This article has been accepted for publication in a future issue of this journal, but has not been
fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2017.2648509, IEEE
Transactions on Smart Grid
Volt-VAR Control through Joint Optimization of
Capacitor Bank Switching, Renewable Energy, and
Home Appliances
Mosaddek Hossain Kamal Tushar, Member, IEEE and Chadi Assi, Senior Member, IEEE
Abstract—Today, the evolution of smart grid, electric vehicles
(EV) with V2G (voltage to the grid) mode, and deployment
of renewable energy sources (RES) are bringing revolutionary
changes to the existing electrical grid. Volt-VAR optimization
(VVO) is a well-studied problem, for bringing solutions to reduce
the losses and demand along the distribution lines. The current
VVO, however, does not acknowledge the role of elastic and
inelastic loads, EVs, and RESs to reduce the reactive power losses
and hence the cost of generation. We propose a mathematical
model Volt-VAR and CVR Optimization (VVCO)/Optimal En-
ergy Consumption Model (OECM) to solve the VVO problem
by considering load shifting, EV as the storage and carrier
of the energy, and use of RES. The VVCO/OECM not only
reduces the reactive load but also flatten the load curve to
reduce the uncertainty in the generation and to decrease the
cost. The system also considers the efficiency of the electrical
equipment to enhance the lifetime of the devices. We develop a
non-cooperative game to solve the VVCO/OECM problem. To
evaluate the performance, we simulate the VVCO/OECM model
and compare with the existing VVO solution. We found that our
method took almost a constant time to produce a solution of VVO
regardless of the size of the network. The proposed method also
outperform the existing VVO solution by reducing the generation
cost and flatten the load and minimizes the uncertainty in the
power generation. Results have shown that exploiting RES will
reduce the voltage drop through reducing the injection of reactive
power to the system.
I. I NTRODUCTION
Recent advances in technologies such as M2M (machine to
machine) and D2D (device to device), are reshaping today’s
electricity transmission and distribution as well as the electric-
ity market. Electric transmission and distribution networks are
considered as a critical and essential infrastructure for modern
societies, any failure on any transmission or distribution line
can cause major power outages [1]. Studies showed that
the world electricity consumption has increased about 3.1%
annually from 1980 to 2006, and according to the US Energy
Information Administration (EIA), the power consumption is
expected to grow from 3,836 to 4,797 billion kWh from 2013
to 2040 [2], [3]. Currently, about 10% of the energy is lost
during transmission and distribution to the consumer. About
40% of the total loss occurs in distribution networks [3].
In 2012, the total energy losses and distribution losses were
1,835 and 733 billion kWh, respectively [4], [3]. Hence, a
10% reduction of distribution losses would, therefore, save
This work has been supported by NSERC Discovery Grants and Concordia
University. The statements made herein are solely the responsibility of the
authors.
1949-3053 (c) 2016 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.
1
73 billion kWh [4], [3]. Currently, electric power systems
use (and for many years have used) Volt-VAR (volt-ampere
reactive) Optimization (VVO) to reduce the distribution losses
and increase efficiency as well as to reduce the electricity peak
demand [5], [6]. The primary goal of VVO is to maintain an
acceptable voltage at all points of the distribution system. VVO
is an advanced process which periodically responds to the
operators real-time demand using a two-way communication
network and adjust the voltage regulator and reactive compen-
sation elements for energy delivery. Proper control of capacitor
banks and voltage regulators may yield in reactive power com-
pensation, which improves voltage regulation, power factor,
and quality as well as loss reduction [7].
Few years back, the American Electric Power in Ohio
took several initiatives to increase the efficiency and improve
the service of the electricity delivery system. As part of
the initiative, a Coordinate Volt-VAR Optimization (CVVO)
system is deployed to decrease the amount of energy necessary
to satisfy the customers need with the quality of service [8].
One of the objectives of the CVVO is to reduce energy use and
peak load by operating at the lower end of ANSI C84.1 band-A
standard. Another objective is to adjust the capacitors to keep
the power factor of a substation near unity [8]. In [6], the
author discussed the impact of AMI smart meter, distributed
energy sources, and demand response (DR) on the Integrated
Volt-VAR Optimization (IVVO). This investigation presents a
significant opportunity and benefit to the IVVO and imposes
additional constraint to the energy management system. In [9],
the authors proposed VVCDDR (Volt/VAR Control and Dis-
tributed Demand Response), which is an integrated Volt/VAR
DR control scheme to improve the reliability and efficiency
of the distribution network. The authors modified the original
Integrated Volt/VAR Controller (IVVC) of GridSpice to show
that the IVVC with an eligible DR event can tighten the volt-
age profile and facilitate a more effective voltage conservation.
A coordination scheme for DR and VVC is developed and
simulated on American Electric Power distribution feeders in
[10]. A varying level of DR and VVC for the different types
of loads show that the integration of DR and VVC in real time
can reduce the load and feeder voltage through redistribution.
[11] demonstrated an Evolutionary Algorithm using the Mod-
ified Teaching-Learning-Algorithm to solve scenario based
multiobjective VVC problem in a distribution network which
is powered by various energy sources. In [12], the authors
presented a mechanism to use the bi-directional charges and
V2G function of EVs to compensate the reactive power of
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2017.2648509, IEEE
Transactions on Smart Grid
2

the distribution network. Here, the authors define a three- Transmission (138kV − 768kV )
phase inverter topology together with DC/DC bi-directional Sola
r Pa
nel
converter which has the interface with EV battery.

Substation 1
Sm
art PEV
M
ete
II. P ROBLEM D ESCRIPTION AND C ONTRIBUTION r

It is evident from the above discussion that VVO works by

Primary Distribution (26kV − 69kV )
adjusting the feeders and substation components in response to
the operator’s demand to reduce losses. Further today, the ad- Step Down Transformer
dition of renewable energy sources and electric vehicles (EVs), f1
and home appliances with flexible consumption, featured the OLTC

2.4kV − 46kV
power grid with a new load dimension. Such load dynamism

has become an attractive feature, triggering activities to reduce
peak load and adjust the demand, according to the generation.
Moreover, studies show that many electrical devices operate
more efficiently at reduced voltage [6]. However, beyond
a certain minimum operating voltage, the efficiency of the
device drops. Unfortunately, all the solutions for VVO/VVC
select the lower voltage from ANSI C84.1 standard without
considering the performance of the appliances. With the recent
technology advancement, a cost-effective fine-grain solution
for the VVO could be achieved while maintaining maximum
efficiency of the devices. Moreover, such solution will provide
a quality electricity service without reducing the lifespan of
the equipment but with reduced generation and cost.
Unlike previous work, we investigate the electricity distri-
bution network (shown in Fig. 1) that connects a variety of
renewable and non-renewable energy sources and microgrid
customers with the existing grid. Each of the customers has a
set of appliances (including EVs) which are connected to the
HEMS (home energy management system) and have elastic
or inelastic consumption pattern. Moreover, we assume that
the HEMS has the capability to identify the active (currently
in operation) non-intelligent appliances (e.g., through the
Eigenload approach) [13]. A community energy management
system (CEMS) collects the consumption pattern, the amount
of energy generated from the local renewable sources (from
HEMS), and optimize the electricity use to minimize the
electricity cost. All the distribution feeders of the existing
grid are equipped with capacitor banks, step down transformer
with on-load tap changer (OLTC) and service-drop transformer
(with OLTC) which connects customers (shown in Fig. 1).
We formulate the problem as a non-cooperative game be-
tween the communities connected to the feeders (Fig. 1).
Each of the communities adjusts the consumption pattern and
service drop OLTC tap (to regulate the terminal voltage) to
minimize its electricity cost (or maximizes payoff) according
to the price signal obtained from the utility (known as optimal
energy consumption model, OECM in Fig. 4). As a result,
the utility operator adjusts the capacitor banks and substation
transformers tap and recalculate the cost of electricity, known
as Volt-VAR and CVR (conservative voltage regulation) opti-
mization model (VVCO in Fig. 4). The interplay between these
two schemes results in a non-cooperative mixed strategy game
which will terminate when there is no change in electricity
cost for any change of consumption pattern. This condition is
known as the Nash equilibrium state of the non-cooperative
game. The proposed system presents a fine-grain solution for
Secondary Feeders
r1,1 r1,2 r1,m1
f2
r2,1 r2,2 r2,m2
fn
rn,1 rn,2 rn,mn
Fig. 1. Electricity Transmission and Distribution System
the VVO problem which considers (1) micro-level DR model,
(ii) energy efficiency of the equipment, (iii) roof-top solar
or locally installed energy sources, and (iv) G2V and the
V2G control modes of the electrical vehicles. To evaluate
the performance of the proposed system, we develop and run
simulations on a large transmission/distribution system similar
to the IEEE test T&D network, and compare the results with
the existing VVO (or VVCO – without community DR model,
EVs’ V2G control mode, and local energy sources) model.
III. S YSTEM M ODEL
We consider a distribution network which is shown in Fig.
1, having n feeders : f1 , f2 , · · · , fn . Each of the feeders, i.e.,
fi (1 ≤ i ≤ n) supplies electricity to a set of neighborhood1
microgrids ri,j ({ri,j |1 ≤ j ≤ mi }),∀i and mi is the number
of communities attached to a feeder fi (see Fig. 1). A service-
drop transformer connects the neighborhood to the feeder.
Each of these transformers serves 120/240V with a maximum
apparent power rating from 10KV A to 200KV A. Customers
are assumed to have roof-top solar panels to serve a fraction
of the electricity needed by the neighborhood.
Each of the customers deploys a HEMS, which sends
the current and projected load, energy, and least operating
terminal voltage to the CEMS. Also, the HEMS is capable
of identifying appliances and controlling the consumption of
the equipment with the control information received from
the CEMS. On the other hand, the CEMS (one for each
ri,j ) takes the terminal voltage of the community, adjusts the
voltage regulator (or OLTC) and sends the demand to the
substation energy management system (EMS). Let t and τ
be the time (hour) and total hours of a day, where t ∈ τ .
The EMS at the substation regulates the generation, fixes the
OLTC transformer TAP, adjusts the capacitor bank switches
1A neighborhood is also considered to be a microgrid.

1949-3053 (c) 2016 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 publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2017.2648509, IEEE
Transactions on Smart Grid
3

and calculates the electricity cost µ(Egt ) (total generation at Vst

t is Egt ) and reduces the distribution losses. Next, the EMS
t
Ici,j δv
t X
Ii,j L
sends the electricity cost µ(Egt ) to the CEMS. Upon obtaining i,j

the electricity cost, each of the community improves the

θ θ

consumption pattern and transfer the control to the HEMS. t

Ii,j t R
Ii,j i,j

We assume that HEMS connects the appliances of a cus-

tomer through an HAN (home area network), and a NAN Vdt
(neighborhood area networks) connects the HEMS to the
i,j

Vdt
CEMS. Also, CEMSs are connected to the substation EMS i,j

using a WAN (wide area network) to transfer control and Ri,j XL

i,j
I0 t
t
Ii,j
i,j to (ri,j+1 )
measurements between them. Besides, all other components t
of the grid (PMU, PDC, etc.) are dedicated to monitoring
Ii,j
t t
Ic

ri,j
Xc
and reporting measurements for the stability and the fault- t
vs
i,j i,j t
vi,j

free operation of the electrical network according to the

decision made by the EMS. Based on the mentioned system
architecture, our proposed system is composed of two major
Fig. 2. Distribution Feeder and Phasor Diagram
energy management schemes: (i) VVCO and (ii) OECM which
are shown in Fig. 3 which interact to achieve the minimum
energy generation cost and billing of the customer while where η is a very small positive number near to zero. Next,
satisfying the demand. we assume the PMUs at communities κi,j send the phasor
measurement data such as, voltage (Vi,j t
), current (Ii,j
t
), and
IV. VOLT-VAR AND CVR O PTIMIZATION M ODEL phase (θ). If a PMU is not available for some or most of the
(VVCO) communities, then data from RTU and SCADA (Supervisory
The primary objective of our proposed VVCO system is control and data acquisition) can be used to calculate the
to minimize the electricity generation cost by adjusting the phase. Note that, the received data from PMUs or RTUs may
capacitor bank switches and OLTC transformer TAP (at the have errors which must go through a filtering and screening
substation) to serve the customers demand at time t. The process to increase the accuracy of the measurements to an
objective can be accomplished by minimizing losses along the acceptable level [14]. We assume the power system has such
distribution feeders. capabilities.
Now, by applying the measurements, we can easily evaluate
the terms of the above equation (2) in real time. The power
A. Distribution Losses and Volt-VAR balance equation for the communities κi,j of feeder fi is,
Let the capacitor banks and PMUs (Phasor Measurement q
Unit) be placed on the distribution feeders (i.e., fi ) to compen- t
Pi,j = (PRt i,j )2 + (PXt
L
)2 (4)
i,j
sate for the reactive power losses and obtain the measurements
(voltage, current, and phase). Let, κi,j be the set of commu- where, Pi,j
t
, PRt i,j , PX
t
are the apparent, real, and reactive
Li,j
nities for which a capacitor bank j is installed, where Ri,j power of the feeder at time t for communities κi,j . Let PX t
and XLi,j are the resistance and inductance of the feeder fi at Ci,j
be the energy supplied by the capacitor to compensate the
location j (see Fig. 2). Further, assume that the phase between
reactive power at κi,j then,
voltage and current at j is θi,j t
. Let Ii,j
t
be the amount of
current flowing through the feeder to serve the demand of the t
PX t
= PX (5)
communities κi,j during time slot t. Then, for the communities L i,j C i,j

load, the voltage drop Vdti,j (magnitude) on the feeder fi at j from eqs. (3) and (5), it is evident that the reactive power
can be calculated using the phasor diagram in Fig 2 (above) of the feeder will only be compensated when the capacitive
as, power is equal to the induction (reactive) power, then,
Vdti,j = Ii,j
t t
Ri,j cos θi,j t
+ Ii,j t
XLi,j sin θi,j (1)
Now, assume a shunt capacitor is added to compensate for the XLt i,j = XCt i,j (6)
inductive loss; hence the voltage drop Vdti,j is modified and
Assume the capacitor bank contains a series of identical
equation (1) becomes,
capacitors which can be switched on/off electronically (e.g.,
Vdti,j = Ii,j
t t
Ri,j cos θi,j t
+ Ii,j t
XLi,j sin θi,j using thyristors). Assume the capacitor bank has l capacitors
(2) and st switching state, where, st = 0 means all capacitors are
−Icti,j Xcti,j sin θi,j
t
,
switched off at t, and st = k (1 ≤ k ≤ l) indicates that k of
where Xcti,j is the impedance of the capacitor bank at j on l capacitors are switched on. Now, let the capacitance of each
the feeder fi . The value of Xcti,j can be adjusted to minimize capacitor in the bank at (i, j) be ci,j , and then at t,
the loss, more specifically the loss is reduced when,
st
t
Ii,j XLi,j t
sin θi,j − Icti,j Xcti,j t
sin θi,j = η. (3) XCt i,j = , (7)
2πf ci,j

1949-3053 (c) 2016 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 publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2017.2648509, IEEE
Transactions on Smart Grid
4

where the value of variable st is defined as, 0 ≤ st ≤ l, B. Electricity Cost

f is the frequency of the line (50 or 60Hz) and ci,j is the Let µ(Egt ) be the cost function of the electricity generation
capacitance of each identical capacitor of the bank. at t. We assume that µ(Egt ) is increasing and strictly convex.
To ensure the quality electricity service, besides capacitor Therefore, if Egt > Ẽgt then µ(Egt ) > µ(Ẽgt ) [15] and the cost
bank switching, the substation EMS must ensure that the function is a quadratic function as [16],
voltage of the furthest service-drop of a feeder is greater than
or equal to the minimum primary voltage of the community µ(Egt ) = a(Egt )2 + bEgt + c (12)
(ri,mi ). Assume the OLTC of the distribution feeder (fi )
where a > 0 and b, c ≥ 0 and Egt is the amount of electricity
has ψi taps, indicated as T AP1 , T AP2 , · · · , T APψi . Each
generated by the energy source or generator. The electricity
T APi represent the ratio of primary and secondary windings
cost is solely dependent on the total demand of the customers
of the distribution transformer. Now, let the primary (at the
and various losses throughout the distribution system.
substation) voltage be Vpt and the secondary feeder input
voltage for feeder fi at t be Vst . To ensure uninterrupted
services and quality of power, the following must be satisfied, C. VVCO Mathematical Model
mi
The mathematical model is depicted as follows:
X
Vst = t
Vi,m + Vdti,j , ∀i, (8) τ
X
Objective: min µ(Egt ) (13)
i
j=1
t=1

Vst ≥ Vpt × T APk , 1 ≤ k ≤ ψi (9) subject to, Eq. (2), (3), (7), (8), (9), (11), and (12)
After solving the above VVCO model, the utility determines
where Vi,mt
is the terminal voltage of the last customer on the amount of energy to be generated (Egt ) and the cost µ(Egt ).
i
feeder fi . Now, the energy loss on the distribution system at Note that this was decided according to the input obtained
from the CEMS (such as Ii,j t
), a shown in Fig. 3. The rate
µ(E t )
of the electricity πt , ∀t (πt = E tg ) is now sent back to the
g
Capcitor Control Command Feeders CEMS, so that each community can solve an optimal energy
Banks OLTC consumption model (OECM) to reschedule local loads, decide
the charging/discharging state of EVs as well as the use of
Substation generation & Measurement renewable energy. The objective here is to redistribute the load,
Feeder Controller (EMS) Collector using the current price and send back to the utility for a new
decision from the VVCO model.
VVCO πt
measurements
V. O PTIMAL E NERGY C ONSUMPTION M ODEL (OECM)
Neghborhood Community (ri,j ) adjust
In our proposed model, each community is assumed to
CEMS/Controller (microgrid)
Local
be connected to the distribution feeder through a step-down
Optimization transformer, known as a service drop transformer. The trans-
OECM tap position OLTC
Control Command
HEMS
former has two hot legs (Hot L1 and L2) output voltage
(+120V, -120V) with opposite phase and a center tap known
as neutral. The multi-wire (L1, L2 and neutral) branch circuit
Terminal Reschedule Energy Store/Use
supplies 120/240V for the residential appliances. The hot leg
Voltage Dev. (RES) energy
L1/L2 and neutral are used to supply 120V and for the 240V
appliances are connected to L1 and L2. Some appliances (like
measurements EV, electric motor, etc.) are connected to 240V circuit to get
better performance. Devices such as, light bulb, television,
Fig. 3. VVCO and OECM interactions, Volt-VAR and energy management. microwave oven, dishwasher, dryer, etc. are connected to
the 120V circuit. We assume a voltage regulator connected
t is (using eq. (2)), to 120V circuit to adjust the voltage level (ANSI C84.1
standard between 108 to 120V) according to the requirement
j
mi X
n X
X determined by the CEMS.
t
Eloss = Vdti,j Ii,j
t
, (10)
i=1 j=1 k=1
A. Customer Load
Next, let the terminal (primary) voltage of each of the com- A customer may have elastic appliances (such as an EV, heat
munities at j of feeder fi be Vi,j
t
, then the total amount of water tank, heating/air conditioning system, washing machine,
electricity which needs to be supplied/generated at t is, dishwasher, etc.,) and inelastic appliances (such as light bulbs,
mi
n X electric oven, electric iron, etc.) [17], [18]. Contemporary tech-
nology advances are emerging as one of the growing trends
X
Egt = t
Ii,j t
Vi,j t
+ Eloss , ∀t (11)
i=1 j=1
for home appliances. Evolution of such new devices and IEEE

1949-3053 (c) 2016 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 publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2017.2648509, IEEE
Transactions on Smart Grid
5

802.15.4, ZigBee, are enabling the HEMS with easy access to Now, let the terminal voltage of community ri,j at t be Vi,j
t
;
appliances information and consumption control. Moreover, then,
h
numerous studies have been done and are ongoing to recognize X
the non-intelligent home appliances consumption patterns [19],
t
Vi,j = V k αw,a
k,t
, ∀t, ∀w, ∀a (18)
[13], [20], [21]. Once the operating appliances are detected, the k=1

HEMS can pull the efficiency voltage specification from the where h = |V | and αw,a
k,t
is a binary variable which determines
local database. Next, the HEMS sends the measurement (such the status of the load a of w at t. αw,a
k,t
= 1 indicates that the
as minimum terminal voltage, consumption duration, admit- load is consuming energy at t, otherwise it remains idle. For
tance, etc.) to the CEMS. The CEMS adjusts the consumption energy efficiency and long life operation of the equipment, the
pattern, and the terminal voltage according to the energy cost following constraint should be satisfied,
received from the substation EMS. Moreover, to reduce the t
Vi,j l
≥ vw,a k,t
αw,a , ∀w, ∀a (19)
consumption from the grid, the CEMS may schedule V2G on
EV, and energy use from renewable. Then, the CEMS sends the and the energy consumption of the load a of w at t is
demand (current, voltage) to the substation EMS to improve h
X
the energy cost. t
Ew,a = k
Ew,a k,t
αw,a , ∀t (20)
Let w (w ∈ ri,j ) be a customer, Aw be set appliances k=1
of customer w, and a (a ∈ Aw ) be an appliance. Now, for The VVCO system must ensure that each of the elastic loads
simplicity an elastic and inelastic load of a customer can be must be scheduled to consume electricity for one slot within
represented by a tuple as follows, the duration (Tw,a ) specified in the tuple (eq. (14)). Therefore,
h
l
Lw,a = (V, Ew,a , vw,a , Tw,a ) (14) X X
k,t
αw,a =1 (21)
where V is a voltage set (within ANSI C84.1 utilization k=1 t∈Tw,a
voltage range) with interval of 1 volt, Ew,a is a set of power Therefore, the total electricity consumption of the community
(ZIP power) consumed by a load (appliance) a of customer w ri,j at t can be expressed as,
for an operation voltage of V , vw,a
l
is the minimum operating X X
voltage, below this voltage the efficiency is low [22], [6],
t
Ei,j = t
Ew,a (22)
which may cause a shorter lifespan of the appliances. We w∈ri,j a∈Aw

assume that vw,a

l
is a value between the standard voltage range B. Electric Vehicle (EV)
of ANSI C84.1. Tw,a is the acceptable period of operation such
that an appliance a of customer w may consume electricity Assume each customer has one or a set of electric vehicles
exactly for one slot2 in Tw,a (|Tw,a |> 0. When |Tw,a |= 1, EVw which participate in both V2G and G2V operation. Let
then the load is an inelastic load, otherwise it is an elastic the target charging (energy) of the electric vehicle be Ew,e ,
load. Let Pw,a
k
and Qkw,a be the active and reactive power of and initial energy stored in the EV be Ew,e i
, where e ∈ EVw ,
a load a of w at operating voltage Vk (where Vk ∈ V ). Then the constant charging rate be Cw,e and maximum discharging
for the constant current, power and impedance, the ZIP model rate be Dw,e
max
. Also, let the set of slots of the EV at home be
[23], [24] for the active and reactive power of the load can be Tw,e , and discharging rate at t be dtw,e then,
expressed as, G2V Mode : ctw,e = αw,e
t
Cw,e φc (23)
 
Vk Vk
k
Pw,a = P0 Zp ( )2 + Ip + Pp , ∀Vk ∈ V, (15) V2G Mode : (1 − αw,e
t
)Dw,e ≤ dtw,e φd ≤ 0 (24)
V0 V0
X
and i
Ew,e ≤ Ew,e + t
Ew,e max
≤ Ew,e (25)
  t∈Tw,e
Vk 2 Vk
Qkw,a = Q0 Zq ( ) + Iq + Pq , ∀Vk ∈ V, (16)
V0 V0 t
Ew,e = ctw,e + dtw,e (26)
where P0 and Q0 are the active and reactive power consumed
ti
at nominal voltage V0 . Zp ,Ip and Pp are the ZIP coefficients X
min
Ew,e i
≤ Ew,e + t
Ew,e max
≤ Ew,e ; ∀ti ∈ Tw,e (27)
for active power and Zq ,Iq and Pq are the ZIP coefficients for
t=ts
reactive power [24]. Therefore the magnitude of the consumed
energy by load a of w is, where φc and φd are the charging and discharging efficiency of
an EV. Ew,e
min
and Ew,e
max
are the discharging and charging limit
1
k
Ew,a k
= [(Pw,a )2 + (Qkw,a )2 ] 2 (17) of the EV, and ts is the starting slot in set Tw,e . The term αw,e
t

in equations (23) and (24) is a binary variable which indicates

For each voltage V k ∈ V , the corresponding power consump- that EV e of w is charging its battery when αw,e t
= 1 (and
tion of each a of w can be determined. Thus, a set Ew,a is hence only constraint (23) is a valid one) and otherwise (i.e.,
determined and Ew,a
k
∈ Ew,a corresponds to the operating t
αw,e = 0) is discharging at time slot t [25][26]. Now, for the
voltage V k . community ri,j , EV load at t can be expressed as,
X X
2 Multiple slots operation of the appliance can easily be extended from this t
Ei,j,e = t
Ew,e (28)
model. w∈ri,j e∈EVw

1949-3053 (c) 2016 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 publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2017.2648509, IEEE
Transactions on Smart Grid
6

C. Residential Energy Sources Proof. Let the average consumption in each time slot t be Ēg .
Assume that some customers may have roof-top solar panel Therefore, the cost of electricity at each time slot t is,
or micro wind turbine to meet the partial demand. Let Ew,s
t
be π̄ = a(Ēg )2 + bĒg + c (34)
the electricity produced by the renewable source of customer
w at t and the maximum capacity be RESw . Then, then, the total cost for the whole time span is,
t
0 ≤ Ew,s ≤ RESw , (29) πT = π̄T = (a(Ēg )2 + bĒg + c) · T (35)
and for the community, the total renewable energy at t is, Now, let a player shift its consumption (smallest possible) from
0
t
Ei,j,s =
X
t
Ew,s (30) time slot t to t , thus the utility need to decrease generation
0

w∈ri,j
∆Eg at t and increase ∆Eg (where ∆Eg > 0) at t , then the
0
total electricity πT is,
D. OECM Mathematical Model 0
πT = (a(Ēg )2 + bĒg + c) · (T − 2) + a(Ēg + ∆Eg )2
To minimize the energy price for a duration T , each of the
+b(Ēg + ∆Eg ) + c) + a(Ēg − ∆Eg )2
communities must solve the following model, (36)
+b(Ēg − ∆Eg ) + c)
= (a(Ēg )2 + bĒg + c) · T + 2a(∆Eg )2
X
Objective: min t
πt (Ei,j t
+ Ei,j,e t
− Ei,j,s )
(31)
t ∈T > πT
Subject To: eqs. from (17) to (30)
Therefore, the payoff defined in eq. (32) will decrease for the
where πt is the unit cost of the electricity at t which is deviated player. A higher deviation further will decrease the
received from the EMS of the substation. We assume that each payoff of the involved players. Thus, the optimal solution of
community always consumes a certain amount of electricity the OECM game can only be obtained by setting the consump-
from the grid. Therefore, Ei,j
t t
+ Ei,j,e t
− Ei,j,s > 0. tion strategies such that the overall consumption in each slot t
has a minimum deviation from the average consumption.
VI. M IXED S TRATEGY G AME : I TERACTION BETWEEN
VVCO AND OECM Therefore, the target of each community is to achieve an
electricity cost near the cost expressed in the above Eq.
The problem presented in Fig. 4 can be modelled using a
(33). This will result in minimum changes in OLTC trans-
non-cooperative mixed strategy game. Let the payoff of each
former TAP and capacitance of the capacitor bank, flatten
community (player) be β(σi,j , [πt |∀t]) and σi,j be a set of
the generation curve, and increases the life of the distribution
actions (consumption or discharging) for each of the elastic
system components. In practice, however, the load may not be
(including EV) and inelastic load of the community ri,j in
uniformly distributed or might not be achieved by rescheduling
response to the electricity price πt ; ∀t ∈ T . The payoff of the
the variable loads and the discharging of EVs. Without loss of
community is the negative of the cost of total electricity drawn
generality, we assume that the total consumption of the system
from the distribution feeder fi . Thus (from eq. (31)),
X is the same. Therefore, for optimal scheduling of load,
β(σi,j , [πt |∀t]) = max t
πt (Ei,j,s t
− Ei,j t
− Ei,j,e ) (32) T
X T
X
t∈T πt∗ Egt ≥ π̄ E¯gt (37)
Unfortunately, the payoff in Eq. (32) does not guarantee t=1 t=1
a Nash Equilibrium state of the system. At each iteration, where πt∗ is the optimal electricity costs at any t, and the cost
upon receiving the cost, the game forces the community to πt∗ has the minimum average distance from π̄. Therefore, we
reschedule the consumption to those time slots which have redefine the payoff function as,
lower electricity costs. Thus, in the following iteration, the cost
T
of lower slots in the previous iteration will increase, and the X
β(σi,j , [πt |∀t]) = max{π̄ L¯i,j T − t
Si,j } (38)
expenses of the other slots will decrease abruptly. Therefore,
t=1
the payoff of each community may not increase. In fact, the
payoff of a player is dependent on the strategies played by Subject to:
other communities, which are unknown. To design a good
t
Si,j ≥ π̄ L¯i,j − πt Li,j
t
, (39)
payoff function, let us consider the ideal scenario where the
average load of each client is nearly the same, and the loads t
Si,j ≥ πt Li,j
t
− π̄ L¯i,j ; Si,j
t
≥ 0, (40)
are uniformly distributed throughout the hours of a day. In this
ideal scenario, the average cost can be expressed as, Eqs. (17) to (30) where Li,j
t t
= Ei,j t
+ Ei,j,e t
− Ei,j,s , and
T
Li,j
t
P
T
Egt constant L¯i,j = t=1T is the average load. Let the optimal
P

π̄ = µ(E¯gt ), where, E¯gt =

, (33) t=1 payoff of the community be β ∗ (σi,j
∗
, [πt |∀t]) then,
T
T
Lemma 1. Any deviation from the average cost will decrease X
β ∗ (σi,j
∗
, [πt |∀t]) = max{π̄ L¯i,j T − t∗
Si,j } (41)
the payoff of the players of the game VVCO/OECM. t=1

1949-3053 (c) 2016 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 publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2017.2648509, IEEE
Transactions on Smart Grid
7

This is possible when each player plays its best strategy, and the substation controller will send the switching command to
all other strategies will result in less payoff for some or all of all the capacitor bank and set the OLTC tap of the feeder at
the communities. This is known as Nash equilibrium of the the substation. Each of the communities also adjusts voltage
mixed strategy game. The game is a non-cooperative multi- regulator or OLTC according to the terminal voltage selected
player mixed strategy game because the player may draw by the optimization process.
the non-discrete amount of electricity from the feeders, and
discharge continuous (non-discrete) amount of electricity from A. Uncertainty of the community load
the EVs and consume energy from RES. Other than these,
Load variation and uncertainty are common characteristics
the elastic loads have discrete strategies. In each iteration,
for the existing electrical network as well as for the VVO
the communities play the best strategy to minimize (OECM)
system [27], [28], [29]. Indeed, one of the objectives of a
its consumption cost and send the demand to the EMS,
VVO system is to minimize the impact of the uncertainty of
which runs the VVCO scheme and returns the electricity unit
the connected load. The introduction of EVs will however
costs to the community OECM scheme. The interaction will
represent new source of load uncertainty which will pose a
continue until the game ends in a Nash equilibrium state. The
challenge for the centralized control and optimal operation of
community CEMS will act as the player of the game while
the grid[30]. Such uncertainties are the outcome of randomness
the substation serves as the controller of the game.
in factors such as the number of PEVs connected to the grid
Generation Costs πt , ∀t
for charging, the arrival and departure times of PEVs, and
the initial battery State-of-Charge (SoC)[30]. Most existing
solutions consider and try to minimize the forecast errors
Start Game
Controller EMS [27], [28], [29] which is also applicable to our proposed
system. Moreover, we find that the correct response to the
Initialize send CEMS: Receive generation uncertainty of load is to mitigate and adjust the load locally
πt = ∞, ∀t to CEMS cost πt , ∀t & Start
(i.e., at the community level) and minimize the effect on the
electrical network in real time. We should recall here that
Rec. & Update Measurement:
t , V t , θt
Ii,j i,j i,j Is Load/RES No CEMS: Send Load &
our proposed VVCO/OECM is a decentralized method and
∀t∀i∀j Updated ∀t? RES Req. to HEMS
the response time of the system is quite low (e.g., less than
15sec); hence, the community load and the corresponding
Determine New Yes
t & π , ∀t by solving
Eg t
CEMS: Determine
capacitor bank can be adjusted more accurately just before
VVCO model in Sec. IV-C
β ∗ (σi,j
∗ [π |∀t]);
t
optimize OECM model;
HEMS: Send Load &
RES Gen to CEMS
the actual consumption. Further, to reduce the uncertainty,
with Obj in eq. (41) the VVCO/OECM can be run in each operational time slot
Yes Is πt , t ∈ T HEMS: Forecast (just before the time of actual consumption) without any
changed? Load & RES Gen.
modification. A shorter time slot (such as 1 to 5 minutes) will
send measurement

Is payoff
No
Increased? No further decrease the uncertainty to near zero. Alternatively,
Adjust Capacitors & TAP Yes
we believe a local storage (community grade) system may
of Substation OLTC at t Send Old Measurement
t , V t , θt
Ii,j i,j i,j be used to mitigate the change (instantaneous) of load due
Update & Send
to the unpredictable behaviour of users. The charging and
∀t∀i∀j
New Measurement
t , V t , θt
Ii,j
Stop
i,j
∀t∀i∀j
i,j
discharging of the local storage system are similar to that of
EVs without the constraint which is presented by equation
Game Controller Game Player
(25). In this case, the storage system may consume extra
energy or discharge stored energy for the instantaneous change
in the community load.
Fig. 4. Volt-VAR optimization : VVCO/OECM Game.

VII. N UMERICAL E VALUTION

Players: Each CEMS is the player of the game. In each
slot, with a finite number of iterations, community ri,j plays A. Simulation Setup
the game (eq. (38)) to maximize its payoff. The player will We consider a distribution system with several feeders, each
change its strategy σ to maximize the payoff for electricity of which connects 50 to 100 communities by inexpensive
cost (πt ) received from EMS. The player will continue its OLTC transformers. Also, an OLTC transformer connects the
0
play until the payoff reaches its maximum. Let πt , ∀t ∈ T distribution line to the substation. Each community has 20 to
be the one-time cost of the electricity. Then steps (OECM) of 25 residential homes with or without the renewable sources,
the play are shown in Fig. 4. one or more EVs, 5 to 15 inelastic, and elastic loads. The
Game Controller: The substation EMS will act as the consumption corresponds to terminal voltages ranging from
controller of the game. Upon receiving the demand from 108 to 127 (with skip 1 volt) volts and are calculated using
the communities, the EMS calculates the electricity costs the ZIP model (eq. (17)) for each elastic and inelastic load. To
πt , ∀t ∈ T using eq. (13) for the amount of electricity calculate the load, we use the values of ZIP coefficients given
Egt (eq. (11)) and send it to the CEMS. The interactions of in [24]. The renewable sources have the maximum capacity to
EMS (VVCO) with the CEMS (OECM) are shown in Fig. produce 3 to 10 kW electricity. Here, we use ARMA prediction
4. Once the game reached to a Nash equilibrium state, then method to predict the amount of generation from the renewable

1949-3053 (c) 2016 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 publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2017.2648509, IEEE
Transactions on Smart Grid
8

sources for the next 24 hours. In our simulation, the residential

0.8
Electricity Generation Costs
load was chosen between 10 to 20 kWh for a day [4]. We

Electricity Price ($/kWh)

VVCO/OECM
assume that each home has a level 2 charger to charge the EV. VVCO

0.6
The battery capacity of EV is chosen randomly from 18, 24,
60, 70, and 85kWh with charging and discharging efficiency

0.4
80%-95%. For the simulation, we place the capacitor banks
with a maximum of 600 MVAR for 10, 20, 30, 40, or 50

0.2
communities and the impedances of the distribution lines (with
33KV base voltage) are considered which is given in [31].

We implement the VVCO/OECM model in OpenMPI C++
and IBM CPLEX where each process represents a community
and act as the CEMS, and one of the processes serve as the
substation EMS. Each time a process determine its optimal
consumption strategy (OECM) and send the measurements to
the EMS, then the EMS calculates the electricity cost and
sends it to the CEMS. The CEMS accepts the current strategy
if the cost of electricity consumption is less than the previous
cost or increases the payoff (Eq. (38)).
B. Numerical Results
We run the simulation for various configuration of the
distribution system which contains 10, 15, ..., or 60 distribution
feeders or lines (see Fig. 5). We repeat the simulation 50 times
for each distribution system configuration.
12.0
0.0
1.00 pm 5.00 pm 9.00 pm 1.00 am 5.00 am 9.00 am 1.00 pm
Hours in a day
Fig. 6. Compare electricity generation costs determined by VVCO and
VVCO/OECM
and reactive compensation due to the change of consumption
pattern. Fortunately, the process took a constant time and can
process the measurements sent by the several CEMSs.
Fig. 6 shows the electricity generation costs ($/kWh) de-
termined by the VVCO and the proposed VVCO/OECM. We
ran the simulation for all the configuration listed in Fig. 5.
In all the cases, we found that the proposed system flattened
the demand curve as well as the electricity generation cost
of the distribution system. Here, we only present electricity
costs obtained for the distribution system with 50 feeders. It is

also found that the proposed VVCO/OECM results in almost

Execution Time (in seconds)

Feeders Communities Homes

10 763 17156
15 1146 25766 same power generation cost throughout the hours of a day.
11.8

20 1470 33030
25 1875 42194
30 2293 51609
35 2661 59853
On the other hand, the conventional Volt-VAR optimization
system does not flatten the load, and therefore the production
11.6

40 2958 66636
45 3353 75406
varies between the hours of a day. One of the primary goals
50 3774 84857
55 4174 93929
11.4

60 4464 100438
of the proposed system is to flatten the load curve along with
the VVO to minimize the variation of the amount generation
11.2

which is also a prime target for the utility companies. The

VVCO/OECM flatten the load curve by optimally shifting
11.0

10 20 30 40 50 60 the elastic load, determining the optimal EV charging and

Number of feeders discharging strategy, and using (consume or store) renewable
energy. Therefore, the burden in the peak hours is distributed
Fig. 5. Execution time for VVCO-OECM converge to optimal results. to the off-peak hours.

The VVCO/OECM model is a decentralized solution with

34

a controller (substation EMS) which will act as an IESO ( Electricity Generation Cost Saving
Independent Electricity System Operator) to determine and VVCO/OECM Vs VVCO
Cost Savings (%)

control the generation and costs according to the demand.

33

Fig. 5 shows the total time to obtain the optimal results

(peak load shaving, capacitor bank setting, OLTC transformer
32

TAP setting, and evaluating optimal costs) for a distribution

system. Each of the solution process OECM: (i) chooses the
31

optimal consumption strategy (ii) send the message to the

controller process (VVCO) and (iii) receive the electricity
30

cost. The interaction finally ends by determining the optimal 10 20 30 40 50 60

power generation costs and peak demand shaving with a few
iterations (5 to 10). The system took a few seconds (10 to 15)
to converge to an optimal solution irrespective of the size of
the network. This is because CEMS independently play the
VVCO/OECM game to achieve the maximum payoff regard-
less of the solution determined by other processes (CEMSs).
The main burden of the system is the calculation of the costs
Number of Feeders
Fig. 7. Generation cost saving by the proposed VVCO/OECM model.
Fig. 7 shows the percentage of electricity generation cost
saving by the VVCO/OECM compared to the VVCO. The
results tell us that the proposed system not only flatten the

1949-3053 (c) 2016 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 publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2017.2648509, IEEE
Transactions on Smart Grid
9

Substation Secodary Volatge (kV)

demand and cost curves but also reduces the total production

35.0
Substation Secondary Voltage (in kV)
cost compared to the conventional VVO system. We ran VVCO
VVCO/OECM
the simulation for each of the configuration (10, 15, 20,

34.5
Base

..., 60 feeders) by placing compensation capacitors for 10,

34.0
20, ..., 50 communities. In every case, the total generation
cost of the proposed VVCO/OECM system is lower than

33.5
the conventional VVO system which compensates reactive
power without flattening the demand curve. Also, the proposed

33.0
system uses the renewable sources, EV in V2G mode and

32.5
shifting the load which reduces the current flow through the
1.00 pm 5.00 pm 9.00 pm 1.00 am 5.00 am 9.00 am 1.00 pm
distribution line; therefore, the amount of the reactive power
Hours in a day
is reduced. It is also found that the percentage of cost saving
slightly increases with the increasing number of distribution
Fig. 9. Substation Secondary Voltage.
lines (and the communities). The increase of saving is evident
due to the more options available to the proposed system to
shift the load from peak demand duration to off-peak demand 7:00 pm - 11:00 pm) may cause extra energy loss due to the
duration and energy from off-peak to peak demand duration. generation of heat, and may reduce the life of the distribution
line. Moreover, this significant variation of voltage in VVCO
200

may increase the cost of the distribution system equipment

Reactive Compensation
VVCO such as OLTC transformer, breakers, the shunt capacitor, etc.
Standard Deviation

VVCO/OECM
150

2.0
Voltage Drop
(Without load shifting and V2G)
100

With Renewables
1.5
Voltage Drop (in kV)

Without Renewables
50

1.0
0

10 15 20 25 30 35 40 45 50 55 60
0.5

No of Feeders
0.0

Fig. 8. Reactive power compensation. 6 pm 10 pm 2 am 6 am 10 am 2 pm 6 pm

Fig. 8 shows the standard deviation of required capacity of
the capacitor (or compensation device) throughout the day to
minimize the reactive power loss of the distribution system.
In the case of conventional VVO (here VVCO) system, the
deviation between the required capacity of the capacitor varies
over time which is very significant than the variation necessary
for the proposed method. The changes in the proposed method
are minuscule which indicates that a fixed capacitor might
be sufficient (with the small presence of the reactive power)
for the distribution system or capacitor with an inexpensive
switching system is required when the VVCO/OECM is
used. On the other hand, in the case of conventional VVO
system, a significant variation of the capacitance will need
an expensive switching framework. Moreover, the deployment
of the VVCO/OECM system will significantly increase the
lifetime of the capacitor with the mechanical switch.
In Fig. 9 the amount of secondary voltage needed for the
distribution feeders after the compensation of the reactive loss,
are shown for VVCO and proposed VVCO/OECM scheme.
The base voltage of the distribution system is 33KV. The
voltage drop varies in case of VVCO system: increases when
the demand is high and decreases when demand is low. The
voltage drop is mostly due to the resistance of the distribution
feeders, and it will increase or decrease on the increase or
decrease of the demand. The high voltage drop (VVCO at
Hours in a Day
Fig. 10. Use of renewable energy and its effect on the voltage drop.
Fig. 10 shows the significant impact of voltage drop in the
distribution line when customers are using energy from the
renewable sources. It is obvious that voltage drop decreases
when the communities are using renewable energy. This is
because the renewable energy reduces the cumulative load and
therefore reduces the reactive and resistive loss of the system.
Hence, the generator needs to generate less power. Moreover,
the use of renewable energy and EV with V2G mode may add
an extra benefit to reduce the voltage drop in the peak hours.
This is because, in V2G mode, the EV can carry electricity
from the generation period to peak consumption period when
the production of renewable sources are small.
VIII. CONCLUSION
We have proposed a new model (VVCO/OECM) for VVO,
which acknowledges the current technological advancement
of the power grid, evolution of the smart grid, and EVs.
We found that the proposed model and its game-theoretic
solution could solve the VVO problem optimally. The existing
VVO method only solves the problem with the coarse demand
received from the customers. In our solution, we develop an
interactive method which enables the utility to communicate

1949-3053 (c) 2016 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 publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2017.2648509, IEEE
Transactions on Smart Grid
with the customer to flatten the load and ultimately, reduces the
reactive loss and flatten the generation. Thus, VVCO/OECM
can reduce the overall generation cost of the electricity. The
proposed system is scalable and can be implemented for
any size electricity system infrastructure. The system can
be deployed without the intervention of the substation by
calculating the cost at the CEMS premises and adjust the
compensation devices by the local demand.
R EFERENCES
[1] H. Farhagi, “Path of the smart grid,” IEEE Power and Energy Magazine,
vol. 8, no. 1, pp. 18–28, Jan./Feb. 2010.
[2] U.S. Energy Information Administration (EIA), “Annual energy outlook
2015,” April 2015. [Online]. Available: http://www.eia.gov/forecasts/
aeo/pdf/0383(2015).pdf
[3] X. Feng, W. Peterson, F. Yang, G. M. Wickramasekara, and J. Finney,
“Smarter grids are more efficient: Voltage and var optimization
reduces energy losses and peak demands,” ABB, Review Report
9AKK104295D7260, Mar 2009.
[4] U.S. Energy Information Administration (EIA), “Electric power annual
2012,” Dec 2013.
[5] T. Gönen, Electric power distribution system engineering, ser. McGraw-
Hill Series in Electrical Engineering.
[6] N. Markushevich, “The benefits and challenges of the integrated volt/var
optimization in the smart grid environment,” in Power and Energy
Society General Meeting, 2011 IEEE, July 2011, pp. 1–8.
[7] M. Masoum, M. Ladjevardi, A. Jafarian, and E. Fuchs, “Optimal place-
ment, replacement and sizing of capacitor banks in distorted distribution
networks by genetic algorithms,” Power Delivery, IEEE Transactions on,
vol. 19, no. 4, pp. 1794–1801, Oct 2004.
[8] K. Schneider and T. Weaver, “Volt-var optimization on american electric
power feeders in northeast columbus,” in IEEE PES T&D Conference
and Exposition, May 2012, pp. 1–8.
[9] K. Anderson and A. Narayan, “Simulating integrated volt/var control
and distributed demand response using gridspice,” in SGMS, 2011 IEEE
First International Workshop on, Oct 2011, pp. 84–89.
[10] J. Solanki, N. Venkatesan, and S. Solanki, “Coordination of demand
response and volt/var control algorithm using multi agent system,” in
IEEE PES T&D Conference and Exposition, May 2012, pp. 1–4.
[11] T. Niknam, M. Zare, and J. Aghaei, “Scenario-based multiobjective
volt/var control in distribution networks including renewable energy
sources,” Power Delivery, IEEE Transactions on, vol. 27, no. 4, pp.
2004–2019, Oct 2012.
[12] J. Gallardo-Lozano, E. Romero-Cadaval, V. Minambres-Marcos, D. Vin-
nikov, T. Jalakas, and H. Hoimoja, “Grid reactive power compensation
by using electric vehicles,” in Electric Power Quality and Supply
Reliability Conference (PQ), 2014, June 2014, pp. 19–24.
[13] H. Ahmadi and J. Marti, “Load decomposition at smart meters level
using eigenloads approach,” Power Systems, IEEE Transactions on,
vol. PP, no. 99, pp. 1–12, 2015.
[14] A. P. S. Meliopoulos, G. J. Cokkinides, F. Galvan, B. Fardanesh, and
P. Myrda, “Delivering accurate and timely data to all,” IEEE Power and
Energy Magazine, vol. 5, no. 3, pp. 74–86, May 2007.
[15] M. Greer, Electricity marginal cost pricing : applications in elic-
iting demand responses. Amsterdam ; Boston : Butterworth-
Heinemann/Elsevier, 2012.
[16] A.-H. Mohsenian-Rad, V. Wong, J. Jatskevich, R. Schober, and
A. Leon-Garcia, “Autonomous demand-side management based on
game-theoretic energy consumption scheduling for the future smart
grid,” Smart Grid, IEEE Transactions on, vol. 1, no. 3, 2010.
[17] M. Tushar, C. Assi, and M. Maier, “Distributed real-time electricity
allocation mechanism for large residential microgrid,” Smart Grid, IEEE
Transactions on, vol. 6, no. 3, pp. 1353–1363, May 2015.
[18] M. Tushar, C. Assi, M. Maier, and M. Uddin, “Smart microgrids:
Optimal joint scheduling for electric vehicles and home appliances,”
Smart Grid, IEEE Transactions on, vol. 5, no. 1, pp. 239–250, Jan 2014.
[19] S.-Y. Chen, C.-F. Lai, Y.-M. Huang, and Y.-L. Jeng, “Intelligent home-
appliance recognition over iot cloud network,” in IWCMC, 9th Interna-
tional, July 2013, pp. 639–643.
[20] C.-F. Lai, R.-H. Hwang, H.-C. Chao, and Y.-H. Lai, “A dynamic power
features selection method for multi-appliance recognition on cloud-based
smart grid,” in Computational Science and Engineering (CSE), 2014
IEEE 17th International Conference on, Dec 2014, pp. 780–785.
1949-3053 (c) 2016 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.
10
[21] A. Ruzzelli, C. Nicolas, A. Schoofs, and G. O’Hare, “Real-time recog-
nition and profiling of appliances through a single electricity sensor,”
in SECON, 7th Annual IEEE Communications Society Conference on,
June 2010, pp. 1–9.
[22] B. Uluski. (2011) Volt/var control and optimization concepts and
issues. Electric Power Research Institute (EPRI). [Online]. Available:
http://cialab.ee.washington.edu/nwess/2012/talks/uluski.pdf
[23] L. Toledo, M. Mota, and A. A. Mota, “Load modeling at
electric power distribution substations using dynamic load parameters
estimation,” International Journal of Electrical Power & Energy
Systems, vol. 26, no. 10, pp. 805 – 811, 2004. [Online]. Available:
http://www.sciencedirect.com/science/article/pii/S0142061504000924
[24] A. Bokhari, A. Alkan, R. Dogan, M. Diaz-Aguil, F. de Len,
D. Czarkowski, Z. Zabar, L. Birenbaum, A. Noel, and R. E. Uosef,
“Experimental determination of the zip coefficients for modern residen-
tial, commercial, and industrial loads,” IEEE Transactions on Power
Delivery, vol. 29, no. 3, pp. 1372–1381, June 2014.
[25] Z. Li, Q. Guo, H. Sun, and J. Wang, “Storage-like devices in load
leveling: Complementarity constraints and a new and exact relaxation
method,” Applied Energy, vol. 151, pp. 13–22, 2015.
[26] ——, “Sufficient conditions for exact relaxation of complementarity
constraints for storage-concerned economic dispatch,” IEEE Transac-
tions on Power Systems, vol. 31, no. 2, pp. 1653–1654, 2016.
[27] S. Rahimi, S. Massucco, F. Silvestro, M. R. Hesamzadeh, and Y. Tohidi,
“Volt/var optimization function with load uncertainty for planning of mv
distribution networks,” in 2015 IEEE Eindhoven PowerTech, June 2015,
pp. 1–6.
[28] S. Rahimi, K. Zhu, S. Massucco, and F. Silvestro, “Stochastic volt-var
optimization function for planning of mv distribution networks,” in 2015
IEEE Power Energy Society General Meeting, July 2015, pp. 1–5.
[29] Z. Wang, J. Wang, B. Chen, M. M. Begovic, and Y. He, “Mpc-
based voltage/var optimization for distribution circuits with distributed
generators and exponential load models,” IEEE Transactions on Smart
Grid, vol. 5, no. 5, pp. 2412–2420, Sept 2014.
[30] Mehboob, Nafeesa, “Smart charging of plug-in electric vehicles
in distribution systems considering uncertainties,” 2016. [Online].
Available: http://hdl.handle.net/10012/10445
[31] G. Electric, Distribution Data Book. General Electric Schenectady,
1943. [Online]. Available: http://www.electricalmanuals.net/files/
APP-MANUALS/GE/GET-1008L.pdf
Mosaddek Hossain Kamal Tushar Mr Tushar
received his BSc degree in Applied Physics and
Electronics and his MSc degree in Computer Science
from Dhaka University, Bangladesh. He also re-
ceived his Master in Information Technology (MIT)
from University of New South Wales, Sydney, Aus-
tralia, in 2006. Currently, he is pursuing his PhD
degree in Electrical and Computer Engineering at
Concordia University, Montreal, Canada. Before be-
ginning his PhD studies he was a faculty member of
the Department of Computer Science & Engineering
at University of Dhaka, Bangladesh, from 1997 to 2011. He is a graduate
student member of the IEEE.
Chadi Assi Prof. Assi received his Ph.D. degree
from the City University of New York (CUNY) in
2003. He is currently a full professor at Concor-
dia University, where he also holds a University
Research Chair. He received the prestigious Mina
Rees Dissertation Award from CUNY in 2002 for his
research on wavelength-division multiplexing optical
networks. He is on the Editorial Board of IEEE
Communications Surveys & Tutorials, IEEE Trans-
actions on Communications, and IEEE Transactions
on Vehicular Technologies. His current research in-
terests are in the areas of network design and optimization, network modelling
and reliability, and information security. Dr. Assi is a senior member of the
IEEE.