Bhattacharya, Grey wolf optimisation for
Energy Rev. 13 (6–7) (2009) 1513–1522, doi:http://dx.doi.org/10.1016/j.
optimal sizing of battery energy storage device to minimise operation cost of
rser.2008.09.028.
microgrid, IET Gener. Transm. Distrib. 10 (3) (2016) 625–637, doi:http://dx.doi.
[3] S. Vazquez, S.M. Lukic, E. Galvan, L.G. Franquelo, J.M. Carrasco, Energy storage
org/10.1049/iet-gtd.2015.0429.
systems for transport and grid applications, IEEE Trans. Ind. Electron. 57 (12)
[18] M.R. Aghamohammadi, H. Abdolahinia, A new approach for optimal sizing of
(2010) 3881–3895, doi:http://dx.doi.org/10.1109/TIE.2010.2076414.
battery energy storage system for primary frequency control of islanded
[4] K.C. Divya, J. Østergaard, Battery energy storage technology for power systems
microgrid, Int. J. Electr. Power Energy Syst. 54 (2014) 325–333, doi:http://dx.
– an overview, Electr. Power Syst. Res. 79 (4) (2009) 511–520, doi:http://dx.doi.
doi.org/10.1016/j.ijepes.2013.07.005.
org/10.1016/j.epsr.2008.09.017.
[19] Y. Liu, W. Du, L. Xiao, H. Wang, S. Bu, J. Cao, Sizing a hybrid energy storage
[5] B. Dunn, H. Kamath, J.-M. Tarascon, Electrical energy storage for the grid: a
system for maintaining power balance of an isolated system with high
battery of choices, Science 334 (6058) (2011) 928–935, doi:http://dx.doi.org/
penetration of wind generation, IEEE Trans. Power Syst. 31 (4) (2016) 3267–
10.1126/science.1212741 arXiv: http://www.sciencemag.org/content/334/
3275, doi:http://dx.doi.org/10.1109/TPWRS.2015.2482983.
6058/928.full.pdf.
[20] Decision Adopting Energy Storage Procurement Framework and Design
[6] F. Díaz-González, A. Sumper, O. Gomis-Bellmunt, R. Villafáfila-Robles, A review
Program, 2013. Available online at http://docs.cpuc.ca.gov/PublishedDocs/
of energy storage technologies for wind power applications, Renew. Sustain.
Published/G000/M079/K533/79533378.PDF.
Energy Rev. 16 (4) (2012) 2154–2171, doi:http://dx.doi.org/10.1016/j.
[21] GTM, Distributed Energy Storage 2014: Applications and Opportunities for
rser.2012.01.029.
Commercial Energy.
[7] R. Walawalkar, J. Apt, R. Mancini, Economics of electric energy storage for
[22] J. Neubauer, M. Simpson, Deployment of Behind-the-meter Energy Storage for
energy arbitrage and regulation in New York, Energy Policy 35 (4) (2007)
Demand Charge Reduction Tech. Rep. NREL/TP-5400-63162, National
2558–2568, doi:http://dx.doi.org/10.1016/j.enpol.2006.09.005.
Renewable Energy Laboratory, January 2015.
[8] N. Lu, M.R. Weimar, Y.V. Makarov, C. Loutan, An evaluation of the NaS battery
[23] A. Nourai, et al., Tool Development for Estimating Value of Grid-connected
storage potential for providing regulation service in California, Proc. IEEE
Energy Storage in Bundled Services Tech. rep., KEMA, 2013.
Power Systems Conf. Expos. (2011) 1–9, doi:http://dx.doi.org/10.1109/
[24] Y. Yoon, Y.-H. Kim, Charge scheduling of an energy storage system under time-
PSCE.2011.5772494.
of-use pricing and a demand charge, Sci. World J. (2014), doi:http://dx.doi.org/
[9] A.D. Del Rosso, S.W. Eckroad, Energy storage for relief of transmission
10.1155/2014/937329.
congestion, IEEE Trans. Smart Grid 5 (2) (2014) 1138–1146, doi:http://dx.doi.
[25] T. Simpkins, K. Anderson, D. Cutler, D. Olis, Optimal sizing of a solar-plus-
org/10.1109/TSG.2013.2277411.
storage system for utility bill savings and resiliency benefits, IEEE PES
[10] X. Ke, N. Lu, C. Jin, Control and size energy storage systems for managing
Innovative Smart Grid Technologies Conference (ISGT) (2016) 1–6, doi:http://
energy imbalance of variable generation resources, IEEE Trans. Sustain. Energy
dx.doi.org/10.1109/ISGT.2016.7781237.
6 (1) (2015) 70–78, doi:http://dx.doi.org/10.1109/TSTE.2014.2355829.
[26] D. Wu, M. Kintner-Meyer, T. Yang, P. Balducci, Economic analysis and optimal
[11] S. Teleke, M.E. Baran, A.Q. Huang, S. Bhattacharya, L. Anderson, Control
sizing for behind-the-meter battery storage, Proc. IEEE Power Energy Soc.
strategies for battery energy storage for wind farm dispatching, IEEE Trans.
Gene. Meet., Boston, MA, 2016, doi:http://dx.doi.org/10.1109/
Energy Convers. 24 (3) (2009) 725–732, doi:http://dx.doi.org/10.1109/
PESGM.2016.7741210.
TEC.2009.2016000.
[27] C. Sun, K. Temple, T. Rossi, J.E. Braun, Interaction Between Dynamic Electric
[12] D. Wu, C. Jin, P. Balducci, M. Kintner-Meyer, An energy storage assessment:
Rates and Thermal Energy Storage Control Tech. rep., ASHRAE, 2006.
using optimal control strategies to capture multiple services, Proc. IEEE Power
[28] U.S. Energy Information Administration, Commercial Buildings Energy
Energy Soc. Gene. Meet., Denver, CO, 2015, doi:http://dx.doi.org/10.1109/
Consumption Survey, (2012) .
PESGM.2015.7285820.