1

Economic Analysis of Grid Level Energy

Storage for the Application of Load Leveling
Robert J. Kerestes, Student Member, IEEE, Gregory F. Reed, Member, IEEE,
Adam R. Sparacino, Student Member, IEEE
1

Abstract- The benefits provided by energy storage appear to be

nearly limitless and this is especially true in regards to power
systems. Gas turbine generators are needed to meet peak load
demand but operating them is quite expensive. Energy storage
can be used to flatten the electrical load by charging the storage
when the system load is low and discharging the storage when the
system load is high. If the system load is flat enough, the fast
response time of gas turbine generators wont be needed. This
study investigates the economic feasibility of NaS battery storage
and pumped hydro energy storage used for the application of
load leveling.
Index TermsEnergy storage, load leveling, peak shaving,
battery storage, pumped hydro energy storage, sodium sulfur
(NaS) battery storage, economic dispatch, optimization.

THE

I. INTRODUCTION

last couple of decades have been a great time of

change for the power industry. There are many new and
exciting topics in the field of electric power generation and
distribution that may provide a potential solution to grid
improvement. When looking for a solution to powering the
grid one has to consider more than just the factor of
economics but also feasibility and environmental issues as
well.
Renewable generation is becoming ever more present as
government pressure to reduce greenhouse gas emissions is
being put on industry. One promising form of green energy is
the use of large grid scaled energy storage. Energy storage is
promising due to the multitude of applications that it can be
used for. The application of energy storage which appears to
yield the largest economic gain is the application of load
leveling.
This study explores the economics and feasibility of load
leveling with large grid scaled energy storage systems. There
are several types of energy storage systems that can be used
for load leveling. The advantages and disadvantages of these
energy storage systems will be discussed but this study will
focus on the use large NaS battery farms and pumped storage
facilities for simulation purposes. The charging and
discharging rates are modeled so that the power that is

This work was supported by funding from the PA DCED BFTDA and
Westinghouse. R.J. Kerestes, G.F. Reed, A.R. Sparacino, are with the
Department of Electrical & Computer Engineering and the Power & Energy
Initiative, in the Swanson School of Engineering at the University of
Pittsburgh, Pittsburgh, PA 15210 USA (e-mails: ars88@pitt.edu,
zts6@pitt.edu, reed5@pitt.edu,).

allocated is accurate to real world application.

The layout of this paper is as follows. Section II provides
background on grid level energy storage. Section III discusses
the optimization techniques that were employed in this study.
Section IV sets up and solves the specific problem explored in
this study. Section V provides analysis and results for this
study.
II. GRID LEVEL ENERGY STORAGE
A. Peak Shaving vs. Load Leveling
Peak shaving and load leveling are processes which store
electrical energy when the electrical load is low and discharge
the stored energy when the electrical load is high. In the case
of peak shaving, the energy is stored during a time in which
the system load is low and then discharged to remove only the
peaks of the load. For load leveling, the same process takes
place except the goal is to flatten the load rather than simply
remove the systems peaks. For nearly every load profile the
system demand is low during the early morning hours and is
high in the midday through the evening hours, especially
during rush hour.
Fig. 1 (a) illustrates the use of energy storage for the
application of peak shaving. During the early morning hours
from about 0000 to 0800 the load is slightly raised while the
storage is charging. The stored energy is then discharged
during the peak load hours so that the loads peaks are
removed.
There are many applications for which peak shaving can be
used for, which range from equipment protection to economic
gain. The application in which peak shaving is being used for
determines the size and the type of the storage that is needed.
This paper focuses on the application of load leveling. The
goal of load leveling is to flatten the load as possible. This
technique is very promising when it comes to the economic
benefits that it can yield. Fig. 1 (b) illustrates the
implementation of load leveling.
It should be noted that for load leveling applications, much
more energy storage is required. The charging of the energy
storage system raises the load during the early morning hours.
For load leveling, the load should be raised or lowered to the
systems average load value. It can be seen from Fig. 1 (b) that
the load with the addition of energy storage remains about
constant from hour 2100 to hour 0900. This is the time in
which the storage device is charging and hence the load is

raised due to the power demand that the storage devices

require.

(a)

(b)

Fig. 1. Explanation of peak shaving and load leveling. (a) Daily Load Profile
with the Application of Peak Shaving, and (b) a Daily Load Profile with the
Application of Load Leveling [1]

Fig. 2. Discharge Time at Rated Power vs. System Power Rating for Grid
Level Energy Storage [2]

The stored energy is then discharged during the midday

and early evening hours in an attempt to maintain a flat load
profile. It can also be seen from Fig. 1 (b) that the load has
two peaks. These two peaks present the challenge of deciding
when to discharge the storage devices. One solution to this
problem is to discharge half of the storage device during the
first peak and then discharge the other half of the storage
device during the second peak. However, unless both peaks
are exactly equal in magnitude, which is highly unlikely, the
load will remain uneven after both discharges. A solution to
this problem is to use multiple storage devices and allocate a
greater amount of stored energy for the larger peak and a
lesser amount for the smaller peak.

There are many advantages to using NaS batteries for large

grid scale applications. As far as batteries go, NaS ranks at the
very top along with a couple chemical compositions in terms
of system power capacity. It can be seen from Fig. 2 that NaS
batteries along with flow batteries and lead acid batteries can
discharge at a power rating of up to 10 MW for hours. NaS
batteries have a cycle life of up to 2500 cycles at 100% depth
of discharge and up to 5000 cycles at 90% depth of discharge
[5]. These batteries while operating daily can last as long as
15 years giving them a clear advantage over other large scaled
batteries such as lead-acid. Lastly, NaS batteries are
advantageous due to their high energy density and charging
and discharging efficiency. NaS battery storage has a round
trip ac-to-ac efficiency of 80% [1].
The disadvantages of NaS batteries are that they are limited
to being used only for grid scale applications due to their
operating temperatures which can be as high as 350o Celsius.
This paper is only focused on grid scaled applications so this
is not problematic. Also, NaS batteries, like all batteries, are
expensive.

B. Types of Energy Storage

This section discusses the different types of energy storage
and their applicability for load leveling. Fig. 2 illustrates the
different types of energy storage that can be used as they
range in system power rating and discharge time at rated
power.
For this paper, energy storage is used for the application of
load leveling which requires a storage system with a very
large system power rating and only those types of storage
devices will be discussed.
Sodium Sulfur (NaS) Battery Storage: Sodium-Sulfur (NaS)
batteries are a very promising form of large grid scaled energy
storage. These batteries have been in construction since the
1990s in Japanese businesses and as of the year 2007 could
power the equivalent of 155,000 homes [3].
The NaS cell was developed jointly by the Japanese
companies NGK Insulators Ltd. and the Tokyo Electric Power
Company (TEPCO) [1]. NaS has proven that it can be used as
a large grid scaled energy storage system when it was used to
construct the worlds largest energy storage system at
Futamata in Aomori Prefecture in May 2008 which has a
power rating of 34 MW. This energy storage system was
designed to support a 51 MW wind farm [4]. NGK has also
constructed a 1.2 MW battery that was shipped to the United
States for Distrusted Energy Storage System (DESS) use [5].

Lead-Acid Battery Storage: Lead-Acid (L/A) batteries were

the first rechargeable batteries to be invented. They were
invented by the French physicist Gaston Plant in 1859.
Today L/A batteries range in application from small
applications such as motor vehicle starting engines and
household appliances all the way up to grid level applications
on the megawatt scale.
L/A batteries have the advantage of being the most
technologically mature out of all of the rechargeable battery
chemical compositions. However they have many
disadvantages that make other forms of energy storage a better
choice for grid level applications.
One of the biggest disadvantages of L/A batteries is their
limited cycle life. For grid applications it is highly desirable to
have a storage device that has a very long cycle and calendar
life in order to maximize the economic gain that it provides.
Another disadvantage is that the worlds supply of lead is
limited. At the pace in which lead is being mined and used
today the supply of lead will be exhausted in the year 2049

[6]. Lead-Acid batteries are also very heavy and bulky

making them hard to transport and a poor choice for grid
scaled applications which require transportation of the battery.
Flow Batteries: Flow batteries, also known as redox batteries,
are electrochemical devices which can store electrical energy
with the use of electrolyte tanks. Flow batteries, like NaS
batteries are advantageous because they a large system power
capacity and can discharge at rated power for hours per
discharge. However, unlike NaS batteries, flow batteries have
the disadvantage of their technological maturity. In
comparison to other batteries such as NaS and L/A, flow
batteries are fairly new and in the early stages of their
development. Due to the fact that flow batteries require pumps
for their operation there is the possibility of mechanical failure
which is also a disadvantage. The vanadium redox battery
(VRB) is the most technologically mature out of all of the
flow type batteries. The first successful operation of an all
vanadium redox battery demonstrated in the early 1980s at the
University of South Whales [7]. Flow batteries such as zincbromine (ZnBr) are in the early demonstration and
deployment stages where as other flow batteries such as zincair (Z/air) are still in the R&D stage.
Pumped Hydro Storage: Pumped storage can store massive
amounts of energy and have a system power rating of several
hundreds of megawatts up to gigawatts. The worlds largest
pumped storage facility is the Bath County Pumped Storage
Station which has a system power rating of approximately 2.7
GW [8]. Pumped storage stations are currently the most
efficient way of storing mass amounts of energy [9].
Pumped storage works on the principal that electricity is
used to pump water up a mountain and stores until the energy
is needed i.e. when the system demand is high. The water that
is stored on top of the mountain is released down the
mountain and through a hydro-turbine generator which
generates electricity. Fig. 3 shows a diagram of the Raccoon
Mountain Pumped Storage Plant which has a system capacity
of 1.6 GW [10].

largest batteries available to equal the system power rating.

Pumped storage is also a very efficient and cost effective
means of mass energy storage.
For as many advantages that pumped storage has, it also
has a major disadvantage. The location at which pumped
storage can be used is completely dictated by geography. In
order to have a pumped storage plant there must be a lower
reservoir that can store a large amount of water. A lake is an
ideal lower reservoir because it already has all of the water
that will be needed for storage. There must be an upper
reservoir that is used to store the water pumped from the
lower reservoir. The horizontal distance between the upper
and lower reservoir should be short. This minimizes hydraulic
losses and increases the velocity of the downward flowing
water, increasing response time. The plant must be built on
and around solid rock that can support it and somewhere with
little environmental problems. The plant should also be built
close to existing generation sources so the amount of
transmission losses is kept at a minimum [11]. All of the
geographical requirements for a pumped storage plant limit
the possible locations for construction. The discharge time for
pumped storage can range anywhere from seconds to several
hours [12].
Compressed Air Energy Storage (CAES): Compressed air
energy storage (CAES) is a form of energy storage which
converts electrical energy into mechanical energy by storing
compressed air for later use. CAES technology has been
around for over 40 years. Similar to pumped storage in storage
capacity, CAES has the potential for a system power rating in
the hundreds of megawatts [12].
There are three generations of CAES. The first generation
CAES system uses natural gas which is burnt with air and sent
through a turbine generator. The second generation CAES
systems use the same process as the first except the system is
flexible to meet smart grid. Second generation CAES plants
have from 60-70% green energy [13]. Third generation CAES
plants do not use the gas turbine and likewise do not use any
natural gas. The benefit of third generation CAES plants is
that there are zero carbon emissions. The only generation that
is presently in commercial use is the first generation [2].
III. OPTIMIZATION OF THERMAL GENERATION UNITS AND
ENERGY STORAGE

Fig. 3: Diagram of Raccoon Mountain Pumped Storage Facility

Pumped storage is highly advantageous for applications

such as load leveling which require a very large system power
rating, due to their capacity. For a pumped storage plant such
as the one at Raccoon Mountain it would take hundreds of the

Optimization methods are used to find the most economical

solution to allocate power generated by thermal generation
units. This section focuses on the optimization of thermal
generation units that have to meet a system load. The most
economical solution to dispatching the power generated by the
thermal generators is found by using the economic dispatch
method of optimization. Energy storage is dispatched by using
a highly predictable system load and then allocating the stored
energy with the dynamic programming method of
optimization. These techniques are discussed in detail in this
section.

A. The Economic Dispatch Problem

The economic dispatch problem is stated as an optimization
method which uses Lagrange multipliers along with a function
for each thermal generator called the cost rate to
mathematically find the optimal solution. Each thermal
generator has its own set of constraints that must be followed
when finding the optimal solution. A given daily load must be
broken up into incremental time steps, each having its own
economic dispatch solution.
The economic dispatch problem is configured such that
there are n generators which all have a quadratic cost function.
These generators feed a single point bus which in turn feeds
the system load. The objective is used to find the operating
point for each of the n generators which yield the optimal
solution. In the case of economic dispatch, the solution
minimizes the operating cost of the sum of all n generators.
This configuration can be seen in the form of a one line
diagram represented in Fig. 4 [14].

Equation (1) represents the objective equation to be

minimized for the economic dispatch. The constraint equation
is given as follows
n

Pload Pi 0

(2)

i 1

where Pload is the system demand and Pi is the output power

for the ith generator. Equations (1) and (2) are then used to
develop the Lagrangian equation
n

( P1 , P2 ,..., Pn , ) FT ( P1 , P2 ,..., Pn ) ( Pload Pi ).

(3)

i 1

For n thermal generators, there are n respective fuel cost vs.

output power equations as follows

F1 a1 b1 P1 c1 P12
F2 a 2 b2 P2 c 2 P22

(4)

Fn a n bn Pn c n Pn2

The summation of which is equivalent to the objective

equation in (3). The partial derivative of with respect to
each output power is taken and set equal to zero to find the
optimal operating point for each generator. Yielding the
following system of equations

2c1 P1 b1
2c 2 P2 b2
1

(5)

2c n Pn bn
P1 P2 ... Pn Pload .
The solution to (5) in matrix form is given in the following
equation

Fig. 4: Economic Dispatch to a Load with n thermal generators [14].

The overall cost of this system is given as the summation

of the cost for each generator. The total cost of the system is
given in (1).
n

FT F1 F2 ... Fn Fi ( Pi )

(1)

i 1

where F1, F2,, Fn are the cost functions for each of the n
generators and P1, P2,,Pn are the respective output power
values in MW. Most input-output curves are given in terms of
input heat (Hi) with respect to that source of generation's
output power in MW. In the case in which the input-output
curves are given in terms of input heat with respect to output
power, the fuel cost must be multiplied by these equations to
get the cost functions. These functions are constrained by
minimum and maximum output power and minimum up and
down times. The quadratic equations are used with Lagrange
multipliers to find the most economical solution to the
problem.

P1 2c1
P 0
2


Pn 0
1

0
2c 2

0
1

0
0

2c n
1

1
1

1
0

b1
b
2

bn
Pload

(6)

B. Allocation of Energy Storage with Dynamic Programming

Richard Bellman had originally used dynamic programming
which would later be recognized by the IEEE as a systems
analysis and engineering topic [15]. Bellman and Dreyfus
coined the Principal of Optimality which states An optimal
policy has the property that whatever the initial state and
decision are, the remaining decisions must constitute an
optimal policy with regard to the state resulting from the first
decision. [16]. Dynamic programing starts from a feasible
solution to the problem at hand and backtracks to find lowest
cost solution.
Dynamic programing has been used for many engineering
applications and is particularly useful in electric power

engineering. With the increasing amount of renewable

generation that is tied into the grid comes the task of
allocating the power that is generated from these renewables
effectively. All energy must either be used as it is generated or
stored. Dynamic programming with the help of scientific
computing can be used to find the most cost effective method
of allocating energy through storage devices.
An example of dynamic programming for the allocation of
energy is given in Fig. 5. This example was taken from [17]
and represents two energy storage units which must allocate
their stored energy by the most cost effective methods. The
state of the storage is represented by a two bit binary number.
For storage unit R and storage unit Q there is a possibility of
three different next states possible. The next state which has
the cheapest minimum cost path will be chosen for each of the
storage units. This process continues on for the next twenty
three hours of the day. In this example one day is covered.
The initial state is when t is equal to zero and the final state is
when t is equal to twenty four i.e. when t is equal to zero for
the next day.

Fig. 6: Economic Dispatch Problem with Coal, Oil and Gas Thermal
Generators and the Integration of Energy Storage.

One of the goals of this study is to show the economic

benefit of using energy storage to replace the use of gas
turbine generators. Gas turbine generators can react to a
change in system load much faster than other fossil fuel fired
generators and so they are required. Some studies show that
gas turbine generators can actually react to a change in load as
much as ten times faster than other fossil fuel fired thermal
generators [18].
The fuel cost vs. power output curves for this study are
given in the following set of equations from [14]
2
Fcoal ( Pcoal ) 561 7.92 Pcoal 0.001562 Pcoal

Foil ( Poil ) 78 7.97 Poil 0.00482 Poil2

(7)

Fgas ( Pgas ) 545.45 10.909 Pgas 0.0045 P .

2
gas

The load that was used for this study was taken from [19].
The load is broken up into a residential, a commercial and a
street lighting sector. This load is shown in Fig. 7.
Daily Demand Curve
2000
Lighting Load
Residential Load
Commercial Load

1800

Fig. 5: Allocation of Energy Storage by Dynamic Programming [17]

1600
1400

IV. PROBLEM SET-UP AND SOLUTION

1200
Power(MW)

This section employs the techniques discussed in Section

IV to set up the economic dispatch problem both in the case
with grid level energy storage for load leveling and without it.
The problem is set up such that there are three generation
units of different types.
The single point bus that was discussed in Section IV is fed
by a coal fired generator, and oil fired generator and a gas
turbine generator. Where the coal and oil generators can
change their operating point by 150 MW in an hourly time
step and the gas turbine generator can change its operating
point by 450 MW in an hourly time step. Energy storage is
also incorporated into the system. Fig. 4 is adapted to fit the
problem and is illustrated in Fig. 6.

1000
800
600
400
200
0

10

15

20

Hour

Fig. 7: System Daily Demand Curve

Energy storage was used to level the load in Fig. 7 enough

so that there was no change in load greater than 300 MW (the
combined possible change from the coal and oil fired

generators). Two of the five forms of energy storage from

Section II of this report were used. NaS battery storage was
used for a battery type of storage and pumped hydro was used
for a non-battery type of storage. Both NaS battery storage
and pumped hydro energy storage were modeled the same in
terms of charging and discharging. This can be done because
there is quite a bit of freedom in modeling pumped hydro
energy storage. Therefore, the NaS charging and discharging
model will work for the pumped hydro model as well. Fig. 8
illustrates the charging and discharging profiles for a 1 MW
NaS battery. These models were scaled to 10 MW batteries
and connected in parallel to create five 120 MW storage
facilities.

Fig. 8: Charging and Discharging Profile for NaS Battery Storage

For this study the batteries were only discharged to a 90%

depth of discharge rather than a full discharge. The NaS
batteries can either be discharged once per cycle at 90% or
twice per cycle at 45%. The discharging profile for 45%
depth of discharge is illustrated in Fig. 9. This was calculated
by interpolating the data in Fig. 8.
Pdis

2.9 hrs
1 p.u.

case the batteries had 120 MW delivered to each unit for six
hours and then 57.6 MW for one hour to charge the batteries
to their full MWh capacity. The charging profile for the NaS
case was mimicked for the pumped storage case.
The energy that was stored by the storage units was then
allocated using the dynamic programming method of
optimization that was discussed in Section IV. The goal of the
optimization was to minimize the average change in system
demand per hour while maintaining a maximum change less
than or equal to 300 MW

Fig. 10: Discharging Profiles for 90% and 45% Depth of Discharge.

The first and primary goal of the allocation of energy

storage was to reduce the maximum change in power for any
particular time step a value of 300 MW or less. Once the
maximum change in power was reduced to a value less than or
equal to 300 MW the focus of the optimization then shifts to
reducing the average change in power. The flatter that the load
profile is the cheaper the cost of the thermal generating units
is due to the quadratic nature of their fuel cost curves. The less
the speed of the generators has to be changed to meet changes
in the load profile also leads to lower equipment costs. [5]
600

400

3.655 hrs
Fig. 9: Discharging Profile for 45% Depth of Discharge

The charging and discharging profiles in Fig. 8 are

integrated in hourly increments to discretize the power
delivered in the charging and discharging processes. The
discrete power discharge profiles for both 90% depth of
discharge and 45% depth of discharge are illustrated in Fig.
10.
It can be seen from Fig. 7 that the system load is at its
lowest during the early morning hours. These hours are the
opportune time for the charging of storage. For this study the
storage was charged between the hours of 0000 and 1000. The
amount that was charged and the times that the charging took
place were decided using dynamic programming. For the NaS

Energy Storage (MW)

200

-200

-400

-600
0

10

15

20

Hour

Fig. 11: Allocation of Energy Storage

Fig. 11 illustrates the allocation of energy storage where

negative MW represents the charging of the energy storage
system and positive MW represents the discharging of the

energy storage system. The effect that the energy storage had
on the system load is illustrated in Fig. 12. The case in which
the energy storage is discharged twice daily at 45% depth of
discharge achieved the desired result of decreasing the
maximum change in system load to less than or equal to 300
MW. The 90% depth of discharge case actually provided a
flatter load profile overall, however, the maximum change in
the system load was greater than 300 MW and therefore could
not be used to remove the gas turbine generators from the
system.
Daily Demand Curve With No Storage
Power(MW)

2000

Lighting Load
Residential Load
Commercial Load

1000

Fig. 13 (Bottom) illustrates the operating point for the coal

and oil generation units for each hour to supply the daily
system load for the case in which there is energy storage used
for load leveling. This case of course does not use gas turbine
generation. Because gas turbine generators are very
expensive, the elimination of gas turbine generators results in
a much lesser cost of generation for a given day.
It should also be noted that in Fig. 13 (Bottom) the coal and
oil generators have a much flatter operating point profile. This
will also lead to reduced maintenance costs because the
turbine speed will have to change much less.
Power Generation without Energy Storage
1400

10

15
20
Hour
Daily Demand Curve Using Storage With 45% Depth of Discharge

1000

Power(MW)

Power in MW

Power(MW)

2000

1000

10

15
20
Hour
Daily Demand Curve Using Storage With 90% Depth of Discharge

800

600

2000

400

1000

200

Gas Turbine Generation

Coal Fired Generation
Oil Fired Generation

1200

10

15

20

Hour

(8)

15

20

Coal Fired Generation

Oil Fired Generation

1200

1000

800

600

400

1
7.92
0
Pcoal 0.003124

0
0
.
0096
1
7
.
97
oil

1
1
0 Pload [k ] Pstor [k ]

10

Power Generation with Energy Storage

Power in MW

1400

Equation (6) was used to implement the economic dispatch

problem for the case in which there is no energy storage and
the case in which two discharges of 45% depth of discharge
were used. This can be seen in (8) and (9) respectively.
1 7.92
0
0
Pcoal 0.003124
P
1 7.97
0
0.0096
0
oil
Pgas
0
0
0.0090 1 10.909

1
1
1
0 Pload [ k ]

Hour

Fig. 12: Top: System Load without Energy Storage; Middle: System Load
with 45% Depth of Discharge Storage; Bottom: System Load with 90% Depth
of Discharge Storage

200

(9)

V. ANALYSIS AND RESULTS

The economic dispatch was calculated hourly for each hour
of the load in Fig. 12 both for the case in which there was no
energy storage implemented and for the case in which the was
energy storage implemented at two discharges of 45% depth.
Fig. 13 (Top) illustrates the operating point for the coal, oil
and gas generation units for each hour to supply the daily
system load for the case in which there is no energy storage
used for load leveling. It should be noted that the drastic
changes in the load are covered by the gas turbine generator.

10

15

20

Hour

Fig. 13: Top: Hourly Generator Operating Points for Daily Load Profile
without Load Leveling; Bottom: : Hourly Generator Operating Points for
Daily Load Profile with Load Leveling

Due to the fact that the curves used in (7) were from 1984
some calculations had to be made to update them to the year
2011. Data dating back to 1991 was used to calculate the
inflation cost for coal, oil and gas. The results of the
calculations are given in (10), (11) and (12) as follows [20].
The units are in dollars per dollar.

Coal Price
Inflation:
Oil Price
Inflation:
Gas Price
Inflation:

$130 per metric ton in 2011

3.250
$40 per metric ton in 1991
$107 per barrel in 2011
3.567
$30 per barrel in 1991
$160 per cubic meter in 2011
3.720
$43 per cubic meter in 1991

(10)
(11)
(12)

The economic dispatch for the daily load was calculated

using the inflation rates in (10)-(12). TABLE I illustrates the
daily cost of generation for the case without energy storage
and the case with energy storage used for load leveling. The
difference in the cost of generation for one day with energy
storage and without energy storage is equal to $61,911. Over
the course of one year this leads to a $22,962,515 savings.
TABLE I
COST OF GENERATION WITH AND WITHOUT ENERGY STORAGE

Coal
Generation
Oil
Generation
Gas
Generation
Total
Generation

Without Storage

With Storage

$685,547

$770,127

$246,838

$261,360

$162,120

$0

$1,094,398

$1,031,487

The cost of both the NaS battery storage facility and the
pumped hydro storage facility had to be calculated to check if
these energy storage systems are a feasible solution for load
leveling. The calculation for the NaS battery storage is given
in (13) and the pumped hydro storage facility is given in (14).
The prices used in this calculation were taken from [5,10].
$1500 1000KW 120 MW

6 batteries $900,000,000
1 KW
1 MW 1battery

(13)

$300000000 $3.47in 2011

600MW $390,000,000
1600 MW
$1 in 1979

(14)

The question is then a question of economic feasibility. The

maximum possible calendar life of NaS battery storage units is
15 years [5]. At a $22,962,515 per year savings with the use
of energy storage for load leveling the NaS batteries only
yield a $344,437,725 savings. This means that current NaS
battery technology is not economically feasible due to the
$900,000,000 cost. On the other hand the pumped storage
facility has an installation cost of $390 million and therefore
yields approximately a 17 year return on investment.
Currently the Raccoon Mountain pumped storage facility
has been in operation for 33 years [10]. If a pumped storage
facility had been in operation as long as the Raccoon
Mountain pumped storage facility it would have not only
returned on investment but it would have nearly saved
$390,000,000 in the cost of generation.

VI. CONCLUSION
The potential economic benefit that grid level energy
storage can provide is quite clear from this study. This can be
seen by comparing the economic dispatch for the case in
which there is no energy storage with the economic dispatch
for the case where there is energy storage. There is the
potential to save millions and even billions of dollars.
When used for load leveling, pumped storage shares nearly
all of the benefits that batteries have without the
comparatively large initial capital investment. Using pumped
storage as a form of energy storage is relatively cheap and
pumped storage has a massive maximum power capacity that
can range all the way up to the multi-gigawatt level. It
provides a response time that can be as fast as seconds. With
new developments in pumped storage such as the variable
speed pumped storage unit the charging and discharging rate
can be controlled. The money that can be saved by using
pumped storage for the application of load leveling can yield a
return on investment in a relatively short amount of time. The
main downfall that pumped storage has is geographic
limitations. However, there is a great deal of locations where
new construction is underway.
Currently pumped storage is the only form of large grid
level energy storage that is economically feasible for the
application of load leveling. The locations that can support the
installation of a pumped storage facility should be maximized
in order to provide the greatest economic gain possible.
Although at the present moment battery technology is not
where it needs to be to provide this economic gain, the
economic benefit that it can potentially have is clear as long as
its efficiency and cycle life are increased and its cost is
decreased. New battery types and chemical compositions
should also be explored so the best possible battery option is
found for the use of grid level applications.
VII. ACKNOWLEDGEMENT
The electric power and energy research group for grid
infrastructure (EPERGI) would like to extend a special thanks
to the Commonwealth of PA Ben Franklin Technology
Development Authority (BFTDA) and Westinghouse for their
support of this work.
VIII. REFERENCES
[1]
[2]
[3]
[4]
[5]

[6]
[7]

NGK Insulators LTD, NaS Batteries, http://www.ngk.co.jp/english/

products/power/nas index.html
EPRI, Energy Storage Technology Options, A White Paper on
Applications, Cost and Benefits, Final Report, December 2010
P. Davidson, New Battery Packs Powerful Punch, USA Today,
[online], Available: http:// www.usatoday.com/money/industries/energy/
2007-07-04-sodium-battery_N.htm
NGK Insulators LTD, Video on NaS Battery Storage, [online],
Available: http://www.ngk. co.jp/english/products/power/nas/movie/
movie_nas_english.html
A. Nourai, Installation of the First Distributed Energy Storage System
(DESS) at American Electric Power, A Study for the DOE Energy
Storage Systems Program, Sandia Report, SAND2007-3580,Unlimited
Release, Printed June 2007
New Scientist: The global science and technology weekly, How Long
Will It Last, May 26, 2007, ISN 0262-4079
M. Skyllas-Kazacos, An Historical Overview of the Vanadium Redox
Flow Battery Development at the University of New South Wales,

[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]

[19]
[20]
[21]

Australia, University of New South Wales, [Online], Available:

http://www.ceic.unsw.edu.au/centers/vrb/overview.htm
Dominion, Bath County Pumped Storage Station, [Online], Available:
http://www.dom.com/about/stations/hydro/bath-county-pumped-storagestation.jsp
Janning, J.; Schwery, A.; , "Next generation variable speed pumpstorage power stations," Power Electronics and Applications, 2009. EPE
'09. 13th European Conference on , vol., no., pp.1-10, 8-10 Sept. 2009
Tennessee Valley Authority, The Mountain Top Marvel, [online],
Available: http://www.tva.gov/heritage/mountaintop/index.htm
Cochran, A.M.; Isles, D.E.; Pope, I.T.; , "Development of pumped
storage in a power system," Electrical Engineers, Proceedings of the
Institution of , vol.126, no.5, pp.433-438, May 1979
ClimateTechWiki, A clear technology platform, Energy Storage:
Compressed
Air
(CAES),
[Online],
Available:
http://climatetechwiki.org/technology/jiqweb-caes
M. Nakhamkin, CAES Bottom Cycle Concept, 2010, [online],
Available: http:\\www. espcinc.com
Wood, Allen J., Wollenberg Bruce F., Power Operation, Generation,
and Control, Wiley, 1984, pp. 29 - 90
Dynamic Programming, October 2002, http://en.wikipedia.org/wiki/
Dynamic _programming
R. Bellman, Dynamic Programming 1957, republished by Dover press,
2003
G. L. Kusic, Allocation of Power in Medium Voltage DC Distribution
Circuits, University of Pittsburgh
O'Connor, R.; Reed, G.; Zhi-Hong Mao; Jones, A.K.; , "Improving
renewable resource utilization through integrated generation
management," Power and Energy Society General Meeting, 2010 IEEE ,
vol., no., pp.1-6, 25-29 July 2010
Gnen, Turan, Electric Power Distribution System Engineering,
Taylor and Francis Group, LLC, 2008, pp. 37-38
Index
Mundi,
Commodity
Prices,
[online],
Available:
http://www.indexmundi.com/commodities/
E. Ingram, Balancing the Grid with Pumped Storage, [online],
Available:
http://www.renewableenergyworld.com/rea/news/article/
2011/08/balancing-the-grid-with-pumped-storage

IX. BIOGRAPHIES
Robert Kerestes (M2011) was born in Pittsburgh Pennsylvania. He served
in the United States Navy from 1998 to 2002 as an interior communication
electrician on board the U.S.S. Constellation. From 2002 to 2006 Robert
served in the United States Naval reserves as an electrician in the construction
battalion. He went on to attend the Community College of Allegheny County
from 2005 to 2007 where he later transferred to the University of Pittsburgh.
In 2010 Robert graduated Magna Cum Laude from the University of
Pittsburgh with a bachelors degree in electrical engineering and with a
concentration in electrical power systems. Robert will graduate with a
masters degree in electrical engineering in the fall of 2011 with a
concentration in electrical power systems and intentions of pursuing a PhD in
the same field. Robert was awarded the first ever Siemens T&D Service
Solutions Graduate Power and Energy Scholarship in September of 2011.
Gregory F. Reed (M1985) received his B.S. in Electrical Engineering from
Gannon University, Erie PA; his M. Eng. in Electric Power Engineering from
Rensselaer Polytechnic Institute, Troy NY; and his Ph.D. in Electrical
Engineering from the University of Pittsburgh, Pittsburgh PA. He is the
director of the Power and Energy Initiative in the Swanson School of
Engineering, associate director of the Center for Energy, and associate
professor in the Electrical and Computer Engineering Department at the
University of Pittsburgh. He has 25 years of electric power industry
experience, including utility, manufacturing, and consulting at Consolidated
Edison Co. of NY, Mitsubishi Electric, and KEMA Inc.
Adam R. Sparacino (M2009) was born in New Castle, Pennsylvania.
Currently, he is finishing his Masters degree in electrical engineering from
the University of Pittsburgh with a concentration in electric power
engineering. He received his B.S. in Electrical Engineering from the
University of Pittsburgh, Pittsburgh PA. Adams research interests include

energy storage, renewable integration, power electronics and control

technologies.