Electrical Power and Energy Systems 64 (2015) 10171024

Contents lists available at ScienceDirect

Electrical Power and Energy Systems

journal homepage: www.elsevier.com/locate/ijepes

Energy storage model with gridable vehicles for economic load dispatch
in the smart grid
Uttam Kumar Debnath a,, Iftekhar Ahmad a, Daryoush Habibi a, Ahmed Yousuf Saber b
a
b

Smart Energy Research Group, School of Engineering, Edith Cowan University, Joondalup, WA 6030, Australia
R&D Department, Operation Technology, Inc., ETAP, Lake Forest, CA 92630, USA

a r t i c l e

i n f o

Article history:
Received 8 June 2014
Received in revised form 24 August 2014
Accepted 4 September 2014
Available online 20 September 2014
Keywords:
Battery lifetime
Energy storage
Gridable vehicles
Participation rate
Renewable energy
Smart grid

a b s t r a c t
The intermittent nature of renewable energy sources (RESs) and unpredictable variable load demands
have necessitated the inclusion of energy storage devices in the smart grid environment. Electric vehicles
(EVs) and plug-in hybrid electric vehicles (PHEVs), with vehicle-to-grid capability, referred to as gridable
vehicles (GVs), have become an option as storage devices. However, unsupervised use of GVs as storage
devices presents new challenges due to concerns over battery lifetime and cost effectiveness of this twoway power transfer. These issues reduce the participation rate of GVs in the vehicle-to-grid discharge
program. In this study, we present a system model, for GVs to act as distributed storage devices, which
mitigates concerns over battery lifetime, and provides GV owners with a transparent cost-benet analysis
of their participation in the vehicle-to-grid discharge program. With this model in place, fuel and emissions cost has been reduced, as shown using particle swarm optimization. Simulation results show that
up to 52% of GVs might be discharging at a net loss. In contrast, our proposed system ensures that no GVs
are at loss when discharging to the grid. This factor alone is expected to increase the participation rate of
GVs by a signicant margin, and ensure economic load dispatch.
2014 Elsevier Ltd. All rights reserved.

Introduction
Sustainable energy systems have become a priority for almost
all countries in the world. A smart grid is meant to provide such
an energy system by integrating extra operational capabilities
and renewable energy sources (RESs) into the power system. RESs
such as wind and solar power are well accepted sustainable energy
sources. However, unreliability in continuous energy production
from RESs has been a major problem. A straightforward solution
to this problem is to use conventional energy storage systems, such
as electrochemical accumulators, which tend to be very costly. A
variety of energy storage provisions have been proposed to atten
the cost [1], although achieving an acceptable cost of storage is still
a very active area of research. In recent times, gridable vehicles
(GVs) have emerged as a signicant contender, for dealing with
the uncertainty of RESs, in order to keep the utility grid unaffected.
Research trends indicate that the use of GVs as energy storage
devices will bring about a breakthrough in the way that RESs are
managed.
Corresponding author. Tel.: +61 431530999, +61 863045318; fax: +61
863045811.
E-mail addresses: u.debnath@ecu.edu.au (U.K. Debnath), i.ahmad@ecu.edu.au
(I. Ahmad), d.habibi@ecu.edu.au (D. Habibi), aysaber@ieee.org (A.Y. Saber).
http://dx.doi.org/10.1016/j.ijepes.2014.09.004
0142-0615/ 2014 Elsevier Ltd. All rights reserved.

Recently, researchers [2] have been working on modeling the

benets of using vehicle energy. Others are working on the charging and discharging algorithms for GVs [3]. Modeling of charging
demand from GVs and the impact of charging load have been
addressed in some recent research [4,5]. Some other researchers
[6] have addressed the economic load dispatch (ELD) problem with
GVs as storage devices. Two major limitations of these works are
the implied assumptions that: (i) there exists a linear relationship
between a charging/discharging cycle and the lifetime of a GV battery; and (ii) GVs will discharge at any time required by the grid.
While such assumptions may be valid in an ideal situation, in practice, they do not accurately reect real-world conditions. The reasons include: (i) GVs are dependent on the performance of their
batteries, and battery lifetime depends on its remaining capacity
after frequent use as a source and load; (ii) the cost of owning
and maintaining a GV is still high; and (iii) the cost of a single
charging/discharging cycle for each GV is dependent on the rate
of capacity degradation. None of the research thus far has considered battery capacity degradation in the cost model, or in the
objective function of the economic load dispatch model.
While GVs solve a major problem regarding storing and delivering energy, battery capacity degradation and the actual cost of
energy from GVs in the real-time, put a limit on how often a GV
can be used as a storage device. Participation rate of GVs in the grid

1018

U.K. Debnath et al. / Electrical Power and Energy Systems 64 (2015) 10171024

Nomenclature

Wdep/Wpre
Wmin/Wmax
f
FCi()/ECi()

wi
wc/we
D(t)
H
N
NGV(t)
NGV-Dsch(t)
NGVmax
Pi(t)
Pwind(t)
Ppv(t)
Pmax
/Pmin
i
i

departure/present state of charge

min./max. state of charge
system efciency (battery)
fuel/emission cost function of unit i
emission penalty factor of unit i
weight factors for fuel/emission cost
load demand at time t
scheduling hours
number of thermal generating units
no. of vehicles connected to the grid at hour t
no. of vehicles discharging to the grid at hour t
total vehicles in the system
power of unit i at time t
power from wind farm at time t
power from solar photovoltaic panels at time t
max./min. output limit of unit i

Pvj
R(t)
Ncycle
CDremaining
CDstorage
CDcycling
Ccycle
Copp/Cdgdn
wopp/wdgdn
Vc(t)
Es(t)
ps(t)
pbest/gbest
Ite/MaxIte

discharge program is still low due to this limitation, which hinders

the realization of a sustainable energy system. An intelligent system is thus required to relate battery capacity with the charging/
discharging cycle in order to use GVs as a viable storage option;
which provides the motivation for this research.
Problem description
In the smart grid, along with conventional thermal generators,
the system consists of RESs, mainly wind and solar power sources,
and GVs such as plug-in hybrid electric vehicles (PHEVs) and electric vehicles (EVs). RESs are considered to be the compulsory
sources of generation to reduce emissions and running costs along
with the thermal generators, and GVs are considered to be distributed storage devices to help balance loads. After considering all the
sources including GVs, an optimization method is used to generate
an intelligent schedule for cost and emissions reduction.
One of the representative objective functions for cost-emission
optimization is [6], subject to some constraints:

min

N X
H
X
wc FC i Pi t we wi EC i Pi t

!
1

i1 t1

Two essential constraints of this optimization model, giving

consideration to load balancing and providing adequate spinning
reserves, are given below in (2) and (3).
With GVs as sources of energy the load balance equation with
reserve capacity is given as [6]:
NX
N
GV t
X
Pmax
t Ppv t
fPv j Wpre  Wdep
i
i1

RESs, and achieving a load balance condition. A high participation

rate of GVs is thus necessary for the successful implementation of
a smart grid with RESs. However this cannot be achieved unless a
model is developed which provides transparent information, and
instills enough condence in the owners that they will not end up
losing out if they participate in the grid-discharge program.
The aim of this work is to design an intelligent system model,
taking both the battery condition and capacity degradation issues
into consideration, and making the actual cost of GV energy transparent to the owners; assisting them to make a decision on discharging. Such a model is expected to signicantly increase the
participation rate, and to create a valuable contribution towards
the realization of a sustainable smart grid system.
Proposed system model for efcient and economic use of GVs
The main contributions of this paper are two-fold. First, we
model the capacity degradation cost of a battery, based on its numbers of cycles that have already been spent, and include this in the
cost of a single charging/discharging cycle of the GV, at various
stages of its lifetime. This model is then used to make a trade-off
between the cost involved in a charging/discharging cycle, and
the real-time power available from a GV; in order to decide
whether it should discharge or not, and in what amount. Second,
we develop an economic load dispatch model, where the objective
function takes the cost of using the GVs energy into account, in
conjunction with the cost of fuel and emissions for thermal generators. By using these two models, we ensure a cost-effective use of
the battery energy, which is expected to enhance the participation
rate of the GVs.

j1

Pwind t P Dt Losses Rt

And with GVs as loads or storages the load balance equation

with reserve capacity is given as [6]:
N
X
Pmax
t Ppv t Pwind t P Dt Losses Rt
i
i1

capacity (power) of vehicle j

system reserve requirement at hour t
cycle number of a battery
capacity degradation as a percentage of the initial
capacity
capacity degradation due to storage fade
capacity degradation due to cycling fade
cost per charging/discharging cycle
opportunity/capacity degradation cost per cycle
weight factors for Copp/Cdgdn
total cost of energy from vehicles
energy supplied from s-th vehicle at time t
price of energy of s-th vehicle at time t
particles best position/global best position
current iteration/maximum No. of iterations

NX
GV t

fPv j Wdep  Wpre

j1

From (2) and (3), it is evident that power transfer to/from the GVs is
the only determining factor for ensuring maximum utilization of

A. Modeling capacity degradation and actual cost of using GVs as

storages
Battery capacity degradation depends on a number of factors,
such as spent depletion cycles, age, temperature at which it has
been used, size and type, and battery chemistry. As the vehicles
continue using the batteries, capacity degradation is believed to
fall in two categories; namely cycling degradation and storage degradation. Research shows that these two broad categories of loss
cover most of the factors concerning capacity degradation, and
can be represented as a function of the number of spent cycles
[7]. If the current cycle number of a battery is Ncycle, the capacity

1019

U.K. Debnath et al. / Electrical Power and Energy Systems 64 (2015) 10171024

degradation as a percentage of the initial capacity, CDremaining, can

be given as in (4), further details being described in the following
sections.

The real-time energy available from a GV is justied against the

current cycle number, to make a trade-off, so that a GV does not
pay more in terms of its capacity degradation.
The actual cost of GV energy consists of the capacity degradation cost, and the opportunity cost, both of which are described
below. Opportunity cost is the initial cost of purchasing the battery. Capacity degradation cost is the cost incurred for the actual
degradation of the battery capacity due to energy discharging from
battery to the grid. A combination of both these costs has been
used in this study to account for the initial capital cost as well as
the capacity degradation cost that is variable to individual energy
transfer scenarios. With both these cost items included in the battery energy cost calculation, vehicle owners are given more freedom to set their own battery energy price that would ultimately
encourage them to accept this business model towards adopting
the grid discharge program.
If this actual cost is benecial with respect to the current energy
pricing, a GV will discharge; otherwise the GV can deny
discharging.
Capacity degradation cost
The capacity loss of batteries has been observed by many
researchers and been stated that the fraction of capacity loss can
be measured per cycle of charging and discharging [8]. Degradation
in current cycle number has thus been used in this study as one of
the representative measure of the capacity losses during continuous cycling of batteries, as it has been shown on various experimental results [7]. To make a closer approximation of the
conventional Li-ion battery capacity degradation trajectory, representative experimental results have been taken [7] to demonstrate
how to determine the capacity degradation from the current cycle
number. Each vehicle is expected to charge and discharge only
once a day, after giving consideration to both the battery lifetime
issues and the chargingdischarging time. Urban driving prole
has been considered to reect the driving prole of a usual vehicle
owner.
In order to relate capacity degradation to cycle number, we use
the experimental results graphs [7] showing capacity degradation
during storage, and cycling against time and cycle numbers. Ultimate capacity fade is determined from the minimum of these
two capacity fading effects [7]. Reformulating the graphs (cycle
numbers in the X-axis) and assuming each battery will discharge
at most once a day, we get (5) and (6).
For storage, the capacity fade equation becomes

CDstorage 2  106 N2cycle  0:008Ncycle 100:8

Using 4000 cycles as a benchmark, and taking the minimum value

of capacity from (5) and (6), we nd (7) and the corresponding tted graph in Fig. 1 (R2 = 0.9985). We have used R2, which is the
Coefcient of Determination to quantify the goodness of t of
a nonlinear model with an experimental result. R2 is computed from
the sum of the squares of the distances of the points from the bestt curve determined by nonlinear regression.

CDremaining 2  1016 N5cycle  2  1012 N 4cycle 5  109 N3cycle

 5  106 N2cycle  0:0038Ncycle 99:961

98

96

94

92

90

88
0

500

1000

1500

2000

2500

3000

3500

4000

Cycle Number
Fig. 1. A plot of the total capacity fading of a battery against the number of cycles
both from [7] and the tted Eq. (7). The quality of the t is given by R2 = 0.9985.

To nd out the number of cycles already spent by the battery, we

use the battery internal management system records. We use the
capacity degradation versus cycle number graph in Fig. 1 to nd
the capacity degradation at that particular cycle number. This information can now be used to calculate the real-time degradation as a
percentage of the original capacity.
Costing for the degradation caused by the current charging-discharging cycle can be calculated using established battery performance data. As a battery cannot be used for vehicle-to-grid
power transfer once the remaining capacity drops below 80% of
the initial capacity [9,10], cost of the battery should be distributed
to the range between 80% and 100% of the original capacity.
When using the GV batteries, the end-of-life (EOL) requirements of the batteries must also be addressed. When discharging
the batteries, the state-of-charge (SOC) window should not cross
a certain limit [11] to ensure the expected longevity and safety
of the battery. Because of the EOL requirements of the battery,
owners are always concerned about the depth-of-discharge
(DOD) while discharging. The owners also look at the DOD range
when selecting their weight factors for different cost items; as
described in the following sections.
Battery opportunity cost
The cost to manufacture lithium-ion batteries depends on the
time, size, and volume of the production run. Currently the opportunity cost is approximately $1000/kW h [12,13].
Both the opportunity cost and the degradation cost are converted into a cost per cycle, and are then added together to nd
the per cycle charging-discharging cost. The cost per charging/discharging cycle, Ccycle, can thus be modeled as:

C cycle wopp C opp wdgdn C dgdn

For cycling, the capacity fade equation becomes

CDcycling 4  108 N2cycle  0:003Ncycle 99:82

Curve for Source Data

Curve for Fitted Equation

Remaining Capacity (%)

CDremaining f Ncycle

100

Weighting factors (ranging from 0 to 1) are selected by the owner,

depending on the current cycle number and the DOD the batteries
are required to operate up to. Before discharging a vehicle, an
owner/owners agent can look at this costing per cycle and corresponding battery capacity to analyze the revenue and battery lifetime. If the current price of selling energy to the grid exceeds
Ccycle, the GV may decide to discharge; otherwise it should decline
discharging. The owchart for the decision making process is given
in Fig. 2.
Selecting the weight factors is the owners choice as to which
cost concerns them the most, and they should be given the freedom of choosing the weight factors that may help reduce their revenue concerns.

1020

U.K. Debnath et al. / Electrical Power and Energy Systems 64 (2015) 10171024

per cycle. From previous research [7], we can relate the DOD to
the cost per cycling of a battery. A DOD corresponding to discharge
cycles of either less/more than that specied, require an adjustment factor of more/less than unity, respectively. The capacity degradation cost then becomes a direct product of the cost and the
adjustment factor.
C. Proposed optimization model considering cost of vehicle energy
Wind and solar energy are emission free and their operating
costs are negligible. Fuel cost of a conventional thermal generator
is expressed as a quadratic function of the units generated power
as follows:

FC i Pi t ai bi Pi t ci P2i t

10

where ai, bi and ci are positive fuel cost coefcients of unit i.

Emissions cost is expressed as another quadratic function of the
units generated power as follows:

EC i Pi t ai bi Pi t ci P2i t

11

where ai, bi and ci are emission coefcients of unit i.

With GVs as sources, the load balance equation is:
NGVDsch
N
X
X t
Pi t Ppv t
fPv j Wpre  Wdep Pwind t
i1

j1

Dt Losses

12

With GVs either acting as a load or providing storage capacity, the

load balance equation becomes:
Fig. 2. Flowchart for discharging decision by the owner of GVs in the smart grid
using proposed model.

N
X
Pi t Ppv t Pwind t Dt Losses
i1

It is possible to nd an objective choice for these parameters,

but that might push the vehicle owners towards a confusion as
to who decides on the use of their own assets. A range of values
for each of these parameters have been proposed in this study.
Copp and Cdgdn in Eq. (8) are computed in three steps: rst, total
opportunity cost and total capacity degradation cost are determined for the whole automotive lifetime of a vehicle battery. Then
useful life of the battery is determined on the basis that the battery
will cease to be suitable for automotive purpose once the capacity
degrades below 80% of the original capacity. This lifetime is converted into a number of cycles the battery can charge and discharge. And nally, the total opportunity cost and total capacity
degradation costs are divided by the number of operating cycles
to nd Copp and Cdgdn.
All the necessary parameters (including cost, and amount of
energy from all the vehicles that discharge at a particular hour),
are calculated to nd the total cost of energy from the vehicles,
Vc(t) as follows:

V c t

N GVDsch
X t

Es tps t

s1

B. Variation in per cycle charging/discharging cost and analysis of the

vehicle-to-grid economics
The cost of capacity degradation and consequently of per cycle
discharging, changes according to the variation of the vehicle price,
type of vehicle, and rate of discharge. Further, the cost of energy
sold to the grid depends on the variation in the amount of energy
actually sold, as this is directly related to the depth-of-discharge
(DOD), which ultimately affects the rate of capacity degradation

N GVDsch
X t

fPv j Wdep  Wpre

13

j1

Load balance equations with adequate spinning reserves are given

by Eqs. (2) and (3), where NGV(t) is replaced by NGV-Dsch(t). Each thermal generator has its own generation range, represented as:

Pmin
6 Pi t 6 Pmax
i
i
Pimax

14
min

where
and Pi
are the maximum and minimum generation
limits of the i th unit.
Charging/discharging up to certain maximum/minimum levels
is ensured to prevent battery failure, and is given by:

Wmin Pv j 6 Pv j t 6 Wmax P v j

15

where Wmin and Wmax are the minimum and maximum levels of
charge respectively, of the individual vehicles. Vehicles that have
been registered for charging/discharging from/to the grid NGVmax,
can take part during a predened scheduling period; and are given
by:
H
X
NGV t Nmax
GV

16

t1

Minimizing generation cost, emissions cost, and cost of buying

energy from the vehicles are considered to be the primary objectives of the smart grid; and load balance, reserve, power generation
limit, and charging/discharging limits are considered to be the
constraints.
The objective function for cost-emission optimization, including
the cost for vehicle energy, is therefore:

min

N X
H
X
wc FC i Pi t we wi EC i Pi t V c t
i1 t1

!
17

1021

U.K. Debnath et al. / Electrical Power and Energy Systems 64 (2015) 10171024

Optimizing fuel and emissions cost

Results and discussion

An efcient optimization method is required to minimize fuel

and emissions cost in a system consisting of thermal generators,
RESs, and GVs. Particle swarm optimization (PSO) [14] is used in
this study to schedule the power intelligently. PSO provides a population based search procedure- where each individual called a
particle, is represented by its position (state) and velocity, and particles move around in a multidimensional search space. Each particle adjusts its position according to both its own experience,
and its neighboring particles experience, making use of the best
positions discovered by itself and its neighbors.
In PSO, the velocity and position of each particle is calculated
iteratively as follows:

The system described in this study includes thermal generators,

RESs, and GVs in the smart grid environment. An on-board GV
interface system and the parking station computer system communicate with all registered vehicles for collecting information on the
vehicles battery condition. This is how the owners will know the
current rate of capacity degradation, and are able to decide if discharging will make revenue for them. GVs that discharge to the
grid are eligible for charging at a subsidized rate for a period determined by the operator. An independent system operator (ISO) of a
6-unit system with 50,000 registered GVs has been simulated in
this study. Unit characteristics taken from [16], and emission coefcients taken from [6] are given below in Tables 1 and 2. For GVs,
the following parameter values were considered: vehicle battery
capacity S = 15 kW, H = 24 h, minimum Wdep = 40%, and f = 85%.
For PSO, swarm size = 50, iterations = 1000, c1 = c2 = 2, Range = 0.5,
wi = 25 $/ton, and wc = we = 1.
GVs arrive at parking stations randomly, so the number of GVs
available may not meet the real-time requirement of the grid. Planning is thus necessary to provide a match between the number of
GVs and the real-time demand. An availability planning model [17]
does this matching by scheduling the GVs to discharge to the grid
only when the grid needs them the most. The availability planning
model provides a distribution of number of GVs that will discharge
energy to the grid at different times throughout the peak periods.
To reect the real-time mobility of the vehicles during the 24 h
daily cycle, a mobility factor, m, which is a measure of the percentage
of vehicles that are on the road (not at the parking station) at a certain period, has been considered [17]. In the availability planning
model we divide the 50,000 registered vehicles into a group of
20,000 and a group of 30,000 for the two peak periods, pk1 (9 am
4 pm) and pk2 (4 pm11 pm), respectively. We have set the mobility
factors m1 and m2 as 0.2 and 0.3, corresponding to pk1 and pk2,
respectively. We also assumed that around 0.2% of the GVs would
not show up at all, giving us the proposed availability planning
Vdisch for each of the 24 h of the day: Vdisch = [0 0 0 0 0 0 0
0 16 336 2176 5456 5456 2176 336 37 441 2856 7161 7161 2856 441 21 0]. All GVs are assumed to be charged during each 24-h period,
and thus the total energy required to charge them is distributed over
the hours when the demand is close to the base load demand. Economy of the charging load distribution over individual hours is
dependent on the load demand, RESs generation, time-of-use pricing, generator parameters, and balance between cost and emissions.
An intelligent coordination of all these factors has been performed,
which gives us the following daily distribution of GV charging,
where Vchrg = [20668 13333 6666 0 0 0 0 0 0 0 0 0 0 0 0 0 1667 0 0 0 0 0
1000 6666].
To study the effect of the proposed cost models with respect to
cost and emissions reduction, these two models have been incorporated in the economic load dispatch (ELD) problem. PSO was
used to minimize fuel and emissions costs for the economic dispatch. For RESs, solar insolation data, wind speed data, and generated power, and demand data over the 24-h period have been
taken from [6]. The time-of-use pricing of energy (in $/MW h) for

v pq k 1 v pq k c1 r1 pbestpq k  xpq k
c2 r2 gbest q k


Range
 xpq k 1
Ite  1
MaxIte
xpq k 1 xpq k v pq k 1

18

19

where velocity v, position x, accelerating parameters c1 and c2, random numbers r1 and r2, particle number p, problem dimension q,
and iteration index k, are standard terms of PSO [15]. The owchart
for the minimization of fuel and emissions cost with RESs and GVs
in a smart grid, using our proposed models is given in Fig. 3. If at
hour t, the schedule is; [P1(t), P2(t),. . ., PN(t), NGV(t), Ppv(t), Pwind(t)]T,
then power supplied to/from vehicles is fNGV(t)Pvi (Wpre  Wdep).
The sign of this expression will indicate whether it is a load or
source; and the rest of the load demand, given by the expression;
[D(t) + fNGV(t)Pvi (Wpre  Wdep)  Ppv(t)  Pwind(t)] will be met from
the conventional thermal units.

Table 1
Generating unit capacity and coefcients.

Fig. 3. Flowchart for fuel and emissions cost minimization with RESs and GVs in the
smart grid using proposed models.

Unit

Pmin (MW)

Pmax (MW)

a ($)

b ($/MW)

c ($/MW2)

1
2
3
4
5
6

100
50
80
50
50
50

500
200
300
150
200
120

240
200
220
200
220
190

7.0
10.0
8.5
11.0
10.5
12.0

0.0070
0.0095
0.0090
0.0090
0.0080
0.0075

1022

U.K. Debnath et al. / Electrical Power and Energy Systems 64 (2015) 10171024

Table 2
Generator emissions coefcients.

a (ton/h)

b (ton/MW h)

c (ton/MW2 h)

1
2
3
4
5
6

10.33908
32.00006
32.00006
30.03910
32.00006
30.03910

0.24444
0.38132
0.38132
0.40695
0.38132
0.40695

0.00312
0.00344
0.00344
0.00509
0.00344
0.00509

a typical day is as follows [18]: from hours 7 to 17, price is 320.30;

from hours 18 to 23, price is 332.00; and during all other hours,
price is 145.90.
For calculating per cycle charging-discharging costs, the opportunity cost has been taken from the $800$1200 per kW h range
[12,13] (assuming a cycling capacity of around 4000 cycles). Capacity degradation cost is measured throughout the entire life of a battery until it is suitable for discharging to the grid. By taking battery
costs from [12,13], capacity degradation cost for a 4000 cycle battery can be taken as $6000$8000 throughout its discharging
capacity lifetime. A random distribution of these costs has been
considered, within the specied range, for all 50,000 vehicles.
Owners are free to choose the weighting factors in (8) that best
represent the cost of their vehicles energy. Although vehicles discharge to supply the grid, they also discharge while being driven
for everyday purposes, which accounts for a signicant portion of
the various cost items. Depending upon the situation, capacity degradation costs may account fully for each discharge cycle. A range
of different weighting factors have been studied. From this study,
the gross numbers of vehicles experiencing a loss, and the percentage with respect to the available vehicles have been calculated as
shown in Figs. 4 and 5.
Figs. 4 and 5 show that up to a 52.47% of vehicles can be in a loss
condition, depending on the specic choices made by each owner.
Even with a grid-friendly weight factor selection, which is unlikely
to happen from the owners perspective, at least 4.76% of the vehicles would be operating at loss.
To reect the differences in weighting factor selection from different GV owners, the calculations have been performed again
allowing the weighting factors to vary randomly within a certain
range, and illustrated in Figs. 6 and 7. Figs. 6 and 7 show that even
though different GV owners select a variety of weighting factors
from within a given range, up to a 36.30% of vehicles can still be
operating at a loss.

Bar 1 - For weights (0.35, 0.95)

Bar 2 - For weights (0.35, 1.00)
Bar 3 - For weights (0.40, 0.90)
Bar 4 - For weights (0.40, 1.00)
Bar 5 - For weights (0.45, 0.90)
Bar 6 - For weights (0.45, 0.95)

50

40

30

20

10

Combination of weight factors

Fig. 5. Percentage of discharging GVs at loss at different hours in a day for different
weight factors (wopp, wdgdn) described in (8), corresponding to Fig. 4.

GVs dicharging for different range of weights (No.)

Unit

Percentage of GVs at loss (%)

60

8000
Expected to discharge
For range of weights
(0.25-0.35, 0.8-0.9)
For range of weights
(0.25-0.35, 0.9-1.0)
For range of weights
(0.30-0.40, 0.8-0.9)
For range of weights
(0.25-0.40, 0.8-1.0)
For range of weights
(0.30-0.40, 0.9-1.0)

7000
6000
5000
4000
3000
2000
1000
0
0

10

15

20

25

Time of a day (Hour)

Fig. 6. Number of discharging GVs at loss at different hours in a day for different
ranges of weight factors (wopp, wdgdn) described in (8).

8000

35

Expected to discharge
For weights (0.35, 0.95)
For weights (0.35, 1.00)
For weights (0.40, 0.90)
For weights (0.40, 1.00)
For weights (0.45, 0.90)
For weights (0.45, 0.95)

7000
6000
5000

Percentage of GVs at loss (%)

GVs discharging for different weights (No.)

40

4000
3000
2000

30

Bar 1-Range of weights (0.25-0.35, 0.8-0.9)

Bar 2-Range of weights (0.25-0.35, 0.9-1.0)
Bar 3-Range of weights (0.30-0.40, 0.8-0.9)
Bar 4-Range of weights (0.25-0.40, 0.8-1.0)
Bar 5-Range of weights (0.30-0.40, 0.9-1.0)

25
20
15
10
5

1000

0
0

10

15

20

25

Time of a day (Hour)

Fig. 4. Number of discharging GVs at loss at different hours in a day for different
weight factors (wopp, wdgdn) described in (8).

Combination of range of weight factors

Fig. 7. Percentage of discharging GVs at loss at different hours in a day for different
ranges of weight factors (wopp, wdgdn) in (8), corresponding to Fig. 6.

1023

U.K. Debnath et al. / Electrical Power and Energy Systems 64 (2015) 10171024
Table 3
PSO results with vehicle energy and costs considered with no price-check.
Time
(H)

Unit 1
(MW)

Unit 2
(MW)

Unit 3
(MW)

Unit 4
(MW)

Unit 5
(MW)

Unit 6
(MW)

Solar
(MW)

Wind
(MW)

Vehicles
(MW)

Demand
(MW)

Loss
(MW)

Total cost
($)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24

170.45
168.57
180.40
192.98
204.41
231.45
255.29
273.87
323.11
412.60
450.69
483.57
377.49
315.50
279.80
216.45
206.38
227.15
257.80
412.58
330.81
232.00
189.01
174.46

152.41
150.84
161.13
172.08
182.05
200.00
200.00
200.00
200.00
200.00
200.00
200.00
200.00
200.00
200.00
192.46
183.74
200.00
200.00
200.00
200.00
200.00
168.64
156.00

160.38
158.78
169.10
180.08
189.89
213.21
233.61
249.47
291.20
300.00
300.00
300.00
300.00
284.77
254.51
200.30
191.61
209.51
235.79
300.00
297.67
213.69
176.51
163.98

104.90
103.79
110.89
118.48
125.38
141.58
150.00
150.00
150.00
150.00
150.00
150.00
150.00
150.00
150.00
132.58
126.57
138.98
150.00
150.00
150.00
141.90
116.10
107.34

149.80
148.22
158.37
169.12
178.87
200.00
200.00
200.00
200.00
200.00
200.00
200.00
200.00
200.00
200.00
189.16
180.54
198.18
200.00
200.00
200.00
200.00
165.69
153.32

101.68
100.55
107.67
115.21
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
112.80
104.12

0.00
0.00
0.00
0.00
0.00
0.00
0.09
17.46
31.45
36.01
38.06
35.93
36.78
31.59
9.70
12.92
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00

10.54
22.27
25.50
25.50
25.50
25.50
25.50
25.50
25.50
25.50
25.50
25.50
25.50
24.82
20.74
14.62
25.50
19.04
25.50
18.02
25.50
21.42
0.00
2.55

131.76
85.00
42.5
0
0
0
0
0
+0.1020
+2.1420
+13.8720
+34.7820
+34.7820
+13.8720
+2.1420
+0.2359
7.8157
+18.2070
+45.6514
+45.6514
+18.2070
+2.8114
6.2411
42.50

700.00
750.00
850.00
950.00
1000.00
1100.00
1150.00
1200.00
1300.00
1400.00
1450.00
1500.00
1400.00
1300.00
1200.00
1050.00
1000.00
1100.00
1200.00
1400.00
1300.00
1100.00
900.00
800.00

18.40
18.02
20.56
23.45
26.11
31.74
34.49
36.30
41.36
46.25
48.13
49.78
44.56
40.55
36.89
28.73
26.53
31.07
34.74
46.25
42.19
31.81
22.51
19.27

17967.96
17707.78
19420.21
21362.12
23158.92
27092.55
29286.39
30849.98
35524.14
42317.36
48783.07
57904.84
50047.09
39027.93
32034.26
25050.59
24321.10
32248.91
43644.59
55803.82
41926.79
28017.35
20772.68
18552.13

Total Cost (Fuel, Emission and Vehicle Energy) = $782822.56

Table 4
PSO results with vehicle energy and costs considered with price-check.
Time
(H)

Unit 1
(MW)

Unit 2
(MW)

Unit 3
(MW)

Unit 4
(MW)

Unit 5
(MW)

Unit 6
(MW)

Solar
(MW)

Wind
(MW)

Vehicles
(MW)

Demand
(MW)

Loss
(MW)

Total cost
($)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24

170.45
168.57
180.40
192.98
204.41
231.45
255.29
273.87
323.13
413.65
457.20
500.00
394.13
319.00
280.34
216.46
206.70
228.79
265.27
452.85
333.81
232.29
189.00
174.46

152.41
150.84
161.13
172.08
182.05
200.00
200.00
200.00
200.00
200.00
200.00
200.00
200.00
200.00
200.00
192.50
184.00
200.00
200.00
200.00
200.00
200.00
168.64
156.00

160.38
158.78
169.10
180.08
189.89
213.21
233.61
249.47
291.21
300.00
300.00
300.00
300.00
287.73
254.99
200.34
191.86
210.94
242.13
300.00
300.00
213.96
176.48
163.98

104.90
103.79
110.89
118.48
125.38
141.58
150.00
150.00
150.00
150.00
150.00
150.00
150.00
150.00
150.00
132.60
126.72
139.95
150.00
150.00
150.00
142.08
116.10
107.34

149.80
148.22
158.37
169.12
178.87
200.00
200.00
200.00
200.00
200.00
200.00
200.00
200.00
200.00
200.00
189.17
180.86
199.60
200.00
200.00
200.00
200.00
165.68
153.32

101.68
100.55
107.67
115.21
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
120.00
112.86
104.12

0.00
0.00
0.00
0.00
0.00
0.00
0.09
17.46
31.45
36.01
38.06
35.93
36.78
31.59
9.70
12.92
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00

10.54
22.27
25.50
25.50
25.50
25.50
25.50
25.50
25.50
25.50
25.50
25.50
25.50
24.82
20.74
14.62
25.50
19.04
25.50
18.02
25.50
21.42
0.00
2.55

131.76
85.00
42.5
0
0
0
0
0
+0.0765
+1.1411
+7.6946
+18.4429
+18.9401
+7.7775
+1.1730
+0.1339
9.0461
+13.0560
+32.5571
+33.0289
+13.1899
+2.1038
6.2539
42.50

700.00
750.00
850.00
950.00
1000.00
1100.00
1150.00
1200.00
1300.00
1400.00
1450.00
1500.00
1400.00
1300.00
1200.00
1050.00
1000.00
1100.00
1200.00
1400.00
1300.00
1100.00
900.00
800.00

18.40
18.02
20.56
23.45
26.11
31.74
34.49
36.30
41.36
46.30
48.45
50.62
45.35
40.92
36.95
28.74
25.50
31.38
35.46
46.90
42.50
31.85
22.51
19.27

17967.96
17707.78
19420.21
21362.12
23158.92
27092.55
29286.39
30849.98
35517.73
42081.92
47375.61
54302.50
46242.31
37488.73
31783.01
25022.97
23985.69
30864.90
40205.66
52848.47
40673.34
27827.76
20769.16
18552.13

Total Cost (Fuel, Emission and Vehicle Energy) = $762387.67

Resource scheduling has been performed with vehicles as

sources, storages, and loads. Results for cost and emissions reduction in a smart grid system, incorporating the proposed availability
planning model and the battery capacity degradation model, with
real-world costing for wind, solar, and GVs are shown in Tables 3
and 4. As a likely real-world scenario, weighting factors representing the fourth bar in the bar chart of Fig. 7 have been taken to select
the number of discharging vehicles. Table 3 represents the economic load dispatch for a conventional smart grid model, and the
planned availability distribution. Table 4 represents the same

scenario for weighting factors of (0.250.40, 0.81.0) that leads

to 35.85% of discharging vehicles operating at a loss, and hence
being restricted from discharging.
As a result, 35.85% of the expected discharging vehicles have been
saved from operating at a loss. The GV discharging distribution in
this case is V = [0 0 0 0 0 0 0 0 12 179 1207 2893 2971 1220 184 21
248 2048 5107 5181 2069 330 19 0]. Signs for the GV energy in Tables
3 and 4 represent GVs as source (+) and load ().
It is evident from Tables 3 and 4 that fuel and emissions costs
have been reduced with our proposed models. As a considerable

1024

U.K. Debnath et al. / Electrical Power and Energy Systems 64 (2015) 10171024

Table 5
Overall benets of using our proposed models.
Items

With conventional
model

With our proposed

model

GVs at loss
Expected GV
participation rate
Total cost
Potential for RESs
integration
Basis of discharging
decision

Up to 52%
As low as 48%

Close to none
Close to 100%

More
Less due to participation
rate
Non-transparent

Less
More
Transparent and
objective

amount of power has been supplied from the vehicles, utilization

of more renewable sources has been and would be possible, if it
were available. Similar results with other combination of weighting factors justify the benets of using our proposed models.
A summary of the benets of using our models are given in
Table 5.
For an ISO implementing V2G, total storage capacity potentially
available from the GVs is dependent on the number of participating
GVs and the effective discharging capacity of the GV batteries. Our
proposed system model maintains the owners condence in their
vehicles operating conditions, as well as maintaining revenue outcomes against battery wear. At the same time, owners have the
freedom to stop discharging to the grid should they be concerned
that they are not earning any revenue, which would help provide
an incentive to participate and remain in the grid discharge program. Because the individual owners are convinced of their revenue
outcomes, the system operator can in turn be condent of the availability of a considerable number of GVs for discharge. As a result,
both the vehicle owners and the system operator have their own
freedom to choose any combination of buying and selling energy
to/from the grid. Providing such a exible arrangement will encourage more owners to participate in the grid discharge program,
which is imperative if the GV integration is to be a success.
In an effort to compare the approach taken by other research
appearing in literature [19,20], especially for what concerns the
inclusion of degradation costs into the optimization models, our
results show that we have taken a different approach to the inclusion of degradation costs into the optimization model. While the
mentioned research [19,20] use the capacity degradation costs to
compare the effectiveness of their proposed models, our research
demonstrates how to calculate the degradation cost, why is it
important to include it in the optimization model and how the
vehicle owners can save themselves from revenue losses from discharging to the grid. Moreover, our research deals with the fundamental issue of decision making as to whether or not a vehicle will
discharge at the grid operators request.
Conclusion
Using GV batteries as energy storage units for dealing with the
variable RESs and loads in the smart grid environment has been an
undesirable option for each individual operator. The low participation rate in the V2G discharge program has been a major impediment to the successful implementation of this concept. The main
obstacles being the owners anxiety about battery lifetime, and
concern over the ultimate benet arising from using GVs as energy
storage units. In this paper, we have proposed a model that considers the issues associated with battery degradation and relevant
costs, and have provided the owners of the GVs with a transparent

tool for estimating the real-time cost of discharging to the grid,

eradicating concerns that they will incur a loss over the long run.
We have also proposed an economic load dispatch model that
includes the cost of using GV energy in the objective function,
along with the fuel and emissions cost of the thermal sources.
Our proposed models will save discharging vehicles from experiencing a loss, which is expected to signicantly increase V2G participation rate, to integrate more RESs, and ensure improved levels
of sustainability. In addition, the models provide an assurance to
the operators that GV energy is available, enabling them to deal
with more variable generations and loads, and to maintain economic load dispatch with GVs.
Acknowledgement
This work is supported by a Commonwealth of Australia
Scholarship.
References
[1] Makarov YV, Du P, Kintner-Meyer MCW, Jin C, Illian HF. Sizing energy storage
to accommodate high penetration of variable energy resources. IEEE Trans
Sustainable Energy 2012;3(1):3440.
[2] Ma Y, Houghton T, Cruden A, Ineld D. Modelling the benets of vehicle-togrid technology to a power system. IEEE Trans Power Syst
2012;27(2):101220.
[3] Sortomme E, El-Sharkawi MA. Optimal charging strategies for unidirectional
vehicle-to-grid. IEEE Trans Smart Grid 2011;2(1):1318.
[4] Li G, Zhang X. Modelling of plug-in hybrid electric vehicle charging demand in
probabilistic power ow calculations. IEEE Trans Smart Grid 2012;3(1):4929.
[5] Liu C, Wang J, Botterud A, Zhou Y, Vyas A. Assessment of impacts of PHEV
charging patterns on wind-thermal scheduling by stochastic unit
commitment. IEEE Trans Smart Grid 2012;3(2):67583.
[6] Saber AY, Venayagamoorthy GK. Efcient utilization of renewable energy
sources by gridable vehicles in cyber-physical energy systems. IEEE Syst J
2011;4(3).
[7] Smith K, Markel T, Pesaran A. PHEV Battery Trade-Off Study and Standby
Thermal Control. In: 26th International Battery Seminar and Exhibit. Fort
Lauderdale, FL, USA; 1619 March 2009.
[8] Deshpande R, Verbrugge M, Cheng Y, Wang J, Liu P. Battery cycle life prediction
with coupled chemical degradation and fatigue mechanics. J Electrochem
Society 2012;159(10):A17308.
[9] Viswanathan VV, Kintner-Meyer M. Second use of transportation batteries:
maximizing the value of batteries for transportation and grid services. IEEE
Trans Vehicular Technol 2011;60(7):296370.
[10] Neubauer J, Pesaran A. The ability of battery second use strategies to impact
plug-in electric vehicle prices and serve utility energy storage applications. J
Power Sources 2011;196:103518.
[11] Axsen J, Burke A, Kurani KS. Batteries for Plug-in Hybrid Electric Vehicles
(PHEVs): Goals and the State of Technology circa 2008. ITS, Davis, CA, Tech.
Rep. UCD-ITS-RR-08-14; May 2008.
[12] Gondelach SG, Faaij A. Performance of batteries for electric vehicles on short
and longer term. J Power Sources 2012;212:11129.
[13] Vliet OV, Brouwer AS, Kuramochi T, Broek MV, Faaij A. Energy use, cost and
CO2 emissions of electric cars. J Power Sources 2010;196:2298310.
[14] Kennedy J, Eberhart R. Particle swarm optimization. In: Proc. IEEE Int. Conf.
Neural Networks, vol. 4; Aus., 1995, p. 19428.
[15] Valle Y, Venayagamoorthy GK, Mohagheghi S, Hernandez J-C, Harley RG.
Particle swarm optimization: basic concepts, variants and applications in
power systems. IEEE Trans Evolutionary Computat 2008;12(2):17195.
[16] Gaing Z. Particle swarm optimization to solving the economic dispatch
considering the generator constraints. IEEE Trans Power Syst
2003;18(3):118795.
[17] Debnath UK, Ahmad I, Habibi D, Saber AY. Improving Battery Lifetime of
Gridable Vehicles and System Reliability in the Smart Grid. IEEE Systems
Journal (accepted in Nov. 2013), doi: 10.1109/JSYST.2013.2294734, to be
published.
[18] Victoria Government Gazette, No. S 226, 29 June 2012, Australia.
[19] Hoke A, Brissette A, Maksimovic D, Pratt A, Smith K. Electric vehicle charge
optimization including effects of lithium-ion battery degradation. In: Proc.
vehicle power and propulsion conference (VPPC), 2011. IEEE; 2011. p. 69. doi:
10.1109/VPPC.2011.6043046.
[20] Lunz B, Yan Z, Gerschler JB, Sauer DU. Inuence of plug-in hybrid electric
vehicle charging strategies on charging and battery degradation costs. Energy
Policy 2012;46:5119.