You are on page 1of 11

786

IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 4, NO. 3, JULY 2013

Design of a Least-Cost Battery-Supercapacitor


Energy Storage System for Realizing
Dispatchable Wind Power
K. W. Wee, Student Member, IEEE, S. S. Choi, Senior Member, IEEE, and D. M. Vilathgamuwa, Senior Member, IEEE
AbstractA statistical approach is used in the design of a
battery-supercapacitor energy storage system for a wind farm.
The design exploits the technical merits of the two energy storage
mediums, in terms of the differences in their specific power and
energy densities, and their ability to accommodate different rates
of change in the charging/discharging powers. By treating the
input wind power as random and using a proposed coordinated
power flows control strategy for the battery and the supercapacitor, the approach evaluates the energy storage capacities,
the corresponding expected life cycle cost/year of the storage
mediums, and the expected cost/year of unmet power dispatch.
A computational procedure is then developed for the design of a
least-cost/year hybrid energy storage system to realize wind power
dispatch at a specified confidence level.
Index TermsHybrid energy storage system, power dispatch,
wind energy.

I. INTRODUCTION

N RECENT years, wind power generation is seen to enjoy


the most rapid growth among the various renewable energy sources. Unfortunately, wind is generally unsteady and the
generated power from wind turbine generators (WTG) tends
to fluctuate. Strategies to dispatch power from grid-connected
WTG can be broadly classified into either to maximize the harnessing of the wind energy or to realize a more controllable
form of power dispatch. The shortcoming of the former strategy
is that the WTG output power fluctuations could seriously degrade network security. The latter strategy, which often relies
on the actions of the WTG pitch angle control and the associated power converters, does not fully harness the energy from
the wind. Even by taking into consideration the so-called spatial
smoothing effect which reduces considerably the fluctuations in
the wind farm output power [1], it remains a challenging task to
firmly commit the wind farm output power several hours or a
day ahead. Such short-term power dispatchability is an advantage enjoyed by most conventional thermal stations.
In an investigation into the impacts of wind power generation into the ERCOT network, the authors in [2] have estabManuscript received May 01, 2012; revised December 03, 2012; accepted
February 06, 2013. Date of publication March 26, 2013; date of current version
June 17, 2013.
The authors are with the Center for Smart Energy Systems, School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore 639798, Singapore (e-mail: e080005@e.ntu.edu.sg; esschoi@ntu.edu.sg;
emahinda@ntu.edu.sg).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TSTE.2013.2248029

lished that it is essential to have well-designed ancillary service


so that negative impacts of the variable wind power on system
security can be surmounted. Their analysis assumes the utilities are obliged to give priority to wind power productions and
the ratio of conventional dispatchable portfolio mix of generators to the total generation capacity in the ERCOT system will
remain constant even as the wind power capacity grows. To predict how generation mix would evolve in a power system is an
extremely challenging task. However, if recent developments in
several advanced countries in phasing out nuclear power generation within the next 20 years were any guide, the penetration level of wind power would be expected to increase in these
countries. When this occurs, it will be more than likely that wind
farm operators will be pressured to provide firm power dispatch,
in a manner similar to that provided by conventional power stations. One technique to help achieve short-term wind power dispatchability is to incorporate an energy storage system (ESS)
in a wind farm. In power system applications, ESS can play
a role in either energy management or power quality enhancement. See for example, [3][5]. Technical limitations of integrating ESS in large-scale utility applications have been steadily
circumvented through advances made in energy storage and
power electronic technologies. Examples of practical utilityscale ESS include the 2-MW 2-s ride-through supercapacitorbased uninterruptible power supply described in [4], the 34-MW
245-MWh sodium sulfur BESS installed in a 51-MW wind farm
described in [5], and the 26-MW 15-min battery energy storage
system (BESS) shown in [6].
The role of ESS in wind power generation has been actively
researched. For instance, the ESS in [3] has been designed to
gain economic benefit as the wind farm output power becomes
dispatchable. However, the scheme described in [3] requires a
rather complicated ESS switch-over mechanism. ESS is considered in [7] for a standalone system but the need to meet firm
power dispatch has been ignored. Moreover, only the ESS energy capacity is determined but not the ESS power capacity.
The authors of [8] consider the uncertainty in the input wind
power and propose a probabilistic method to determine the ESS
storage capacity. The method requires the ESS power and energy capacities to be treated as independent design variables.
Hence, in determining the power capacity, the authors assume
the ESS energy capacity is unbounded and vice versa. In view
of the shortcomings of [3], [7], and [8], an alternative method of
designing the ESS will be proposed in the present work. Both the
ESS power and energy capacities shall be determined, without
the need to assume the ESS power and energy capacities are independent design variables.

1949-3029/$31.00 2013 IEEE

WEE et al.: DESIGN OF A LEAST-COST BATTERY-SUPERCAPACITOR ENERGY STORAGE SYSTEM

Among the various ESS options available to achieve the firm


power dispatch objective, pumped-hydro storage (PHS), compressed air energy storage (CAES), and battery energy storage
system (BESS) could be suitable contenders for fulfilling the energy management function. As the present work only considers
power dispatchability of a wind-farm which typically has a capacity of up to a few hundred MW, BESS is more appropriate
in terms of its power and energy capacities; PHS and CAES
are intended for applications which require much higher energy storage capacity [9], [10]. Also, BESS is more versatile,
as unlike PHS and CAES, it does not require unique geographic
terrain or significant land area [9]. However, if the BESS were
to be the only ESS media used to smooth the fluctuating wind
farm output power, the number of BESS charging/discharging
cycles would be so high that it can reduce the useful life span of
the BESS to an unacceptable level [11], [12]. A possible solution to this design dilemma is to add a supplementary short-term
ESS to the BESS and form a hybrid ESS (HESS). From [4], [9],
[12][14], possible short-term ESS can be in the form of superconducting magnetic energy storage (SMES), flywheel energy
storage (FES), and capacitor or supercapacitor energy storage
(SC). However, due to the high cost and environmental issues
associated with strong magnetic field, SMES is unlikely to be
suitable. The SC is deemed more suitable as its cost/kWh is
about 3 times lower than that of the FES [9]. The conventional
capacitor has characteristics similar to that of the SC except that
its size is about a hundred times larger and its cycle life is only
half that of the SC [9], [13], [14]. Hence in this investigation,
the SC shall be used in conjunction with the BESS.
In comparing the SC with the BESS, the SC has much higher
specific power density and cycle life, although it is less economical in terms of storing large amounts of energy [11], [12].
Consequently, unlike the approaches of [3], [7], and [8], the
battery-supercapacitor HESS shall be designed to concurrently
address the concern on power dispatch/energy management as
well as that on power quality. The BESS is to participate in the
energy management function and allows power to be committed
several hours or even a day ahead, while the SC is to mitigate
shorter-term power quality issues. Also, the authors of [1] propose to rely on the actions of the WTG pitch angle control and
the associated power converters to reduce the level of wind farm
output power fluctuations. Unfortunately, their technique cannot
ensure maximum wind energy harness. As the SC in the HESS
is designed to handle such short-term power quality issues, the
WTG shall be able to operate at the maximum power harness
state.
In the present investigation, the main challenge is to devise
a suitable coordinated power flow control strategy between the
BESS and SC, and for the designed HESS to be economically
viable. This paper proposes a statistical approach to the design
problem. The wind farm-HESS scheme and the pertinent design considerations are described in Section II. The proposed
power flow control strategy for the two energy storage media is
described in Section III. Section IV shows the method to determine the optimal HESS design, from which the combined wind
farm-HESS power station is able to dispatch power at a given
confidence level. The computational procedure to determine the

787

Fig. 1. Schematic diagram showing real power flows of the grid-connected


wind farm-HESS.

most economical HESS design is developed in Section V. A


case study is given in Section VI to demonstrate the proposed
design methodology.
II. WIND FARMHYBRID ENERGY STORAGE SCHEME
A. Schematic
Fig. 1 shows the real power flows within a rather generic gridconnected wind farm-HESS scheme. The WTG in the wind farm
are aggregated to form the equivalent WTG and produces the
total output power
. Power conversion devices between
the WTG, BESS, and SC have been omitted from the power
flow consideration as losses in these devices tend to be small.
The arrows indicate the assumed positive direction of the power
flows.
and
are the respective SC and BESS output
powers while
is the aggregated power of the HESS. The
role of the HESS is to buffer the grid from the fluctuating
and exercises control on the dispatch power
. In this paper,
it is assumed that the power plant can only export power to the
grid. From Fig. 1,
,
, and
are governed by the
relationship
(1)
Note that the HESS considered in Fig. 1 obviates the need of
the rather complicated BESS switch-over scheme described in
[3].
B. BESS and SC Service Lifetimes
Expected lifetime of BESS is at most 15 years [15], [16].
Usage factors that affect the battery life span include the depth
of discharge (DOD) and rate of change of the charging/discharging power (ramp-rate) [15]. For renewable energy applications, the typical DOD of the batteries ranges from 50%90%.
Even when the battery operates below its rated power level,
the rate of change in the charging/discharging power should be
controlled. Subjecting the battery to charge/discharge ramp-rate
faster than what the chemicals can react causes localized overcharge/over-discharge conditions inside the battery [15]. Hence
one main consideration in the latter sections is to include into
the HESS design the optimum maximum rate of change of the
BESS charge/discharge power, measured in terms of certain performance index.
The life span of an SC is affected by its operating temperature and voltage [17]. Since the SC used in the wind farmHESS scheme is likely to be housed indoors where temperature
control can be expected, and the interposing power conversion

788

IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 4, NO. 3, JULY 2013

Unfortunately, due to the inevitable forecasting error,


would not be identical to the actual wind power
,
. Hence, by setting
into
(1), the mismatched power component between
and
would have to be catered for by the HESS in order to
meet the dispatch commitment. The challenge is then to design
the HESS, subject to the uncertainty in predicting
.
B. Coordinated BESS-SC Power Flows Control Scheme

Fig. 2.
from WTG and
detailed plot showing HESS charging

obtained from the forecasted


and discharging
intervals.

and the known


, a new coordinated
Based on
power flow control scheme is now developed. Fig. 2 shows that
within any particular interval , the areas labeled as correspond to instances when
is larger than
. Hence,
the HESS is being charged by the WTG over these instances.
Conversely, areas
signify instances when the HESS is discharging. One can then rewrite (1) as
(3)

equipment between the SC and the grid would actively exercise


voltage control, the effects of both the operating temperature
and the voltage on the SC lifetime have been omitted.
III. POWER FLOWS COORDINATION BETWEEN BESS AND SC
A. Determination of Dispatch Power Reference
When designing the HESS for a wind farm, it is reasonable
to assume sufficient historical data on
is available. For
the HESS design problem in hand, it is necessary to consider
how the wind farm-HESS shall operate in order to meet the
short-term dispatch commitments
. Unlike a conventional
power station, the input wind power
is random and therefore, in planning
, the wind farm operator will need to
forecast
. To emulate a possible short-term dispatch planning procedure undertaken by the operator, a portion of
,
, is used to provide the wind power forecast
for the following period , which is typically of several hours
(see Fig. 2). The exact method of obtaining
is not the
focus of this paper; rather it is assumed any one of the currently
available forecasting methods such as that described in [18] has
been used. The forecasted short-term wind power
for
the period
, can, therefore, be generated.
The short-term dispatch power
is then derived from the
forecasted wind power
. Often
is intended to be
constant on an hourly or even longer duration: the exact duration
shall depend on the prevailing dispatch policy of the power
market the wind farm operates in. The constant dispatch level
is denoted as
, the reference quantity for
for the
interval , as shown in Fig. 2. Within each ,
shall be
determined from
by ensuring that there is no net change
in the HESS stored energy level, in accordance to the forecasted
, i.e.,
(2)
is then set equal to its reference level
The dispatch power
:
becomes the wind farm committed dispatch
power.

in (3) denotes the reference mismatched power between


and
. With spatial smoothing effect as described
in [1], the level of power fluctuation in
and therefore in
will be reduced considerably.
will include both
the BESS output power reference, denoted as
, and that
of the SC, denoted as
, viz.
(4)
Next, unlike the approach of [12], the new power flows coordination method begins by first considering the BESS lifetime. Section II-B explains that the BESS lifetime can be significantly affected by the rate of change of the battery charging/discharging power. Suppose
is increasing and to achieve
a reasonable BESS lifetime, the maximum ramp-up rate of the
BESS power is set at p.u. MW/s. Hence, the BESS shall meet
the power demand on the HESS on its own if the ramp-up rate
of change of
is lower than . Then in (4),
shall
thus be set equal to
while
. However, should
the rate of change of
exceed , both the BESS and SC
shall share in meeting
: the BESS shall operate at the
maximum ramp-up rate while the SC shall take the balance between
and
. Hence, the BESS will only cater for
a change in
at a rate not exceeding .
A more detailed analysis of the above proposed strategy is
shown in Fig. 3 where
is seen to vary linearly over a
sufficiently small time step
. From Fig. 3(a), the HESS is
shown being charged by the WTG and between the interval
to ,
is to increase at the maximum rate of and the SC
is tasked to handle the difference between
and
.
Thus,
and
are set as shown in (5), at bottom of
the next page.
For the interval
, the rate of change in
is
less than the maximum BESS ramp-up rate . Hence

(6)
Conversely, suppose the BESS has the maximum ramp-down
rate of p.u. MW/s. Consider Fig. 3(b) when the HESS happens
to be discharging. In the interval
, the BESS
discharging power is set to decrease at the maximum rate of

WEE et al.: DESIGN OF A LEAST-COST BATTERY-SUPERCAPACITOR ENERGY STORAGE SYSTEM

Fig. 3. Examples of power flows coordination between the BESS and SC.

and SC is tasked to take up the difference between


and
. Thus we have (7), shown at bottom of the page.
In the interval
,
as ramp-down
rate of
is less than . Hence,
and
are

(8)
Let the BESS and SC power ratings be denoted as
and
, respectively. Then in (5)(8), clearly it is necessary
and
. However, both
and
are unknown at this stage and the procedure
to determine them shall be described in Section IV-A. Furthermore, the BESS state of charge is directly dependent on
and the BESS energy capacity. Denote the BESS state of charge
and its energy capacity, respectively, as
and
.
Thus,
is given by
(9)
is the BESS initial state of charge. As stated
where
in Section II-B, the BESS should not operate beyond a certain depth of discharge (DOD). Denote the BESS maximum
allowable DOD as
, then
in (5)(8) must ensure
at any instance must lie within the range
.
Similar reasoning also applies to the SC. Denote the respective SC state of charge and its energy capacity as
and
. Thus
is given by
(10)

789

is the SC initial state of charge. Let the SC


where
maximum allowable DOD be
, then the SC power
in (5)(8) must ensure
must lie within the range
. Note that both
and
are also unknown at
this design stage and the procedure to determine them shall be
described in Section IV-B.
Apart from the scenarios shown in Fig. 3, (5) and (6) will also
apply to cases when the HESS is ramping up as it is discharging
while (7) and (8) will be equally applicable when the HESS is
ramping down as it is being charged.
Thus the proposed power flow coordination scheme constrains the respective BESS power ramp-up and ramp-down
rates within every
to be no more than and p.u. MW/s.
The more rapid fluctuating power components are to be dealt
with by the SC. With a judicious selection of and values,
and the method to do so shall be described in Section V-D, it
is intended the larger proportion of the energy contained in
shall be dealt with by the BESS, as compared to that
handled by the SC, although the SC shall have to undergo a
larger number of charge/discharge cycles. In this manner, the
proposed scheme is to exploit the desirable attribute of the SC
in that it has a much higher cycle life as compared to the BESS.
By diverting the rapid fluctuating power components to the SC
therefore prolongs the life span of the BESS while the BESS,
with its inherently larger specific energy density, handles the
larger proportion of the charging/discharging energy in
.
Another important consideration is the value selection for .
There are two pertinent factors concerning the selection of
and these are the dynamic response characteristics of the WTG
and the wind speed at the wind farm. For the case of the variable speed WTG, the power converter allows the rotor generator
speed to be varied and in the event of wind gusts, the WTG can
better compensate the fast variations in the wind power [19].
Hence, it would be several seconds before one could expect any
appreciable change in
and for the need of
to be
recomputed using (3). However, in the case of fixed speed pitch
angle-regulated WTG, due to the relatively slower response of
the pitch mechanism to avoid the power fluctuations in the wind
gusts [19], [20], therefore, a shorter duration of
would be required. In selecting the appropriate , one could refer to [12]
and [21]: the sampling interval
for
is stated as 0.2 s
in [12] when the authors investigate the adequacy of a control
strategy for an HESS whereas
is set at 1 s in [21] for the design of a dispatch scheme using a BESS. A computer simulation
study could be conducted using the wind speed data collected at
the wind farm. With the inclusion of suitable WTG model and
by a process of trial-and-error, appropriate
can be obtained

(5)

(7)

790

to produce a suitable HESS scheme. In the present work,


4 s has been found to be satisfactory.

IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 4, NO. 3, JULY 2013

of

or
(12b)

IV. STATISTICAL METHOD TO DETERMINE HESS POWER AND


STORED ENERGY CAPACITIES
The next task is to determine the HESS power and stored energy capacities based on the proposed power flows coordination
strategy. Unlike [8], the method shown below does not assume
the HESS power and energy capacities are independent design
variables.
A. BESS and SC Power Capacities
From the procedures described in Section III-A,
is determined and then using (3), evaluate the HESS power reference
. Based on the power coordination strategy described
in Section III-B, suppose
and
are determined
using (5)(8) over the complete study period of (say)
days.
Recall instances when the BESS is being charged are indicated by positive
whereas negative
indicates
the battery is discharging.
can, therefore, be statistically
quantified: Take the absolute values of
and compute
its probability density function (pdf) and cumulative probability
function (cdf). From the cdf of
, the battery rated power
capacity
could be chosen to be the value of
which corresponds to the cdf of 1 so that the BESS can meet all
the possible charging/discharging power requirements. Unfortunately,
so determined may be unacceptably large and
uneconomical. To overcome this dilemma,
can be limited to a lower value which has the corresponding cdf of ; i.e.,
the BESS can be expected to successfully meet the power demand on it with the probability of . With this BESS power
rating,
can no longer be met at those instances when
exceeds
unless the SC power capacity
is
made sufficiently large. Unfortunately, it is also prudent to limit
in order for the HESS scheme to be viable. Hence, the
extent of successful power dispatch would depend on the selection of both power ratings
and
. In view of the
above, denote
(11)
where
denotes the probability of condition
being
met. Therefore,
and
denote, respectively, the conditions
and
being met.
As aforementioned, successful power dispatch can only be
achieved when
does not exceed
and
is less than
, i.e., conditions and are both satisfied.
Therefore, let
denotes the probability of successful
power dispatch, then

denotes the conditional probability


In (12a) and (12b),
of is successful subject to being successful. Given either
(12a) or (12b), the order of the BESS and SC power ratings selection will be governed by the unconditional probability: e.g.,
in (12a),
will be selected prior to that of
. Equation (12a) will be used in the present investigation because the
primary motive of using the HESS is to exploit the inherently
larger specific energy density of the BESS whereas the SC plays
the role of prolonging the BESS life span by handling the less
energy-intensive but rapid microcycles during the charge-discharge process. Hence the BESS power capacity shall be selected before that of the SC.
The process to determine
and
and the corresponding probability of successful power dispatch
will
now be described. Firstly the cdf of
is used to set the
BESS power rating
. This means that over , the probability of event being successful is , i.e.,
(13)
Then, in each of the time-steps
such as that shown in
Fig. 3, consider only those intervals when condition is satisfied, i.e.
. Using the th time-step in Fig. 4(a)
as an example, the corresponding
is calculated using
(4) and because condition
is satisfied over the whole timestep,
over this
is to be included in the evaluation of
. In Fig. 4(b) however, condition is only satisfied
between
because beyond ,
and the BESS shall be overloaded. Thus in computing
using (12a), only the corresponding
between the interval
needs be considered.
This tracking procedure of
is repeated for all
in
but only during those instances when condition , i.e.,
is satisfied. From the tracked
,
construct the pdf and cdf of
, much in the same manner
as for
. Note that the obtained pdf and cdf of
is only applicable for the selected BESS power rating
.
Again similar to the manner of selecting
, the SC power
rating
is decided based on the confidence level one
wishes to achieve with respect to the final successful power
dispatch
. So from the cdf of
, suppose one
selects an SC power rating of
and it corresponds to a
confidence level of , i.e.,
(14)
Based on (12a), therefore, the probability of successfully
meeting the committed dispatch power with the HESS is

(12)

(15)

As can be seen from (3) and (4), the determination of


is
dependent on the value of
. Thus, the two conditions
and are dependent and
can be rewritten either as

Thus in summary, the above statistical method determines the


BESS and SC power capacities to yield the successful power
dispatch probability given by (15). In other words, the committed dispatch power can be met with the capacities of the
HESS so determined, with the probability
.

(12a)

WEE et al.: DESIGN OF A LEAST-COST BATTERY-SUPERCAPACITOR ENERGY STORAGE SYSTEM

791

capacities are determined from


and
, based on
the BESS power ramp-up and ramp-down rates and . This
section describes how to determine
to achieve the most
economical HESS design, measured in terms of the annualized
life cycle cost of the HESS and the expected penalty cost for
failing to meet the power dispatch obligations.
Fig. 4. Tracking of
.

(a) over complete time-step

and (b) between

B. BESS and SC Stored Energy Capacities


To ensure the HESS so determined is capable of accommodating the amount of energy it has to store, the HESS energy
capacity needs to be evaluated. With regard to the BESS energy
capacity and with respect to an arbitrary reference time
,
at time , the BESS stored energy is changed by the amount
(16)
In the event when
,
.
Suppose over , the range of
is between a maximum value of
and a minimum value of
. Then
will represent the minimum BESS energy
capacity required to meet the committed power dispatch based
on the BESS power rating
. However, as stated in
Sections II-B and III-B, the BESS should not operate beyond
the depth of discharge
. Therefore, the rated BESS
stored energy capacity is then determined as
(17)
To determine the SC energy capacity, note that the energy
stored in the SC is proportional to the square of its terminal
voltage. Let
and
denote the minimum and nominal SC voltage, respectively. It can be readily shown that the
SC maximum allowable DOD
is given by
(18)
Also, from the known profile of
and in a similar
manner as described for the BESS, calculate the maximum
and minimum
in the SC stored energy levels
over . The rated energy capacity
of the SC, with the
power rating
is, therefore,

(19)

V. DETERMINATION OF BESS POWER RAMP-RATES


The proposed method described in Section IV allows one to
determine the BESS and SC power and energy capacities to
yield the successful power dispatch probability
. The

A. Expected Lifetimes of the BESS and SC


Generally, manufacturers express BESS cycle life in terms of
the number of charge-discharge cycles the BESS can undertake
before the BESS has to be replaced. Typically the cycle life is
obtained by discharging the BESS to a specified DOD level and
recharging it to its full capacity at rated current. The process
is repeated until the BESS is unable to be so discharged and
recharged. Unfortunately, the charging-discharging cycles are
irregular in the present scheme due to the unsteady input wind
power. Hence, the number of life cycles the BESS has undergone at any given time cannot be so readily determined.
The above-mentioned BESS life cycle estimation problem
could be eluded using the BESS ampere-hour (Ah) throughput
measure. This approach has also been adopted in [16] in photovoltaic applications. In the present investigation, consider
only those instances when the BESS is discharging. This is
reasonable as the charging and discharging Ah throughputs of
the BESS will be the same if
is sufficiently long. Suppose
the BESS terminal voltage does not vary greatly from its rated
value, then from
, the discharge current
can be
readily calculated. The Ah throughput with respect to the th
. By summing
discharge interval is
all the discharging
obtained over
, the total BESS
discharge Ah throughput is obtained. The ratio of the total
BESS discharge Ah throughput to the BESS rated Ah, denoted
as
, yields the BESS equivalent service cycle life
over , viz.
(20)
Since the BESS cycle life at rated conditions
ally provided by the manufacturers, the expected life
the BESS can be expressed as

is usuof

(21)
where
has the unit of time.
SC manufacturers also specify SC cycle life in terms of the
number of charge-discharge cycles the SC can undertake before
it needs to be replaced. SC can be fully cycled or irregularly
cycled without affecting its cycle life [22]. Thus, the SC lifetime can be estimated from its discharged energy throughput.
Accordingly and similar to the case of the BESS, only consider
those intervals when the SC is discharging. The SC discharged
energy throughput for the th discharge interval
is obtained by integrating
between
and . Next, all the
discharging energies
obtained over are summed to yield
the total SC discharged energy. Divide the total SC discharged

792

IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 4, NO. 3, JULY 2013

energy by the SC rated energy capacity


vice cycle life
over the period

, the equivalent seris obtained, viz.


(22)

Again, using the SC cycle life


provided by the manufacturers, the expected life
of the SC is given as
(23)
The

has the unit of time.

B. Annualized Life Cycle Cost of the HESS


The life cycle cost of the HESS is now considered. The capital cost of the BESS includes both its power converter and
the battery. Hence, the BESS capital cost can be expressed as
, where
and
are the
respective BESS power and energy capacities determined in
Section IV.
and
are expressed in terms of the cost per
MW and cost per MWh, respectively, and they are obtainable
from the manufacturers or the literature [9], [14], [23]. Similar reasoning applies to the SC: The SC capital cost becomes
, where
and
are the cost per
MW and cost per MWh, respectively, for the SC, given the SC
power rating of
and energy capacity of
. Thus the
total HESS capital cost (TC) is expressed as
(24)
Let to be the interest rate and
denotes the service life
span of the wind farm, the annualized HESS capital cost (AC)
can be calculated through the determination of the capital recovery factor (CRF) where
(25)

C. Cost for Failing to Meet Dispatch Obligations


A tiered approach has been used to penalize the generator that
fails to meet its dispatch obligation so as to ensure good scheduling practices by power generators. For example, the regulatory rule contained in [24] subjects the generator to a lower tier
penalty of a certain fixed generation imbalance cost if its generation deviates within the band of 1.5% 7.5% from the hourly
scheduled energy; a higher tier penalty shall be imposed if the
deviations exceed 7.5%.
With the HESS in place, the wind farm-HESS is designed
to dispatch power much like a conventional power station. Suppose the probability of successful dispatch
is selected to
be less than 1. By selecting this dispatch probability level, which
in turn determines the HESS capacities, the incurred penalty
cost in which the committed dispatch power
cannot be
successfully met is evaluated in the following way. Adopting a
similar practice as [24], let
and
(expressed in $/p.u.
MWh) denote the higher and lower tiered fix generation imbalance charge, respectively. Suppose the penalty bands are set at
percent. Since
is to remain constant over interval , therefore, the amount of energy subjected to the higher
tiered imbalance charge
in is

(29)
Similarly, the amount of energy subjected to the lower tiered
imbalance charge in is

Thus, the annualized HESS capital cost AC is expressed as


(26)
The annualized replacement cost for the BESS

is

(27)
where
is the expected life of the BESS obtained from
(21). The number of times the BESS will need replacement
during the service life of the wind farm will be equal to the
number of terms in the square bracket of (27).
The same reasoning applies to the SC: the annualized replacement cost for the SC
is determined as

(28)
where

is the SC expected life obtained from (23).

(30)
The total amount of energy which is subjected to the higher and
lower tiered imbalance charges is the summation of all
and
in . If the period of the study is one year, then the
total annual penalty cost
incurred due to the unmet power
dispatch is
(31)

WEE et al.: DESIGN OF A LEAST-COST BATTERY-SUPERCAPACITOR ENERGY STORAGE SYSTEM

793

D. Procedure to Determine the Optimum BESS Ramp Rates


Sections IIIV-C have been devoted to developing the procedures to determine the BESS and SC power and energy capacities so that the committed power dispatch can be met with
probability
. The capacities are seen to be dependence on
the selected BESS power ramp rate limits
. It is proposed
herewith that the selection of the most suitable
shall be
based on the minimization of the following performance index
, viz.
(32)
The index in (32) takes into consideration the HESS annualized life cycle cost, which includes both its annualized capital
and replacement costs, plus the expected penalty cost shown
in Sections V-B and V-C. Thus index represents the total
cost/year due to the inclusion of the HESS in the wind farm and
the expected cost the committed power schedule
cannot
be met.
A computational procedure is needed to search for
so
that is minimized. From the optimal
, one can then determine the power and energy capacities of the BESS and SC.
A linear search method is utilized in this research to achieve
this design objective. In this method, (32) shall be the objective function and the variables to be searched for are the BESS
charge/discharge power ramp rates
. Fig. 5 provides the
detailed search procedure for obtaining the optimum
.
Starting with the calculated
, the search begins with an
initial set of
. The procedure described in Section III-B is
then used to obtain the corresponding
and
. Note
that in Section III-B, although there are explicit power and energy limits placed on both the BESS and SC, these limits are
unknown at this stage. Therefore, the only constraint is the rate
of change in the BESS power
in (5)(8) must not exceed
the ramp-up and ramp-down rates
. Then with the procedures described in Section IV, establish the cdf of
and
determine the BESS power capacity
for a selected confidence level . Next, obtain the corresponding
. Following
which establish the conditional cdf of
and determine
for the confidence level and the corresponding
.
Evaluate the expected lives of the BESS and SC, and the HESS
annualized life cycle cost using the results of Sections V-A and
V-B, respectively. Subsequently, evaluate the expected annual
penalty cost using (31) to account for the cost of not meeting
the dispatch obligations. Finally, calculate the index using
(32). Repeat the above procedure for different sets of
and
search until the minimum is obtained. The corresponding set
of
provides the least-cost HESS design in meeting the
committed dispatch with probability
.
VI. ILLUSTRATIVE EXAMPLE
A hypothetical 100-MW wind power station is used in this
section to illustrate the proposed HESS design methodology.
A 1-year wind power data (May 2009April 2010) measured
at the authors campus was used. The wind data was sampled
at 4-s interval and was scaled to take into account the spatial
smoothing effects to obtain
.

Fig. 5. Linear search method to obtain the optimum BESS ramp-rates.

To begin the HESS design process,


must first be determined. This in turn requires the knowledge of
which
is derived from the forecasted wind power
. As the exact
method of obtaining
was not the focus of this study, it
was assumed that the error in the forecast
had zeromean and was normally distributed. The error was generated
and added to
to form
. The root-mean-square of
the error was assumed to be 15%, a level typical in wind power
forecast [18]. Using the method from Section III,
was
then obtained from
to yield constant hourly dispatch. It
is interesting to note that average daily
is some 20 MW
or 20% of the rated wind farm capacity, a level typical for a wind
farm.
A. Optimal BESS Ramp-Rates
From
,
, and
, calculate
using
(3). The linear search procedure described in Section V-D can
be used to determine the optimal
to yield the least-cost
BESS-SC design. In this example, assume
and the following results were obtained to meet the probability
. Fig. 6 shows the required BESS and SC power and energy
ratings at the various power ramp rate . At
MW/s which
corresponds to BESS not in service, all the HESS power flows
are those to/from the SC. However, as increases, the BESS
handles an increasingly larger proportion of
while that

794

IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 4, NO. 3, JULY 2013

Fig. 6. Required BESS and SC power and energy ratings at various BESS
.
power ramp-rates:

Fig. 9. Performance index


values:
.

Fig. 7. Expected life of BESS and SC at various BESS power ramp-rates:


.

Fig. 8.
and

and

for
and
at
: plotted against various BESS power ramp-rates.

dealt with by the SC is correspondingly reduced. Hence, the required SC power rating (also the SC energy capacity) reduces
with the increase in while the BESS power rating (also its energy capacity) increases. The SC appears to play an insignificant
role in the power flow control beyond of about 1.2-MW/s: the
BESS appears to be the only ESS required to handle the wind
power perturbations.
Fig. 7 shows the expected life of the BESS decreases as
increases while the opposite trend is observed for the SC. This
is due to the increase in the BESS usage. The BESS expected
life ranges from 554 to 1130 days while that of the SC is from
1456 to 2743 days.
Fig. 8 shows the total unserved energy over the 1-year period.
For comparison purpose, cases for
and
are shown, corresponding to the penalty bands set at
and
. As the tiered imbalance charge rates
and
are not specified in [24],
and
(in US$/MWh)
are assumed to be 500 and 1000, respectively, in this example.
It is seen from Fig. 8 that the unserved energy remains relatively constant for a given dispatch confidence level despite
the difference in . And as expected, the unserved energy reduces with the increase in the probability of successful dispatch
.

at the various BESS power ramp-rates and

The total cost/year index


expressed as (32) for various
is shown in Fig. 9. It shows the sensitivity of with which
leads to ranges of HESS capacities, annualized life cycle cost,
and the expected penalty cost as shown in Figs. 68. The interest
rate and compounding periods for determining (25) were taken
from [14] at
and
of 20 years. The nominal BESS
and SC cycle life values
, and the possible ranges
of cost per MW
in US$/MW and cost per MWh
in US$/MWh were taken from [9] and [23]:
cycles
and
cycles,
,
,
,
and
.
The HESS cost/year index is impacted by the uncertainty
in the cost parameters
of the HESS. Fig. 9 illustrates this
sensitivity through the inclusion of three sets of the cost parameters. From Fig. 9, it is seen that is much larger at
MW/s
as compared to those of other BESS power ramp-rates. This reinforces the fact that SC is not viable as a long-term storage device, as alluded to in Section I. It is also seen that the impact of
on is not significant for
MW/s for
,
, when varies from
and
varies from
. This is due to the much
larger power and energy capacities of the BESS over that of the
SC when evaluating the annualized HESS capital and replacement costs in (32). It is observed from Fig. 9 that is a convex
function with respect to : there exists a minimum cost/year
point close to
MW/s. Other studies for various ranges
of
, , and
also lead to the same conclusion. Thus,
it is reasonable to select
MW/s. With the identification of the optimum and through the search for the
minimum , the least-cost HESS design is thus obtained. The
corresponding BESS and SC power and energy capacities can
be readily inferred from Fig. 6 at
MW/s. The design
meets the committed dispatch power at the confident level of
.
Fig. 10 shows a sample hourly profile of
and
for
MW/s. The respective numbers of discharge cycles that the BESS and SC have undergone for a day
were also computed using (20) and (22). The BESS is seen
to have undergone the equivalent of 0.71 cycles while that of
the SC is 22.54 cycles. Hence, the results indicate that the proposed HESS power coordination method has indeed exploited

WEE et al.: DESIGN OF A LEAST-COST BATTERY-SUPERCAPACITOR ENERGY STORAGE SYSTEM

Fig. 10. Hourly plots of the resulting


rate of 1.0 MW/s.

Fig. 11. CDF of

and

and

using BESS ramp-

MW/s,

the much larger cycle life of the SC. In the process, the SC has
prolonged the BESS lifetime.
B. HESS Power and Energy Capacities
Fig. 11 shows the BESS and SC power cdf for
MW/s.
With
MW/s and
, it is seen that the
required BESS power capacity
is 11 MW and that of the
SC is 530 kW. The corresponding energy capacities are
MWh while
kWh. Hence, the probability of
successful power dispatch
is 0.95 0.95 or 0.9025.
Although not shown here, the case of
MW/s,
has also been evaluated. The corresponding BESS
power and energy capacities are 12.5 MW, 49.5 MWh and that
of the SC are 1000 kW, 1.5 MWh. At the expense of higher
HESS capacities and cost/year index
,
is improved
to 0.98 0.98 or 0.9604.
C. Energy Production Cost Increase Due to the HESS
It will be interesting to assess the increase in the cost of energy production due to the addition of the HESS. In this example, from the
obtained over the one-year study period, the total yearly harnessed wind energy for
was determined to be some 170 820 MWh. Assuming
,
,
, and
, the cost/year
was determined to be
US$5 487 300 for
. Divide by the net yearly
dispatched wind energy which is the difference between the
total yearly harnessed and the unserved energies, the increase
in cost due to the HESS installation is about US$0.033/kWh for
or
. Increasing the certainty
of successful power dispatch to
or 0.9604
increases the cost to US$0.034/kWh. On the other hand, for
absolute certainty of power dispatch, i.e.,
,
there would be an increase of to US$5 502 000. However, in
this case, all the harnessed wind energy of 170 820 MWh would

795

be dispatched and the resulting cost/kWh energy production is


slightly lower at US$0.032/kWh.
As a comparison, present-day average electricity production cost of large central thermal power stations is between
US$0.05/kWh and US$0.10/kWh while that of onshore wind
farms is about US$0.06/kWhUS$0.15/kWh [25]. Hence, the
electricity production cost from the example wind farm-HESS is
between US$0.092 and US$0.184/kWh for various confidence
levels of successful dispatch. While the energy production
cost of the wind farm-HESS power station is seen to be about
2 times that of the thermal power stations, the combined wind
farm-HESS scheme results in much less environmental damage
and hence attracts no environmental costs, e.g., carbon-tax.
Also, with continuing advances being made in energy storage
technology, the cost of the HESS is expected to reduce in
the future. Therefore, it is very hopeful the proposed wind
farm-HESS scheme can be an attractive alternative form of
reliable power generation, especially in locations where the
wind resource is bountiful.
VII. CONCLUSION
A hybrid energy storage system comprising of a battery and
supercapacitor has been proposed for use in a wind farm to
achieve firm power dispatch objective, while maximizing the
wind energy harness. In this scheme, a power-flows coordination strategy between the BESS and SC has been proposed
where the BESS power ramp-up and ramp-down rates are constrained so as to achieve an acceptable BESS life span. The more
rapid fluctuating wind power components are dealt with by the
SC. Furthermore, due to the random input wind power, a statistical approach is used to design the HESS so that the wind farm
power dispatch can be achieved at a specified confidence level.
The method allows the HESS power and energy capacities as
well as the BESS power ramp-rates to be determined. This is
done through minimizing the sum of the annualized capital and
replacement costs of the HESS, and the expected penalty cost
for not meeting the dispatch obligations.
From the evaluation of the cost/year index
of the HESS, it
is confirmed that SC is not viable as a long-term storage device.
Furthermore, numerical examples show that the capacities of
the HESS increases with the confidence level needed to achieve
firm power dispatch. Hence the proposed statistical approach
is a useful tool in the design of the HESS for it balances the
objective of meeting the firm power dispatch against the cost of
the HESS.
REFERENCES
[1] P. Li, H. Banakar, P. K. Keung, H. G. Golestani Far, and B. T. Ooi,
Macromodel of spatial smoothing in wind farms, IEEE Trans. Energy Convers., vol. 22, no. 1, pp. 119128, Mar. 2007.
[2] V. Banunarayanan, A. Chahal, L. Freeman, J. Martinez, N. Miller, D.
V. Zandt, M. Walling, and R. A. Walling, Analysis of Wind Generation
Impact on ERCOT Ancillary Services Requirements [Online]. Available: http://www.uwig.org/AttchB-ERCOT_A-S_Study_Final_Report.pdf
[3] D. L. Yao, S. S. Choi, T. J. Tseng, and T. T. Lie, A statistical approach
to the design of a dispatchable wind power-battery energy storage
system, IEEE Trans. Energy Convers., vol. 24, no. 4, pp. 916925,
Dec. 2009.
[4] J. Kondoh, Stationary applications. Load levelling, in Industrial Applications of Batteries, M. Brousselyon and G. Pistoia, Eds. Amsterdam, The Netherlands: Elsevier, 2007, p. 462, 465, 470477.

796

IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 4, NO. 3, JULY 2013

[5] N. Kawakami, Y. Iijima, Y. Sakanaka, M. Fukuhara, K. Ogawa, M.


Bando, and T. Matsuda, Development and field experiences of stabilization system using 34 MW Nas batteries for a 51 MW wind farm,
in Proc. IEEE Int. Symp. Industrial Elect., Jul. 2010, pp. 23712376.
[6] T. DeVries, The GVEA BESSChoosing a Multi-Million Dollar
System [Online]. Available: http://www.battcon.com/PapersFinal2002/DeVriesPaper2002.pdf
[7] J. P. Barton and D. G. Infield, A probabilistic method for calculating
the usefulness of a store with finite energy capacity for smoothing electricity generation from wind and solar power, Power Sources, vol.
162, no. 2, pp. 943948, Nov. 2006.
[8] H. Bludszuweit and J. A. Dominguez-Navarro, A probabilistic
method for energy storage sizing based on wind power forecast
uncertainty, IEEE Trans. Power Syst., vol. 26, no. 3, pp. 16511658,
Aug. 2011.
[9] H. Chen, T. N. Cong, W. Yang, C. Tan, Y. Li, and Y. Ding, progress
in energy storage system: A critical review, Prog. Natural Sci., vol.
19, no. 3, pp. 291312, Mar. 2009.
[10] Energy Storage Association Website [Online]. Available: http://www.
electricitystorage.org/site/technologies
[11] F. Liu, J. Liu, and L. Zhou, A novel control strategy for hybrid energy storage system to relieve battery stress, in Proc. IEEE Int. Symp.
Power Elect. for Dist. Gen. Sys., Jun. 1618, 2010, pp. 929934.
[12] C. Abbey, K. Strunz, and G. Joos, A knowledge-based approach for
control of two-level energy storage for wind energy systems, IEEE
Trans. Energy Convers., vol. 24, no. 2, pp. 539547, Jun. 2009.
[13] F. A. Farret and M. G. Simes, Integration of Alternative Sources of
Energy. Hoboken, NJ, USA: Wiley, 2006, pp. 267294.
[14] S. M. Schoenung, Long vs Short-Term Energy Storage Technologies
AnalysisA Life-Cycle Cost Study, Sandia Lab. [Online]. Available:
http://prod.sandia.gov/techlib/access-control.cgi/2003/032783.pdf
[15] D. Linden, Lindens Handbok of Batteries. New York, NY, USA: McGraw-Hill, 2011, p. 3, 13.15, 15.16.
[16] D. J. Spiers and A. A. Rasinkoski, Limits to battery lifetime in photovoltaic applications, Solar Energy, vol. 58, no. 4, pp. 147154, Oct.
1996.
[17] C. J. Kaiser, The Capacitor Handbook. Olathe, KS, USA: Commission Junction Pub., 1995, pp. 2126.
[18] G. Sideratos and N. D. Hatziargyriou, An advanced statistical method
for wind power forecasting, IEEE Trans. Power Syst., vol. 22, no. 1,
pp. 261264, Feb. 2007.
[19] A. D. Hansen, Generators and power electronics for wind turbines,
in Wind Power in Power Systems, T. Ackermann, Ed. Hoboken, NJ,
USA: Wiley, 2005, pp. 5378.

[20] L. Munteanu, A. L. Bratcu, N. Cutululis, and E. Ceanga, Optimal Control of Wind Energy Systems. London, U.K.: Springer-Verlag, 2008,
pp. 71108.
[21] S. Teleke, M. E. Baran, A. Q. Huang, S. Bhattacharya, and L. Anderson, Control strategies for battery energy storage for wind farm dispatching, IEEE Trans. Energy Convers., vol. 24, no. 3, pp. 725731,
Sep. 2009.
[22] M. Cohen, Using Ultracapacitors in System Designs [Online].
Available: http://www.maxwell.com/products/ultracapacitors/docs/
200904_Whitepaper_UltracapacitorsTopTenReasons_MCohen.pdf
[23] P. Poonpun and W. T. Jewell, Analysis of the cost per kilowatt hour
to store electricity, IEEE Trans. Energy Convers., vol. 23, no. 2, pp.
529534, Jun. 2008.
[24] FERC Regulations for Transmission Service Discrimination Prevention [Online]. Available: http://www.ferc.gov/whats-new/comm-meet/
2007/021507/E-1.pdf
[25] G. M. Masters, Renewable and Efficient Electric Power Systems.
Hoboken, NJ, USA: Wiley, 2004, p. 6, 352, 364.

K. W. Wee (S09) received the B.E. degree in 2004 and M.Sc. degree in power
engineering in 2008 from Nanyang Technological University, Singapore, where
he is currently working toward the Ph.D. degree.

S. S. Choi (M03SM07) received the B.E. and Ph.D. degrees from the University of Canterbury, New Zealand, in 1973 and 1976, respectively.
He is now a Professor in the School of Electrical and Electronic Engineering,
Nanyang Technological University, Singapore. His current research interests include renewable energy systems and energy storage.

D. M. Vilathgamuwa (M93SM99) received the B.Sc. and Ph.D. degrees in


electrical engineering from the University of Moratuwa, Sri Lanka, and Cambridge University, Cambridge, U.K., in 1985 and 1993, respectively.
He is now an Associate Professor in the School of Electrical and Electronic
Engineering, Nanyang Technological University, Singapore. His current
research interests include electrical drives and power quality.

You might also like