This article has been accepted for inclusion in a future issue of this journal.

Content is final as presented, with the exception of pagination.

IEEE TRANSACTIONS ON SMART GRID 1

Smart Scheduling and Cost-Benefit Analysis

of Grid-Enabled Electric Vehicles for
Wind Power Integration
Mahmoud Ghofrani, Amirsaman Arabali, Mehdi Etezadi-Amoli, and Mohammed Sami Fadali

Abstract—This paper proposes a stochastic framework to mit- Penalty cost for wind power imbalances ($).
igate the effects of uncertainty and enhance the predictability of
wind power using the vehicle-to-grid (V2G) capabilities of elec- Penalty cost for wind power over-production
tric vehicles (EVs). An Auto Regressive Moving Average (ARMA) ($).
wind speed model forecasts the wind power output. Using Fuzzy
C-Means (FCM) clustering, EVs are grouped into 6 fleets of similar Penalty cost for wind power under-production
daily driving patterns. A Genetic Algorithm (GA) is used in combi-
nation with a Monte Carlo simulation (MCS) to optimize charging ($).
and discharging of the EVs. The optimization scheme minimizes Discount rate (%).
the sum of the penalty cost associated with wind power imbalances
and V2G expenses associated with purchased energy, battery wear Self-discharge rate for the EV battery (%).
and capital costs. The proposed method provides a collaborative
strategy between the wind participants and EV owners to increase DoD Depth of discharge for the EV battery (%).
their revenues and incentives. A cost-benefit analysis assesses the
economic feasibility of V2G services for wind power integration. Capacity of the EV battery (kWh).
The coordinated charging/discharging scheme optimally utilizes
the V2G capacities of EVs and compensates for power imbalances Discharging energy of the EV battery (kWh).
due to random variations of wind power.
Revenue for the discharged V2G energy ($).
Index Terms—ARMA, battery storage, driving patterns, elec-
tric vehicles, genetic algorithm, Monte Carlo simulation, optimal PDF for the simulated random variable.
charging/discharging, stochastic modeling, vehicle-to-grid, wind
integration. Generation of the gas-fired generator (kWh).
Functional relation between the input and
NOMENCLATURE output.

Capital cost for the EV battery ($). HR Turbine heat rate for gas-fired generator
(Btu/kWh).
Capital cost for V2G service ($).
Number of time steps.
Cost of equipment degradation due to V2G
service ($/kWh). Battery lifetime in cycles (cycle).

Cost per energy unit discharged through V2G Battery lifetime in energy throughput (kWh).
($/kWh). Battery lifetime in years (year).
Cost of generation for gas-fired generator ($). Number of charging/discharging cycles per
day.
Annualized capital cost for V2G service
($/year). Order of the AR model.
Cost of natural gas ($/MBtu). Charging power of the EV battery (kW).
Cost of operation and maintenance Discharging power of the EV battery (kW).
($/kW-year).
Net power generated by wind (kW).
Cost for V2G service ($).
Rated wind power (kW).
Cost of purchased energy for V2G service
Actual wind power (kW).
($/kWh).
Scheduled wind power (kW).

Manuscript received August 16, 2013; revised January 15, 2014; accepted Charging and discharging power rating (kW).
May 25, 2014. Paper no. .
The authors are with the Electrical Engineering Departments at Uni-
Production tax credit associated with V2G
versity of Washington, Bothell and University of Nevada, Reno. (e-mail: ($).
[email protected]; [email protected]; [email protected]; [email protected]).
Order of the MA model.
Digital Object Identifier 10.1109/TSG.2014.2328976

1949-3053 © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

2 IEEE TRANSACTIONS ON SMART GRID

Total revenue for V2G service ($). [5] and [6] used fast-response units such as natural gas turbines
or hydro generators to provide the reserve capacity and the fast
State of charge of the EV battery (kWh). ramping requirements for high wind penetrations. This solution
Simulated wind speed (m/h). increases the operation cost and decreases the efficiency of the
power system [7]. In addition, utilization of thermal units for re-
Minimum and maximum battery capacity newable integration increases air pollution which works against
(kWh). the greenhouse gas reduction agenda targeted by the RPS. More-
Tax credit ($/kWh). over, using hydro generators to compensate for the difference be-
tween predicted wind power and the actual output can result in
Wind speed (m/h). greater maintenance requirements.
, , Cut-in, cut-out and rated wind speed (m/h). Energy storage technologies provide the required flexibility for
the net wind power delivery to more accurately follow the sched-
Sample for the scenario and simulation ules cleared in the market [8]–[11]. This is accomplished by dis-
instant. charging the stored energy in excess of the cleared schedule to
avoid the penalties associated with producing less energy than
stochastic input variable. forecast. Utilization of storage systems to reduce forecast errors
stochastic output variable. and enhance wind power predictability was investigated in [8]
and [9] where the storage sizing is either arbitrary or based on
Roundtrip efficiency of the EV battery (%). simple trial and error, with no attempt at optimizing their param-
eter values. Several references proposed probabilistic methods
Charging efficiency of the EV battery (%).
for energy storage sizing based on wind power forecast uncer-
Discharging efficiency of the EV battery (%). tainty [10], [11]. A dynamic daily sizing of the necessary storage
is presented in [10] to minimize the cost of power imbalances
Moving average coefficient of the ARMA
based on the degree of risk taken by the power producer. How-
model.
ever, the simplifications made in storage modeling and the market
, Mean and standard deviation for the wind information may not appropriately reflect the real world condi-
speed. tions. [11] uses a persistence approach to model the wind fore-
cast scenarios based on which an energy storage system is de-
Market clearing price ($). signed to reduce forecast uncertainty. Although the persistence
Autoregressive coefficient of the ARMA forecasts give an appropriate estimate of the storage power, they
model. have limitations for the storage capacity calculation. These lim-
itations may result in an underestimate even for the worst-case
persistence forecasts where the storage capacities are less than
I. INTRODUCTION those obtained based on real world forecasts. Deployment of en-
ergy storage for renewable integration depends on its economic

T HE widespread adoption and insertion of plug-in hybrid

electric vehicles (PHEVs) into the power grid provides an
opportunity to use their energy storage for renewable integration.
feasibility, particularly in a market system where the installation
decision is based on a cost-benefit analysis that takes into account
market opportunities such as energy arbitrage. None of these ref-
This is particularly true for grid-enabled vehicles (GEVs) where erences provides a justification for the storage cost.
their V2G capabilities offer a wide range of benefits and rev- V2G can potentially provide an appropriate solution for wind
enues such as load leveling or other forms of arbitrage, hedge power integration as an alternative to other storage technologies.
against forecast uncertainty, energy and ancillary services. These The capital cost of adding V2G capability to a PHEV is less than
services include frequency regulation, voltage control, spinning the capital cost of energy storage systems (such as battery bank)
and non-spinning reserve [1]. V2G enables bidirectional power which makes V2G more economical for grid support [12], [13].
flow between a vehicle’s energy storage and the electric grid. In addition, EVs as distributed storage provide technical bene-
EVs adoption is facilitated by this technology as targets of 20 fits such as better wind power utilization, transmission and dis-
million EVs by the end of 2020 have been set by the global elec- tribution upgrade deferral, and significantly reduce the adverse
tric vehicle initiative [2]. However, transportation electrification impacts of charging/discharging on the system when compared
may pose some adverse impacts on the electric grid if EVs are to centralized storage technologies [14]. The societal benefits of
not scheduled intelligently. An energy management system is re- V2G deployment include reduced electric grid costs, inexpensive
quired to coordinate charging/discharging of EV fleets and de- storage and support for renewable integration [15].
velop a sustainable integrated electricity and transportation in- Several studies investigated V2G applications for wind power
frastructure. The need for a smart scheduling strategy becomes integration [16]–[20]. A multi-paradigm modeling approach was
obvious when considering the high penetration of intermittent used in [16] to evaluate the V2G service as operating reserves re-
renewable resources mandated by renewable portfolio standards quired to deal with wind forecasting inaccuracies in California.
(RPS). Wind power in excess of demand was used to charge EVs in a
Wind energy resources require more complicated strategies as specified pattern more appropriate for a flat-rate electricity tariff
compared to PV generation [3]. Greater fluctuations of wind gen- environment. This is not applicable for power systems where the
eration impose a typical forecast error in the final output power wind power does not exceed the demand and/or the electricity
in the range of 30%–50% [4]. Penalties associated with wind rate is based on market clearing prices. In addition, these as-
power imbalances increase wind integration costs. Several solu- sumptions avoid the efficient utilization of V2G capacities for
tions have been proposed to comply with the wind power sched- spinning reserves and result in an underestimation of the V2G
ules cleared in the market and hedge against forecast uncertainty. services. [17] investigated the V2G integration in the Western
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

GHOFRANI et al.: SMART SCHEDULING AND COST-BENEFIT ANALYSIS OF GRID-ENABLED ELECTRIC VEHICLES 3

Danish power system. The V2G regulation service was evalu- TABLE I
ated for typical days with high and low wind production. This CHARACTERISTICS OF THE EVS AND THEIR BATTERIES [25],[26]
results in an overestimation as the analysis is based on the de-
terministic modeling of the worst-case scenarios without consid-
ering the wind power uncertainties. In addition, the aggregated
battery storage model considered for the V2G system neglects the
driving pattern and availability of the individual EVs for battery
aggregation. [18]–[20] evaluated the benefits of using V2G to
integrate intermittent renewable generation in distribution grids.
Wind power was the focus of these studies as it typically exhibits
greater fluctuations compared to other uncertain factors and dom-
inates the renewable generating technology.
The practices proposed in the literature are concerned with
the operation of power systems with EVs. This requires the use erful short-term forecasting method that effectively characterizes
of available generation technologies as well as V2G capacities random variations of wind generation [22], [23]:
to provide regulation and ancillary services [16]–[20]. The need
for operating reserves can be significantly reduced if the wind (1)
power participants comply with their cleared schedules. This re-
quires developing an appropriate strategy for the wind producers
where is the white noise with zero mean, zero autocorrelation,
to compensate for the wind forecast uncertainties and submit
and a variance of . Historical hourly wind speed data are used
more accurate generation bids while providing economic incen-
to calculate the autoregressive and moving average coefficients of
tives for EV owners to participate in V2G services. [3] and [21]
the corresponding ARMA model. The time series ( ) is shifted
conclude that V2G has the potential to level out the fluctuations
using the mean and scaled using the standard deviation of the
of wind power without the need for other storage and generation
historical data to obtain simulated wind speed at time [22]:
technologies. The V2G potential, if fully realized to improve the
dispatchability of wind power, increases revenues by minimizing (2)
the penalty cost for wind power under-/over-production and V2G
expenses. The increased revenue provides an incentive for EV The wind power output is calculated using the wind power curve
owners and increases their participation in V2G services. The as [24]:
utilization of wind power can be significantly increased which
provides an additional revenue stream in the form of renewable
production tax credits (PTC). To the best of our knowledge, the (3)
collaboration of wind producers and EV owners to enhance the
predictability of wind power and increase the revenues and in-
centives for these participants has not been addressed in the liter- B. Battery Model for EV
ature. Hence, there is a need for a stochastic framework to opti- The state of charge (SOC) of the EV battery system at each
mize the use of EVs to mitigate the effect of random wind power time is calculated as [25]:
variations.
This paper proposes a GA-based MCS to optimally coordinate (4.a)
the EV fleets for their V2G services and minimize the penalty
cost associated with wind power imbalances. The wind power in for charging when
excess of the scheduled output is optimally used to charge V2G
capacities and then reduce power imbalances during wind power (4.b)
under-production periods. The paper uses a stochastic model and
more realistic assumptions than earlier studies to better evaluate and for discharging when .
the potential of V2G services. In addition, unlike earlier studies The EV cannot be charged and discharged simultaneously.
which emphasized the operation perspective, the main objective The SOC of the EV battery at instant satisfies the minimum,
here is to provide a V2G strategy that realizes the full potential of , and maximum, , battery capacity requirements if:
V2G services and increases the revenues and incentives for both
wind participants and EV owners. (5)
Section II explains the ARMA model for wind speed and the
where is determined based on the depth-of-discharge (DoD)
battery and economic models for the EVs. It also describes the
of the battery:
clustering of EVs into fleets of similar driving patterns. Section II
also presents the GA optimization and MCS. Simulation results (6)
are given in Section III together with a cost-benefit analysis that
incorporates the economic characteristics of EV batteries and the For each hour, the charging/discharging power of the EV bat-
costs of V2G service. Conclusions are presented in Section IV. tery must not exceed the charging/discharging power rating as
follows:
II. METHODOLOGY
(7.a)
A. ARMA Model for Wind Speed (7.b)
An ARMA ( -order AR, -order MA) time series model Table I provides the specifications of the EVs used for this paper.
is used to forecast the wind power output because this is a pow-
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

4 IEEE TRANSACTIONS ON SMART GRID

TABLE II A discount rate of 10% and production tax credit of 2.3

COST PARAMETERS OF THE EV BATTERY are considered for our study. Table II provides the cost parameters
for the EV.

D. Driving Patterns for EV

Driving and plug-in patterns of private cars are fairly pre-
dictable as many drivers have a fixed schedule of working hours
and leisure time activities. This is particularly true for the EVs
that are parked for most hours of the day. Statistical data on
transportation behavior of such EVs are used to obtain driving
C. Economic Model for EV patterns [30], [31]. Each pattern represents the distance driven
for each hour of the daily driving [32]. FCM clustering is used
The economic value of V2G is the difference between the as- to merge EVs with similar driving patterns into fleets [33]. The
sociated revenue and costs. The revenue of selling V2G energy number of clusters is determined by the elbow method [34].
is the product of price and discharging energy [15]: FCM, in combination with the elbow method, categorizes the
EVs into 6 fleets of similar 24-hour driving patterns.
(8)

The V2G application for hedging against wind forecast uncer- E. Monte Carlo Simulation
tainties enhances the predictability and utilization of wind power.
Therefore, an additional revenue stream is provided as the renew- MCS is an accurate technique to estimate probability density
able electricity PTC [27]: functions (PDFs) using historical data. This paper uses MCS of
the uncertainties of the system including the wind power vari-
(9) ability and stochastic nature of driving patterns [35]. The ARMA
model of wind generation and driving patterns of the EV fleets
The total revenue for the V2G service is the summation of the are used to simulate the system states. The functional relation be-
revenues for the discharged energy and PTC: tween the output and input random vectors is given by:

(10) (18)

The cost of V2G consists of purchased energy, battery wear MCS generates a sample ( ) for each input random vari-
and capital costs. The general formula for the cost calculation on able ( ) using the wind ARMA model and EV driving patterns.
an annual basis is [15]: calculates the output for each sample. This process is re-
peated for a specified number of scenarios ( ) that ensures con-
(11) vergence over a simulation period. The sample outputs are used
to simulate the system behavior.
The cost of energy discharged thorough V2G ( ) includes the
purchased energy and equipment degradation (wear) costs [15]:
F. GA Optimization
(12)
GA-based optimization is used to solve the non-smooth
problem of optimally charging/discharging the EV fleets to
where is the roundtrip efficiency given by:
minimize the cost associated with wind forecast uncertainties.
(13) The conditional equations of (4) and (21) make our objective
function discontinuous and the optimization problem cannot be
The degradation cost of the EV battery ( ) is: solved by traditional derivative-based optimization methods. In
addition, the random uncertainties of renewable generation and
(14) driving patterns as well as the Monte Carlo randomness in the
search procedure requires a stochastic optimization method for
the solution of our problem. GA is used in this paper because
where is the capital cost of battery including replacement labor
it considers all feasible states and not just the expected values
cost and is the battery lifetime in energy throughput given by:
of the states, as do other non-heuristic stochastic optimization
(15) methods such as stochastic linear programming.
The GA is initialized using a set of randomly selected solu-
is the lifecycle which is determined by the depth-of-discharge tions. The solutions are ranked based on their fitness function
of the battery [28]. The battery lifetime is calculated by: values. Recombination, crossover, selection, and mutation are
then used to evolve the population. The process halts upon the
(16) satisfaction of a termination criterion. This population-based
evolutionary algorithm is appropriate to solve the coordinated
V2G charging/discharging problem where classical optimization
The annualized capital cost for V2G service is calculated by uni-
methods cannot handle the non-convexity and discontinuity of
formly distributing the total capital cost over the life time as:
the problem. In addition, the uncertainties of the system make
our optimization problem a stochastic program that cannot be
(17)
solved by traditional derivative-based optimization methods.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

GHOFRANI et al.: SMART SCHEDULING AND COST-BENEFIT ANALYSIS OF GRID-ENABLED ELECTRIC VEHICLES 5

G. GA-Based MCS Optimization

Fig. 1 shows the flowchart for the developed method. The vec-
tors of input and output random variables are:
(19)
(20)

Historical wind speed and transportation data are used to char-

acterize the uncertainties of wind generation and transportation
behavior by the wind ARMA model and driving patterns of EV
fleets. Each scenario includes 7 sets of 24-hour samples for the
scheduled wind power and driving patterns of six EV fleets. The
V2G capacity for each EV fleet is the difference between the
total battery capacity of the fleet and the capacity required to
make the daily trip. The V2G capacity for each hour is updated
based on the driven distance which releases the associated ca-
pacity for V2G usage. The calculated V2G capacity is available
for the hours when the EV is parked, i.e., the driving distance
is 0. The wind power is scheduled based on the simulated wind
speed (2) and wind power curve (3). For each hour, the sched-
uled wind power sample is compared with the actual wind power
to determine the charge/discharge status for the available V2G
capacities. The wind power in excess of the scheduled output is
used to optimally charge the available V2G capacities, which are
then released to compensate the wind power deficit when the ac-
tual power is less than the scheduled output.
The wind power producer incurs energy imbalance charges if
energy production differs from advance schedules. The energy
deviations are settled at real-time MCPs [36]. The penalty cost
for wind power under-/over-production is calculated based on
real-time prices [37], [38]:

(21)

The Federal Energy Regulatory Commission (FERC) requires

that intermittent resources imbalances are settled at 90% of decre-
mental costs and 110% of incremental costs as [36], [38]:
(22)
(23)

The GA is used to coordinate the V2G charging/discharging of

the EV fleets. The solution minimizes the sum of the penalty cost
for the wind power mismatch and V2G expenses. The objective
function of the proposed optimization method is:

(24)
Fig. 1. Flow chart for the proposed method.

The GA is randomly initialized and then evolved through sev- The charging and discharging power of EVs, the number of
eral operators to optimize the solution and minimize the cost for charging/discharging cycles per day and depth-of-discharge of
a specified number of scenarios ( ) over a simulation period. the battery are the decision variables for the optimization. The
The children are generated by crossover and mutation operators. battery lifetime is a dependent variable calculated by (16).
Selection and recombination are then used to produce the next The GA is terminated once the process is repeated for a suffi-
generation. For each sample, a large penalty coefficient ( ) is en- cient number of generations (150). The number is determined by
forced on the violated constraints to assign a higher cost to an trial and error to ensure convergence.
infeasible solution [39]. The combination is given by:
III. CASE STUDIES
(25)
The proposed method is used to optimally charge and dis-
where , , and are the objective function, charge the V2G capacity of 484 EVs for minimizing the sum
constraints, dependent and decision variables, respectively. of the imbalance cost of 10 MW wind power capacity and V2G
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

6 IEEE TRANSACTIONS ON SMART GRID

Fig. 2. Driving patterns of EV fleets.

Fig. 5. Wind power imbalances for the 24-hour simulation period.

Fig. 3. Predicted wind power for the 24-hour simulation period.

Fig. 6. Penalty cost associated with wind power imbalances for the 24-hour
simulation period.

TABLE III
SIMULATION RESULTS FOR CASES I AND II

scheduled output. This offers potential revenues for the wind pro-
ducer and eliminates the penalty cost associated with wind power
over-production. The stored energy is then discharged through
V2G to reduce wind power under-production. The V2G trade
is settled at real-time MCPs. This agreement will provide the
Fig. 4. SOC distribution of the EV fleets for the 24-hour simulation period. EV owners with economic incentives to participate in V2G ser-
vices. The wind speed and MCP are from historical hourly data
costs of the EVs. A maximum of 20 000 cycles is considered for from the Mesonet and California ISO [41], [42]. Two cases are
the battery lifetime [28]. It is assumed that the EVs have the fast studied: Case I, where the wind power is scheduled based on the
charging capabilities to charge for driving application at a power forecast with no scheme to reduce the power imbalances; and
rating of 50 kW or more in 15–30 minutes [26]. This capability Case II, where the proposed method is used to coordinate the
makes the residual capacity of the EV available for V2G services EVs charging/discharging and minimize the wind forecast uncer-
whenever parked and not in use for transportation. This provides tainty.
the required flexibility to reduce the wind forecast error to the Simulation results are summarized in Figs. 2–6 and
minimum possible extent using the V2G capacity of the EVs. An Tables III–IV.
aggregator coordinates the EV fleets for optimal charging and dis- Fig. 2 shows the driving patterns of the EV fleets. The asso-
charging [40]. A bilateral contract between the wind supplier and ciated numbers of EVs are 105, 50, 83, 90, 81, and 75 for fleets
EV aggregator is used to purchase the wind power in excess of the 1–6.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

GHOFRANI et al.: SMART SCHEDULING AND COST-BENEFIT ANALYSIS OF GRID-ENABLED ELECTRIC VEHICLES 7

TABLE IV TABLE V
SIMULATION RESULTS FOR COST-BENEFIT ANALYSIS OF THE V2G SERVICE SIMULATION RESULTS FOR ECONOMIC COMPARISON OF DIFFERENT
FOR THE 24-HOUR SIMULATION PERIOD TECHNOLOGIES USED FOR THE PROPOSED APPLICATION (24-HOUR
SIMULATION PERIOD)

ation and power capacities of 21.84 MWh and 1.281 MW for the
Fig. 3 shows the forecast of hourly wind power with the fore- gas-fired alternative. A capacity of 6.157 MWh and power rating
cast ranges for 3000 scenarios generated by MCS. These ranges of 1.539 MW are required for the battery storage alternative to
provide the extents that the predicted wind power deviates from achieve the same penalty cost of $137.32 for wind power imbal-
the actual output. The probability that the predicted wind power is ances. The cost of the battery storage system is composed of the
within a forecast range gives the chance of generated wind power costs for energy (500 $/kWh), power (175$/kW), and replace-
coverage for all scenarios. For example, 20% of the generated ment (500$/kWh) [43]. The cost of gas-fired conventional gener-
samples are captured within the first forecast range. The chance ators is the combination of the investment and generation costs.
of sample coverage increases to 100% for the last forecast range The investment cost is 695 $/kW and the generation cost includes
with probability one. the maintenance and fuel costs using (26). The heat rate (HR),
Fig. 4 shows the SOC of the EV fleets for the scenarios gen- cost of natural gas ( ), and operation and maintenance cost
erated by MCS. For each fleet, the percentage of EVs within a ( ) of the gas-fired generators are 8000 Btu/kWh, 5 $/MBtu
certain SOC interval is calculated. For example, 30% of EVs of and 5 $/kW-year, respectively [44]. Table V provides the cost of
Fleet 1 have SOC values less than 0.32 MW for hour 14. This different technologies used to enhance the predictability of wind
value increases to 0.55 MW, 0.80 MW, 1.12 MW, 1.36 MW, 1.69 power. The cost of V2G technology is the sum of the costs for
MW, and 2.52 MW for 40%, 50%, 60%, 70%, 80%, and 100% of fleets 1–6. The results demonstrate the improved efficiency of
EVs of Fleet 1, respectively. This figure provides the SOC distri- using EVs over other alternatives.
butions of the EV fleets for the 24-hour simulation period.
Fig. 5 shows the absolute values of the average wind power im- (26)
balances for Cases I and II. The power imbalance is the difference
between the scheduled wind power and the actual output. The av- IV. CONCLUSION
erage power imbalance is calculated for the generated scenarios
over the 24-hour simulation period. The optimal EV charging and This paper proposes a stochastic framework to optimally
discharging of Case II significantly reduces wind power imbal- charge and discharge EVs to mitigate the effects of wind power
ances when compared to Case I (no V2G service). uncertainties. The proposed algorithm offers a V2G service to
Fig. 6 shows the penalty costs and MCPs for the simulation minimize the sum of wind power imbalance cost and V2G ex-
period. The penalty cost is the average of the cost of wind power penses. Wind power is scheduled for the simulation period based
under-/over-production for the generated scenarios. The cost is on an ARMA model of the wind speed. FCM is used to cluster
significantly reduced for the optimal scheduling of V2G services the EVs into fleets with similar daily driving patterns. MCS
of Case II. simulates the system behavior by generating samples of system
Total wind power imbalances and the associated penalty cost states using the wind ARMA model and EVs driving patterns.
for Cases I and II are summarized in Table III. Wind power The proposed method is used to evaluate the usefulness of V2G
utilization is also calculated for the 24-hour simulation period, in mitigating the effects of random wind power variations. The
which demonstrates the ability of the proposed method to en- economic characteristics of automotive battery technologies and
hance the performance of wind integration. costs of V2G service are incorporated into a cost-benefit analysis
Table IV gives the revenue and cost of V2G service as well which evaluates the economic justification of the proposed V2G
as the DoD of different fleets for improving wind power pre- application. Simulation results demonstrate that the developed
dictability. Simulation results for the cost-benefit analysis show algorithm enhances wind power utilization and reduces the
that the V2G revenues justify the cost of V2G services for the pro- penalty cost for wind power under-/over-production. This offers
posed application. The proposed strategy provides the EV owners potential revenues for the wind producer. Simulation results of
with economic incentives to trade their V2G power for the pro- the cost-benefit analysis show that V2G revenues justify the cost
posed service. Minimizing the cost function of (24) requires re- of V2G services for the proposed application. This provides EV
ducing the V2G cost to the minimum possible extent because it is owners with economic incentives and makes V2G economically
the most significant term in the cost function. This is equivalent viable. The simulation results also show the economic merits
to maximizing the battery lifetime ( ). The maximum lifetime of the V2G technology compared to the gas-fired and battery
of 20 000 cycles corresponds to the DoD of 65% based on the storage alternatives.
cycle life testing of [28]. This is evident from the optimization
results of Table IV where a DoD of 65% is obtained for all fleets REFERENCES
as one would intuitively predict. [1] E. Sortomme and M. A. El-Sharkawi, “Optimal charging strategies for
To evaluate the economics of different technologies for wind unidirectional vehicle-to-grid,” IEEE Trans. Smart Grid, vol. 2, no. 1,
pp. 131–138, Mar. 2011.
power integration, battery storage systems and gas-fired genera- [2] Fact Sheet: Electric-Vehicles-Initiative, Clean Energy Ministerial, U.S.
tors are used as alternatives for the V2G technology to decrease Department of Energy, Jul. 2010 [Online]. Available: http://www.en-
the wind power imbalance cost to $137.32. This results in gener- ergy.gov/news/documents/Electric-Vehicles-Initiative-Fact-Sheet.pdf
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

8 IEEE TRANSACTIONS ON SMART GRID

[3] W. Kempton and J. Tomic, “Vehicle-to-grid power implementation: [26] M. Yilmaz and P. T. Krein, “Review of battery charger topologies,
From stabilizing the grid to supporting large-scale renewable energy,” charging power levels, and infrastructure for plug-in electric and hybrid
J. Power Sources, vol. 144, pp. 280–294, 2005. vehicles,” IEEE Trans. Power Electron., vol. 28, no. 5, pp. 2151–2169,
[4] J. Garcia-Gonzalez, R. M. R. de la Muela, L. M. Santos, and A. M. Gon- May 2013.
zalez, “Stochastic joint optimization of wind generation and pumped- [27] T. Das and J. D. McCalley, Compressed Air Energy Storage. Ames,
storage units in an electricity market,” IEEE Trans. Power Syst., vol. IA, USA: Iowa State Univ., 2012.
23, no. 2, pp. 460–468, May 2008. [28] Energy storage for commercial and off-highway vehicles [On-
[5] S. Kamalinia, M. Shahidehpour, and A. Khodaei, “Security-con- line]. Available: http://www.a123systems.com/solutions-transporta-
strained expansion planning of fast-response units for wind integra- tion-commercial-vehicles.htm
tion,” Elect. Power Syst. Res., vol. 81, pp. 107–116, 2011. [29] B. Canis, “Battery manufacturing for hybrid and electric vehicles:
[6] T. Lee, “Optimal spinning reserve for a wind-thermal power system Policy issues,” CRS, Report for Congress, R41709, 2013, .
using EIPSO,” IEEE Trans. Power Syst., vol. 22, no. 4, pp. 1612–1621, [30] Model Center [Online]. Available: http://www.modelcenter.transport.
Nov. 2007. dtu.dk/
[7] P. Denholm, E. Ela, B. Kirby, and M. Milligan, “The role of energy [31] K. Capion, “Optimized charging of electric drive vehicles in a market
storage with renewable electricity generation,” Tech. Rep. NREL/TP- environment,” M.Sc., Tech. Univ. Denmark, RisØ National Laboratory
6A2-47187, 2010. for Sustainable Energy, Lyngby, Denmark, 2009.
[8] M. Korpaas, A. T. Holen, and R. Hildrum, “Operation and sizing of [32] M. Pantos, “Exploitation of electric-drive vehicles in electricity mar-
energy storage for wind power plants in a market system,” Int. J. Elect. kets,” IEEE Trans. Power Syst., vol. 27, no. 2, pp. 682–694, May 2012.
Power Energy Syst., vol. 25, no. 8, pp. 599–606, Oct. 2003. [33] J. C. Bezdek, Pattern Recognition with Fuzzy Objective Function Al-
[9] G. N. Bathurst and G. Strbac, “Value of combining energy storage and gorithms. New York: Plenum Press, 1981.
wind in short-term energy and balancing markets,” Elect. Power Syst. [34] D. J. Ketchen and C. L. Shook, “The application of cluster analysis
Res., vol. 67, no. 1, pp. 1–8, Oct. 2003. in strategic management research: An analysis and critique,” Strategic
[10] P. Pinson, G. Papaefthymiou, B. KlÖckl, and J. Verboomen, “Dynamic Management J., vol. 17, pp. 441–458, 1996.
sizing of energy storage for hedging wind power forecast uncertainty,” [35] R. Y. Rubinstein, Simulation and the Monte Carlo Method. New
in Proc. IEEE PES Gen. Meet., Calgary, AB, Canada, 2009. York: Wiley, 1981.
[11] H. Bludszuweit and J. A. Dominguez-Navarro, “A probabilistic [36] J. Rogers and K. Porter, “Wind power and electricity markets, a living
method for energy storage sizing based on wind power forecast summary of markets and market rules for wind energy and capacity
uncertainty,” IEEE Trans. Power Syst., vol. 26, no. 3, pp. 1651–1658, in North America” Utility Wind Integration Group, Reston, VA, USA,
Aug. 2011. 2011.
[12] C. Quinn, D. Zimmerle, and T. H. Bradley, “An evaluation of state-of- [37] A. Ingram, “Storage options and sizing for utility scale integration of
charge limitations and auction signal energy content on plug-in hy- wind energy plants,” in Proc. Int. Solar Energy Conf. 2005, Orlando,
brid electric vehicle, vehicle-to-grid reliability, and economics,” IEEE FL, USA, Aug. 2005, pp. 843–851.
Trans. Smart Grid, vol. 3, no. 1, pp. 483–491, Mar. 2012. [38] T. K. A. Brekken, A. Yokochi, A. V. Jouanne, Z. Z. Yen, H. M. Hapke,
[13] T. Mosher, “Economic valuation of energy storage coupled with pho- and D. A. Halamay, “Optimal energy storage sizing and control for
tovoltaic: Current technologies and future projections,” M.Sc., Mass. wind power applications,” IEEE Trans. Sustain. Energy, vol. 2, no. 1,
Inst. Technol., Dept. Aeronautics and Astronautics, , 2010. pp. 69–77, Jan. 2011.
[14] M. Ghofrani, A. Arabali, M. Etezadi-Amoli, and M. S. Fadali, “A [39] T. Weise, Global Optimization Algorithms—Theory and Application
framework for optimal placement of energy storage units within a 2009 [Online]. Available: http://www.it-weise.de/projects/book.pdf,
power system with high wind penetration,” IEEE Trans. Sustain. 2nd
Energy, vol. 4, no. 2, pp. 434–442, Apr. 2013. [40] T. K. Kristoffersen, K. Capion, and P. Meibom, “Optimal charging of
[15] W. Kempton and J. Tomic, “Vehicle-to-grid power fundamentals: Cal- electric drive vehicles in a market environment,” Int. J. Appl. Energy,
culating capacity and net revenue,” Elsevier J. Power Sources, vol. 144, vol. 88, pp. 1940–1948, Dec. 2010.
pp. 268–279, 2005. [41] ISU Soil Moisture Network [Online]. Available: http://mesonet.agron.
[16] B. S. Hodge, S. Huang, A. Shukla, J. F. Pekny, and G. V. Reklaitis, iastate.edu/agclimate/info.phtml
“The effects of vehicle-to-grid systems on wind power integration in [42] California ISO [Online]. Available: http://www.caiso.com/about/
California,” in Proc. 20th Eur. Symp. Comput. Aided Process Eng. (ES- Pages/default.aspx
CAPE20). [43] S. M. Schoenung and W. V. Hassenzahl, “Long- vs. short-term energy
[17] J. R. Pillai and B. Bak-Jensen, “Integration of vehicle-to-grid in the storage technologies analysis: A life-cycle cost study” A study for the
Western Danish power system,” IEEE Trans. Sustain. Energy, vol. 2, DOE energy storage systems program, Sandia National Laboratories,
no. 1, pp. 12–19, Jan. 2011. SAND2003-2783, 2003.
[18] J. A. P. Lopes, P. M. R. Almeida, and F. J. Soares, “Using vehicle-to- [44] H. E. Shaalam, “Generation of electric power,” in Handbook of Electric
grid to maximize the integration of intermittent renewable energy re- Power Calculations. New York: McGraw-Hill, ch. 8.
sources in islanded electric grids,” in Proc. Int. Conf. Clean Electr.
Power Renewable Energy Resources Impact, Capri, Italy, Jun. 2009,
pp. 290–295. Mahmoud Ghofrani (S’11) is currently an assistant professor at the School
[19] T. Wu, Q. Yang, Z. Bao, and W. Yan, “Coordinated energy dispatching of Science, Technology, Engineering and Mathematics (STEM), University of
in microgrid with wind power generation and plug-in electric vehicles,” Washington, Bothell, WA, USA. His research interests include power systems
IEEE Trans. Smart Grid, vol. 4, no. 3, pp. 1453–1463, Sep. 2013. operation and planning, renewable energy and smart grid systems, electricity
[20] J. A. P. Lopes, F. J. Soares, and M. R. Almeida, “Integration of electric market, and electric vehicles.
vehicles in the electric power system,” Proc. IEEE, vol. 99, no. 1, pp.
168–183, Jan. 2011.
[21] W. Kempton and A. Dhanju, “Electric vehicles with V2G storage for Amirsaman Arabali (S’11) received the Ph.D. degree in electrical engineering
large-scale wind power,” Windtech Int., vol. 1, no. 2, Mar. 2006. from University of Nevada, Reno, NV, USA in 2014. His research interests in-
[22] R. Karki, D. Dhungana, and R. Billinton, “An appropriate wind model clude power systems operation and planning, energy management, and renew-
for wind integrated power systems reliability evaluation considering able energy systems.
wind speed correlations,” Appl. Sci., vol. 3, pp. 107–121, Feb. 2013.
[23] W. Wangdee and R. Billinton, “Considering load carrying capability
and wind speed correlation of WECS in generation adequacy assess- Mehdi Etezadi-Amoli (LSM’11) is a Professor of Electrical Engineering at
ment,” IEEE Trans. Energy Convers., vol. 21, no. 3, pp. 734–741, Sep. the University of Nevada, Reno, NV, USA. His interests include power system
2006. protection, large-scale systems, fuzzy control, and renewable energy systems.
[24] S. Roy, “Market constrained optimal planning for wind energy con- Dr. Etezadi is a registered professional engineer in Nevada.
version systems over multiple installation sites,” IEEE Trans. Energy
Convers., vol. 17, no. 1, pp. 124–129, Mar. 2002.
[25] J. Tant, F. Geth, D. Six, P. Tant, and J. Driesen, “Multiobjective bat- Mohammed Sami Fadali (SM’91) is a Professor of Electrical Engineering at
tery storage to improve PV integration is residential distribution grids,” the University of Nevada, Reno, NV, USA. His research interests include robust
IEEE Trans. Sustain. Energy, vol. 4, no. 1, Jan. 2013. control and fault detection.