An Energy Storage Assessment Using Optimal Control Strategies To Capture Multiple Services

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

An Energy Storage Assessment: Using Optimal

Control Strategies to Capture Multiple Services


Di Wu, Chunlian Jin, Patrick Balducci, and Michael Kintner-Meyer

Abstract—This paper presents a methodology for evaluating ben- I. I NTRODUCTION


efits of battery storage for multiple grid applications, including
energy arbitrage, balancing service, capacity value, distribu- Using an energy storage system (ESS) for grid applications
tion system equipment deferral, and outage mitigation. In the is not a new concept. Even before 1980s, due to the limited
proposed method, at each hour, a look-ahead optimization is
first formulated and solved to determine battery base operating flexibility of steam plants and projected nuclear power capacity
point. The minute by minute simulation is then performed to additions, power system engineers in the U.S. began studying
simulate the actual battery operation. This methodology is used to energy storage technologies. As a result, several ESS projects
assess energy storage alternatives in Puget Sound Energy System. were built, primarily pumped hydro plants. Wind and solar
Different battery storage candidates are simulated for a period power have been developed rapidly in recent years world-
of one year to assess different value streams and overall benefits,
as part of a financial feasibility evaluation of battery storage wide, with the growing challenges to integrate these variable
projects. energy resources into power systems on a large scale. Recent
developments and advances in energy storage and power
Index Terms—Battery, energy storage system, optimization, power
system economics. electronics technologies are making the application of ESS
technologies a viable solution for increasing flexibility and
N OMENCLATURE improving reliability and robustness of power systems.
Many studies have been devoted to various ESS tech-
Es Battery energy capacity.
nologies for grid applications. The technical and economic
Lk , Lk Lower and upper bounds of state of charge (SOC)
characteristics of an array of ESS technologies are reviewed
for during hour k, respectively.
and compared in [1]–[3], including pumped hydro, battery
lk Battery SOC at the end of hour k.
storage, flywheel, compressed air, superconducting magnetic
pk Power exchange between battery storage and grid
energy storage, and advanced capacitors. The identified grid
(measured at the grid connection point) during
applications include energy arbitrage/load leveling, frequency
hour k, which is positive when injecting power
regulation, load following, spinning and non-spinning reserves,
into grid, i.e., using generator convention.
T&D deferral etc. References [4] and [5] specifically study en-
p+ −
k , pk Power injection/withdrawal into/from grid during
ergy storage for wind power balancing and integration. Studies
hour k, respectively.
[6] and [7] are dedicated to various battery technologies and
pbatt
k Rate of change of energy stored in the battery at
methods of assessing their economic viability and impacts on
the end of time hour k, which is positive when
power systems. Besides these reviews and high level studies
battery is discharged.
of ESS, many research work proposes method to allow more
Pkreq Minimum power output requirement during
detailed analysis. The economics of NaS batteries for energy
hour k for capacity value and distribution deferral
arbitrage and flywheels for regulation services are evaluated
applications.
for New York Independent System Operator and PJM in [8],
p+ −
max , pmax Maximum power (measured at the grid connec-
which is based on fixed utilization factors. The study in [9]
tion point) that can be injected and withdrawn
incorporates realistic CAISO regulation signals and battery
into/from grid, respectively.
responses to these signals to achieve more accurate results.
rk+ , rk− Balancing up and down capacity during hour k,
In fact, a battery storage device can be operated to provide
respectively.
multiple services at the same time. Considering the cost of
βk+ , βk− Balancing up and down price of hour k, respec-
the battery under today’s technology, value streams from mul-
tively.
tiple applications are extremely important for battery storage
+ −
k , k Energy reserve per MW balancing up and down
projects to be financially viable. Pacific Northwest National
service of hour k, respectively.
Laboratory (PNNL), Puget Sound Energy (PSE), and Primus
η+ , η− Discharging and charging efficiency of the battery
Power recently collaborated on a project funded by the U.S.
storage, respectively, including components such
Department of Energy (DOE) and the Bonneville Power Ad-
as conductor, power electronics, and battery.
ministration (BPA) to assess energy storage alternatives in the
λk Energy price of hour k.
PSE system considering multiple grid applications. This paper
This material is based upon work supported by the Energy Storage Systems presents the battery storage evaluation methodology developed
Program of the U.S. Department of Energy and the Technology Innovation at PNNL. The methodology described in this paper is novel. It
Grant Program of the Bonneville Power Administration. formulates the value creation as an optimization problem that
D. Wu, P. Balducci and M. Kintner-Meyer are with Pacific Northwest
National Laboratory, Richland, WA. maximizes value streams over a period of time. The recent
C. Jin is with Electric Reliability Council of Texas, Austin, TX.

978-1-4673-8040-9/15/$31.00 ©2015 IEEE


Authorized licensed use limited to: Institut Teknologi Sepuluh Nopember. Downloaded on October 24,2022 at 18:01:43 UTC from IEEE Xplore. Restrictions apply.
proceedings of the California Public Utility Commission indi- Optimization time horizon to determine base ༃
cates the difficulty assessing ESS providing bundled services. operating point for hour 1
To the best of our knowledge, this paper documents for the first Simulate actual operation for minute 1 in hour 1
time an optimization approach that seeks a control strategy that ༄

captures multiple services simultaneously. To identify cost-
effective markets for the nascent battery storage technologies, Optimization time horizon to determine base ༃
it will become necessary to capture not only one value but operating point for hour 2
multiple values at the same time.
Simulate actual operation for minute 1 in hour 2
The rest of this paper is organized as follows: Section II

describes different applications and key assumptions in this …
study. Section III presents the value assessment methodology.


In Section IV, evaluation results are presented and discussed
for battery storage in the PSE system. Finally, Section V time
concludes the paper. Fig. 1. Optimal control strategy for battery storage.

II. A PPLICATION D ESCRIPTION AND K EY A SSUMPTIONS


depends on how batteries are operated. In this work, we
An analytical framework for evaluating the net benefits of first propose an optimal control strategy for battery storage.
battery storage has been developed at PNNL. The following The corresponding battery benefits are then assessed. In the
applications are considered: proposed strategy, at each hour, a look-ahead optimization
• Energy arbitrage: Battery storage is charged during off-peak is formulated based on energy and balancing prices, and
hours and discharged during peak hours to pursue revenue the power requirements for capacity value and distribution
from energy trading. The economic reward is the price deferral services. The length of the look-ahead window is 24
differential between buying and selling electrical energy, hours in this study. The optimization problem is solved to
minus the cost of losses during the full charging/discharging determine the base operating point (power exchange between
cycle. In this study, it is assumed that the hourly energy battery and grid at AC side) and balancing service capacity
prices are available within the look-ahead windows. on an hour-by-hour basis. These optimal solution (provided in
• Balancing service: Battery storage can also provide balanc- hourly increments) together with balancing signals and outage
ing service by responding to system signals in real-time, information are used to simulate minute by minute battery
and therefore help to balance system generation and load. charging and discharging operation through the operating hour.
The hourly balancing prices are input to the optimization to This process repeats every hour, resulting in a rolling window,
determine balancing capacity from the battery. The response as illustrated in Fig. 1.
of the battery to balancing signal is then simulated to
determine the actual power output. A. Optimization Formulation
• Capacity value and distribution system deferral: The basic
assumption governing the capacity value and distribution This section presents the mathematical formulation of multi-
deferral analysis is that an ESS could offset part of in- hour optimization to be solved at each hour. Due to the
vestment in generation, substation, or distribution circuit, maximum power limit, the scheduled power output level affects
given the forecast system peaks. To receive the value from the balancing capability of battery storage. In addition, since
providing system capacity or deferring local distribution there is only limited energy that can be stored in the battery, the
system investment/upgrade, the battery storage must exceed operation of battery in different hours is interdependent. For
certain power output level during peak hours. example, injecting more energy into the grid in one hour in-
• Outage mitigation: Battery storage can reduce outage du-
creases the revenue for that hour, but results in less energy left
ration and therefore reduce the cost associated with out- for future use, and therefore may reduce the overall revenue.
ages. Two scenarios are considered when evaluating outage Hence, the battery base operating point for each hour needs to
mitigation values. In the outage-without-foresight scenario, be determined by solving a multi-hour optimization problem,
battery operation is scheduled without foresight of outages. as formulated in (1), where capacity value, distribution system
Therefore, the optimization proceeds the same as without deferral, and outage mitigation services are also reflected.
outages. When outages occur, the battery immediately re- In general, capacity and distribution deferral value per MW
sponds to pick up the local load until the battery is fully is much higher than energy arbitrage and ancillary services.
discharged or the system recovers from outages. In the In addition, these two services only require the battery to
outage-with-foresight scenario, it is assumed that battery is meet certain power output levels during a few peak hours
scheduled with foresight of outages. The battery is operated through out a year. Therefore, their power requirements are
to maintain full SOC before outages occur and mitigate formulated as constraints, with their service values estimated
outages as much as possible. This scenario provides an upper exogenously. As for outage mitigation, because neither the
bound of outage mitigation value. timing nor duration of outage is known when optimization
is performed in the outage-without-foresight scenario, outage
III. A SSESSMENT M ETHODOLOGY does not affect optimization at all. In the outage-with-foresight
There are different ways to charge and discharge batteries scenario, the battery is scheduled to maintain full SOC before
over time, and the total value from various applications really outages. Therefore, SOC should be set to 1 previous to each

Authorized licensed use limited to: Institut Teknologi Sepuluh Nopember. Downloaded on October 24,2022 at 18:01:43 UTC from IEEE Xplore. Restrictions apply.
K
  
max λk pk + βk+ rk+ + βk− rk− (1a)
p+ − + −
k ,pk ,pk ,pk ,lk ,rk ,rk
batt
k=1
subject to:
Power injection limit: 0 ≤ p+ +
k ≤ pmax , ∀k = 1, ··· ,K (1b)

Power withdrawal limit: 0 ≤ pk ≤ p − max , ∀k = 1, ··· ,K (1c)
+ −
Either charging or discharging (unnecessary): pk pk = 0, ∀k = 1, ··· ,K (1d)
Power transfer between battery and grid: pk = p + −
k − pk , ∀k = 1, ··· ,K (1e)
req
Power support requirement: Pk ≤ p k , ∀k = 1, ··· ,K (1f)
p+k − −
Rate of change of energy in battery: k = + − pk η ,
pbatt ∀k = 1 , · · · , K (1g)
η
1 batt
Dynamics of SOC: lk = lk−1 − p , ∀k = 1 , · · · , K (1h)
Es k
State of charge level limits: Lk ≤ lk ≤ L k , ∀k = 1 , · · · , K (1i)
Balancing up capacity: rk+ ≤ p+ max − pk , ∀k = 1 , · · · , K (1j)
Balancing down capacity: rk− ≤ pk + p− max , ∀k = 1 , · · · , K (1k)
+ +
 r
SOC limits with balancing up energy: Lk ≤ lk − k+ k ≤ Lk , ∀k = 1 , · · · , K (1l)
η Es
− r − η −
SOC limits with balancing down energy: L k ≤ lk + k k ≤ Lk , ∀k = 1 , · · · , K (1m)
Es
Desired SOC at the end of time horizon: lset ≤ lK . (1n)

outage. follow balancing signals sometimes. In this study, we found


The objective expressed in (1a) includes revenue from that there is only about two percent of the time when the
energy trading and balancing services within the look-ahead battery fails to follow balancing signals. Of course, one can
window. Constraints (1b) and (1c) enforces bounds of power always increase the energy reserve + −
k and k in (1l) and
injection and withdrawal into/from grid, respectively. Con- (1m) to improve balancing service performance. However, this
straint (1d) is nonlinear, which ensures charging and discharg- will reduce the amount of balancing services that the battery
ing does not take place at the same time in the mathematical can provide. As for an outage, when it occurs, the battery
model. However, when η + or η − is strictly less than 1 and λk is is programmed to stop following the operation schedule and
positive, this constraint is not necessary because a solution with start immediately to support local system until the battery runs
p+ −
k and pk both nonzero cannot be optimal. Constraint (1e)
out of energy or the outage ends. The required battery output
expresses power transfer between battery and grid. Constraint during an outage is equal to local system load. The reduced
(1f) ensures that battery power output meets the requirement outage duration (because of battery support) is calculated
from capacity value and distribution deferral applications. and compared with the original duration (without battery) to
Constraint (1g) calculates rate of change of energy stored in evaluate the outage mitigation value.
the battery. Constraint (1h) describes the dynamics of SOC.
Constraint (1i) restricts the battery SOC level to be between IV. C ASE S TUDY R ESULTS
its lower and upper bounds. Constraints (1j) and (1k) calculate There were four candidate sites for battery storage installa-
balancing up and down capability, respectively. Constraints tion in the PSE system. Six criteria were used for preliminary
(1l) and (1m) ensure enough energy is reserved to provide screening: i) application values, ii) travel time from PSE
balancing services. Constraint (1n) specifies SOC at the end operations/maintenance centers to the site, iii) site preparation
of the schedule time horizon. and permitting, iv) the extent that the project aligned with key
operational priorities at PSE, v) the additional expense to PSE
B. Minute by Minute Simulation to ensure communication with the ESS, and vi) the likelihood
In minute by minute simulation, the amount of balancing of positive community engagement. After preliminary screen-
power is first calculated for each minute, which is equal to the ing, Bainbridge Island (BI) and Baker River were selected for
product of the balancing signal per MW and the balancing further analysis using the method presented in Section III. Due
service capacity determined in optimization. These values to space limitations, this paper only presents the evaluation
together with the scheduled power exchange level are used to results for BI. One of the battery storage candidates is a
determine the actual battery output and SOC at each minute. 4 MW/16 MWh Zinc bromide flow battery from Primus
Once the SOC reaches the upper or lower bound, the desired Power. The charging and discharging efficiencies (including
output will not be deployed. Therefore, the battery may fail to both battery and inverter) are 0.836 and 0.807, respectively.

Authorized licensed use limited to: Institut Teknologi Sepuluh Nopember. Downloaded on October 24,2022 at 18:01:43 UTC from IEEE Xplore. Restrictions apply.
TABLE I Arbitrage Balancing T&D deferral Capacity value Outage Mitigation
C APACITY VALUE AND D ISTRIBUTION D EFERRAL P OWER R EQUIREMENT
1200

Application Hour Power req. (MW) 1000

Capacity value 1/5/18 8:00 AM 4 800

$thousand
Distribution 1/12/18 8:00 AM 1.4
600
deferral 1/12/18 9:00 AM 0.88
400

200

A. Input Data 0

In order to evaluate the battery storage benefits, the follow- 0 5 10 15 20 25


Price index
30 35 40 45 50

ing data are required to run the optimization and simulation:


i) minute by minute system balancing signal, ii) hourly energy Fig. 2. Annual value streams under 50 prices in the outage-without-foresight
scenario.
and balancing prices, iii) power requirement for capacity value
and distribution deferral, and iv) outage information. Arbitrage Balancing T&D deferral Capacity value Outage Mitigation
• Balancing signals: The stochastic methodology presented in 1200

[10] is used to determine the system reserve requirement and 1000


balancing signals in the PSE system, based on one minute
800

$thousand
system load data, wind generation data, statistics of current
load and wind forecast errors. 600

• Energy and balancing prices: Aurora is used by PSE to min- 400

imize the production cost and generate the corresponding 200

energy prices. A number of Monte Carlo (MC) simulations 0

have been performed in order to capture uncertainties. One 0 5 10 15 20 25 30 35 40 45 50


Price index
can follow the approach in [11] to determine the number MC
samples that is sufficient enough for obtaining output dis- Fig. 3. Annual value streams under 50 prices in the outage-with-foresight
scenario.
tribution. In this study, because of time consuming process
in running production cost model, fifty MC simulations are
performed as the trade-off between outcome space coverage
is based on a detailed proforma which is built to estimate
and simulation time. Based on the results from Aurora
the revenue requirements for combustion turbine. The deferral
and system reserve requirement generated by PNNL, PSE
value is calculated as the difference in the net present value
internal mixed-integer LP model is used to determine the
requirements between building the new substation as planned
balancing services and prices.
by PSE versus deferring it for nine years. Please refer to
• Capacity value and distribution deferral: For capacity value,
[12] for detailed economic analysis. Considering the round-
the battery storage system on BI is assumed to offset the
trip efficiency of 0.674, the price difference is not significant
same amount of equivalent of a peaking turbine. For distri-
enough for energy arbitrage most of the time. There is only
bution deferral, PSE has considered many options for adding
around 15% of 8,760 hours when the battery is scheduled
additional capacity on BI but currently favors adding a
for energy arbitrage. Hence, the value from arbitrage only
new substation near the existing Winslow substation. Based
accounts for a fraction of the total benefits. It should be noted
on the forecast load information and available system/local
that the response of battery to system balancing signals also
resource capacity, the power requirement from these two
incurs energy exchange between battery and grid, which may
applications is listed in Table I.
decrease revenue from energy arbitrage but is necessary for
• Outage: To evaluate the benefits from outage mitigation,
balancing service. In this study, the energy associated with
historical outage events have been examined. These events
balancing service is also settled into the arbitrage component.
are defined by the timing, duration, and number of customers
Therefore, arbitrage values are small negative numbers for
affected by each outage. The average annual number of
some of the prices. For all the prices, power requirement from
outages are then selected from 2000 to 2012 for simulation1 .
capacity value and distribution deferral can be met. Therefore,
B. Simulation Results these value streams are the same for all 50 prices. The
balancing service value varies much because of variation of the
Using the input data, the optimization and simulation are
input balancing prices. The outage mitigation value depends
performed through 8,760 hours using each set of 50 energy
on the SOC of battery before outages, which also varies with
& balancing prices from PSE. The annual value streams from
prices. However, in the outage-with-foresight scenario, since
different applications in the outage-without-foresight scenario
the battery is operated to maintain full SOC, outage mitigation
are plotted in Fig. 2, with Lk = 0 (lower bound of SOC) and
value does not vary with price, as shown in Fig. 3.
Lk = 1 (upper bound of SOC). Herein, the capacity value
In order to increase the value from outage mitigation in
1 There were 16 transmission-level outages that affected BI system during outages-without-foresight scenario, one can increase the lower
the 2000-2012 time frame, and that averages to 1.2 outages per year, which bound of SOC (Lk ) in (1). On the other hand, this will
we rounded down to 1. There were 59 distribution-level outages that affected
BI system during the 2010-2012 time frame, and that averages 19.7 outages reduce the balancing capacity that the battery storage can
per year, which we rounded to 20. provide. However, if this lower bound only increases when the

Authorized licensed use limited to: Institut Teknologi Sepuluh Nopember. Downloaded on October 24,2022 at 18:01:43 UTC from IEEE Xplore. Restrictions apply.
TABLE II
A NNUAL VALUE S TREAMS IN BASE C ASE ($000) simultaneously. The proposed control strategy and evaluation
method are used to assess battery storage alternatives in the
Arbitrage Balancing Cap. value Deferral Outage mitig. Total PSE system. The results of battery storage on BI are presented
5.336 290.251 568.843 622.631 1066.852 2553.914 and discussed. It is found that multiple grid applications can
Power (MW)
be successfully exploited from a battery storage device.
Scheduled hourly power Schedule power+balancing signal Acutally power minute by minute In this study, an optimal control strategy is proposed to
4
outage mitigation maximize the total benefits for a given battery size. In future
2
H

0
 work, we plan to extend the existing methodology for optimal
−2 sizing of battery storage.
−4
06/09/18−00:00 03:00 06:00 09:00 12:00 15:00 18:00 21:00 06/10−00:00
SOC
1
ACKNOWLEDGMENT
0.8
The authors are particularly thankful to Dr. Imre Gyuk,
0.6

0.4
manager of the Energy Storage System Program of DOE,
0.2 Office of Electricity Delivery and Energy Reliability, and
0
06/09/18−00:00 03:00 06:00 09:00 12:00 15:00 18:00 21:00 06/10−00:00
Mr. Terry Oliver, Chief Technology Innovation Officer of the
Technology Innovation Grant Program of BPA for co-funding
Fig. 4. Power output and SOC for a sample day. this project. We are also thankful to Mr. Patrick Leslie and
Mr. Charles Daitch from PSE and Dr. Andrew Marshall from
balancing up prices (corresponding to needs of discharging) are Primus Power for their valuable inputs and discussion. We also
zero, the revenue from balancing service will not be affected wish to acknowledge the oversight and direction provided by
much. In our study, 40% is used as SOC lower bound when Mrs. Melanie Smith of BPA and Mrs. Kari Nordquist of ACME
battery does not provide balancing services. It is found that the Business Consulting, LLC. Finally, we are thankful to Mr. Tom
outage mitigation value significantly increases compared with Stepien of Primus Power for his leadership and insights into
Fig. 2, and is close to its upper bound in Fig. 3. Meanwhile, the operation and management of the Primus Power energy
the other value streams remain almost the same. storage system.
It is found that case No. 24 provides total benefits that are
closest to the mean value of 50 cases. This case is selected as
the base case, and the corresponding value streams are listed in R EFERENCES
Table II. The battery power output information together with [1] P. F. Ribeiro, B. L. Johnson, M. Crow, A. Arsoy, and Y. Liu, “Energy
storage systems for advanced power applications,” Proc. IEEE, vol. 89,
SOC for a sample day (June 9, 2018) are plotted in Fig. 4. no. 12, pp. 1744–1756, 2001.
As can be seen, the actual power deviates from the scheduled [2] I. Hadjipaschalis, A. Poullikkas, and V. Efthimiou, “Overview of current
power because the battery responds to balancing signals. For and future energy storage technologies for electric power applications,”
Renew. Sust. Energy Rev., vol. 13, no. 6-7, pp. 1513–1522, 2009.
this sample day, the actual output is the same as the sum [3] S. Vazquez, S. M. Lukic, E. Galvan, L. G. Franquelo, and J. M. Carrasco,
of scheduled power and balancing power, which means the “Energy storage systems for transport and grid applications,” IEEE Trans.
battery is able to follow all the balancing signals. There is an Ind. Electron., vol. 57, no. 12, pp. 3881–3895, 2010.
[4] S. Faias, P. Santos, F. Matos, J. Sousa, and R. Castro, “Evaluation of
outage starting from 7:59 to 9:38 AM. The battery responds energy storage devices for renewable energies integration: Application
to pickup the local load when the outage occurs. to a Portuguese wind farm,” in Proc. IEEE Elec. Market EEM 5th Int.
In the financial feasibility analysis, we have compared the Conf. Eur., Lisboa, May 2008.
[5] F. Díaz-González, A. Sumper, O. Gomis-Bellmunt, and R. Villafáfila-
total benefits of battery storage for PSE system in different Robles, “A review of energy storage technologies for wind power
cases (using different credible capacity value, outage mitiga- applications,” Renew. Sust. Energy Rev., vol. 16, no. 4, pp. 2154 – 2171,
tion assumption, and other battery sizes) and the associated 2012.
[6] K. C. Divya and J. Østergaard, “Battery energy storage technology for
revenue requirements from storage installation. Due to space power systems–An overview,” Electric Power Sys. Research, vol. 79,
limitations, only the analysis results corresponding to the base no. 4, pp. 511–520, 2009.
case in Table II are presented here. The total battery storage [7] B. Dunn, H. Kamath, and J.-M. Tarascon, “Electrical energy storage for
the grid: A battery of choices,” Science, vol. 334, no. 6058, pp. 928–935,
capital cost is about $14.8 million, which includes battery 2011.
storage device cost (accounts for about $9.2 million without [8] R. Walawalkar, J. Apt, and R. Mancini, “Economics of electric energy
consideration of of sales taxes or overhead charges), and storage for energy arbitrage and regulation in New York,” Energy Policy,
vol. 35, no. 4, pp. 2558–2568, 2007.
other non-battery cost, such as electrical, thermal manage- [9] N. Lu, M. R. Weimar, Y. V. Makarov, and C. Loutan, “An evaluation
ment, site/civil, installation, communications, IT, overheads, of the NaS battery storage potential for providing regulation service in
WA sales tax, and contingency. Total revenue requirement in California,” in Proc. IEEE Power Systems Conf. Expos., 2011, pp. 1–9.
[10] Y. Makarov, S. Lu, B. McManus, and J. Pease, “The future impact of
present value is $20.3 million assuming 20-year life of the wind on BPA power system ancillary services,” in Proc. IEEE Power
ESS with discount rate equal to 7.77%. The revenue from the Energy Soc. Transm. Distrib. Conf. Expos., 2008, pp. 1–5.
base case is about $26.6 million, which represents a return on [11] Y. Xu, Q. Hu, and F. Li, “Probabilistic model of payment cost mini-
mization considering wind power and its uncertainty,” IEEE Trans. Sust.
investment (ROI) of $6.5 million and ROI ratio of 1.32. Energy, vol. 4, no. 3, pp. 716–724, July 2013.
[12] P. Balducci, C. Jin, D. Wu, M. Kintner-Meyer, P. Leslie, C. Daitch,
V. C ONCLUSION AND F UTURE W ORK and A. Marshall, “Assessment of energy storage alternatives in the
This paper presents a novel optimization approach that Puget Sound Energy system vol1: Financial feasibility analysis,” Pacific
Northewest National Laboratory, Tech. Rep., 2013.
captures multiple value streams of energy storage systems

Authorized licensed use limited to: Institut Teknologi Sepuluh Nopember. Downloaded on October 24,2022 at 18:01:43 UTC from IEEE Xplore. Restrictions apply.

You might also like