1

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.

I. INTRODUCTION

HE 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

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

T

2

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]

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.

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].

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

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.

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

3

[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 zinc-bromine (ZnBr) are in the early demonstration and deployment stages where as other flow batteries such as zinc-air (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].

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

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

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.

4

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].

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

iiinT PFFFFF

121 )(... (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.

Equation (1) represents the objective equation to be minimized for the economic dispatch. The constraint equation is given as follows

01

n

iiload PP (2)

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

).(),...,,(),,...,,(1

2121

n

iiloadnTn PPPPPFPPP (3)

For n thermal generators, there are n respective fuel cost vs. output power equations as follows

2

2222222

2111111

nnnnnn PcPbaF

PcPbaFPcPbaF

(4)

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

....2

22

21

1

222

111

loadn

nnn

PPPPbPc

bPcbPc

(5)

The solution to (5) in matrix form is given in the following equation

load

nnn

Pb

bb

c

cc

P

PP

2

11

2

1

2

1

01111200

10201002

(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

5

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

IV. PROBLEM SET-UP AND SOLUTION 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.

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]

2001562.092.7561)( coalcoalcoalcoal PPPF 200482.097.778)( oiloiloiloil PPPF (7)

.0045.0909.1045.545)( 2gasgasgasgas PPPF 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.

0 5 10 15 200

200

400

600

800

1000

1200

1400

1600

1800

2000Daily Demand Curve

Hour

Pow

er(M

W)

Lighting LoadResidential LoadCommercial Load

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

6

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.

2.9 hrs

3.655 hrs

1 p.u.

Pdis

t

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

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]

0 5 10 15 20

-600

-400

-200

0

200

400

600

Hour

Ene

rgy

Sto

rage

(MW

)

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

7

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.

0 5 10 15 200

1000

2000Daily Demand Curve With No Storage

Hour

Pow

er(M

W)

0 5 10 15 200

1000

2000Daily Demand Curve Using Storage With 45% Depth of Discharge

Hour

Pow

er(M

W)

0 5 10 15 200

1000

2000Daily Demand Curve Using Storage With 90% Depth of Discharge

Hour

Pow

er(M

W)

Lighting LoadResidential LoadCommercial Load

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 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.

][909.1097.792.7

011110090.000100096.00100003124.0 1

kPPPP

load

gas

oil

coal

(8)

][][97.792.7

01110096.0010003124.0 1

kPkPPP

storload

oil

coal

(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.

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.

0 5 10 15 200

200

400

600

800

1000

1200

1400Power Generation without Energy Storage

Hour

Pow

er in

MW

Gas Turbine GenerationCoal Fired GenerationOil Fired Generation

0 5 10 15 200

200

400

600

800

1000

1200

1400

Hour

Pow

er in

MW

Power Generation with Energy Storage

Coal Fired GenerationOil Fired Generation

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.

8

Coal Price Inflation: 250.3199140$

2011130$ intonmetricperintonmetricper (10)

Oil Price Inflation: 567.3199130$

2011107$ inbarrelperinbarrelper (11)

Gas Price Inflation: 720.3199143$

2011160$ inmetercubicperinmetercubicper (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

Without Storage With Storage Coal

Generation $685,547 $770,127

Oil Generation $246,838 $261,360

Gas Generation $162,120 $0

Total Generation $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].

000,000,900$61120

11000

11500$ batteries

batteryMW

MWKW

KW (13)

000,000,390$600

19791$201147.3$

1600300000000$ MW

inin

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

products/power/nas index.html [2] EPRI, Energy Storage Technology Options, A White Paper on

Applications, Cost and Benefits, Final Report, December 2010 [3] 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

[4] NGK Insulators LTD, Video on NaS Battery Storage, [online], Available: http://www.ngk. co.jp/english/products/power/nas/movie/ movie_nas_english.html

[5] 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

[6] New Scientist: The global science and technology weekly, How Long Will It Last, May 26, 2007, ISN 0262-4079

[7] M. Skyllas-Kazacos, An Historical Overview of the Vanadium Redox Flow Battery Development at the University of New South Wales,

9

Australia, University of New South Wales, [Online], Available: http://www.ceic.unsw.edu.au/centers/vrb/overview.htm

[8] Dominion, Bath County Pumped Storage Station, [Online], Available: http://www.dom.com/about/stations/hydro/bath-county-pumped-storage-station.jsp

[9] Janning, J.; Schwery, A.; , "Next generation variable speed pump-storage power stations," Power Electronics and Applications, 2009. EPE '09. 13th European Conference on , vol., no., pp.1-10, 8-10 Sept. 2009

[10] Tennessee Valley Authority, The Mountain Top Marvel, [online], Available: http://www.tva.gov/heritage/mountaintop/index.htm

[11] 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

[12] ClimateTechWiki, A clear technology platform, Energy Storage: Compressed Air (CAES), [Online], Available: http://climatetechwiki.org/technology/jiqweb-caes

[13] M. Nakhamkin, CAES Bottom Cycle Concept, 2010, [online], Available: http:\\www. espcinc.com

[14] Wood, Allen J., Wollenberg Bruce F., Power Operation, Generation, and Control, Wiley, 1984, pp. 29 - 90

[15] Dynamic Programming, October 2002, http://en.wikipedia.org/wiki/ Dynamic _programming

[16] R. Bellman, Dynamic Programming 1957, republished by Dover press, 2003

[17] G. L. Kusic, Allocation of Power in Medium Voltage DC Distribution Circuits, University of Pittsburgh

[18] 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

[19] Gnen, Turan, Electric Power Distribution System Engineering, Taylor and Francis Group, LLC, 2008, pp. 37-38

[20] Index Mundi, Commodity Prices, [online], Available: http://www.indexmundi.com/commodities/

[21] 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.