Journal of Energy Storage 15 (2018) 336–344

Contents lists available at ScienceDirect

Journal of Energy Storage

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

The benefits of grid-scale storage on Oahu

James F. Ellison* , Lee J. Rashkin* , Joseph Serio, Raymond H. Byrne
Sandia National Laboratories, PO Box 5800, Albuquerque, NM 87185-1140, United States

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

Article history:
Received 28 August 2017 The Hawaiian Electric Company intends to procure grid-scale Battery Energy Storage System (“BESS”)
Received in revised form 6 December 2017 capacity. The purpose of this study is to determine whether providing contingency reserve or time-of-day
Accepted 7 December 2017 shifting is of more benefit to the Oahu grid, and to better understand the relationship between BESS size
Available online 23 December 2017 and level of benefit. This is an independent study by Sandia, and is not being used to support the
regulatory case for BESS capacity by Hawaiian Electric. The study team created a production cost model of
Keywords: the Oahu grid using data primarily from the Hawaiian Electric Company. The proposed BESS supplied
Energy storage contingency reserve in one set of runs and time-of-day shifting in another. Supplying contingency reserve
Reserves
led to larger savings than time-of-day energy shifting. Assuming a renewable reserve and a quick-start
Production cost model
reserve, and $15/MMBtu for Low-Sulphur Fuel Oil, the 50-MW/25-MWh, 100-MW/50-MWh, and 150-
Hawaii
MW/75-MWh systems supplying contingency reserve provided, respectively, savings of 9.6, 15.6, and
18.3 million USD over system year 2018. Over the range of fuel prices tested, these cost savings were
found to be directly proportional to the cost of fuel. As the focus is the operational benefit of BESS
capacity, the capacity value of the BESS was not included in benefit calculations.
© 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND
license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction A BESS becomes more expensive as the energy storage

The Hawaiian Electric Company has requested proposals for
grid-scale storage on Oahu [1]. Given that Hawaii has adopted a
100% Renewable Portfolio Standard for electric utilities by 2045 [2],
it is clear that the amount of variable generation1 will only
increase. Given its usefulness in integrating variable generation [4],
the opportunities for energy storage to play a role in Hawaii will
only increase.
Given Hawaiian Electric's intent to procure Battery Energy
Storage System (“BESS”) capacity, Sandia's2 motivation for this
study was to explore what BESS service would most benefit the
system, and to explore how BESS size impacts the amount of
benefit. This is an independent study by Sandia, and is not being
used to support the regulatory case for BESS capacity by Hawaiian
Electric.
* Corresponding authors.
E-mail addresses: jelliso@sandia.gov (J.F. Ellison), lrashki@sandia.gov
(L.J. Rashkin).
1 integration in Hawaii [6], as well as the value of energy storage on
Variable generation, such as wind and solar photovoltaic generation, is not
dispatchable. For an overview of variable generation in power systems, please see
[3].
2
In this paper, ‘Sandia’ refers to the authors of this paper. It does not imply a
viewpoint or position held by the leadership of Sandia National Laboratories.
requirement increases [5]. Therefore the ideal function for a BESS
to perform would be one that adds a great deal of value while
requiring a small energy storage component. One such function
may be the provision of Contingency Reserve. A Contingency
Reserve is set aside in the event of an unforeseen unit trip (often
called a ‘forced outage’), and is typically a spinning reserve. It
usually takes hours to repair a unit that has experienced a forced
outage. A BESS with a small energy storage capacity could provide
this reserve, provided that additional generation could be brought
on line before the battery's reserves are exhausted. This generation
would need to be brought on line fairly quickly, hence we term it a
“Quick Start” Reserve.
Another possibility for savings is having the battery act to
provide time-of-day energy shifting, which is often termed
“Arbitrage.” Generally energy is more costly to produce on-peak
than it is off-peak, as less fuel-efficient units may be dispatched at
peak times. Having the battery generate on-peak and charge off-
peak may save on operational costs, provided that the gap in on-
peak and off-peak generation cost is sufficient to overcome the
BESS round-trip efficiency losses.
This study builds on previous work studying solar and wind
Maui [7].
Our hypothesis is that Contingency Reserve supplied by the
combination of the BESS and a Quick Start Reserve would be less

https://doi.org/10.1016/j.est.2017.12.009
2352-152X/© 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
 J.F. Ellison et al. / Journal of Energy Storage 15 (2018) 336–344
expensive than employing conventional units to provide the
reserve. In addition, the expected benefits from using the BESS (in
conjunction with a Quick Start Reserve) to provide Contingency
Reserve are greater than from using the BESS to provide Arbitrage.
2. Material and methods
To test this hypothesis, a production cost model of the Oahu grid
was created using data supplied by Hawaiian Electric as well as
solar data generated at Sandia.
A production cost model captures the costs of operating a fleet
of generators, and can consider the transmission constraints of a
power system. An optimization is performed to identify the least-
cost dispatch given the system load, unit characteristics, fuel costs,
and renewable generation [8]. In this case, a Mixed-Integer Linear
Programming (“MILP”) model was used. A MILP model will
formulate the problem where some variables, such as unit
commitment (whether a unit is to be on or off), are integer
values, whereas other variables (such as heat rate) are not
constrained to be integers [9].
As this study assumes an energy storage facility is operational in
2018 [10], the model was initialized to reflect generation and load
as projected for 2018. When conducting a study on a system in the
future, it is expected that certain assumptions (such as fuel prices
and reserve requirements) may prove to be inaccurate. Therefore,
the study team sought to do sufficient scenario analysis to
understand what the value of a BESS might be under a variety of
conditions. The model runs in this study were conducted in
September 2016.
In this study, PLEXOS1 (production cost modelling software by
Energy Exemplar1) was used. The study team input unit
characteristics, load, variable generation output, fuel types and
costs, and reserve types and required amounts into PLEXOS1. The
software uses this input to formulate the optimization problem,
and then passes this problem off to a solver to come up with the
solution.
The optimization problem is the least-cost provision of energy
and reserve, subject to constraints. The total cost which the model
seeks to minimize (while serving load and observing all other
constraints) is shown in Eq. (1), where FuelCosti, StartCosti, and
VariableOperatingCosti represent the total fuel, start, and variable
operating costs, respectively, in each hour across all units.
8X
760
TotalCost ¼ FuelCosti þ StartCosti
i¼1
þ VariableOperatingCosti ð1Þ
1
Strictly speaking, PLEXOS does not attempt to minimize Total
Cost in each hour for the whole year at once. Instead, it
approximates this through the use of integrated simulation phases
that start out with looking at the entire year, and at the default
setting end up with performing hourly dispatch (the “dispatch
interval”) that is optimized over a 24-h period (the “step size”).
While the default setting is used for this analysis, the user can
specify both the step size and the dispatch interval. Dispatch can
take place every 5 min if desired. Fig. 1 illustrates the integrated
simulation phases that the PLEXOS1 model performs.3 Informa-
tion from each simulation phase is passed down to the next.
First, in the Projected Assessment of System Adequacy (or PASA)
phase, the model looks over the entire year, but not for the
purposes of hourly dispatch. It is in this phase that the random
forced outages of generators (using user-specified probabilities)
3 4
A discussion of the PLEXOS simulation phases is contained in the user manual
(only available to subscribers) at: https://wiki.energyexemplar.com.
337
Fig. 1. PLEXOS1 flowchart.
are assigned. Taking both the assigned generator forced outages
and scheduled maintenance into account, the model chooses the
optimal time for distributed maintenance (where a certain amount
of maintenance must be performed, but a schedule is not
specified). Having done this, the model then calculates a Loss of
Load Probability (LOLP).
Second, the Medium Term Schedule (or MT Schedule) uses
temporal simplification to greatly speed execution time. The main
purpose of this phase is to decompose medium-term constraints
and objectives so that they can be fully accounted for in the
chronological Short Term Schedule simulation. An example in this
model is that several units have annual minimum fuel consump-
tion amounts.
Last, the Short Term Schedule (or ST Schedule) is a chronologi-
cal Unit Commitment and Economic Dispatch (UCED) model based
on mixed-integer programming. This phase co-optimizes energy
and reserves in order to meet the load and reserve specifications
for each time period. The window of optimization (or step size)
used in this study is 24-h, meaning that unit commitment and
dispatch are solved for each 24-h period as one problem. This
means that the model has perfect foresight over the 24-h period.
An hourly dispatch interval was specified for this analysis.
Reserve Requirements. A Contingency Reserve is power produc-
tion capacity that is set aside so that it may be used in the event of a
forced outage at a generator. It should be able to supply power long
enough to either repair the unit with an outage or get a
replacement unit on-line. Here we define Contingency Reserve
as on-line reserve that can reach nominated output within 10 min
and maintain that level of output for at least 2 h. Any conventional
unit that is on-line can supply this reserve. The largest single unit
on Oahu is the 180-MW AES coal-fired plant. For the purposes of
this study, the Contingency Reserve is set at this size. Assuming
that 25 MW of demand response will be available at all times
reduces the net Contingency Reserve requirement to 155 MW,
which is what the study team has modeled.
In order to achieve the same level of reliability as conventional-
unit Contingency Reserve, BESS-supplied Contingency Reserve
should be able to supply the required duration of output (in this
case, two hours). Since the BESS we analyze here is able to supply
30-min of power at full ouput, there should be a Quick Start reserve
to provide the power output duration that the BESS lacks.4 Here we
have defined Quick Start Reserve as off-line power generation
capacity that can be brought on-line at nominated output within
10-min. We assume that all of the combustion turbines and
In cases with a Quick Start Reserve, we have constrained BESS Contingency
Reserve participation to the capacity available in Quick Start Reserve.
338 J.F. Ellison et al. / Journal of Energy Storage 15 (2018) 336–344
reciprocating engines on the Hawaiian Electric system can supply
Quick Start Reserve, with the exception of the Kalaeloa units,
which are primarily used for baseload generation.
There is an opportunity cost to specifying a Quick Start Reserve
– the units in that reserve will no longer be available to supply peak
power. In order to reduce costs, the system could be operated (less
reliably) without a Quick Start Reserve and therefore the scenario
where there is no Quick Start Reserve backing up a Contingency
Reserve-supplying BESS was also explored.
By 2018, a reserve to compensate for unexpected drops in solar
and wind production (referred to herein as a “Renewable Reserve”)
may be required. This reserve would also be on-line reserve that
can reach nominated output within 10 min and maintain that
output for at least 2 h, and can be supplied by any conventional unit
that is on-line.5 The amount of Renewable Reserve required will
depend on both the source of the variable energy and on the
amount dispatched. For analysis purposes the formula considered
in this effort is in Eq. (2).6
Renewable Reserve ¼
MINðWindGeneration; 0:5  WindCapacityÞ
þ MINðUtilityPVGeneration; 0:6  UtilityPVCapacityÞ
þ MINðDistributedPVGeneration; 0:2  DistributedPVCapacityÞ
ð2Þ
In Eq. (2), the actual Renewable Reserve is a function of the
actual output, but is capped at 50% of the installed capacity for
wind, 60% for utility PV, and 20% for distributed PV. Given the
assumptions used for 2018 generation, this comes out to 60 MW
for wind, 90 MW for utility PV, and 80 MW for distributed PV.7
To be clear, the model does not allow the BESS to provide
Renewable Reserve. The hourly model used here is insufficient to
understand how the battery would actually be used to firm wind
and solar generation. Frequent intra-hour discharges of energy
would likely be required to firm variable generation to a forecast.
Each time the BESS is used in this way, it would be unavailable for
either Renewable Reserve or Contingency Reserve until it is
recharged. Without modeling at a finer time scale, it is not possible
to examine the trade-offs involved in using the BESS to supply
Renewable Reserve and Contingency Reserve simultaneously. The
only category of reserve the BESS is allowed to provide in this study
is therefore Contingency Reserve.
At the same time, having a Renewable Reserve in the model is
important, as it reflects how the system may be operated with
greater amounts of variable generation. Having larger reserve
requirements puts more stress on the system. Whether the BESS
serves as Contingency or Renewable Reserve it still functions as a
replacement for spinning reserve, and thus reduces the stress on
the system.
BESS capacity and characteristics. A battery size sensitivity was
performed in this study, examining 50 MW/25 MWh, 100 MW/
50 MWh, and 150 MW/75 MWh.
The largest capacity was chosen because 150 MW is close to the
maximum Contingency Reserve that the BESS in this study could
provide, given a net Contingency Reserve requirement of 155 MW.
The other two sizes were selected to be spaced evenly apart, and to
provide a wide enough range for a useful comparison. A 30-min
5
For a general discussion of operating reserves for wind power integration,
please see [11].
6
This Renewable Reserve equation reflects thinking at Hawaiian Electric at the
time this study was conducted. A newer version was articulated in the December 23,
2016 PSIP.
7 11
Please see the Hawaii Solar Integration Study [6] for a discussion of how
different types and amounts of variable generation impact the amount of operating
reserves required.
storage volume (at maximum power output) was chosen as it
matches the Hawaiian Electric storage proposal request of “up to
30 minutes” [1].
While a 12-h battery (at maximum output) may seem small for
an Arbitrage application, it should allow the BESS to provide power
when it is most valuable. It is expected that each additional hour of
storage would provide diminishing returns. One caveat here is that
the model will dispatch the battery for the most valuable hour in
the day, as opposed to 12-h, as the finest time interval in the model is
an hour. Therefore, the value of this battery in the time-of-day
shifting application is likely to be understated.8
We assume that this BESS will be a lithium-ion battery, as this
an established energy storage battery technology and is well-
suited for the power and storage capacity ratings examined here. A
90% round-trip efficiency for the BESS is assumed when providing
Arbitrage. This represents the typical efficiency of a Lithium Ion
battery system [5].
In this study, the BESS is modeled as a pumped-hydro system,
with the power and storage capacity of the BESS being analyzed.
The battery model is thus two stocks – upper and lower reservoirs
– with flow between them. To pump 90 MWh into the upper
reservoir, 100 MWh must be drawn from the grid – yielding the
required 90% round-trip efficiency. In the Contingency Reserve
case, the battery is not actually activated – it simply remains in
reserve. In the Arbitrage case, the battery is deployed.
The reason this model's BESS does not actually supply power
when supplying Contingency Reserve is that this is an hourly
model with perfect foresight within a 24-h period. This model
cannot represent things that happen within an hour, such as an
outage at 1:34 pm and spinning reserve immediately being called
upon to compensate. Also, the model “knows” that a unit will
experience a forced outage in, say, hour 15 of a given day, and will
dispatch accordingly. The model neither needs a Contingency
Reserve nor is it able to call upon it. The closest we can come to the
real world in an hourly production cost model is to specify a
Contingency Reserve that must be set aside at all times – and never
activated.
3. Data
The conventional units included in the model, along with their
installed capacities, are in Table 1. All units use conventional fuel
oil, except for Schofield which will primarily use biodiesel. Both the
independently-owned dispatchable units and the variable genera-
tion in the model are listed in Table 2.9
Hawaiian Electric provided the following elements of data: unit
heat rate curves, fuel type10 and projected cost,11 startup costs (fuel
and maintenance), minimum and maximum generation points,
maintenance schedule, and forced outage rate (EFOR). Each unit
has its own probability of a forced outage, and each forced outage is
8
By half-hour battery, what is meant is that the battery has enough storage to
sustain its maximum power rating for half an hour of output. This means that a 100-
MW battery would have 50 MWh of storage. If the model were operating on a half-
hourly basis, the battery could be dispatched at 100 MW for half an hour, which
could show more value to the system than the hourly dispatch does. (However, the
battery could be dispatched differently, say at 25 MW for 2 h.)
9
For information on Hawaiian Electric's current generation fleet and their plans
going forward, please see their December 2016 PSIP (Power Supply Improvement
Plan) [12].
10
In these tables, ST stands for Steam Turbine, CT for Combustion Turbine (or gas
turbine), and RE stands for Reciprocating Engine. Also, LSFO stands for Low Sulphur
Fuel Oil, Oil for Fuel Oil #2, Bio for Biodiesel, and KALA Oil for the fuel used by the
Kalaeloa generator.
The projected 2018 fuel cost used in the model is (in USD/MMBtu): Coal – 1.93,
LSFO – 15.18, KALA Oil – 15.69, Fuel Oil – 17.94, and Biodiesel – 34.95. Please note
that Oahu LSFO prices on January 1, 2017 were about 9.30 USD/MMBtu.
 J.F. Ellison et al. / Journal of Energy Storage 15 (2018) 336–344 339

Table 1 Table 3
Hawaiian Electric Company's dispatchable units in model. Total generation cost savings (in million USD).

Unit Type Fuel Capacity Unit Type Fuel Capacity BESS size Renewable reserves No renewable reserves
Kahe 1 ST LSFO 82 MW Waiau 5 ST LSFO 55 MW Arb CR + QS CR Arb CR + QS CR
Kahe 2 ST LSFO 82 MW Waiau 6 ST LSFO 54 MW
50 MW/25 MWh 2.9 9.6 12.0 3.2 4.8 7.3
Kahe 3 ST LSFO 86 MW Waiau 7 ST LSFO 83 MW
100 MW/50 MWh 4.7 15.6 20.7 4.8 7.1 11.6
Kahe 4 ST LSFO 86 MW Waiau 8 ST LSFO 86 MW
150 MW/75 MWh 5.9 18.3 25.5 5.5 7.5 13.6
Kahe 5 ST LSFO 135 MW Waiau 9 CT Oil 53 MW
Kahe 6 ST LSFO 134 MW Waiau 10 CT Oil 50 MW
CIP CT CT Oil 112 MW Schofield RE Bio 49 MW

reserves results in significant added costs. The introduction of a

BESS has a greater impact when these costs are higher.
Table 2 If the BESS provides Contingency Reserve and a Quick Start
Independent dispatchable units in model | variable generation in model. Reserve is employed, the gas turbine units are not available to
Unit name Type Fuel Capacity Plant name Type Capacity
provide peak power to the system. This leads to higher operational
costs than in the scenarios where the Quick Start Reserve is not
AES ST Coal 180 MW Kahuku Wind 30 MW
HPOWER ST Waste 69 MW Kawailoa Wind 69 MW
employed.
Kalaeloa 1 CT KALA Oil 90 MW New Wind Wind 24 MW This same information for the Renewable Reserves cases is
Kalaeloa 2 CT KALA Oil 90 MW Utility Solar PV 148 MW presented in graphical form in Fig. 2. Increasing the capacity of the
Kalaeloa ST ST Waste 28 MW Distributed PV 400 MW BESS results in declining marginal utility.
Heat Solar
150-MW BESS, Contingency Reserve scenario. In order to
Airport DSG RE Biodiesel 8 MW
understand better how having a BESS provide Contingency Reserve
leads to reduced operational costs for the system, the Base Case
with Renewable Reserve is compared with the 150-MW BESS
assumed to last 24 h. As a result, in these model runs dispatchable
providing Contingency Reserve Case (simply referred to as the
generation capacity is offline due to forced outages for 1.5% of the
‘’150-MW BESS” case in this section).
year. Variable generation is assumed not to experience forced
Fig. 3 shows the cost increase or decrease for each unit for the
outages.
150-MW BESS Case as compared to the Base Case. Positive
Hawaiian Electric also provided hourly load, and hourly wind
numbers show an increase in cost as compared to the Base Case,
and utility solar generation – all for 2014. Sandia generated the
and negative numbers show a decrease in cost as compared to the
hourly distributed PV output for 2014 using satellite weather data,
Base Case.
as well as the location and concentration of distributed PV, as
Fig. 3 does not show a net difference in cost, but instead breaks
inputs [13–15].
the difference down into four components. Power Generation
Change indicates the change in cost due to increased (positive) or
4. Results
decreased (negative) generation. Average Heat Rate Change
indicates the change in cost due to an increase (positive) or
Table 3 illustrates the projected cost savings for 2018 for each
decrease (negative) in average heat rate. Variable O&M Change
scenario studied. In this table, “Arb” means time-of-day shifting
shows the change in cost due to more (positive) or less (negative)
through Arbitrage, “CR+QS” means Contingency Reserve together
operational costs. Start Cost Change shows the change in cost due
with a Quick Start Reserve, and “CR” means Contingency Reserve
to an increase (positive) or decrease (negative) in start costs.
without a Quick Start Reserve. The BESS scenarios with a
Most of the savings come from the gas turbines (Waiau 9 and 10,
Renewable Reserve are compared to a Base Case with a Renewable
and CIP CT1), though Kalaeloa 2 and Waiau 5 also produce
Reserve, and the scenarios without a Renewable Reserve are
substantial savings. The results also show increased costs from the
compared to a Base Case without a Renewable Reserve.
Kahe units, as well as Waiau 7 and 8. Most cost savings are from
Providing Arbitrage yields significantly less value than provid-
ing Contingency Reserve for all three battery sizes, whether or not
the system is required to carry Renewable Reserves. When
providing Arbitrage, the BESS is being dispatched to act as a
peaking plant, thus reducing the number of unit starts.12 The
specification of a 1:2 energy storage capacity to power ratio for
each BESS limits its utility as a peaking plant.
At the same time, the Arbitrage savings calculated here do not
take into account the capacity value of the BESS system. Adding
BESS capacity for Arbitrage provides additional peaking capacity,
which could allow Hawaiian Electric to defer adding new
generation to deal with peak demand. This additional capacity
has a value. If this were taken into account, the value of a BESS
performing Arbitrage would be higher.
The provision of Contingency Reserve is more attractive when
the system must carry Renewable Reserves. With the Renewable
Reserve, marginal plants must be dispatched more often because
of the need to carry higher levels of reserve. The high level of

12
Reduced unit start costs were responsible for 77% of the $5.9 million in savings
resulting from the use of the 150-MW BESS for Arbitrage. Fig. 2. Total generation cost savings, renewable reserve scenario.
340 J.F. Ellison et al. / Journal of Energy Storage 15 (2018) 336–344
Fig. 3. Difference in total generation cost savings by unit.
lower fuel costs, with about 60% of fuel cost savings due to changes
in generation (less frequent dispatch of marginal units), and 40%
due to improved heat rate (greater efficiency in those units that are
dispatched).
The Schofield units operate at roughly the same cost in both
cases, but produce 8.3 GWh more power in the 150-MW BESS Case.
This is because they were dispatched primarily for reserve in the
Base Case, operating at minimum load – and minimum efficiency.
In the 150-MW BESS Case, they are used more for peak power,
operating at higher loads (and therefore higher efficiency).
Fig. 4 shows the difference in annual generation by unit
between the two cases in GWh, with positive numbers represent-
ing an increase (and negative numbers a decrease) in generation
from the Base Case. Comparing Figs. 3 and 4, it can be seen that the
same units showing the largest cost decreases operate less.
Tables 4 and 5 provide more insight into how the gas turbines
and reciprocating engines are being dispatched in each case.
In the Base Case, almost all of these units are operating at their
minimum, as can be seen from their average unit output in Table 4.
They seem to be dispatched not for power output, but for spinning
reserves.13
However, with the 150-MW BESS case, the picture is much
different. As shown in Table 5, all of the quick-start units are
dispatched at higher output levels. Schofield's average output is
near its maximum, reflecting the fact that it is now being used for
power output rather than reserves. The three combustion turbines
are used very little, as one would expect given that they are in a
quick-start reserve. This lower level of usage, combined with
higher set points (and higher efficiencies) when they are used,
accounts for much of the savings provided by the BESS providing
Contingency Reserve.
After having examined annual operational data, it is useful to
study a specific time period for additional insight. Fig. 5 illustrates
the Base Case (with Renewable Reserve) for three days, from June
1–3. The vertical bars show power output by plant, and the solid
contours at the bottom show the provision of spinning reserve. The
sum of the vertical bars is equal to the load, and the sum of the solid
13
At the time of the study, 2 MW was the estimated minimum unit capacity at
Schofield. As of May 2017, Hawaiian Electric estimates the minimum capacity per
unit at Schofield to be 4 MW. The capacity of Schofield to provide spinning reserve,
therefore, may be less than was assumed in this study.
Fig. 4. Difference in generation by unit.
contours is equal to the required spinning reserve from both the
Contingency Reserve and the Renewable Reserve.
Notable here is that from roughly 2 pm to 9 pm on each of the
three days Waiau 9 + 10 as well as Schofield are dispatched. The
decline in solar generation over this period means that conven-
tional generation must increase to fill in the void. While the on-line
steam turbine units at Kahe and Waiau could generate the needed
power, the system must dispatch additional units to obtain the
relatively high level of spinning reserves required.
Fig. 6 illustrates the 150-MW BESS Contingency Reserve Case
for the same three days. The BESS supplies 150 MW of spinning
reserve in each hour. Note that Waiau 9 + 10 and Schofield are not
dispatched over this period. Neither reserves nor power from these
units is needed. The system achieves savings by eliminating Waiau
9 + 10 starts, not operating these marginal units at partial load
(which makes them even less efficient), and by operating the Kahe
and Waiau steam turbine units at higher levels, making them more
efficient.
Fig. 7 shows the provision of reserve (which here we define as
Contingency Reserve plus Renewable Reserve) by unit for the Base
Case versus the 150-MW BESS case. Each unit providing reserve in
the Base Case provides a lower amount of reserve in the 150-MW
case. The BESS itself, not shown on this figure, provides
1,290,000 MWh of reserve in the 150-MW case (averaging out to
147 MW of reserve provision in each hour of the year).
While the 150-MW BESS providing Contingency Reserve has
been the main focus in this section, it is also instructive to
understand how the BESS operates if instead it operates to provide
Arbitrage. Fig. 8 illustrates how the 150-MW BESS providing
Arbitrage operates over the same three days examined in Figs. 5
and 6. Over this timeframe, the BESS tends to charge at around
11 pm or around noon when solar output is plentiful, and discharge
to help with the evening or morning peaks (when there is little or
no solar output).
Finally, the savings due to the BESS were highly dependent on
the fuel price assumed. Model sensitivities were performed with
the BESS providing both Contingency Reserve (with the Renewable
Reserve and Quick Start Reserve) and Arbitrage (with the
Renewable Reserve), reducing liquid fuel costs by 20% and
increasing them by 20%. With the BESS providing Contingency
Reserve, a 20% reduction in liquid fuels prices led to a 20%
reduction in the annual savings, and a 20% increase in liquid fuels
prices led to a 20% increase in the annual savings. With the BESS
providing Arbitrage, however, a 20% reduction (or increase) in
liquid fuels prices led to a 10% reduction (or increase) in annual
savings.
 J.F. Ellison et al. / Journal of Energy Storage 15 (2018) 336–344 341

Table 4
Quick-start units: base case.

Plant Number of units Avg. unit output Max unit capacity Min unit capacity Unit hrs operation Unit hrs above min
Schofield 6 2 8 2 10,205 620
Airport DSG 4 1 2 1 3353 38
Waiau 9 1 12 53 6 1085 671
Waiau 10 1 21 50 6 1600 1596
CIP 1 41 112 41 1009 4

Table 5
Quick-start units: 150-MW BESS case.

Plant Number of units Avg. unit output Max unit capacity Min unit capacity Unit hrs operation Unit hrs above min
Schofield 6 7.6 8 2 3,744 3598
Airport DSG 4 1.4 2 1 2,24 1153
Waiau 9 1 19 53 6 20 20
Waiau 10 1 25 50 6 78 78
CIP 1 66 112 41 270 178

5. Discussion
This study has examined one year of operations of the Oahu grid
in detail. Understanding this year gives insight into the different
factors impacting the value of a BESS.
In the base case, all dispatchable units provide some level of
spinning reserve (which is what Contingency Reserve and
Renewable Reserve actually are) – see Fig. 7. Reserve provision
can sometimes be the outcome of economic dispatch. In other
cases, it must be provided for by deviating from economic dispatch.
Output on efficient units can be decreased to supply headroom.
Marginal units can be brought on-line in order to provide the
needed spinning reserve. Or both of these can occur, depending on
what provides the lowest cost.
When a BESS provides Contingency Reserve, this allows the
most efficient units to operate at full capacity, making them even
more efficient. The generation from these units will increase. And it
allows the least efficient units to be utilized less, incurring lower
start and fuel costs. When a BESS provides Arbitrage, it operates as
a peaker. This allows for conventional peaking units to be started
less frequently, reducing start costs. It helps to avoid dispatching
peaking units at low levels of output, where they are least efficient.
Using a BESS to provide Contingency Reserve or Arbitrage,
therefore, allows the conventional generation fleet to be dis-
patched more efficiently. The amount of savings achieved depend
on how much more efficient the system becomes, as well as on the
cost of fuel. The more expensive fuel is, the greater the savings
achieved by a given increase in efficiency. It is important to keep in
mind that the efficiency, flexibility, and cost of generation will
change as new plants are added, old plants retired, and fuel prices
change – and that this will have an impact on the level of benefits
the BESS provides to the system.

Fig. 5. Base case with renewable reserve: June 1–3.

342 J.F. Ellison et al. / Journal of Energy Storage 15 (2018) 336–344
Fig. 6. 150-MW BESS contingency reserve case: June 1–3.
The low BESS storage capacity to power ratio (1:2) limits the
utility of using the BESS for Arbitrage, where it is used primarily as
a peaker plant. It would be instructive to test larger storage
capacities to determine whether the added benefits outweigh the
added costs. At the same time, the Arbitrage savings calculated in
this study do not take into account the capacity value of the BESS
system. Adding BESS capacity for Arbitrage provides additional
peaking capacity, which could allow Hawaiian Electric to defer
adding new generation to deal with peak demand. This additional
capacity has a value. If this were taken into account, the value of a
BESS performing Arbitrage would be higher. At the same time, the
battery undergoes more cycles when providing Arbitrage as
opposed to Contingency Reserve, shortening its longevity. This
should also be taken into account.
At the beginning of each 24-h period, the model has “perfect
foresight” – it has information as to the load and renewable
generation in each hour of that day. To the extent that Arbitrage
value depends on variable generation output, this could inflate the
value of BESS Arbitrage provision, as grid operators will not be able
to forecast variable generation with complete accuracy. To the
Fig. 7. Reserve provision by unit.
extent that Arbitrage value depends on changes in load, the
model's “perfect foresight” may distort the outcome very little, as
load can generally be predicted with high accuracy.
As variable generation on Oahu increases, it is unclear whether
providing Contingency Reserve will remain the highest-value
application the BESS can perform. Allowing the BESS to firm
variable generation is an alternative application that should be
studied. As the study team examined a year where most power is
from conventional generation, and as the model's one-hour time
resolution did not allow an investigation into how the BESS would
need to respond in order to firm variable generation, this issue
should be kept in mind for a future study. The study team notes,
however, that as the capacity of variable generation on Oahu
continues to increase, the need for both spinning reserve and
flexible generation capacity will only increase – both of which
point to an increase in value in the services provided by a BESS.
The study team also notes that the capital cost of the BESS was
not considered in this study. This study focuses on the operational
benefits of additional BESS capacity. For the reader's benefit,
however, a rough estimate of what a BESS such as this might cost is
1000 USD/kW [5].14
6. Conclusions
The Hawaiian Electric Company intends to procure grid-scale
Battery Energy Storage System (“BESS”) capacity. The purpose of
this study is determine whether providing Contingency Reserve or
Arbitrage is of more benefit to the Oahu grid, and to better
understand the relationship between BESS size and level of benefit.
This is an independent study by Sandia, and is not being used to
support the regulatory case for BESS capacity by Hawaiian Electric.
The study team created a production cost model of the Oahu
grid using data primarily from the Hawaiian Electric Company. The
14
This estimate does not take into account the expense of transporting the BESS to
Hawaii, nor does it take into account any differences in the cost of labor or land.
 J.F. Ellison et al. / Journal of Energy Storage 15 (2018) 336–344
Fig. 8. 150-MW BESS Arbitrage case: June 1–3.
proposed BESS supplied Contingency Reserve in one set of runs and
Arbitrage in another. A sensitivity was performed on Contingency
Reserve, where it was examined with and without a Quick Start
Reserve.
We examined these cases under two different scenarios: one
where a Renewable Reserve is required and another where it is not.
Our results are in Table 3.15
When providing Arbitrage, the BESS is being dispatched to act
as a peaking plant, thus reducing the number of unit starts. When
providing Contingency Reserve, the BESS is never dispatched –
instead, it remains on standby at a full state-of-charge, reducing
the amount of Contingency Reserve that must be supplied by
spinning reserve.
The most value is obtained by having the BESS serve as
Contingency Reserve, decreasing the need for spinning reserve and
thereby making conventional unit dispatch more efficient. The
BESS is more valuable when a Renewable Reserve is required
because marginal units must be dispatched more often to provide
this additional reserve – and dispatching marginal units increases
costs. The BESS supplying Contingency Reserve provides less value
when a Quick Start Reserve is required – this is because peaker
plants are required to be in reserve and cannot provide peak power
– even when the system would find it economical for them to do so.
One must consider that the BESS was limited to having a 1:2
energy storage capacity to power ratio in all of the scenarios
studied. A BESS with a larger storage capacity would be more
valuable as a peaker plant – and perhaps provide more value than
the provision of Contingency Reserve. At the same time, the
Arbitrage savings calculated here do not take into account the
capacity value of the BESS system. Adding BESS capacity for
Arbitrage provides additional peaking capacity, which could allow
Hawaiian Electric to defer adding new generation to deal with peak
demand. This additional capacity has a value. If this were taken into
account, the value of a BESS performing Arbitrage would be higher.
15
In this table, “Arb” means time-of-day shifting through Arbitrage, “CR+QS”
means Contingency Reserve together with a Quick Start Reserve, and “CR” means
Contingency Reserve without a Quick Start Reserve.
343
Over the range of fuel prices tested, the annual savings in
Table 3 were found directly proportional to the price of fuel. It is
important to note that while this study used projected 2018 LSFO
pricing of 15.18 USD/MMBtu, Oahu LSFO pricing on January 1, 2017
was about 9.30 USD/MMBtu. This suggests the possibility that 2018
LSFO pricing will be lower than what was used in this study, which
in turn would mean that the 2018 benefits from using a BESS as
Contingency Reserve would be lower than the estimates in Table 3.
At the same time, if LSFO prices go above 15.18 USD/MMBtu in
future years, benefits from a BESS providing Contingency Reserve
would likely be higher than those reported in Table 3.
7. Acknowledgements
We gratefully acknowledge the support of this study's sponsor,
the Department of Energy's Energy Storage Program. Led by Dr.
Imre Gyuk, the Energy Storage Program supports both technical
and analytical work to advance energy storage. The analytical
work, such as this study, is often done to support a utility in order
to provide benefit to both the utility and to the broader community.
We appreciate the willingness of Hawaiian Electric Company
personnel to supply us data and help us better understand their
system. Victor Bolin, Creighton Liu, Daniel Lum, Ross Sakuda,
Kenton Suzuki, and Robert Uyeunten were especially helpful.
Hawaiian Electric did not participate in the analysis presented
here. The conclusions in this report are not endorsed by the
Hawaiian Electric Company. This report is an independent study by
Sandia, and has not been/is not being used by Hawaiian Electric to
determine the size of the BESS being procured, to evaluate system
needs, or to support any regulatory filings related to adding BESS
capacity. Any errors or omissions are the responsibility of the
authors alone.
Sandia National Laboratories is a multimission laboratory
managed and operated by National Technology and Engineering
Solutions of Sandia, LLC., a wholly owned subsidiary of Honeywell
International, Inc., for the U.S. Department of Energy's National
Nuclear Security Administration under contract DE-NA-0003525.
344 J.F. Ellison et al. / Journal of Energy Storage 15 (2018) 336–344
References
[1] Hawaiian Electric Company Request for Proposal (RFP#072114-01) for 60 to
200 MW of Energy Storage for Oahu, (2014, May) . (accessed 13.07.16) https://
www.hawaiianelectric.com/clean-energy-hawaii/request-for-proposals-
energy-storage-system.
[2] EIA, Hawaii and Vermont Set High Renewable Portfolio Standard Targets,
(2015, June) . http://www.eia.gov/todayinenergy/detail.cfm?id=21852.
[3] G.M. Masters, Renewable and Efficient Electric Power Systems, John Wiley &
Sons, Inc., Hoboken, NJ, 2004.
[4] P. Denholm, E. Ela, B. Kirby, M. Milligan, The Role of Energy Storage with
Renewable Electricity Generation, Tech. Rep., National Renewable Energy
Laboratory, 2010, January. http://www.nrel.gov/docs/fy10osti/47187.pdf.
[5] A.A. Akhil, et al., DOE/EPRI Electricity Storage Handbook in Collaboration With
NRECA, Tech. Rep., Sandia National Laboratories, 2015, Feb, doi:http://dx.doi.
org/10.2172/1170618. http://www.osti.gov/scitech/servlets/purl/1170618.
[6] K. Eber, D. Corbus, Hawaii Solar Integration Study: Executive Summary, Tech.
Rep., National Renewable Energy Laboratory, 2013, June. http://www.nrel.gov/
docs/fy13osti/57215.pdf.
[7] J. Ellison, D. Bhatnagar, B. Karlson, Maui Energy Storage Study, Tech. Rep.,
Sandia National Laboratories, 2012, December, doi:http://dx.doi.org/10.2172/
1088089. http://www.osti.gov/scitech/servlets/purl/1088089.
[8] E. Kahn, Regulation by simulation: the role of production cost models in
electricity planning and pricing, Op. Res. 43 (3) (1995) 388–398.
[9] M. Carrion, J.M. Arroyo, A computationally efficient mixed-integer linear
formulation for the thermal unit commitment problem, IEEE Trans. Power
Syst. 21 (3) (2006).
[10] D. Shimogawa, Hawaiian Electric Pushes Back Major Energy Storage Plan by a
Year, Pacific Business News, 2015, June.
[11] E. Ela, et al., Evolution of operating reserve determination in wind power
integration studies, Power and Energy Society General Meeting, IEEE (2010) 1–
8.
[12] Hawaiian Electric Company, Power Supply Improvement Plans Update Report,
(2016, December) . http://selectedpucfilings.hei.com/phoenix.zhtml?
c=101675&p=irol-special-puc-filings.
[13] M. Lave, Simulation of Distributed PV in Oahu, HI, Tech. Rep., Sandia National
Laboratories, 2016, August. http://prod.sandia.gov/sand_doc/2016/167832.
pdf.
[14] M. Lave, PV Power all Oahu, (2016, August) . http://prod.sandia.gov/sand_doc/
2016/167831r.pdf.
[15] M. Lave, PV Power by Neighborhood, (2016, August) . http://prod.sandia.gov/
sand_doc/2016/167830r.pdf.