Distributed Energy Storage System Control for ...

2 downloads 0 Views 977KB Size Report
Apr 5, 2011 - Level Volt (V) Current (A) Max Power (kW). 1 ... Max AC Voltage. V ..... [23] A. Nourai, R. Sastry, T. Walker, ”A vision & strategy for deployment of.
1

Distributed Energy Storage System Control for Optimal Adoption of Electric Vehicles R. Arghandeh, Student Member, IEEE, A. Onen, Student Member, IEEE , R. P. Broadwater, Member, IEEE  Abstract—In recent years, automotive manufactures have introduced plug-in electric vehicles (PEV) into the automotive market. This market is expected to expand rapidly in the near future. Depending on the charging rates and charging times of PEVs, distribution networks can face significant stress. In this paper, distributed energy storage (DES) systems are proposed to improve the flexibility of residential distribution networks to support PEV adoption. Operating these DES systems optimally not only facilities PEV adoption but also reduces operating costs by taking advantage of the real-time energy market to profit from charging and discharging the DES units. This paper presents a Discrete Ascent Optimal Programming algorithm for optimizing energy market participation. Index Terms—Electric Vehicle, Energy Storage, Electricity Market, Optimization, Control.

I. INTRODUCTION

E

concerns, global warming and rising gas prices are creating a shift in the expectations of consumers and industries to move toward more electrical systems. The concerns are mainly raised by the transportation sector. To come up with challenges on the growth of transportation sector, automotive manufactures have introduced plug-in electric vehicles (PEV) into the market. A PEV is an electric vehicle that has a battery of 4kW or more and can be driven for at least 10 miles with stored electricity [1]. Governments across the world are advocating clean vehicle technologies and providing tax credit incentives for owners and car manufacturers. These incentives and other factors are expected to push the market share of PEVs in the U.S. to 25% by 2020 [2]. Depending on the circuit and charging time of PEVs, distribution networks can face significant stress. It is crucial to improve the flexibility of residential distribution networks to support PEV adoption. In addition, utility companies have to pay different prices for electricity during different times of the day due to the dynamic electricity market. Therefore, the PEV charging time and charging length may be adjusted with electricity prices. NVIRONMENTAL

R. Arghandeh is with with Virginia Tech – ECE Department, Blacksburg, VA 24061, USA. (e-mail: [email protected]). Ahmet Onen is with Virginia Tech – ECE Department, Blacksburg, VA 24061, USA. (e-mail: [email protected]). Robert Broadwater is Professor of Virginia Tech – ECE Department, Blacksburg, VA 24061, USA. (e-mail: [email protected]).

Some of the current research and literature in this field are focused on load control methods to prevent transformers from overloading. The author in [3] proposed a demand response program based on the time of use (TOU) rates. Authors in [4] suggest cutting the PEV charging during the peak hours. Demand response (DR) can play a crucial role in peak shedding, but there are barriers to implementing DR programs. DR depends on the customers’ participation. There are always uncertainties in the number of customers who agree to participate in DR [5]. Moreover, DR can affect life style and satisfaction of customers. The other barrier refers to low adoption of advanced metering infrastructures (AMI). Without an accurate AMI, DR programs are impossible [5]. In this paper, a distributed energy storage system (DES) is proposed to address the challenges of PEV adoption. A DES system is a fleet of batteries connected to the secondary side of distribution transformers. The DES system control not only facilities PEV adoption, but also has potential to improve the capacity, efficiency, and reliability of electricity distribution networks. Currently, the high price of batteries prevents widespread application of DES systems. But, raising interest in intermittent renewable resources and dynamic prices of electricity provide more incentives for using batteries. Also, reusing electric   vehicles’ batteries for grid support applications has potential for economic benefits. Allocating PEV batteries for second-use applications leads to environmental benefits by delaying disposing and recycling of PEV batteries [6]. The objective of this paper is to represent the DES optimal control system for relieving the stress of PEV adoption on distribution networks. It has to provide mutual benefits for utility companies and customers. The DES control system needs to know 1) when each vehicle starts to charge, 2) how much energy the PEV required, 3) How much the cost of electricity is [7]. Transportation surveys address the first two questions. Real-time electric energy markets managed by independent system operators (ISO) and regional transmission organizations (RTO) answer the third question. Economic operation of an energy storage system is a complex problem because battery capacity depends on the previous time steps. Research on optimal control of distributed generation is represented in [8] using the particle swarm optimization method. In [9], an intelligent control system is introduced to decrease the microgrid operation cost. Reference [10] introduces an optimal allocation for energy storage system

2 inside the micorgrid using a genetic algorithm. The previous papers are generally focused on energy storage applications for loss reduction and peak shedding based on heuristics and artificial intelligence techniques. In this paper, decision making directly depends on the market condition and network contingency with considering locational marginal price (LMP) in battery scheduling. The mathematical approach in this paper is the discrete ascent optimal programming (DAOP) algorithm. It has simplicity for implementation in programming code. One advantage of the DAOP approach over previous works is its assurance for convergence after a finite number of computational iterations [11]. The rest of this paper is organized as follows: section II is a discussion about PEV loads. Section III describes the distributed energy storage (DES) system together with its specifications. Section IV presents the control system architecture for DES. The optimization algorithm and mathematical representation of problem are also covered in section IV. In sections V and VI case studies and simulation results are discussed, respectively. II. PLUG-IN ELECTRIC VEHICLES Plug-in electric vehicles refer to a class of vehicles that can use both fuel and electricity, independently or dependently. PEV can be considered as a Battery Electric Vehicle supplied with an internal combustion engine to increase the vehicle driving range. Moreover, PEV is like a conventional Hybrid Electric Vehicle with an extended electric driving range and battery charging capability [4]. PEV are classified based on the all-electric range (AER) that they can be driven only on battery. PEV-X is a representation for X miles AER [7]. The PEV-20, which is used in case study, refers to a PEV with 20 miles driving range on battery.

grid. One of the factors extracted from the NHTS survey is the daily miles driven by each vehicle. The most common daily travel in the US is in the range of 25-30 miles. Fig. 1 shows the percentage of vehicles vs. daily travel distance in US [7]. The other important factor in the PEV load profile is the charging start time. Some literature assumes that owners will start to charge PEV once they arrive home [12,13]. An average vehicles home arrival time can be extracted from the NHTS survey. With the help of Monte Carlo simulations on the census data and the average of arrival time to home, a probability distribution of the PEV charging start time is achieved (see Fig. 2) [14].

Fig. 2. Probability distribution of start time for PEV charging [14].

The type of PEV has a major effect on the load profile. The 2001 NHTS user guide classified vehicles to seven vehicle type groups [15]. This paper considers vehicle types 1 to 4, because most cars in residential areas are among these types 1 to 4. Same classification is used for PEVs (see table I). To clarify required energy for each type of PEV, the total battery capacity for different types of PEV-20 are presented in Table I [7]. TABLE I PEV TYPES AND CAPACITY [7,15] Vehicle Type 1 2 3 4

Description Car Van SUV Truck

Capacity (kWh) 6.5 7.5 8.7 10.1

There are a number of charging levels in previous research [16,17]. This research uses the charging standard from the National Electric Code (NEC) [18]. TABLE II NEC CHARGING LEVELS [18]

Fig. 1. Distribution of daily trip distance in US based on NHTS survey [7].

To evaluate the impact of PEV on distribution circuits, the availability of grid connected electrical vehicles to the grid must be taken into account. The duration of PEV availability can be determined based on the National Household Travel Survey (NHTS) database. With time information about the departure and arrival times of vehicles, it is possible to estimate where, when, and how long PEV are connected to the

Level

Volt (V)

Current (A)

Max Power (kW)

1 2 3

120 240 480

15 (12) 40 -

1.44 3.3 (15 A limit) 60-150

Three levels of charging are defined in the NEC standard (see Table II). Higher charging levels lead to higher peak load and shorter charging time. This paper considers the level one charging because the 110V/15A outlets are available in all houses. Level one charging equipment is assumed [18]. Based on level one charging (Table II) and PEV capacity (Table I), typical charging scenarios for all PEV types are illustrated in Fig. 3. It is assumed that the battery is fully discharged at the start of charging time.

3

Fig. 3. Typical charging profile for different types of electric vehicles.

The PEV type 3 is considered for simulation (blue column) in this paper, because this type (SUV) is more popular than the PEV type 4 (Truck) in most residential areas [7]. III. DISTRIBUTED ENERGY STORAGE SYSTEM (DES) A. DES System Overview DES refers to the fleet of battery-based energy storage units connected to the secondary side of distribution transformers [19, 20]. DES can increase capacity, efficiency and reliability of distribution circuits due to the functions illustrated in Fig.4 [21]. Peak shaving, renewable resource integration and ancillary services have higher financial interest for utility companies [22]. Placing   energy   storage   at   customers’   sites provides flexibility for voltage control and reliability aspects. DES can deal with rapid changes in customer load. As more PEV are added to the grid, load variations are expected to increase. In addition to the advantages mentioned thus far, DES units are more scalable and flexible than large substation batteries [23]. Meeting these challenges at the customer level is in accordance with the smart grid evolution

Other business models involving third-party ownership such as   “battery   leasing”   and   “battery   as   a   service”   may result in reduced DES prices. Moreover, standardizing the DES units can effectively keep manufacturing costs down [19,21]. This paper supposes DES units that are owned and operated by a utility to optimize performance of distribution circuits. Utility ownership allows aggregation of storage units for efficient dispatching purposes. Utilities can profit from load diversity associated with DES units. The load diversity factor refers to the difference between coincident and non-coincident peak demands of different house loads that are connected to a distribution transformer [23]. B. DES Specifications The DES units are connected to the secondary side of distribution transformers to support 120/240 volt circuits. Each transformer serves a small group of houses. Reference [19] proposed a guideline to design distributed energy storage systems. In this paper, the Lithium-Ion battery is the selected technology for DES batteries. The Lithium-Ion battery has a high energy density and a long life cycle. It can be manufactured in different shapes and sizes [24]. The outage supporting time varies with battery size. This paper considers batteries capable of storing 50kWh of energy and discharging at rates up to 25kWh. Table III shows the specifications of the DES units considered here. The DES unit has to provide the rated   power   and   energy   level   after   1000   cycles   in   25   ˚C   ambient temperature. During islanding, the DES has the capability to discharge up to 2.5 times the rated capacity for 3 seconds to serve motor inrush current. The minimum isolating contactor current rating of DES should be 400 Amps. The fault tolerance of DES unit restricts operation to transformer sizes less than or equal to 100KVA [20]. TABLE III DES B ATTERY/I NVERTER SPECIFICATIONS Item Battery Type Power Capacity Min AC Voltage Max AC Voltage Normal AC Voltage Charging Rate Discharging Rate

Unit kW kWh V V V kW kW

Default Li-Ion 25 50 115 125 117 25 12.5

IV. DES CONTROL AND OPTIMIZATION APPROACH Fig. 4. DES advantages at grid and customer levels.

There are several key factors that affect the successful deployment of DES systems. The high cost of batteries, lack of standards, and infrastructure needed for controlling DER units seem to be barriers. Some strategies can help to provide a sustainable growth for PEV adoption and DES deployment. Reusing PEV batteries for stationery energy storage purposes enables battery manufacturers to increase the production and decrease the price of batteries with the expanding market.

A. DES Control Architecture Each DES unit is deployed on the secondary of a singlephase transformer. They are controlled in two levels, regional control and local control. The DES Local Controller (DLC) manages charge and discharge cycles and local voltage issues. Circuit level control is done with the DES Regional Controller (DRC) for aggregated DES units. Fig. 4 demonstrates a configuration of DES control units. Referring to Fig. 5, the regional controller is responsible for all DES units under one feeder or substation. All DRC have

4 interfaces with the Distribution Control Center (DCC). The DES resource availability and operating parameters are sent from the DRC to the DCC. The DRCs are connected to measurement devices for obtaining load profile and power flow data. The DLC is responsible for managing charging and discharging, islanding control, and meeting power and voltage constraints. It can isolate itself in response to grid anomalies. To accomplish control tasks, the DLC operates based on local set points, local measurements, and commands from the DRC. Information about available battery capacities and charge and discharge cycles are exchanged between the DLC and the DRC. DES unit can participate in the overall control scheme specified by the DRC and the DCC [20, 26].

in feeder losses due to the output of the DES unit at that hour. The sign convention used here associates a positive Pi with discharging and a negative Pi

DESout

DESout

with charging.

The actual output from DES unit will be less than reduction in battery capacity because of the losses internal to the DES unit. Eq.2 relates the actual DES output power to the change in battery capacity ( C ). The internal losses ( Pi

DESLoss

) are a

function of the DES output. (2) C  Pi DESout  Pi DESLoss ( Pi DESout ) The following constraints are imposed on the battery: 1. Battery rate constraint 2. Battery capacity constraint 3. Reserve capacity requirements at each hour 4. Transformer loading constraint at each hour The battery rate constraints can be expressed as follows: Ch DCh (3)  Pmax  PiDESout  Pmax 0  i  23 Ch

DCh

Where Pmax is the maximum charging rate and Pmax

is

the maximum discharging rate, as shown in Table III. The Ch

negative sign in front of Pmax reflects the sign convention used  here:  charging  is  “negative  output.” The battery capacity constraint can be expressed as follows: (4)  Cmin  C  Cmax 0  i  23 Where Cmin and Cmax are the minimum and maximum battery capacities imposed by limitations on the battery state of charge. The reserve capacity requirements further restrict the battery capacity, imposing the following constraints:

Fig. 5. DES system control layout.

B. DES Optimization Overview The authors of [10,25] discuss different approaches for solving energy storage problems. These papers describe algorithms to minimize feeder losses. In this paper, an algorithm is presented which attempts to maximize profit from buying and selling the energy stored in the batteries, rather than attempting to minimize feeder losses (though the value of the feeder losses is incorporated into the profit calculation). In order to determine what the battery should do at the current time, the optimization algorithm determines an estimated schedule for the next 24 hours. Conceptually, the optimization algorithm works as follows: if the current LMP is lower than expected in the near future, the battery will charge, subject to both rate (kW) and capacity (kWh) constraints; similarly, if the current LMP price is higher than expected in the near future, the battery will discharge, subject to the same constraints. The optimum profit schedule is composed of 24 hourly outputs. Thus, the objective function to be optimized is 23 (1) max  LMPi  ( Pi DESout  Pi LossRe d ) i 0

Where LMPi is the price of energy at the ith hour, Pi is the output power at that hour and Pi

LossRe d

DESout

is the reduction

 CRsrvi  C  Cmax

0  i  23

(5)

The reserve capacity, C Rsrvi , is the capacity required to serve the load for two hours starting at the ith hour, per load forecast, as described in section IV. D. below. Finally, the transformer loading constraints require that DES unit be discharged when necessary to prevent overload on the transformer: Trans Pi Load  Pi DESout  Pmax Trans

Where Pmax

the the an (6)

is the kVA rating on the transformer.

Although the reactive component of the load will add to the total kVA loading on the transformer, the DES unit supplies power to the system through an inverter, and that inverter may supply VARs up to a certain limit. Since transformer ratings are not firm ratings (a transformer may be slightly overloaded with minimal impact on reliability and transformer life), this approximation was deemed appropriate. Fig. 6 depicts the data needs in each portion of the optimization architecture. Fig. 6 is discussed further in the following sections.

5

Fig. 6. Data flow of DES system control

C. LMP Pricing The Locational Marginal Price (LMP) is the primary driver of cost and profit (the feeder losses are nearly negligible, as indicated in Table V). The LMP is the incremental cost to supply load to a given region at a given time [27]. In order to truly calculate the schedule for optimal profit, the LMP price would need to be known for many hours in advance. In reality, the real-time LMP prices change rapidly. Energy producers competing in the real-time LMP market must participate in the day-ahead market [28]. The day-ahead prices are used in the optimization algorithm to approximate the prices for the future hours in determining the most profitable battery charging and discharging schedule. D. Load Forecasting The load forecast for the next 24 hours is another important factor for the optimization algorithm. The real-time demand at each transformer with a DES unit is metered locally and provided to the DRC for use in the optimization algorithm. The 24-hour load forecast for each transformer is calculated based on load research statistics as well as the real-time demand measurement at each location [29,30]. The load research statistics provide typical daily load curves for each month or season as well as each type of day (weekday or weekend) for the type of customer supplied by the transformer. Weather data may further help to refine the load forecast. E. Calculation of Feeder Losses In maximizing the profit from charging and discharging the battery,   the   utility   takes   “credit”   for   a reduction in feeder losses when the battery discharges and the utility pays extra for increased feeder losses when the battery charges. Since losses are quadratic, the losses at any given battery output are computed here by interpolating a quadratic polynomial defined by three points relating DES output to feeder losses. Thus, the load flow computation only needs to be done three times per battery to compute the feeder losses for any combination of output powers. F. DES Optimization Algorithm The optimization algorithm attempts to find the charge/discharge quantities for each of the next 24 hours that result in the highest profit without violating the aforementioned constraints. Mathematically, the objective function has 24 independent variables—the kW outputs at each hour—and one dependent variable (the total profit). The algorithm, depicted in Fig. 7, first identifies the

schedule with minimum charging and discharging that satisfies the various constraints at each hour (if there are any infeasible constraints, the reserve capacity constraints are generally violated before the transformer loading constraint is violated). Then, starting from the initial schedule, the algorithm proceeds to add equal amounts (kWh) of charging and  discharging  at  each  iteration,  moving  toward  an  “optimal”   schedule. Initialize Constraints Make Schedule Satisfy Initial Constraints Find Optimal Ch/Dch Pair

Ch/Dch Pair Profit > Min Profit ?

No

Done

Yes Update Schedule and Constraints Fig. 7. DES optimization algorithm flowchart.

At each iteration, charging is added to only one hour and discharging is added to another hour based on price of energy at that hour and subject to the various constraints at that hour. The charging and discharging added at each iteration maintain a consistent time-integrated amount of energy stored inside the battery, which means that, due to internal losses in the batteries and inverter, the actual output and input seen external to the DES unit will be slightly different. Note that the algorithm   is   “greedy”. It does not scale back charging or discharging amounts in any hour during any iteration once a certain charging or discharging amount has been scheduled for that hour. This “greedy”   characteristic   of   discrete   ascent   optimal programming (DAOP) ensures that the algorithm will converge in a finite number of steps. The algorithm will only schedule battery charging and discharging when it can make greater than a certain minimum profit per kWh on the charge/discharge cycle (see Fig. 6). Since   charging   and   discharging   activity   reduces   the   battery’s   life, the utility will want to set the minimum profit margin based on the battery’s cost and life cycle. V. CASE STUDY The circuit used for the DES optimal optimization is presented in Fig. 8. It is a 13.2 kV circuit with residential customers and 1769 kVA annual peak load. The circuit

6 includes multiphase and unbalanced loads. The model of time varying loads is based on the customer billing information, load research data and measurements from the primary sides of distribution transformers.

profitable hours.

Fig. 10. Load profile for different PEV adoption levels. Black line is the load profile with DES.

Fig. 8. Schematic of the case study circuit.

The circuit contains 20 DES units which are 50 kWh Latium-Ion batteries. DES placements are illustrated on Fig.7 with red stars. The DES units are located beside lightly loaded transformers to achieve higher profits. This statement will be proved in the next section. Fig. 9 shows the annual load profile. The heavily loaded month is July and the lightly loaded month is October. In this paper all scenarios are simulated for the month of July.

In Fig. 11 the real-time prices are plotted together with the DES output and the internal capacity. Although the DES might have sold the energy between 2 and 4 P.M. to make a higher profit, it was forbidden from discharging so that enough capacity could be maintained to prevent the overload between 5 and 9 P.M. Of course, the LMP values are still higher between 5 and 9 P.M. than earlier in the day when the DES unit was charged, so the system is still achieving an operating profit while preventing the overload.

Fig. 11. DES capacity with LMP price and load profile. Fig. 9. Total kVA flow of the circuit

VI. SIMULATION AND RESULTS Fig. 10 depicts the effects of the transformer loading constraint on the DES charging/discharging schedule. In Fig. 10, a high PEV adoption level creates a very large load between 5 and 9 P.M on an otherwise lightly loaded transformer. The DES constraints ensure that the battery retains sufficient charge before and during this time frame to be able to prevent a transformer overload during the entire four-hour period in which the PEVs are being charged. The PEV loads therefore affect the charge/discharge schedule in two ways. First, the overload caused by the PEVs causes the reserve capacity constraints to be overridden (the DES unit discharges, even though discharging will reduce the amount of time that the DES unit could supply the load if there were an outage). Additionally, the requirement to discharge during these hours limits the profit that could otherwise be gained by discharging the battery during more

Since transformer overloading will impact DES profit potential, it is worth evaluating what levels of PEV adoption will cause such overloads. Table. IV shows the PEV adoption effects on overloading of distribution transformers spread throughout the circuit. TABLE IV PEV ADOPTION EFFECT ON D ISTRIBUTION TRANSFORMERS Adoption %

0

10

20

30

40

50

Number of Overloaded Transformers

0

3

10

14

18

22

In the case study, 20 DES units are installed on various transformers with various amounts of load. Fig. 12 shows the schedule for all 20 DES units for a weekday in July 2009. The dashed line represents the LMP price and the colorful solid lines are DES units outputs. The outputs vary based on the reserve requirements, which are a direct function of the transformer loading.

7 scheduling because the charging and discharging of batteries are based on the electricity price prediction for the next 24 hours. The realistic LMP prediction, as shown in Fig. 13, decreases the DES scheduling benefit to $107. VII. CONCLUSIONS

Fig. 12. Outputs of 20 DES units.

Table V presents the profits achieved from two different DES optimization approaches with an ideal LMP prediction assumption. The ideal prediction means that the day ahead LMP is the same as the real time LMP. The first approach includes feeder loss reduction as a part of the objective function (see Eq.1). The second approach ignores the feeder losses in optimization, but still incorporates internal battery losses. TABLE V PROFITS FROM DES OPTIMIZATION FOR PEAK D AY IN JULY 2009 WITH IDEAL LMP PREDICTION Objective Function LMP Price + Loss Function LMP Price

In general, charging of PEV can cause voltage deviations, increased losses and circuit overloading. DG components can mitigate the effects of high level of PEV adoption. However, the operating costs of these approaches are significant. This paper proposed application of distributed energy storage (DES) systems providing energy storage flexibility for optimal operation and control. Optimal operation is based on the DAOP method to find the optimal DES schedule for the next 24 hours. The results show DES application cause economic benefits, loss reductions and PEV adoption increases. The applied techniques and methods can be extended to other distributed generation resources with some modifications. ACKNOWLEDGMENT The authors would like to thank Jeremy Woyak of Electrical Distribution Design for his help in devising the optimization algorithm and implementing it in the DEW software.

Profit ($) 185 182

REFERENCES

“Position  Statement:  Plug-in Hybrid Electric Vehicle,”  IEEE USA, 2007. [Online] Available: http://www.ieeeusa.org/policy/events/phev/default.asp [2] S.  W.  Hadley,  “Impact  of  Plug-in  Hybrid  Vehicles  on  the  Electric  Grid”,   ORNL Report, Oct 2006 [Online]. Available: http://apps. ornl.gov/~pts/prod/pubs/ldoc3198_plug_in_paper_final.pdf [3] S.  Shao,  T.    Zhang,  M.    Pipattanasomporn,  S.    Rahman,  “Impact  of  TOU   rates on distribution load shapes in a smart grid with PHEV penetration,”  In proc 2010 IEEE PES Transmission and Distribution Conference and Exposition, 2010, pp: 1 – 6 [4] M. Erol-Kantarci,  J.    Sarker,  H.    Mouftah,  “Communication-based PlugIn  Hybrid  Electrical  Vehicle  load  management  in  the  smart  grid,”  In proc 2011 IEEE Symposium on Computers and Communications (ISCC), 2011 , pp: 404 - 409 [5] M.    Mallette,  G.    Venkataramanan,  “The  role  of  plug-in hybrid electric vehicles  in  demand  response  and  beyond,”  In proc 2010 IEEE PES Transmission and Distribution Conference and Exposition, 2010 , pp: 1 –7 [6] “NREL   Team   Investigates   Secondary   Uses   for   Electric   Drive   Vehicle   Batteries,”   National Renewable Energy Laboratory (NREL), News Release NR-1411, Apr 5, 2011. [Online] Available: http://www.nrel.gov/news/press/2011/959.html [7] Z.  Darabi,  M.  Ferdosi,  “Aggregated  Impact  of  Plug-in Hybrid Electric Vehicles  on  Electricity  Demand  Profile,”    IEEE Transactions on Sustainable Energy, Volume: 2 , Issue: 4 , 2011, pp: 501 - 508 [8] E. Sortomme and M. A. El-Sharkawi,  “Optimal  power  flow  for  a  system   of  microgrids  with  controllable  loads  and  battery  storage,”  in  Proc. IEEE/PES Power Syst. Conf. Expo., 2009, pp: 1–5 [9] C. Marnay, G. Venkataramanan, M. Stadler, A. S. Siddiqui, R. Firestone, and   B.   Chandran,   “Optimal technology selection and operation of commercial-building  microgrids,”  IEEE Trans. Power Syst., vol. 23, no. 3, pp. 975–982, Aug. 2008. [10]   C.   Chen;;   S.   Duan;;   T.   Cai;;   B.   Liu;;   G.   Hu,   “Optimal   Allocation   and   Economic Analysis of Energy Storage System in   Microgrids,”   Power Fig. 13. Real-Time and Predicted (Day ahead) LMP price for peak day in July Electronics, IEEE Transactions on Volume: 26 , Issue: 10, 2011 , pp: 2009 with battery capacity for each price. 2762 – 2773 [11] R. P. Broadwater, P. A. Dolloff, T. L. Herdman, R. Karamikhova, A. F. Sargent   “Minimum   loss   optimization   in   distribution   systems:   discrete   Smaller differences between real time and day ahead LMP ascent optimal   programming,”   Electric Power Systems Research, market prices creates more benefits from DES economic Volume 36, Issue 2, February 1996, Pages 113-121

Table V shows that the feeder losses have a nearly negligible impact on the total profit attainable by optimization. In the analysis here, the real-time price, not the feeder losses, is the primary driver in the economic operation of distributed energy storage systems. There are usually differences between real time LMP and day ahead LMP due to unexpected unit outages, load forecast errors and other factors like weather conditions. The mature electricity markets normally have less LMP volatility in comparison to young electricity markets [31]. Fig. 13 shows the difference between real time and day ahead LMP for the peak day of July 2009.

[1]

8 [12] S.  Shao,  M.  Pipattanasomporn,  and  S.  Rahman,  “Challenges  of  PHEV   penetration  to  the  residential  distribution  network,”  in Proc. IEEE Power Energy Society General Meeting, 2009, pp. 1–8. [13] C.  Camus,  C.  M.  Silva,  T.  L.  Farias,  and  J.  Esteves,  “Impact  of  plug-in hybrid electric vehicles in the portuguese electric utility system,”  in   Proc. IEEE Power Engineering, Energy and Electrical Drives Conf., 2009, pp. 285–290. [14] J. Jung, H. Asgeirsson, T. Basso, J. Hambrick, M. Dilek, R. Seguin, R. Broadwater,  “Evaluation  of  DER  adoption  in  the  presence  of  new  load   growth  and  energy  storage  technologies,”  In Proc 2011 IEEE Power and Energy Society General Meeting, 2011 , pp: 1 – 8. [15] 2001  National  Household  Travel  Survey  User’s  Guide,  NHTS,   Technical Report, 2004. [Online] Available: http://nhts.ornl.gov/2001/usersguide/UsersGuide.pdf [16] C.Farmer,P.Hines,J.Dowds,andS.Blumsack,“Modelingtheimpact   of   increasing  PHEV  loads  on  the  distribution  infrastructure,”  in Proc. 43rd Hawaii Int. Conf. System Sciences (HICSS), 2010, pp. 1–10. [17] K. Clement-Nyns,   E.   Haesen,   and   J.   Driesen,   “The   impact   of   charging plug-in  hybrid  electric   vehicles  on  a  residential  distribution  grid,”   IEEE Trans. Power Syst., vol. 25, no. 1, pp. 371–380, Feb. 2010. [18] K.   Morrow,   D.   Karner,   J.   Francfort,   “Plug-in Hybrid Electric Vehicle Charging   Infrastructure   Review,”   U.S. Department of Energy Vehicle Technologies Program – Advanced Vehicle Testing Activity, Final report, Idaho National Laboratory, 2008 [19] A.   Nourai   ,“Distributed   Bulk   Storage!“:   The   future   of   batteries   in   grid   applications,”   Power and Energy Society General Meeting, 2011 IEEE, 2011 , pp: 1 [20] “Functional Specification For Community Energy Storage Unit,”   American Electric Power, Rev2.1, 2009 [21] B.   P.   Roberts,   C.   Sandberg,   “The   Role   of   Energy   Storage   in   Development   of   Smart   Grids,”   Proceedings of the IEEE, Volume: 99 , Issue: 6, 2011 , pp: 1139 – 1144 [22] J. Eyer and G. Corey, BEnergy Storage for the electricity grid: Benefits and   market   potential   assessment   guide,’’   Sandia Report. SAND20100815, Feb. 2010 [23] A.  Nourai,  R.  Sastry,  T.    Walker,  ”A  vision & strategy for deployment of energy   storage   in   electric   utilities,”   Power   and   Energy   Society   General   Meeting, 2010 IEEE, 2010 , pp: 1 – 4 [24]    J.  Hickman,  “Clean  Vehicles,”  Renewable Energy Focus, Elsevier Publication, March/April 2009, pp: 33-34 [25] M.  Schulze,  P.C.  Del  Granado,  “Optimization  modeling  in  energy  storage   applied to a multi-carrier   system,”   Power and Energy Society General Meeting, 2010 IEEE, 2010 , pp: 1 – 7 [26] “Functional Specification For Community Energy Storage Control Hub,”  American Electric Power, Rev2.1, 2009 [27]   D.   Gautam,   M.   Nadarajah,   “Influence   of   Distributed   Generation   on   Congestion   and   LMP   in   Competitive   Electricity   Market,”   International Journal of Electrical and Electronics Engineering, Vol 4, Aug 2010, pp: 538-541 [28] E.   Litvinov   ,”Design   and   operation   of   the   locational   marginal   prices   based   electricity   markets,”   Generation, Transmission & Distribution, IET Volume: 4 , Issue: 2, 2010 , pp: 315 – 323 [29] R. P. Broadwater, A. Sargent, A. Yarali, H. E. Shaalan, J. Nazarko,, Estimating substation peaks from load research data Power Delivery, IEEE Transactions on, Volume: 12 , Issue: 1, 1997 , pp: 451 – 456 [30]    T.  E.  McDermott,  I.  Drezga,  R.  P.     Broadwater,  “A  heuristic   nonlinear   constructive method for distribution system   reconfiguration,”   Power Systems, IEEE Transactions on, Volume: 14 , Issue: 2, 1999 , pp: 478 – 483 [31]    M.  Gary,  L.  Gronerating,  “Managing  Risks  of  Generating  Assets  in  LMP   Markets,”  Burns  &  McDonnell  Company  Technical  papers,  Sep  2005

BIOGRAPHIES Reza Arghandeh (S’08   – IEEE) is a Ph.D. candidate in the department of electrical and computer engineering at Virginia Tech – Electrical and Computer Engineering Department. He is member of the Distribution Engineering Workstation (DEW) group. He received his B.S in EE from K.N.T University of Technology, Tehran, Iran in 2005 and M.S in Energy Systems Engineering jointly awarded by University of Manchester, Manchester, UK and K.N.T University of Technology, Tehran, Iran in 2008. His research interests are demand response, energy storage systems for microgrids and distribution networks reliability.

Ahmet Onen (S’08– IEEE) received the M.S degree in Electrical engineering from Clemson, SC in 2010. He is a PhD student in power systems at Virginia Polytechnic Institute and State University. Robert P. Broadwater (M’71– IEEE) received the B.S., M.S., and Ph.D. degrees in electrical engineering from Virginia Polytechnic Institute and State University (Virginia Tech), Blacksburg, VA, in 1971, 1974, and 1977, respectively. He is currently a Professor of electrical engineering at Virginia Tech. He develops software for analysis, design, operation, and real-time control of physical systems. His research interests are object-oriented analysis and design and computer-aided engineering.