Content uploaded by Saqib Kazmi
Author content
All content in this area was uploaded by Saqib Kazmi on May 23, 2022
Content may be subject to copyright.
Available via license: CC BY-NC-ND 4.0
Content may be subject to copyright.
SPECIAL SECTION ON INTELLIGENT SYSTEMS
FOR THE INTERNET OF THINGS
Received September 12, 2017, accepted October 4, 2017, date of publication October 17, 2017, date of current version March 7, 2019.
Digital Object Identifier 10.1109/ACCESS.2017.2763624
Towards Optimization of Metaheuristic
Algorithms for IoT Enabled Smart
Homes Targeting Balanced Demand
and Supply of Energy
SAQIB KAZMI1, NADEEM JAVAID 1, (Senior Member, IEEE), MUHAMMAD JUNAID MUGHAL1,
MARIAM AKBAR1, (Member, IEEE), SYED HASSAN AHMED 2, Member, IEEE),
AND NABIL ALRAJEH3
1COMSATS Institute of Information Technology, Islamabad 44000, Pakistan
2School of Computer Science and Engineering, Kyungpook National University, Daegu, 41566, South Korea
3College of Applied Medical Sciences, King Saud University, Riyadh 11633, Saudi Arabia
Corresponding author: Nadeem Javaid (nadeemjavaidqau@gmail.com)
This work was supported by the Deanship of Scientific Research, King Saud University, under Grant RG-1435-0037.
ABSTRACT Internet of Things enabled smart grid (SG) is one of the most advanced technologies,
which plays a key role in maintaining a balance between demand and supply by implementing demand
response (DR) program. In SG, the main focus of the researchers is on home energy management (HEM)
system, which is called demand side management. Appliance scheduling is an integral part of HEM system as
it manages energy demand according to supply, by automatically controlling the appliances and shifting the
load from peak to off peak hours. In this paper, the comparative performance of HEM controller embedded
with heuristic algorithms, such as harmony search algorithm, enhanced differential evolution, and harmony
search differential evolution, is evaluated. The integration of renewable energy source (RES) in SG makes
the performance of HEM system more efficient. The electricity consumption in peak hours usually creates
peaks and increases the cost but integration of RES makes the electricity consumer able to use the appliances
in the peak hours. We formulate our problem using multiple knapsack theory that the maximum capacity of
the consumer of electricity must be in the range, which is bearable for consumer with respect to electricity
bill. Feasible regions are computed to validate the formulated problem. Finally, simulation of the proposed
techniques is conducted in MATLAB to validate the performance of proposed scheduling algorithms in terms
of cost, peak-to-average ratio, and waiting time minimization.
INDEX TERMS Smart grid, knapsack, enhanced differential evolution, harmony search algorithm, home
energy management system, demand side management.
NOMENCLATURE
Variables and Subscripts Description
tTime interval
UEnergy consumed by
fixed appliances
VEnergy consumed by flexible
appliances
WEnergy consumed by
un-interruptible appliances
ρPower ratings of appliances
Fap Set of fixed appliances
Sap Set of shift-able. appliances
UIap Set of un-interruptible appliances
γON/OFF status of appliances
P(t) Total power consumption of shift-able., fixed and
un-interruptible appliances
C(t) Total electricity bill
λPricing signal (RTP)
EPV Energy generated from RES
ηEfficiency of solar inverter
IrSolar irradiance
APV Area of solar panel
VOLUME 7, 2019
2169-3536 2017 IEEE. Translations and content mining are permitted for academic research only.
Personal use is also permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
24267
S. Kazmi et al.: Toward Optimization of Metaheuristic Algorithms for IoT Enabled Smart Homes
TaOutdoor temperature
CMaximum allowed power consumption
xij Initial population
xlLowest value in population matrix
xuHighest value in population matrix
UjTrial vector
VjMutant vector
FMutant factor
CR Crossover rate
xrl Target vector of EDE
τsch Operation start time of appliances
τot Operation end time of appliances
Hij Harmony memory
HlLowest value in HM
HuUpper limit of values in HM
x0
iNew harmony vector
Vr1Target vector of HSDE
Hji Trial vector of HSDE
NI Maximum no. of iteration
I. INTRODUCTION
Energy demand around the world is increasing day by day.
To fulfil this demand the existing generation is facing a lot of
challenges. It is estimated that total energy demand at the end
of 2020 will increase by 75% as compared to energy demand
in 2000 [1]. This increase may force utilities to reshape
electricity generation and distribution in order to avoid
demanding energy challenges. In order to fulfil energy
demand, different means of electricity generation like renew-
able and sustainable energy resources (RSER) are introduced
in the power system. The integration of renewable energy
sources (RESs) in the existing traditional grid increases
system complexities and dynamics [2]. Smart grid (SG)
is another advancement in the field of power system.
In SG, advanced information and communication technolo-
gies (ICTs) are introduced in traditional grids which provide
customers the ability to interact with utility [3]. Advanced
metering infrastructure (AMI) equips each home with smart
meter that gathers energy demand information from cus-
tomer and uploads to the utility server. SG allows inte-
gration of renewable and distributed energy generation to
diminish the effects of CO2on environment and to optimize
energy consumption. On the other hand demand side manage-
ment (DSM) is a very important aspect of SG that efficiently
manages the energy demand of end users by enabling the
exchange of information between utility and consumers. The
EDE parameters are summarized in Table 2. These programs
aim at improving grids stability by reducing peak forma-
tion [4]. So, utilities and customers can manage the energy
generation and consumption through the implementation of
DSM programs by providing incentives or encouraging the
customers to participate in energy management programs.
The two-way flow of information and energy in SG keeps the
electricity consumers informed about the pricing rates, load
on utility, load shedding schedules and any type of equipment
failure due to any natural or crew cause. It enables the utility
company to monitor and analyze the real time information of
consumers so that responsive actions may be taken according
to utility/end-users demand. End users can take economic
benefits by shifting peak load to off-peak hours using differ-
ent optimization techniques.
In a traditional power system, utility companies manually
shed the selected load of consumers during peak hours to
make their operations safe. On the other hand, load shifting
in SG is accomplished in peak hours to avoid peak forma-
tion. This not only benefits the users but also the utility
company. This strategy in turn increases the reliability of
the grid. Although shifting the load from high peak to low
peak hours reduces the peak load and electricity cost, but
it disturbs user comfort. So, there is a trade-off between
user comfort and cost saving, which cannot be achieved at
the same time. Thus, peak to average ratio (PAR), energy
price signals, daily energy consumption and user comfort are
the constraints needed to be considered. Beside considering
end users desires and needs, utility companies also provide
incentives to motivate consumers to reschedule their load to
mitigate high peaks. Such challenges have motivated the need
for intelligent energy management algorithms that can handle
all types of loads and responds to price variations.
To address the aforementioned challenges, this paper
presents energy management algorithms to schedule
household appliances while meeting the constraints. The
algorithms used in this paper are harmony search algo-
rithm (HSA), enhanced differential evolution (EDE)
algorithm and our proposed hybrid technique; harmony
search differential evolution (HSDE) algorithm. The behavior
of the system embedded with these algorithms is smart
and user satisfactory with respect to billing, PAR, energy
consumption and the waiting time.
The rest of the paper is organized as follows: Section II
discusses the related work. Section III elaborates pro-
posed system model and problem formulation is provided
in Section IV. Proposed schemes are discussed in Section V.
Feasible regions are briefly discussed in Section VI.
In Section VII performance evaluation through simulation
of proposed algorithms is elaborated in detail. Section VIII
concludes this paper.
II. LITERATURE REVIEW
Researchers around the world work to optimally schedule
the appliances to benefit the consumers. Most of the electric
utility companies are investigating and implementing SG to
make the existing power system advanced, reliable, self-
healing and economical. The use of sensors, communication
and computational ability and controlling characteristics in
SG enhance the overall operation of electric power transmis-
sion system [5]. PAR, daily energy consumption, electricity
cost and the hourly energy consumption of shift-able. and
throttle-able appliances of the consumers are the constraints
and the objective function in [6], all of which are to be mini-
mized. This is accomplished by shifting the high energy con-
suming loads to off peak hours which helps to minimize the
24268 VOLUME 7, 2019
S. Kazmi et al.: Toward Optimization of Metaheuristic Algorithms for IoT Enabled Smart Homes
energy consumption in the peak hours. The authors formulate
the initial optimization problem, to minimize the energy cost
of the consumers by determining of the optimal usage of
the power and operational time of throttle-able and shift-
able. appliances. The authors use the distributed algorithms,
which find the near-optimal schedule with minimal infor-
mation exchange between the residential scheduler and con-
sumers. The smart appliances present in smart homes work
in automated fashion that provide consumers high comfort
at less expense. In order to make home energy consumption
more efficient and to control the power supply and demand
numerous researches are presented.
Khosla [7] present a review of different research works
on a wide range of energy management controllers for smart
homes which reduce energy consumption, PAR and energy
wastage. Various HEM schemes, pricing schemes, such as
real time pricing (RTP), critical peak pricing (CPP), time of
use (TOU) and day ahead pricing (DAP) and energy con-
sumption are discussed in detail. Bao et al. [8] propose a
hybrid approach (WDO-DE) from two existing techniques,
i.e., wind driven optimization (WDO) and differential evo-
lution (DE) algorithms. Fifteen benchmark functions are
tested which contain unimodal, multi-modal, low dimen-
sional and high dimensional to check the performance of
proposed algorithm. The experimental results show that the
proposed algorithm can be feasible in both low-dimensional
and high-dimensional cases. The simulation results show that
the performance of WDO-DE algorithm is better than genetic
algorithm (GA), binary particle swarm optimization (BPSO),
WDO, DE, and PSO to reach the optimal solution. The appli-
cation of heuristic algorithms is not confined to scheduling
of energy consumption of appliances, these algorithms are
used in vast number of applications in many other fields such
as image segmentation, cloud resource allocation, multi-level
thresholding etc. as in [9] the authors propose a new quan-
tum wind driven optimization (QWDO) for path planning
of unnamed combat air vehicle (UCAV) in the battlefield
considering the different threats and constraints. Two test
instances are chosen in order to evaluate the performance
of the proposed algorithm. The experimental results depict
that the QWDO algorithm is an appropriate and reliable
technique to solve the UCAV path planning problem and
when compared with other algorithms it shows a better search
performance. The combined pricing schemes of TOU and
inclined block rate (IBR) is used for bill calculation in [10].
In this paper, the authors present an efficient DSM model for
residential energy management system in order to avoid peak
formation while decreasing electricity bill and preserving
user comfort level within acceptable limit. For this purpose,
three heuristic algorithms (GA, BPSO, and ACO) are used
to evaluate the objective function. They suggest that the
GA based EMC is better in term of electricity bill reduction
and PAR minimization and maximization of user satisfaction
than BPSO and ACO. However the computational time of
the algorithm is higher. Arafa et al. [11] reduce the com-
putational time by introducing evolutionary algorithm in an
enhanced way, that improves the performance (convergence
rate and accuracy) of DE called EDE for load scheduling
in homes. It has less number of parameters to control. The
algorithm is tested on 47 benchmark functions. All the steps
are followed as in DE only the number of trial vectors is
increased to minimize the chance of repetition of selection
reduces. Many research works are still in progress to improve
the performance of the algorithm in order to make them more
compatible with increasing demands.
Moon and Lee [12] study a society based load scheduling
problem with different classes of appliances in the grid. While
designing the optimization algorithm the overall society0s
satisfaction is kept under consideration. They design the
smart grid with various constraints such as minimize energy
consumption, alter peak formation and limit the budget. The
overall society electricity usage pattern is observed keenly
and sum of net consumption of all homes and their electric-
ity cost are compared which give near-optimal scheduling.
On the base of collected data the lower and upper bounds of
the objective function are formulated. The simulation results
show that the algorithm is effective in treating heterogeneous
residences.
In [13], the negligence of user comfort in the previous
papers is given consideration along with electricity cost sav-
ing and PAR reduction. They propose GA based algorithm
for DSM. Five types of appliances are taken for scheduling
and their mathematical models are formulated by considering
thermal and comfort constraints. The pricing scheme used in
this paper is RTP. When there are peak hours they integrate
micro grid with the traditional grid so that the bill can be
reduced and also user comfort can be increased.
The algorithms enable the consumers to pursue the best
consumption benefits within consumption range. To improve
the financial benefits of the electricity consumers, a novel
concept of cost efficient load scheduling framework is intro-
duced in [14]. The authors merge the two pricing techniques
RTP and DAP by using fractional programming approach.
They have explained the effect of simple power shifting of
specific appliances on the consumption cost, to show the
direct relationship between consumption load pattern and
cost. Further proposed algorithm allows the consumers to
fully utilize the electricity with remarkable savings of bill.
Another contribution of the paper for the cost saving and
minimization of CO2emission in environment is integration
of RES into the grid. This RES along with the fractional
programming reduce the electricity cost efficiently. How-
ever, the PAR minimization and user comfort are not dis-
cussed in the paper. Ahmad et al. [15] propose an optimized
HEM system (OHEMS) to minimize the electricity bill in
response to dynamic pricing by scheduling the household
appliances using four heuristic algorithms; GA, BPSO, WDO
and bacteria foraging optimization (BFO) algorithm. More-
over they offer a new hybrid GA-PSO (HGPO) algorithm
which incorporates the positive features of GA and PSO
algorithms in a single algorithm. RES and ESS are integrated
to encourage the consumers to take part in DSM.
VOLUME 7, 2019 24269
S. Kazmi et al.: Toward Optimization of Metaheuristic Algorithms for IoT Enabled Smart Homes
TABLE 1. Categories and parameters of appliances.
A new meta-heuristic, population based algorithm mimick-
ing the improvisation process by musicians was developed
by Geem et al. [16] in 2001. It is simple concept based,
less parameterized and easily implementable algorithm. It has
been applied to various engineering and non engineering opti-
mization problems successfully. Unlike GA, HSA generates
a new optimal vectors considering all the available vectors in
the search space, whereas, in GA only two parent vectors are
selected to produce one better offspring. Therefore the flex-
ibility and accuracy of HSA is better than other algorithms.
We have further enhanced its performance by hybridizing it
with EDE in this paper.
The fixed value of pitch adjustment ratio (PAR) and arbi-
trary distance bandwidth (bw) in improvisation step of the
traditional HSA cause poor performance and increase the
number of iterations needed to find the optimum solution.
Mahdavi et al. [17] introduce a variable PAdR and bw
values in the improvisation step to overcome the draw-
back. Moreover to check the effectiveness of improved
HS, they apply this to different constrained and uncon-
strained benchmark functions. In [18], unit commitment
problem is solved using HSA. The total production cost
is minimized by optimizing the controllable parameters
within the limits. The HSA has fast convergence time
and is economical than conventional and improved GA.
Fesanghary and Ardehali [19] present a novel meta-heuristic
approach based on HSA to solve economic dispatch (ED)
problem to minimize the total power generation cost. They
formulate two approaches; swarm intelligence concept and
hybrid harmony search quadratic programming (HSQP)
to improve the quality and convergence rate of HSA.
Karthigeyan et al. [20], the authors compare the performance
of HS, bio-geography based optimization (BBO) algorithm
and improved harmony search (IHS) algorithms for solv-
ing constraint economic dispatch power in power system.
Twenty generating units are tested through these algorithms
with ramp rate limits and valve point loading constraint. The
improved HSA gives minimum fuel cost and good conver-
gence characteristics as compared to HSA and BBO. The
integration of RES in SG, its economical benefits in term of
cost reduction and challenges while integration is discussed
in [21]. Different types of RESs are available for integration
FIGURE 1. Proposed system model.
with SG in market nowadays. Phuangpornpitak and Tia [22]
discuss in detail, the various sources of renewable energy,
their utilization and trends of technological advancement in
RES integration with SG. The security issues and impact of
RES on environment are also elaborated.
III. PROPOSED SYSTEM MODEL
DSM in a smart grid makes operation of the grid more reliable
and stable. In smart home it manages and controls the energy
usage by scheduling the appliances according to the scheduler
embedded in the HEM system [23]. The smart meter allows
two way flow of information between consumer and utility,
i.e., pricing signal and load demand. The information is sent
to the EMC by smart meter and EMC accordingly schedules
the appliances in the smart home based on pricing signal,
load demand and user preferences. Simple architecture of
HEM system is shown in Fig. 1.
A. LOAD CATEGORIZATION
We classify appliances into three categories; fixed, shift-
able. and uninterruptible appliances according to consumer
usage behavior. The power rating of appliances and time of
operation in a day is summarized in Table 1. Details of these
categories are given below.
24270 VOLUME 7, 2019
S. Kazmi et al.: Toward Optimization of Metaheuristic Algorithms for IoT Enabled Smart Homes
1) FIXED APPLIANCES
These are called fixed appliances because their operation
length cannot be modified. The scheduler has to sched-
ule these appliances between user defined time-slots. These
appliances are light, AC, refrigerator and oven. We represent
set of fixed appliances as Fap and its power consumption as U.
faFap has power rating of ρfa, then total power consumed
by fixed appliances in each timeslot is calculated using the
following equation,
U(t)=
T
X
t=1X
faFap
ρfa×γfa(t),(1)
where, a= {1,2,...,n},T=24 for one day, γfa(t) is
ON /OFF state of the appliance in the respective time-slot.
2) SHIFT-ABLE APPLIANCES
These appliances are those which can be shifted to any time-
slot and when required can be interrupted during operation.
These appliances include vacuum cleaner, water pump, water
heater, and fans. As the AC and fans are both used for cool-
ing purpose, the scheduler will schedule them accordingly
so that user can get maximum comfort. We represent shift-
able. appliances as Sap and its power consumption by V.
saSap has power rating of ρsa, then the total power consumed
by shift-able. appliances in each time-slot is calculated in the
following equation,
V(t)=
T
X
t=1X
saSap
ρsa×γsa(t),(2)
where, γsa(t) is the ON /OFF state of the appliance in that
hour. Our focus is to minimize the per hour cost of each
appliance, as a result, the overall cost will be reduced.
3) UN-INTERRUPTIBLE APPLIANCES
These appliances can be delayed or schedule earlier but once
started operation cannot be interrupted until the operation
completes. Washing machine, cloth dryer and dishwasher are
included in this category. UIap is the set of un-interruptible
appliances such that uiaUIap and ρuiais the power rating
VOLUME 7, 2019 24271
S. Kazmi et al.: Toward Optimization of Metaheuristic Algorithms for IoT Enabled Smart Homes
of each appliance. The total power consumption Wof this
category appliances can be calculated as,
W(t)=
T
X
t=1X
uiaUIap
ρuia×γuia(t).(3)
B. INTEGRATION OF RES
The roof of smart home is fitted with rooftop photo-
voltaic (PV) energy source. The smart home is not fulfilled
whole of its electricity demand by the PV source rather only
when the utility is providing electricity with peak pricing, the
PV source operates. The EMC integrates the PV source with
the utility grid when required. In each time-slot the power
generated by PV source is EPV (t).
The PV source provides the energy for the time-slots
tα−tβ,so the total energy generated each day by the
PV source can be calculated as,
EPV =
tβ
X
t=tα
EPV (t),(4)
The PV source can be integrated only if it is meeting a
minimum capacity defined below,
EPV ≥Emin
PV .(5)
In the smart home forecasting device of the expected
PV generation is installed according to which the scheduler
schedules the appliances. The forecasted output power of
PV source is affected by many factors like solar irradiance Ir,
area of the solar panel APV , the outdoor temperature at that
time Ta(t) and inverter efficiency η. The generated output
power of the PV source can be calculated by, [24].
EPV =ηPV ×APV ×Ir(1 −0.005(Ta(t)−25)),(6)
C. ENERGY CONSUMPTION MODEL
The total energy consumption of all the appliances in each
hour can be calculated using the following equations,
PT(t)=W(t)+V(t)+U(t),(7)
PT(t)=X
faFap 24
X
t=1
ρfa×γfa(t),
+X
saSap 24
X
t=1
ρsa×γsa(t),
+X
uiaUIap 24
X
t=1
ρuia×γuia(t).(8)
To calculate the total energy consumed (demand of con-
sumer) in a day, the per hour energy consumption is calcu-
lated and added.
D. ENERGY COST MODEL
The electricity cost is calculated by multiplying pricing signal
with energy consumed by appliances.
CT=
T
X
t=1
(P(t)×λ(t)),(9)
where, λis pricing signal used in our work. We have taken
RTP scheme which has per hour changing behavior w.r.t price
and remains unchanged in that hour. The price or rate at which
electrical energy is supplied to consumers is called tariff.
Numerous electrical tariffs are available to define the energy
pricing over a day such as ToU, DAP, RTP, CPP etc.
IV. PROBLEM FORMULATION
Formulation of objective function is a key step in optimization
problem. In this paper, the objective function is defined as
electricity consumption cost minimization to achieve maxi-
mum user comfort. The smart home is equipped with smart
meter which sends the consumer’s energy demand and pref-
erences to the utility company. The utility accordingly offers
DR signal which contain necessary load scheduling and
optimization. The smart meter can directly communicate
with EMC and grid. The EMC defines operating sched-
ule of household appliances and communicates with appli-
ances using communication technologies. Home appliances
are categorized based on operating time and energy con-
sumption requirement for efficient management of energy.
We formulate our problem as knapsack that the total elec-
tricity consumption must not exceed the maximum capacity
defined, i.e.,
max Xpi×xi,(10)
s.t.,
n
X
i=1
wi×xi≤C,(11)
where, pishows the profit of each item, wirepresents weight
of each item and xirepresents the binary number 1 and 0
means ON/OFF state of each appliance. (11) shows that the
collective weight after considering maximum profit must not
exceed the capacity C of the knapsack. We consider appli-
ances as items and weight as power ratings of appliances.
The operational cost of an appliance is taken as its profit.
Our objective is to minimize the electricity cost by getting
maximum profit,
min
T
X
t=1
m
X
i=1
PT×γi(t),(12)
s.t.,
T
X
t=1
m
X
i=1
PT×γi(t)×λ(t)<C,(13)
and,
τsch =τot ,(14)
where, γi(t) is the ON/OFF state of the appliance at the time-
slot t,τsch,τot represent the operation start and ending times
24272 VOLUME 7, 2019
S. Kazmi et al.: Toward Optimization of Metaheuristic Algorithms for IoT Enabled Smart Homes
TABLE 2. Parameters of EDE.
of the appliance and Cis the maximum allowable power to
be consumed. Violation of this constraint may lead to peak
formation and system stability issues. For this purpose, the
scheduling algorithms must follow the knapsack constraint.
When the amount of consumed power exceeds the maximum
limit, the EMC computes new optimal allocation using an
algorithm and sends control message to the appliances. Once
the power consumption constraint is satisfied the scheduler
gives a consumption pattern, now accordingly we formulate
our energy consumption cost.
V. PROPOSED SCHEMES
The appliance scheduling problem formulated in section IV
is evaluated using three heuristic algorithms (HSA, EDE
and our proposed technique HSDE). Contrary to classical
optimization techniques like linear programming (LP) [25],
integer linear programming (ILP) [26], and mixed integer
linear programming (MILP) [27], heuristic algorithms poses
fast convergence rate and simple steps to reach the optimum
solution. The proposed algorithms are discussed in detail
below.
A. EDE
The EDE algorithm is advanced form of differential evolu-
tion algorithm introduced by Storn and Price [28] in 1995.
It has now become one of the most common technique to
solve the scheduling problem. This algorithm has only three
parameters; mutation, crossover, and selection. The tuning
control parameters are size of the population, scaling factor
of mutation, and crossover rate. Like all other algorithms the
first step is the random population generation. The generation
of population in EDE algorithm is simple given as,
xij =xl+rand(1) ×(xu−xl),(15)
where, xij is the initial population and xland xuare lower and
upper limits of the values in the population. This population
is in the form of real numbers between the upper and lower
limits which is given in the parameters defined. Once the
population is generated it is converted to binary number by
any function like sigmoid function. The three operations of
EDE are briefly discussed below.
1) MUTATION
It expands the search space of the problem. For mutation three
vectors are selected randomly. One vector is taken as target
vector and the difference of other two vectors are multiplied
with mutation factor Fand the result is added to target vector
to get the mutant vector,
Vj=xr1+F×(xr2−xr3),(16)
where, Vjis a mutant vector, Xr1,Xr2and Xr3are the target
and other two randomly selected vectors to produce mutant
vector.
2) CROSSOVER
It incorporates successful solution from the previous gener-
ation. Once the mutation step is completed, EDE algorithm
performs crossover operation to produce trial vectors. The
trial vector is generated from the variables of mutant vectors
and variables of target vectors,
Uj1=(Vj,if rand(b)≤CR1,
xr1,if rand(b)>CR1,(17)
where, Uj1is the trial vector, CR represents crossover rate.
The variable to be selected for trial vector will be chosen from
target and trial vector while keeping (15) in consideration.
The difference between DE and enhanced DE (EDE) lies here
that in DE only one trial vector is generated to replace the
target vector selected from the population while in EDE five
trial vectors are generated. Three of which are produced at
different crossover rates as shown in the following equations,
Uj2=(Vj,if rand(b)≤CR2,
xr1,otherwise,(18)
Uj3=(Vj,if rand(b)≤CR3,
xr1,otherwise.(19)
The two trial vectors are generated below,
Uj4=rand(b)∗xr1,(20)
Uj5=rand(b)∗Vj+(1 −rand (b)) ∗xr1.(21)
3) SELECTION
The trial vectors and target vector are test by fitness function.
Among six vectors; five trial vectors and one target vector,
the one which is fittest will replace the target vector in the
population,
xr1=(Uji,if F(Uji )>F(xr1),
xr1,otherwise.(22)
Once the mutation, crossover and selection is completed the
algorithm will continue search for next fittest value until
maximum iteration defined reaches.
B. HSA
HSA is introduced by Geem et al. [16]in 2001. It is formed
from the concept of the improvisation process of a musician
in which the musician always search for perfect state of
harmony. The musician tries to find pleasing harmony just as
optimization techniques search for global best solution. The
musician makes various combination of pitches stored in the
VOLUME 7, 2019 24273
S. Kazmi et al.: Toward Optimization of Metaheuristic Algorithms for IoT Enabled Smart Homes
TABLE 3. Parameters of HSA.
harmony memory. If all the pitches make a good harmony
one by one, those pitches are stored in the library/memory
and the chance of producing better harmony in the next try
increases. Same process happens in the engineering optimiza-
tion where each decision variable chooses initial values from
the search space and makes single solution vector. If this
solution vector is fitter than the previous one, then search
for more better solution (global best) continues until stopping
criteria reaches. The parameters used in HSA are defined in
Table 3.
The HS algorithm requires less mathematics and no initial
value setting of the decision variables. The main steps of
HS algorithm are
1) Harmony memory generation
2) Improvisation of new harmony from HM
a) Harmony memory consideration rate (HMCR)
adjustment
b) PAdR adjustment
c) Random selection
3) Update HM
4) Repeat step 2 and 3 until stopping criteria reaches.
1) HARMONY MEMORY GENERATION
Initial HM is generated randomly,
Hij =Hl+rand(b)∗(Hu−Hl),(23)
where, Hij represents the set of random values in the HM,
Hland Huare the lower and upper limit of values in the HM.
2) IMPROVISATION OF NEW HM
A new harmony vector x0is produced from HM consid-
ering HMCR, PAR and random selection. Each value in a
new harmony vector is taken by comparing with HMCR
and PAdR values. The value of x0can be chosen from
x0
1,x0
2,x0
3, ...., x0
nwith probability of HMCR and 1-HMCR,
these values can be taken from the entire feasible region xij.
x0=(x0
i∈[x0
1,x0
2,x0
3· ··,x0
n],with prob HMCR,
x∈Hij with prob 1 −HMCR.
(24)
Each value in the new harmony vector x0is checked whether
it should be pitch adjusted or not.
pitch adj for x 0
i=(yes,with prob PAdR,
no,with prob 1 −PAdR.(25)
The value of probability (1-PAdR) shows that no adjustment
is required for the variable but if the pitch adjustment check
is satisfied, then the variable x0
iis adjusted below,
x0
i=x0
i+rand(b)∗bw,(26)
where, bw is the arbitrary distance bandwidth.
3) UPDATE THE HM
The new generated harmony vector x0=[x0
1,x0
2,· · ··,x0
n]
is compared with the worst harmony vector in HM using
objective function, in our case minimum cost is fittest one.
If the new vector is better than worst one then new one will
take place of the worst one in the memory.
4) STOPPING CRITERIA
Once the maximum number of iterations are completed, the
algorithm stops its execution.
C. HYBRID HSDE ALGORITHM
The hybrid HSDE has the common features of both EDE
and HSA in it. The random selection step of HS algorithm
is replaced by the fitness check steps of EDE which provides
a good result in terms of user comfort and PAR reduction.
The steps of HSDE are same as HS algorithm except the
improvisation of harmony the random selection is replaced
by fitness evaluation of the new harmony vector and ran-
domly selected target vector. The fittest vector is stored in the
harmony memory and then search for the next fittest vector
starts until some stopping criterion is reached. The working
principle of HSDE algorithm is summarized in Algorithm 1.
The parameters used in HSDE are defined in Table 4.
The target vector for evaluation of fitness is selected in the
following equation,
Hr1=Vr1+F×(Vr2−Vr3),(27)
where, Hr1is mutant vector, Vr1is target vector, Fis mutant
factor, Vr2and Vr3are other two randomly selected vectors
from HM. After mutation, crossover step is followed same
as in EDE. Here only three trial vectors are generated for
evaluation so the computational time is reduced,
Hj1=(Hr1,if rand(b)≤CR1,
Vr1,otherwise,(28)
Hj2=(Hr1,if rand(b)≤CR2,
Vr1,otherwise,(29)
Hj3=(Hr1,if rand(b)≤CR3,
Vr1,otherwise.(30)
24274 VOLUME 7, 2019
S. Kazmi et al.: Toward Optimization of Metaheuristic Algorithms for IoT Enabled Smart Homes
TABLE 4. Parameters of HSDE.
The selection of trail vector in HSDE will be same as in EDE
that the fittest value among the trial vectors, target vector and
new generated harmony vector will be saved in the memory.
Hij =
x0,if F(Hj)<F(x0)>F(Vr1),
Hj,if F(Vr1)<F(Hj)>F(Vr1),
Vr1,if F(Hj)<F(Vr1)>F(Vx0).
(31)
VI. FEASIBLE REGIONS
The feasible region contains the set of possible solutions
which satisfy all the constraints of a system. In this work
the feasible regions for cost vs waiting time termed as user
comfort and cost vs load are defined. The scheduler ensures
that the power consumption should be done in such a way that
the cost does not exceed the range defined in feasible region.
The feasible region for waiting time and cost have trade-off
behavior as shown in Fig. 2. When the delay is maximum,
the electricity bill lies in the range between 28.05 −57.54
cents for minimum and maximum loads as shown by P2and
P3respectively. While the electricity cost is maximum when
the waiting time is zero with maximum and minimum power
consumption as shown by points P5and P1respectively.
Fig. 3 shows feasible region for electricity cost and elec-
tricity consumption with the help of P1,P2,P3,P4and
P5forming a trapezoidal shape. Point P1represents the
electricity bill when minimum possible energy consumption
(minimum load) is scheduled with minimum price value in
the RTP signal. Point P2shows the electricity bill when
minimum load is scheduled with maximum price value in the
RTP signal. Similarly points P4and P5represent the elec-
tricity bills when maximum load (all operating appliances)
are scheduled with minimum and maximum price values
in the RTP signal. As the pricing signal is always set by
the utility company and the consumer can never change or
modify it, the consumer can only schedule their appliances
accordingly so that maximum saving could be achieved.
We have put a constraint that the scheduler must always
schedule the appliances of the smart home in such manner
that the cost at any time-slot must not exceed 350.6 cents.
The point P3is the point of constraint given to the scheduler.
VII. SIMULATION RESULTS
In this section, we discuss the simulation results and ana-
lyze the performance of the scheduling algorithms in term
of electricity cost savings, user comfort and PAR. We take
eleven different appliances with different energy demands
Algorithm 1 HSDE Algorithm
1Initialize algorithm parameters (HMCR, HMS, CR,
PAdR, bw, F, I)
2Generate the harmony memory
3for {p=1:NI} do
4Find f(xworst )
5Generate a new harmony vector (x0
i)
6for x0
i=1:HMS do
7if rand (1)< HMCR then
8x0=HM[i][j] where i ∈(1, 2, . . . , HMS) if
rand(1) < PAdR then
9x0
i=x0
i* rand(1) * bw
10 end if
11 end if
12 else
13 Select three harmony vectors to generate
mutant and trial vectors
14 Mutation
15 Vj=xr1+F×(xr2−xr3)
16 Crossover
17 Uj1(Vjif rand(b)≤CR1
xr1Otherwise
18 Check for the fittset one among trial vector
if f(Vj)<f(x0)then
19 Replace xworst with trial vector Else
Replace xworst with x0
20 end if
21 end if
22 end for
23 Update the harmony for next iteration
24 end for
25 Continue until termination criteria reaches
FIGURE 2. Feasible region cost vs waiting time.
and different operation times normally used in the homes.
RTP pricing signal has been used for billing purpose. The
simulation is performed for the time period of 24 hours.
VOLUME 7, 2019 24275
S. Kazmi et al.: Toward Optimization of Metaheuristic Algorithms for IoT Enabled Smart Homes
FIGURE 3. Feasible region cost vs load.
FIGURE 4. RTP.
FIGURE 5. Forecasted daily temperature.
Fig. 4 and Fig. 5 are forecasted input pricing and temper-
ature signals given to the HEM system from utility company
and METEONORM 6.1 for Islamabad region of Pakistan.
The forecasted outdoor temperature is formulated to inform
FIGURE 6. Solar irradiance.
FIGURE 7. PV system generation.
the scheduler how much generation is possible for that day.
The RTP signal is made by utility on the average consumption
behavior of the consumers for the last three months. The price
in each hour is changing and from 7 :00 a.m to 3 :00 p.m it
is comparatively high and expensive. Similarly temperature
forecasted for 24 hours is high at noon.
The conversion efficiency of generator, area of genera-
tor, solar irradiance and outdoor temperature are the factors
effecting the generation of PV system as modeled in (6).
The 90% of estimated power generated by PV system is
utilized by the consumers at day time to reduce their elec-
tricity bill and the remaining 10% is utilized by system to
facilitate integration complexities. Fig. 6 and Fig. 7 present
the solar irradiance and the estimated electricity generation
from PV system.
Fig. 8a shows the hourly electricity consumption without
RES and ESS integration. Peaks are formed in unscheduled
scenario and the three algorithms have optimized the con-
sumption by uniformly distributing the load over the schedul-
ing horizon. Though EDE and HSA has shifted maximum
24276 VOLUME 7, 2019
S. Kazmi et al.: Toward Optimization of Metaheuristic Algorithms for IoT Enabled Smart Homes
FIGURE 8. Energy consumption per hour. (a) Energy consumption per
hour without RES. (b) Energy consumption per hour with RES.
load to the ending time-slots, however, their peaks are still
lower than peaks of unscheduled pattern. Peaks are formed
in unscheduled scenario in the time-slots 8 and 15 touching
17.8, and 18.9 kWh, respectively. The hybrid technique, i.e.,
HSDE shows moderate behavior throughout the scheduling
horizon. HSA shows some peaks in the ending time-slots 15,
16 and 18, it has maximum electricity consumption
of 16.2 kWh. It has shown minimum and negligible consump-
tion in the starting hours, i.e., 1 −6. EDE based scheduling
technique reduces the peaks and provides moderate consump-
tion pattern. It has minimum electricity consumption in time-
slot 1 and 3, i.e., 4.6 and 5.1 kWh, respectively, and maximum
value of consumption in any time-slot is 15.6 kWh in the
time-slots 19 −24.
The hourly energy consumption patterns of unscheduled
and scheduled load along with RES and ESS integration are
presented in Fig. 8b. The unscheduled load pattern has a
peaks of 12.9, 11.95 and 14 kWh in time-slots 3, 6 and 22,
respectively. It has minimum consumption in time-slot 9 and
zero consumption from utility in time-slots 1, 11 and 24.
Rest of the time-slots are showing the moderate consump-
tion. Using HSA technique with RES and ESS, the peaks
in unscheduled pattern are reduced up to 7.5 kWh in time-
slot 16. As compared to unscheduled energy consumption
pattern, the per hour energy consumption in HSA is optimum.
The consumption is maximum, i.e., 7.5 kWh in time-slot 16
and minimum in time-slots 2 −6, and 8, i.e., 1.7 and
2.7 kWh respectively. In time-slots 1 and 24 there is zero
energy consumption. Rest of the time-slots have average
consumption pattern. EDE based scheduling shows some
peaks in the starting and ending hours but the overall elec-
tricity consumption pattern is moderate. The consumption
pattern of EDE shows peaks in time-slots 8 and 17 −24.
The minimum consumptions is in time-slots 9, 13 and 14
are 1.5, 2.4 and 2.4 kWh, respectively, and in time-slots
8 and 11 negligible amount of electricity is consumed from
the utility. Our proposed hybrid technique, i.e., HSDE also
has peaks at the starting and ending time-slots but when
RES and ESS are integrated the consumption pattern shows
minimum and negligible behavior in time-slots 7 −17. The
maximum electricity consumption of 11.5 kWh is shown in
time-slot 23 which are followed by other peaks in time-slots
3, 6, 18 and 24. The energy consumption in these time-slots
is 9, 9, 9.8 and 10.9 kWh, respectively.
The hourly electricity bill of unscheduled and scheduled
load without RES and ESS is shown in Fig. 9a. Results show
that the bill of heuristic algorithms (HSA, EDE and HSDE)
based scheduling remains within the feasible region. From
the figure it can be seen that unscheduled scenario results in
maximum bill. In time-slot 8, the electricity cost is 475 cents
which is maximum among all. The three proposed techniques
(HSA, EDE and HSDE) have reduced this electricity cost
considerably. In the same time-slot the maximum cost of each
algorithm falls which is 28.84%, 38.72% and 49.89% less
than unscheduled cost. In the starting time-slots 1 −7, the
electricity cost is comparatively lower than last time-slots.
The electricity bill of scheduled load is less than unscheduled
load.
The hourly electricity bill of unscheduled and scheduled
loads with RES and ESS is shown in Fig. 9b. This shows
that the electricity bills of scheduling algorithms (HSA, EDE
and HSDE) is less as compared to unscheduled cost. The
overall cost in all the time-slots is also less than that in
Fig. 9a. The energy consumption pattern of unscheduled
scenario is most expensive in the graph e.g. in time-slot 9
the cost of unscheduled pattern is 195 cents. This expensive
peak is reduced by all the three algorithms up to 45, 40
and 60%. HSDE has maximum cost in 6th hour of a day, i.e.,
116.3 cents that is 43% less than the peak of unscheduled
cost. HSA has maximum cost in 7th hour of a day, i.e.,
90 cents that is 55% less than unscheduled maximum cost
and EDE has maximum cost of 100 cents in 4th hour of a day
VOLUME 7, 2019 24277
S. Kazmi et al.: Toward Optimization of Metaheuristic Algorithms for IoT Enabled Smart Homes
FIGURE 9. Electricity cost per hour. (a) Bill per hour without RES. (b) Bill
per hour with RES.
that is 49.5% reduced than unscheduled cost. HSDE has zero
cost in 8th and 9th time-slots which means in these hours the
total energy is consumed from RES and ESS. Similarly, EDE
has zero consumption in time-slots 8 and 11.
Peak formation is a major drawback in traditional electric
power system as it causes customer to pay high electricity
bill as well as challenges the stability of grid. Performance
of all the designed models (HSA-EMC, EDE-EMC and
HSDE-EMC) with respect to PAR reduction is shown in
Fig. 10a. It shows that PAR is significantly reduced
by HSA-EMC, EDE-EMC and HSDE-EMC than the
unscheduled case because these are designed to avoid peak
formation.
When RES and ESS are integrated the load pattern
becomes smooth and the peaks are reduced. The smart user
does not fully rely on the utility grid, instead prefer the use
of RES and ESS when available. By doing this both cost
and PAR are reduced resulting in grid stability. In Fig. 10b
the PAR values after integration of RES and ESS are shown
which are lower than Fig. 10a.
FIGURE 10. PAR. (a) PAR without RES. (b) PAR after RES.
TABLE 5. Performance trade-off comparison.
The comparison of overall daily electricity bill of the
unscheduled and scheduled load without RES and ESS is
shown in Fig. 11a. The daily electricity cost in unscheduled,
and in scheduled load scenarios using HSDE, HSA, and
EDE algorithms are 3652.63, 3163.39, 3000.02, and 3227.61
cents respectively. The comparison of total electricity cost
shows that HSDE, HSA, and EDE algorithms based HEM
system reduces the electricity bill by 13.2%, 17.86%, and
11.5% respectively. In Fig. 11b, daily electricity costs after
integration of RES and ESS are shown. Compare to Fig. 11a
the electricity bill bars of Fig. 11b are shorter. HSA along
with RES and ESS gives minimum cost than others.
A. TRADE-OFF BETWEEN PARAMETERS
The trade-off behavior explained in Section VI is summarized
in Table 5 as shown in Figs. 11 and 12. HSA has maximum
24278 VOLUME 7, 2019
S. Kazmi et al.: Toward Optimization of Metaheuristic Algorithms for IoT Enabled Smart Homes
TABLE 6. The overall comparison of the algorithms.
FIGURE 11. Daily cost. (a) Daily cost without RES. (b) Daily cost with RES.
waiting time (less user comfort) which has minimum cost
among the three algorithms, i.e., 3000.02 cents in Fig. 11
while EDE entertains the user with maximum comfort giving
less saving in electricity bill, i.e., 425 cents. Our proposed
HSDE algorithm presents less waiting time than HSA.
The electricity bill offered by this algorithm is also expensive
then HSA. Results reveal that integration of RES and ESS
minimizes the electricity bill up to 52% daily. Peaks in the
electricity consumption pattern are smoothen and the per
hour consumption of all three algorithms is invulnerable to
utility and consumer. HSDE shows balanced load pattern than
HSA and EDE which increases the power system stability.
FIGURE 12. Waiting time.
Reduction in PAR by HSDE, HSA and EDE is 17.247%,
32.871% and 16.586% respectively. Furthermore our simu-
lation verify the feasible regions which is provided before
the simulation for optimal solution. On the other hand, there
exists a trade-off between electricity cost and user waiting
time which is depicted in our work.
The overall performances of the algorithms before and
after RES integration is shown in Table 6.
VIII. CONCLUSION
Due to lack of energy sources and aging of existing power
systems, demand for smarter and efficient power system has
increased. The concept of SG has been introduced for this
reason in which the appliances are made smart in such a way
that they can coordinate through EMC and even control the
power consumption of smart home. The integration of RES
and ESS into SG maximizes the user comfort by economi-
cally consuming electricity in the peak hours. In this paper a
new hybrid algorithm HSDE is proposed by hybridizing two
existing heuristic algorithms (HSA and EDE). Their perfor-
mance on the bases of PAR, electricity cost, user comfort and
energy consumption is evaluated. To tackle the intermittent
behavior of RES and implementing RES with future work is
another direction of our work.
REFERENCES
[1] (2010). Deploying a Smarter Grid Through Cable Solutions and Ser-
vices. Accessed: Jan. 31, 2016. [Online]. Available: http://www.nexans.
com/Corporate/2010/WHITE-PAPER-SMART-GRIDS-(2010).pdf
VOLUME 7, 2019 24279
S. Kazmi et al.: Toward Optimization of Metaheuristic Algorithms for IoT Enabled Smart Homes
[2] Y. Guo, M. Pan, and Y. Fang, ‘‘Optimal power management of residential
customers in the smart grid,’’ IEEE Trans. Parallel Distrib. Syst., vol. 23,
no. 9, pp. 1593–1606, Sep. 2012.
[3] A. Agnetis, G. de Pascale, P. Detti, and A. Vicino, ‘‘Load scheduling for
household energy consumption optimization,’’ IEEE Trans. Smart Grid,
vol. 4, no. 4, pp. 2364–2373, Dec. 2013.
[4] H. Nilsson, ‘‘The many faces of demand-side management,’’Power Eng. J.,
vol. 8, no. 5, pp. 207–210, 1994.
[5] S. L. Nalbalwar, J. D. Ruikar, and S. R. Sakpal, ‘‘Smart grid: A modern-
ization of existing power grid,’’ Int. J. Adv. Eng. Res. Stud., vol. 1, no. 2,
pp. 295–298, 2012.
[6] Y. Liu, C. Yuen, S. Huang, N. Ul Hassan, X. Wang, and S. Xie, ‘‘Peak-
to-average ratio constrained demand-side management with consumer’s
preference in residential smart grid,’’ IEEE J. Sel. Topics Signal Process.,
vol. 8, no. 6, pp. 1084–1097, Dec. 2014.
[7] K. Patel and A. Khosla, ‘‘Home energy management systems in future
smart grid networks: A systematic review,’’ in Proc. IEEE 1st Int. Conf.
Next Generat. Comput. Technol. (NGCT), Sep. 2015, pp. 479–483.
[8] Z. Bao, Y. Zhou, L. Li, and M. Ma, ‘‘A hybrid global optimization algo-
rithm based on wind driven optimization and differential evolution,’’ Math.
Problems Eng., vol. 2015, Jul. 2015, Art. no. 389630.
[9] Y. Zhou, Z. Bao, R. Wang, S. Qiao, and Y. Zhou, ‘‘Quantum wind driven
optimization for unmanned combat air vehicle path planning,’’ Appl. Sci.,
vol. 5, no. 4, pp. 1457–1483, 2015.
[10] S. Rahim et al., ‘‘Exploiting heuristic algorithms to efficiently utilize
energy management controllers with renewable energy sources,’’ Energy
Buildings, vol. 129, pp. 452–470, Oct. 2016.
[11] M. Arafa, E. A. Sallam, and M. M. Fahmy, ‘‘An enhanced differential
evolution optimization algorithm,’’ in Proc. IEEE 4th Int. Conf. Digit. Inf.
Commun. Technol. Appl. (DICTAP), May 2014, pp. 216–225.
[12] S. Moon and J.-W. Lee, ‘‘Multi-residential demand response scheduling
with multi-class appliances in smart grid,’’ IEEE Trans. Smart Grid, vol. 9,
no. 4, pp. 2518–2528, Jul. 2018
[13] M. B. Rasheed et al., ‘‘Real time information based energy management
using customer preferences and dynamic pricing in smart homes,’’ Ener-
gies, vol. 9, no. 7, p. 542, 2016.
[14] J. Ma, H. H. Chen, L. Song, and Y. Li, ‘‘Residential load scheduling in
smart grid: A cost efficiency perspective,’’ IEEE Trans. Smart Grid, vol. 7,
no. 2, pp. 771–784, Mar. 2016.
[15] A. Ahmad et al., ‘‘An optimized home energy management system with
integrated renewable energy and storage resources,’’ Energies, vol. 10,
no. 4, pp. 1–35, 2017.
[16] Z. W. Geem, J. H. Kim, and G. V. Loganathan, ‘‘A new heuristic opti-
mization algorithm: Harmony search,’’ J. Simul., vol. 76, no. 2, pp. 60–68,
Feb. 2001.
[17] M. Mahdavi, M. Fesanghary, and E. Damangir, ‘‘An improved har-
mony search algorithm for solving optimization problems,’’ Appl. Math.
Comput., vol. 188, no. 2, pp. 1567–1579, May 2007.
[18] R. Effatnejad, H. Hosseini, and H. Ramezani, ‘‘Solving unit commitment
problem in microgrids by harmony search algorithm in comparison with
genetic algorithm and improved genetic algorithm,’’ Int. J. Tech. Phys.
Problems Eng., vol. 6, no. 21, pp. 61–65, 2014.
[19] M. Fesanghary and M. M. Ardehali, ‘‘A novel meta-heuristic optimization
methodology for solving various types of economic dispatch problem,’’
Energy, vol. 34, no. 6, pp. 757–766, 2009.
[20] P. Karthigeyan, M. S. Raja, R. Hariharan, S. Prakash, S. Delibabu, and
R. Gnanaselvam, ‘‘Comparison of harmony search algorithm, improved
harmony search algorithm with biogeography based optimization algo-
rithm for solving constrained economic load dispatch problems,’’ Procedia
Technol., vol. 21, pp. 611–618, Nov. 2015.
[21] M. A. Islam, M. Hasanuzzaman, N. A. Rahim, A. Nahar, and
M. Hosenuzzaman, ‘‘Global renewable energy-based electricity generation
and smart grid system for energy security,’’ Sci. World J., vol. 2014,
Aug. 2014, Art. no. 197136.
[22] N. Phuangpornpitak and S. Tia, ‘‘Opportunities and challenges of integrat-
ing renewable energy in smart grid system,’’ Energy Procedia, vol. 34,
pp. 282–290, Jul. 2015.
[23] T. Logenthiran, D. Srinivasan, and T. Z. Shun, ‘‘Demand side management
in smart grid using heuristic optimization,’’ IEEE Trans. Smart Grid, vol. 3,
no. 3, pp. 1244–1252, Sep. 2012.
[24] E. Shirazi and S. Jadid, ‘‘Optimal residential appliance scheduling under
dynamic pricing scheme via HEMDAS,’’ Energy Buildings, vol. 93,
pp. 40–49, Apr. 2015.
[25] J. Y. Lee and S. G. Choi, ‘‘Linear programming based hourly peak
load shaving method at home area,’’ in Proc. IEEE 16th Int. Conf. Adv.
Commun. Technol. (ICACT), Feb. 2014, pp. 310–313.
[26] T. Hubert and S. Grijalva, ‘‘Realizing smart grid benefits requires energy
optimization algorithms at residential level,’’ in Proc. IEEE PES Innov.
Smart Grid Technol. (ISGT), Jan. 2011, pp. 1–8.
[27] K. M. Tsui and S. C. Chan, ‘‘Demand response optimization for smart
home scheduling under real-time pricing,’’ IEEE Trans. Smart Grid, vol. 3,
no. 4, pp. 1812–1821, Dec. 2012.
[28] R. Storn and K. Price, ‘‘Differential evolution—A simple and efficient
heuristic for global optimization over continuous spaces,’’ J. Global
Optim., vol. 11, no. 4, pp. 341–359, 1997.
SAQIB KAZMI received the M.S. degree in electrical engineering from
the Department of Electrical Engineering, COMSATS Institute of Electrical
Engineering, Islamabad, Pakistan, under the supervision of Prof. J. Mughal
and Dr. N. Javaid.
NADEEM JAVAID (S’08–M’11–SM’16) received
the bachelor’s degree in computer science from
Gomal University, Dera Ismail Khan, in 1995,
the master’s degree in electronics from Quid-i-
Azam University, Islamabad, Pakistan, in 1999,
and the Ph.D. degree from the University of Paris-
Est, France, in 2010. He is currently an Associate
Professor and the Founding Director of Communi-
cations over Sensors Research Lab, Department of
Computer Science, COMSATS University Islam-
abad, Islamabad. He has supervised 15 Ph.D. dissertations and 100 master’s
theses. He has authored over 700 articles in technical journals and inter-
national conferences. He is also an Associate Editor of the IEEE ACCESS
Journal and an Editor of the International Journal of Space-Based and
Situated Computing. His research interests include energy optimization in
block chain based smart grids and IoT enabled wireless sensor networks, and
data analytics in smart grids/wireless sensor networks. He was a recipient of
the Best University Teacher Award from the Higher Education Commission
of Pakistan, in 2016, and the Research Productivity Award from the Pakistan
Council for Science and Technology, in 2017.
MUHAMMAD JUNAID MUGHAL is currently
the Chairman of the Department of Electrical
Engineering, COMSATS Institute of Information
Technology, Islamabad, Pakistan.
MARIAM AKBAR (S’13–M’16) received the
M.Sc. and M.Phil. degrees from Quid-I-Azam
University, Islamabad, and the Ph.D. degree in
electrical engineering from the COMSATS Insti-
tute of Information Technology, Islamabad, under
the supervision of N. Javaid. She is currently an
Assistant Professor with the Department of Com-
puter Science, COMSATS Institute of Information
Technology. Her research interests include wire-
less networks and smart grids.
24280 VOLUME 7, 2019
S. Kazmi et al.: Toward Optimization of Metaheuristic Algorithms for IoT Enabled Smart Homes
SYED HASSAN AHMED (S’13–M’17) received
the B.S degree in computer science from the
Kohat University of Science and Technology,
Pakistan, and the master’s and Ph.D. degrees
from the School of Computer Science and Engi-
neering (SCSE), Kyungpook National Univer-
sity (KNU), South Korea. In 2015, he was also
a Visiting Researcher with the Georgia Institute
of Technology, Atlanta, USA. He is currently a
Post-Doctoral Research Fellow with SCSE, KNU,
where he also teaches ’’Design and Analysis of Computer Networks’’ course
at the Graduate School. He authored or co-authored over 90 International
journal articles, conference proceedings, book chapters, and two Springer
brief books. His research interests include sensor and ad hoc networks, cyber
physical systems, vehicular communications and future Internet. He received
the Research Contribution awards by SCSE at KNU, from 2014 to 2016, and
the Qualcomm Innovation Award at KNU, in 2016.
Dr. Hassan is a member of ACM. He is serving as a TPC Member
or Reviewer in more than 50 International Conferences and Workshops,
including IEEE Globecom, IEEE ICC, IEEE CCNC, IEEE ICNC, IEEE
VTC, IEEE INFOCOM, ACM CoNEXT, ACM SAC, and much more. He
has been reviewing papers for more than 30 International Journals, including
the IEEE WIRELESS COMMUNICATIONS MAGAZINE, the IEEE NETWORKS MAGAZINE, the
IEEE COMMUNICATIONS MAGAZINE, the IEEE COMMUNICATIONS LETTERS, the IEEE
SENSORS LETTERS, the IEEE TRANSACTIONSON INDUSTRIAL INFORMATICS,Vehicular
Technologies,Intelligent Transportation Systems Big Data,Mobile Comput-
ing, Elsevier Computer Communications, and Computer Networks.
NABIL ALRAJEH received the Ph.D. degree
in biomedical informatics engineering from
Vanderbilt University, USA. He was a Senior
Advisor for the Ministry of Higher Education, and
his role was in implementing development pro-
grams, including educational affairs, health infor-
mation systems, strategic planning, and research
and innovation. He is currently a Professor with
the Health Informatics, Biomedical Technology
Department, King Saud University. His research
interests include E-health applications, hospital information systems,
telemedicine, intelligent tutoring systems, energy management in smart
grids, and wireless sensor networks.
VOLUME 7, 2019 24281
