See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/276426008

Optimal Coordination of G2V and V2G to Support Power Grids With High
Penetration of Renewable Energy

Article in IEEE Transactions on Transportation Electrification · August 2015

DOI: 10.1109/TTE.2015.2430288

CITATIONS READS
133 1,060

3 authors:

Hoang N.T Nguyen Cishen Zhang

Swinburne University of Technology Swinburne University of Technology
5 PUBLICATIONS 235 CITATIONS 229 PUBLICATIONS 4,221 CITATIONS

SEE PROFILE SEE PROFILE

Md. Apel Mahmud

Northumbria University
299 PUBLICATIONS 5,414 CITATIONS

SEE PROFILE

All content following this page was uploaded by Hoang N.T Nguyen on 18 January 2016.

The user has requested enhancement of the downloaded file.

188 IEEE TRANSACTIONS ON TRANSPORTATION ELECTRIFICATION, VOL. 1, NO. 2, AUGUST 2015

Optimal Coordination of G2V and V2G to Support

Power Grids With High Penetration
of Renewable Energy
Hoang N. T. Nguyen, Cishen Zhang, and Md. Apel Mahmud

Abstract—Electric vehicles (EVs) have recently gained much can become a reality in near future since EVs have recently
popularity as a green alternative to fossil-fuel cars and a feasible gained much popularity as a green alternative for fossil-fuel
solution to reduce air pollution in big cities. The use of EVs can cars. Based on a report from U.S. Department of Energy, there
also be extended as a demand response tool to support high pen-
etration of renewable energy (RE) sources in future smart grid. will be over 1.2 millions EVs just by the end of 2015 [6].
Based on the certainty equivalent adaptive control (CECA) prin- The coordination of the charging [grid-to-vehicle (G2V)]
ciple and a customer participation program, this paper presents and discharging (V2G) of EVs to support RE integration can
a novel control strategy using optimization technique to coordi- be determined by a centralized authority using optimization
nate not only the charging but also the discharging of EV batteries techniques. Such framework requires a bidirectional communi-
to deal with the intermittency in RE production. In addition,
customer charging requirements and schedules are incorporated cation between all participating vehicles and the control center.
into the optimization algorithm to ensure customer satisfaction, Since the rate and amount of data transferred between the two
and further improve the control performance. The merits of this parties are low, existing Internet networks can be utilized for
scheme are its simplicity, efficiency, robustness and readiness for this purpose with relatively low cost of investment.
practical applications. The effectiveness of the proposed control In [7]–[9], an optimization problem is formulated with the
algorithm is demonstrated by computer simulations of a power
system with high level of wind energy integration. objective of reducing cost and emissions in a grid with plug-in
EVs and RE sources. EVs are regarded as either load demand
Index Terms—Centralized control, electric vehicles (EVs), and generator sources. However, the decision variable for them
optimization, renewable energy (RE).
is only the number of vehicles allowed to charge or discharge
at a fixed value, which limits the flexibility of the algorithm.
I. I NTRODUCTION In [10] and [11], stochastic optimization problems are con-
sidered with the objective of minimizing the expected cost of
T HE integration of renewable energy (RE) has been
increasingly promoted throughout the world over the past
few decades. However, there is only a limited amount of RE
the grid over a control period to determine the V2G power
of EVs along with other decision variables. In these cases,
sources (around 25–30%) that existing power grids could accept the problems are formulated as stochastic dynamic mix-integer
without risks of network failures [1]. There are several solu- optimization programming, which is complicated and usually
tions which have been put forward to enhance the integration requires overwhelming computation work. Although advanced
of RE penetration to the power grid [2]. One potential approach solving methods such as particle swarm optimization have been
is employing energy storage systems of electric vehicles (EVs) employed, it still requires an excessively long time to com-
as demand response tools to mitigate the variation in RE gen- pute the final results, which may not be suitable for practical
eration [2]. In this paper, EVs are a general term to describe applications.
all categories of automobiles that have a battery pack and an In our work, a deterministic optimization problem is devel-
electric motor. The electric motor can be the sole source of driv- oped to coordinate the charging and discharging of all EVs
ing force or incorporated with an internal combustion engine in using forecast data of load demand and RE production over a
parallel (hybrid EV) or series design (plug-in hybrid EV) [3], period of time. This optimization will be computed repeatedly
[4]. Notably, EVs present a lot of potentials in supporting RE during the control period to revise the G2V and V2G output of
integration, as they can act as generator sources in vehicle-to- vehicles to deal with errors in prediction data as well as other
grid (V2G) (or discharging) mode to inject power back to the changes in the problem conditions. Being a deterministic prob-
grid when necessary [5]. Furthermore, the proliferation of EVs lem, it can be solved effectively using a normal optimization
software on standard workstation; hence, it may be suitable for
Manuscript received January 09, 2015; revised March 21, 2015; accepted practical applications.
April 19, 2015. Date of publication May 06, 2015; date of current version
August 14, 2015. Another contribution of this paper is that our optimization
The authors are with the School of Software and Electrical Engineering, problem includes with the constraints for customers’ require-
Swinburne University of Technology, Melbourne, Vic. 3122, Australia (e-mail: ments while, to the best of our knowledge, existing work
[email protected]; [email protected]; [email protected]). have not addressed thoroughly. Customers’ requirements are
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. defined as the need to charge their batteries up to a desired
Digital Object Identifier 10.1109/TTE.2015.2430288 state of charge (SOC) before their time of departure. From our
2332-7782 © 2015 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.
NGUYEN et al.: OPTIMAL COORDINATION OF G2V AND V2G TO SUPPORT POWER GRIDS 189

PEVi (t) power consumption of vehicle i at time t. In this

study, EV is considered as a load source; hence, its
value is positive during charging time and negative
when discharging.

If there is no regulation over the power consumed by EVs

(free-charging policy), there is in general a mismatch between
the supply and the demand due to the fluctuation of renewable
power generation, load consumption, and excessive charg-
ing demand of EVs during peak time. Power mismatches are
the root cause of the grid instability issues, and uneconom-
ical operation of power networks [13], [14]. In conventional
power operations, grid operators have to operate power plants
Fig. 1. Microgrid system with five main components. under optimal capacity or call expensive ancillary services from
standby generators to maintain the system balance [15]. Such
point of view, this is one of the most important objectives in consequences are partly the reasons why high level of RE
EVs optimization problem, since customers are usually more and EV penetration is generally not preferable in term of grid
willing to participate the charge-management program only if operations.
their vehicles are fully charged when needed [12].1 To further Therefore, in this study, the control objective is to coordinate
improve the performance of the scheme, a customer participa- the EV charging and discharging such that the power imbalance
tion program is proposed in our work in which EV owners need ΔP in the grid is minimized over a control period subjected
to provide information about all their future charging requests to the physical constraints, and most importantly the customer
for a day ahead. requirements.
This paper is organized as follows. Section II presents the
problem formulation, the customer participation scheme, and
the proposed optimization strategy. Section III presents the sim- B. Customer participation
ulation data and results. Conclusion is presented at the end in To further improve the performance of the optimization prob-
Section IV. lem, information provided by vehicle owners is incorporated
into the control algorithm. This information includes the charg-
ing plans at the beginning of the control period (e.g., a day),
II. O PTIMAL C ONTROL S CHEME FOR EV S such as number of charging requests per day, arrival and depar-
ture time for each charging period, and the SOC or energy level
A. Formulation of the EV Charging and Discharging Problem
at the beginning and desired final SOC.2 At this stage, this infor-
In this study, we consider a microgrid as shown in Fig. 1 mation does not need to be absolutely precise. However, when
which consists of five main components: 1) a conventional cars are already plugged into charging points physically and
power plant; 2) an RE source; 3) a non-EV load demand; 4) a send out charging requests to the control center, it is necessary
group of N EVs; and 5) a standby generator to provide ancillary for owners to provide accurate data about their departure time
service. The power imbalance in the grid at time t is described and the desired SOC [15], [16]. This information will act as a
as follows: commitment or contract between the car owners and the grid
operator.
ΔP (t) = PG (t) + PRE (t) − PD (t) − PEV (t) (1) The set of information that a vehicle owner number i, 1 ≤
i ≤ N needs to provide is as follows:
where 1) number of charging jobs required for this day: Ji ;
PG (t) power output of the conventional power plant at 2) starting SOC of each charging no. j: Es (i, j), 1 ≤ j ≤ Ji ;
time t; 3) desired SOC of each charging job: Ed (i, j), 1 ≤ j ≤ Ji ;
PRE (t) power output of the RE source at time t; 4) commence time of each charging job: Ts (i, j),
PD (t) load demand at time t; 1 ≤ j ≤ Ji ;
PEV (t) the total charging or discharging powers of all 5) departure time of each charging job: Td (i, j) 1 ≤ j ≤ Ji .
participating EVs In return for customers’ contribution, all charging requests
from vehicle owners will be satisfied if 1) they are not physi-

N
cally infeasible3 and 2) their cars remained plugged-in during
PEV (t) = PEVi (t) (2) their committed periods continuously without interruptions
i=1
2 For the SOC at the beginning of the planned charging period, customers
1 According to a survey conducted in Victoria State, Australia on load may alternatively provide information about their traveling distance or even
control response project, up to 80% of the participants would accept charge- their next charging location, which can enable the proposed algorithm to esti-
management of their EVs only if their vehicles are fully charged when needed. mate the SOC based on the average energy consumption per miles of their
Interestingly and surprisingly, nearly 40% are willing to do so if this could pro- cars.
vide a better environment outcome even though there is no financial benefit for 3 An example of a physically infeasible requirement is fully charging an

them. empty car battery within a minute.

190 IEEE TRANSACTIONS ON TRANSPORTATION ELECTRIFICATION, VOL. 1, NO. 2, AUGUST 2015

[15]. Furthermore, there might be some forms of incentives  Td (i,j)

such as lower electricity price or feed-in tariff to encourage the ηi,j PEVi (tk )dt ≥ Ed (i, j) − Es (i, j) − a,
Ts (i,j)
participation from EV owners and compensate for their battery
wears, which overall makes V2G a cost-effective scheme [17]. i ∈ N, n = [1, . . ., Ji ] (10)
 Td (i,j)
ηi,j PEVi (tk )dt ≤ Ed (i, j) − Es (i, j) + a,
Ts (i,j)
C. Optimization Control Scheme
i ∈ N, n = [1, . . ., Ji ] (11)
Assuming that we need to coordinate the charging and dis-
charging of all vehicles for a window of time Tw (e.g., a day), where ηi,j denotes the charging efficient factor of EV no. i
this window Tw is then uniformly divided into M small time in its charging job no. j.
intervals {t1 . . .tm . . .tM }. At each time step tm (1 ≤ m ≤ M ), Constraint (9) means that the charging power of the ith
a deterministic optimization problem is solved with the objec- EV is PEV i = 0 when the vehicle is not plugged-in.
tive of minimizing the total power imbalance ΔP (t) in (1) over Constraints (10) and (11) make sure that all EVs
the period [tm . . .tM ] are charged up to their desired SOC at the end of all
their charging periods. When providing the charging
min ΔP (tk ) (3) plans, customers can arrive late or early at the charging
PEV (tk ),PG (tk ) m≤k≤M
point and with their SOC different from the estimated
  12 value. To deal with this uncertainty, a fuzzy approach
M
where ΔP (tk ) = k=m ΔP (tk )
2
. is employed. Basically, for planned charging jobs, it
The decision variables are the power dispatch of the power is acceptable as long as the results satisfy the upper
plant PG (tk ) and the charging or discharging schedules of all and lower bound [−a; a] constraints. For committed
participating EVs PEVi (tk ) over the period [tm . . .tN ]. In this jobs in which an EV has actually arrived and sent out
study, the curtailment of RE source is not considered, as it is the accurate data, the constraints (10) and (11) become
 T (i,j)
our belief that RE is valuable and should not be curtailed. The a single equality constraint Tsd(i,j) ηi,j × PEV i(t)dt =
above optimization subjects to the following constraints. Ed (i, j) − Es (i, j) (with a = 0), and this job is guaran-
1) Physical constraints of the power plant teed to be completed. It should be noted that during the
actual charging period, the starting energy level is similar
PGmin ≤ PG (tk ) ≤ PGmax (4) to that of the previous step Es (i, j) = Ei (tm−1 ) and the
|ΔPG (tk )| = |PG (tk+1 ) − PG (tk )| ≤ ΔPGmax . (5) starting time is the current time step Ts (i, j) = tm .
In this optimization, we use the real measurements of RE
Constraints (4) and (5) make sure that the operation of the generation PRE (tm ) and load demand PD (tm ) for the current
power plant is within the optimum range [PGmin , PGmax ] time t = tm and the forecast data for all future periods tp (m +
and avoid sudden change in power output, respectively. 1 ≤ p ≤ M ). In practice, forecast data are normally available
2) Physical constraints of the EVs and provided by network operators for a period of time ahead.
For example, the Australian Energy Market Operator (AEMO)

N normally publishes predicted information about one-day-ahead
PEV (tk ) = PEVi (tk ) (6) load demand for different states. It should be noted that forecast
i=1 data are subjected to errors.
Although, for all vehicle i (1 ≤ i ≤ N ) at each time step
−Pmax (i, j) ≤ PEVi (tk ) ≤ Pmax (i, j)
tm , the outcomes of this optimization are control sequences
i ∈ N, t ∈ [Ts (i, j), Td (i, j)], n = [1, . . ., Ji ] (7) {PEVi (tk ), m ≤ k ≤ M }, only the results at the current time
step tm are applied by ordering all vehicles to charge or dis-
Eimin ≤ Ei (tk ) ≤ Eimax ; i ∈ N. (8) charge with their specified values PEVi (tm ). In this study, it is
assumed that when a vehicle receives a charging or discharging
Constraint (7) limits the magnitude of the charging or dis-
command from the optimization, it will regulate its power con-
charging power of the EV no. i in charging job no. j to
sumption or output according to this reference value without
the maximum power Pmax (i, j) of the charging facility.
significant time delay and tracking error [19]. Moving for-
Constraint (8) makes sure that the energy level of vehicle
ward to the next time step tm+1 , the optimization is performed
i Ei (tk ) never breaches the maximum (Eimax ) and min-
again resulting in rescheduled EV charging and discharging
imum (Eimin ) capacity to preserve the batteries’ lifespan
plans to deal with the errors in the forecast data and new
[18]. In addition, the minimum threshold Eimin makes
events in the system. New events can include new measure-
sure that there is always a sufficient amount of energy in
ments of renewable and load demand, availability of more
the car battery in case of emergency.
accurate forecast data, new vehicles begin to plug-in, revision
3) Constraints to guarantee the customers’ requirements
of charging requests, and so on. This process is repeated until
the completion of the control window.
PEVi (tk ) = 0, i ∈ N, tk ∈
/ [Ts (i, j),
Another point that is worth mentioning here is that in
Td (i, j)], n = [1, . . ., Ji ] (9) discrete time, the optimization is a typically standard linear
NGUYEN et al.: OPTIMAL COORDINATION OF G2V AND V2G TO SUPPORT POWER GRIDS 191

Fig. 2. Block diagram of the control process.

programming with root-mean-square (rms) objective function

and linear constraints, which can be solved effectively using
iteration methods [20].
The block diagram of the control process is illustrated in
Fig. 2. In addition, it should be noted that the control time
step should be reasonably small to cope with the fast changes
in the system while taking into account the delays in the
communication network.

III. S IMULATION AND R ESULT

A. Description of Simulation Model
A test system is modeled in MATLAB environment to eval-
uate the performance of the proposed control scheme for 24 h
(1440 min). In this simulation, the capacity of the power plant Fig. 3. Forecast and actual wind generation data [22].
is PG = 5000 kW with additional 500 kW of spinning reserve.
The renewable source is a wind farm representing all wind gen-
erators in Victoria State, Australia, down-scaled by a factor of
100. The average power of the wind farm is PRE = 2000 kW,
so the wind penetration level is PPRE
G
= 2000
5000 = 40%, which rep-
resents a scenario with high RE integration. The load source is
also modeled based on the electricity demand in the Victoria
region and scaled down from GW to kW unit. A set of wind
and load demand data on a random day is selected from the
historical database as the forecast value. In order to model the
uncertainty in the prediction data, 5% and 10% errors are intro-
duced into the load and wind values, respectively, as shown
in Figs. 3 and 4. These modeled errors are within reasonable
ranges according to the average errors in short-term prediction
of wind power and load demand in the existing literature [13],
[21]. In this simulation, the wind power output or load demand
[denoted as U (t)] at time t are stochastically computed using Fig. 4. Forecast and actual load demand data [23].
Gaussian distribution as follows:

U(t) = (1 + R{−1; 1} × ef )Ũ(t) + t (0, ef × Ũ(t)) (12) with mean value μ = 0 and standard deviation σ = ef × Ũ (t)
(error in forecast data).
where Ũ (t) is the forecast value, R{−1; 1} is random selection In this experiment, there are 600 participating cars in total,
between −1 or 1, and t is the a Gaussian distribution noise among which 280 are Mitsubish i-MiEV and the rest are Nissan
192 IEEE TRANSACTIONS ON TRANSPORTATION ELECTRIFICATION, VOL. 1, NO. 2, AUGUST 2015

TABLE I TABLE III

EV U SAGE DATA IN V ICTORIA , AUSTRALIA [12] E XAMPLE OF S CHEDULED S TARTING T IME AND D EPARTURE OF E ACH
C HARGING J OB F ROM V EHICLE N OS . 1 AND 2

TABLE IV
E XAMPLE OF ACTUAL S TARTING T IME AND D EPARTURE OF E ACH
C HARGING J OB F ROM V EHICLE N OS . 1 AND 2
AV, average; SD, standard deviation.

TABLE II
E XAMPLE OF E NERGY L EVEL AT THE B EGINNING AND E ND OF E ACH
C HARGING J OBS F ROM V EHICLE N OS . 1 AND 2

an example of the actual starting and departure times of two

vehicles is given
p + R[−1; 1] × 15 + t (0, 15)
Ta = T (14)
Leaf.4 Each vehicle is assigned with a charging plan including in which Ta is the actual arrival or departure time and Tp is the
number of charging requests per day and requirement informa- planned schedule.
tion for each request. According to the EV driving historical In this case study, if an EV happens to arrive earlier than
data in Victoria State, i.e., the mean and standard deviation its planned time, the control center, upon receiving the early
of EV traveling distance per day as shown in Table I, the charging request, will run two offline optimization problems
energy level Es at the beginning of each charging requirement simultaneously to determine whether to let this vehicle start
is stochastically assigned as in (13) using Gaussian distribution charging at the next time step or delay the charging until the
random number. For the energy level Ed at departure time, it is originally proposed time. In the event of late arrival, the control
assumed that when charging at their working places, customers center will create a new arrival time with a 15-min delay for it.
prefer to remain their batteries unchanged. On the other hand, Although some penalty measures may be introduced to discour-
with the charging request made at home, their preference is to age late arrivals, this issue is currently beyond the scope of this
get the vehicles fully charged. Table II shows an example of the paper.
starting and departure energy level of two vehicles in all their The last assumption in this simulation is that during the
charging requests during the day day, EVs have access to fast charging facilities with Pmax 1 =
7.68 kW at the owners’ workplace car park (equivalent to level
Es = SOCav × Q + t (0, SOCsd × Q) (13) II charging standard). Meanwhile, the maximum power is only
Pmax 2 = 2.88 kW (equivalent to level I charging standard)
in which SOCav and SOCsd are the average SOC at plug-in [24] during the charging period at home.
and the standard deviation, respectively (Table I). Q is denoted The model is simulated in 5-min resolution step for three
as the capacity of vehicle battery. different scenarios: 1) free-charging policy without the pro-
For the timing of each charging events, it is assumed that posed EV control framework; 2) with the proposed EV control
vehicle owners leave home for work in the morning and start framework and the customer participation program; and 3) the
to charge their EVs at their work places at the mean time of proposed EV control framework without the customer par-
8:30 A . M. These cars remain connected until around 5:30 P. M . ticipation program. All simulations have been conducted in
and are replugged at their homes at about 7:00 P. M. For about MATLAB environment with the aid of Convex Optimization
10% of the EV population, owners disconnect their cars to go software package and Gurobi solver on a normal office work-
for lunch at 12:00 noon and return at around 1:00 P. M. An exam- station. The fact that each optimization has been successfully
ple of the scheduled starting and departure times of two vehicles computed almost instantaneously showcases the practical merit
is given in Table III. These timings are computed in the same of the proposed control scheme.
manner as in (13) with a standard deviation of 30 min. Similar
to the prediction data for wind and load value, errors are intro-
duced to model the inaccuracy in the information provided by B. Scenario 1—Without EV Control (Free-Charging Policy)
customers as in (14) with a standard deviation of 15 min. In
In this scenario, there was no control over the charging of
other words, customers can possibly arrive at their charging
EVs and it was assumed that vehicle owners would imme-
points later or earlier than their proposed plans. In Table IV,
diately charge their cars upon arriving home with maximum
4 The proportion of Mitsubishi and Nissan cars resembles the actual propor- power Pmax 2 = 2.88 kW until their battery tanks were full.
tion of these two types in Victoria [12]. The results of this simulation are shown in Fig. 5.
NGUYEN et al.: OPTIMAL COORDINATION OF G2V AND V2G TO SUPPORT POWER GRIDS 193

Fig. 5. Power imbalance in the system without EV control (free-charging Fig. 7. Power imbalance in the system without and with EV control scheme.
policy).

Fig. 8. Charging or discharging power of Vehicle no.1 during the day.

Fig. 6. Total power output of all participating EVs.

imbalance in the system was suppressed significantly to zero
most of the time as in Fig. 7.
During the day time (from 6 A . M . to 6 P. M .), there were three It is noted that there were short periods at the beginning, the
occasions when the main power plant failed to accommodate middle, and the end of the simulation when there were slight
the fluctuation of the wind energy and load demand. Notably, power mismatches in the grid (Fig. 7). This is due to the fact that
during early night time (from minute 690 till 950 in simula- part of the EV pool was still on the road and not yet plugged-in
tion time—roughly 5:30 P. M . to 9:50 P. M . in real time), there during these times. Consequently, there was not enough storage
was a peak load demand exerted on the grid when all vehi- capacity from EVs to minimize the power imbalance.
cles were charging at the same time. From minute 1220 (around As far as customer satisfaction is concerned, all charging
2:30 A . M . next day) to minute 1440 (6 A . M . next day), the net- requests from all participating vehicles were fulfilled success-
work was out of balance again due to sudden increase in wind fully. In other words, all EV batteries were charged up to their
generation and drops in load demand. Unfortunately, now that desired energy level before their departure time. For illustra-
all vehicles had been fully charged earlier, none was available tion, in the case of vehicle 1, its charging request information
to absorb this excessive electricity production. is shown in Tables II and IV. Although the control center
ordered vehicle 1 to charge or discharge with different power
rate (within the maximum power constraint) to support the grid
C. Scenario 2—With EV Control Framework and Customer (Fig. 8), its battery was charged up to the desired energy level
Participation Program Ed at the end of each charging period Td (Fig. 9).
In view of Table V and Fig. 10, the rms of the power imbal-
In this scenario, the charging and discharging power of 600 M
ΔP (t )2
participating EVs was coordinated by the proposed control ance in this scenario is ΔPrms = k=1
M
k
= 144.86 kW,
strategy. It is clearly observed that during the simulation time, which is a significant improvement compared to the previous
plugged-in EVs temporarily discharged their batteries when scenario (without EV control) with ΔPrms = 641.86 kW. The
wind power was low and increased charging when there was simulation was implemented again without introducing errors
a surge in wind production (Fig. 6). As a result, the total power to the forecast data to obtain the best possible rms outcome. It
194 IEEE TRANSACTIONS ON TRANSPORTATION ELECTRIFICATION, VOL. 1, NO. 2, AUGUST 2015

Fig. 9. Battery energy level of Vehicle no. 1 during the simulation5 .

Fig. 11. Power imbalance in the system with and without customer participa-
tion program.
TABLE V
RMS OF THE P OWER I MBALANCE IN S CENARIOS 1 AND 2 AND THE
P ERFECT C ASE D. Scenario 3—EV Control Framework Without Customer
Participation Program
In the previous scenario, customers needed to provide their
charging plans for a day ahead under the proposed customer
TABLE VI
participation program. In this scenario, it was set up that such
RMS OF THE P OWER I MBALANCE IN S CENARIOS 2 AND 3 information was not available to the control center in advance
or in other words, there was no customer participating pro-
gram. Information about the charging requests was only known
when EVs had been physically connected to charging points
and sent out their requirements. The simulation result under this
condition is presented in Fig. 11 and Table VI.
As can be seen in Fig. 11, the power mismatch was slightly
higher during the period from minute 650 to 750 (simulation
time) in the case without customer participation program. As
far as the rms values are concerned, there is an increase of
6.49% in this scenario compared to the previous case with cus-
tomer participation program as shown in Table VI. This result
has showcased the importance of the customers’ contribution in
improving the performance of the control scheme.

IV. C ONCLUSION
In conclusion, our proposed centralized control strategy has
efficiently exploited the EV batteries in supporting a power grid
with high penetration level of RE while still maintains cus-
Fig. 10. RMS of the power imbalance in Scenarios 1 and 2 and the perfect case.
tomer satisfaction. The optimization algorithm is formulated as
a deterministic problem and solved repeatedly along the control
period to update the optimal G2V and V2G schedules for all
can be seen that there is approximately no difference between
vehicles. This optimization problem can be computed quickly
the two rms results from Table V and Fig. 10, which demon-
and efficiently by a normal optimization software in a standard
strates the ability of the control scheme in adapting with the
work station; hence, it is promising for practical applications. In
uncertainties in the system.
addition, the customer participating program plays an important
Another point worth mentioning is that during the day time,
role in improving the performance of the control algorithm.
EVs had access to fast-charging facilities, which allowed them
As far as future work is concerned, although power loss
to provide much more support to the grid than during the night
is not yet considered in this stage, it will be considered with
time (Figs. 6 and 8).
the view to further enhance the system’s efficiency. Moreover,
future work will explore the possibilities of using a hierarchial
5 The gray area means that battery level of Vehicle no. 1 is undefined as it is framework to coordinate the centralized control center with a
not plugged-in. decentralized controller in each individual vehicle.
NGUYEN et al.: OPTIMAL COORDINATION OF G2V AND V2G TO SUPPORT POWER GRIDS 195

R EFERENCES [21] S.-J. Huang and K.-R. Shih, “Short-term load forecasting via ARMA
model identification including non-Gaussian process considerations,”
[1] The European Wind Energy Association. (2005). Large Scale Integration IEEE Trans. Power Syst., vol. 18, no. 2, pp. 673–679, May 2003.
of Wind Energy in the European Power Supply: Analysis, Issues, [22] A. Miskelly. (2013). Wind Farm Performance [Online]. Available: http://
and Recommendations [Online]. Available: http://www.ewea.org/ windfarmperformance.info
fileadmin/ewea_documents/documents/publi-cations/grid/051215_Grid_ [23] Australian Energy Market Operator. (2013). Price and Demand
report.pdf, accessed on Nov. 1, 2014. Graph [Online]. Available: http://www.aemo.com.au/Electricity/Data/
[2] N. Wade, P. Taylor, P. Lang, and P. Jones, “Evaluating the benefits of an Price-and-Demand, accessed on Nov. 1, 2014.
electrical energy storage system in a future smart grid,” Energy Policy, [24] Society of Automotive Engineers (SAE). (2013). SAE Electric Vehicle
vol. 38, no. 11, pp. 7180–7188, 2010. and Plug in Hybrid Electric Vehicle Conductive Charge Coupler
[3] R. Curtin, Y. Shrago, and J. Mikkelsen, Plug-in Hybrid Electric Vehicles. [Online]. Available: http://standards.sae.org/j1772_201001, accessed on
Reuters/University of Michigan, Surveys of Consumers, 2009 [Online]. Nov. 1, 2014.
Available: http://ns.umich.edu/Releases/2009/Oct09/PHEV_Curtin.pdf,
accessed on Mar. 15, 2015.
[4] C. Mi, M. A. Masrur, and D. W. Gao, “Plug-in hybrid electric vehicles,”
Hybrid Electric Vehicles: Principles and Applications With Practical Hoang N. T. Nguyen received the B.Eng. degree
Perspectives, Chichester, U.K.: Wiley, 2011, pp. 107–138. in electrical and electronics engineering from Hanoi
[5] B. K. Sovacool and R. F. Hirsh, “Beyond batteries: An examination of University of Technology, Hanoi, Vietnam, in 2007,
the benefits and barriers to plug-in hybrid electric vehicles (PHEVs) and a and the Master’s degree in sustainable energy
vehicle-to-grid (V2G) transition,” Energy Policy, vol. 37, no. 3, pp. 1095– and electrical engineering from RMIT University,
1103, 2009. Melbourne, Vic., Australia, in 2012. He is currently
[6] U.S. Dept. of Energy. (2011). One Million Electric Vehicles by 2015, pursuing the Ph.D. degree at Swinburne University of
U.S. Dept. of Energy, Washington, DC, USA. [Online]. Available: Technology, Melbourne, Vic., Australia.
http://www1.eere.energy.gov/vehiclesandfuels/pdfs/1_million_electric_ He was involved in many Energy Efficiency
vehicles_rpt.pdf, accessed on Nov. 1, 2014. and Renewable Energy projects in 2007–2009. His
[7] A. Saber and G. Venayagamoorthy, “Resource scheduling under uncer- research interests include power system operation,
tainty in a smart grid with renewables and plug-in vehicles,” IEEE Syst. renewable energy, and electric vehicle integration in smart grid.
J., vol. 6, no. 1, pp. 103–109, Mar. 2012.
[8] A. Saber and G. Venayagamoorthy, “Plug-in vehicles and renewable
energy sources for cost and emission reductions,” IEEE Trans. Ind.
Electron., vol. 58, no. 4, pp. 1229–1238, Apr. 2011.
[9] A. Saber and G. Venayagamoorthy, “Efficient utilization of renewable Cishen Zhang received the B.Eng. degree in com-
energy sources by gridable vehicles in cyber-physical energy systems,” puter engineering from Tsinghua University, Beijing,
IEEE Syst. J., vol. 4, no. 3, pp. 285–294, Sep. 2010. China, in 1982, and the Ph.D. degree in electrical
[10] M. Khodayar, L. Wu, and M. Shahidehpour, “Hourly coordination of engineering from Newcastle University, Callaghan,
electric vehicle operation and volatile wind power generation in SCUC,” N.S.W., Australia, in 1990.
IEEE Trans. Smart Grid, vol. 3, no. 3, pp. 1271–1279, Sep. 2012. He was with the Department of Electrical and
[11] W. Su, J. Wang, and J. Roh, “Stochastic energy scheduling in microgrids Electronic Engineering, University of Melbourne,
with intermittent renewable energy resources,” IEEE Trans. Smart Grid, Melbourne, Vic., Australia, as a Lecturer, Senior
vol. 5, no. 4, pp. 1876–1883, Jun. 2014. Lecturer, Associate Professor, and Reader from 1989
[12] Department of Planning, Transportation and Local Infrastructure. to 2002, and with the School of Electrical and
(2013). Creating a Market: Victorian Electric Vehicle Trial Mid-term Electronic Engineering and School of Chemical and
Report [Online]. Available: http://www.transport.vic.gov.au/projects/ev- Biomedical Engineering, Nanyang Technological University, Singapore, from
trial, accessed on Mar. 15, 2015. 2002 to 2010. Since November 2010, he has been the Professor with the
[13] S. Soman, H. Zareipour, O. Malik, and P. Mandal, “A review of wind Department of Electrical and Electronic Engineering, Swinburne University of
power and wind speed forecasting methods with different time horizons,” Technology, Melbourne, Vic., Australia. His research interests include control,
in Proc. North Amer. Power Symp. (NAPS), Sep. 2010, pp. 1–8. signal processing, and medical imaging.
[14] L. Xie et al., “Wind integration in power systems: Operational challenges
and possible solutions,” Proc. IEEE, vol. 99, no. 1, pp. 214–232, Jan.
2011.
[15] H. Nguyen, C. Zhang, and M. A. Mahmud, “Smart charging and discharg-
ing of electric vehicles to support grid with high penetration of renewable Md. Apel Mahmud received the B.E. degree
energy,” in World Congr., vol. 19, no. 1, 2014, pp. 8604–8609. in electrical engineering from Rajshahi University
[16] C. Jin, J. Tang, and P. Ghosh, “Optimizing electric vehicle charging: of Engineering Technology (RUET), Rajshahi,
A customer’s perspective,” IEEE Trans. Veh. Technol., vol. 62, no. 7, Bangladesh, in 2008, and the Ph.D. degree in elec-
pp. 2919–2927, Sep. 2013. trical engineering from the University of New South
[17] C. Zhou, K. Qian, M. Allan, and W. Zhou, “Modeling of the cost of EV Wales, Sydney, N.S.W., Australia, in 2012.
battery wear due to V2G application in power systems,” IEEE Trans. Currently he is working as a Lecturer with the
Energy Convers., vol. 26, no. 4, pp. 1041–1050, Dec. 2011. Department of Electrical and Renewable Energy
[18] K. W. E. Cheng, B. Divakar, H. Wu, K. Ding, and H. F. Ho, “Battery- Engineering, Deakin University, Geelong, Vic.,
management system (BMS) and SOC development for electrical vehi- Australia. He also worked as a Lecturer with the
cles,” IEEE Trans. Veh. Technol., vol. 60, no. 1, pp. 76–88, Jan. 2011. Department of Electrical and Electronic Engineering,
[19] M. C. Kisacikoglu, M. Kesler, and L. M. Tolbert, “Single-phase on-board Swinburne University of Technology, Melbourne, Vic., Australia; a Research
bidirectional PEV charger for V2G reactive power operation,” IEEE Fellow with the University of Melbourne, Melbourne, Vic., Australia; and
Trans. Samrt Grid, vol. 6, no. 2, pp. 767–775, Mar. 2015. a Research Publication Fellow with the University of New South Wales.
[20] S. Boyd and L. Vandenberghe, Convex Optimization. Cambridge, U.K.: His research interests include different aspects of microgrids, power system
Cambridge Univ. Press, 2004. stability and control, and nonlinear control theories.

View publication stats