Demand side management using meta-heuristic techniques and ToU in smart grid

Abstract

In this paper, we perform performance evaluation of home energy management system (HEMS) for demand side management (DSM) in smart grid. In this work, smart home is equipped with HEMS, smart meter, and smart appliances for two-way communication between utility and consumer. HEMS performs scheduling of smart appliances based on meta-heuristic techniques to balance load for whole day to avoid peak creation in any hour. Smart meter performs electricity cost calculation for consumed energy based on time of use (ToU) pricing signal provided by utility. Our focus is to efficiently handle user demand, reduction in peak-to-average ratio (PAR) and electricity cost minimization. The implemented meta-heuristic techniques in this work are: Enhanced differential evolution (EDE), harmony search algorithm (HSA), bacterial foraging algorithm (BFA), and genetic algorithm (GA). The simulation results show the performance of HEMS based on optimization techniques using ToU.

Full text

Demand side management using meta-heuristic
techniques and ToU in smart grid
Sajeeha Ansar1, Wajeeha Ansar2, Kainat Ansar1, Mohammad Hashir Mehmood3,
Muhammad Zabih Ullah Raja1and Nadeem Javaid1,∗
Abstract In this paper, we perform performance evaluation of home energy man-
agement system (HEMS) for demand side management (DSM) in smart grid. In this
work, smart home is equipped with HEMS, smart meter, and smart appliances for
two-way communication between utility and consumer. HEMS performs scheduling
of smart appliances based on meta-heuristic techniques to balance load for whole
day to avoid peak creation in any hour. Smart meter performs electricity cost calcu-
lation for consumed energy based on time of use (ToU) pricing signal provided by
utility. Our focus is to efficiently handle user demand, reduction in peak-to-average
ratio (PAR) and electricity cost minimization. The implemented meta-heuristic tech-
niques in this work are: Enhanced differential evolution (EDE), harmony search al-
gorithm (HSA), bacterial foraging algorithm (BFA), and genetic algorithm (GA).
The simulation results show the performance of HEMS based on optimization tech-
niques using ToU.
1 Introduction
Smart grid is bi-directional communication between user and utility by installing
smart meter and HEMS for DSM. This bi-directional communication is useful for
energy optimization, load balancing, electricity cost reduction and minimizing PAR
[2], [3], [4]. Load balancing is basically efficient management of energy consump-
tion by balancing load in on-peak hours and off-peak hours [13], [14]. User tries to
minimize electricity cost by shifting load from on-peak hour to off-peak hours. In
addition, this load shifting creates peak in any other hour which ultimately affects
PAR. Thus, load balancing is an efficient way to avoid PAR for complete operational
1COMSATS Institute of Information Technology, Islamabad 44000, Pakistan
2National University of Sciences and Technology, Islamabad 44000, Pakistan
3Roots Millennium School, Islamabad 44000, Pakistan
∗Correspondence: www.njavaid.com, nadeemjavaidqau@gmail.com
1

2 Sajeeha Ansar et al.
time of smart appliances. However, utility wants reduction in PAR for efficient sup-
ply of energy to fulfil user demand in any hour. In terms of user comfort, there is
always tradeoff between electricity cost and waiting time [4]. While, performing
load balancing through HEMS by optimization techniques, this ultimately affects
user defined scheduled operational time of smart appliances.
Smart meter takes utility pricing schemes and appliances schedule as input. Con-
sequently, smart meter performs electricity cost calculation based on energy con-
sumption by smart appliances schedule provided by HEMS. Smart meter helps in
reducing electricity cost by providing calculations to HEMS and utility [2], [3], [4].
HEMS is also known as scheduler in smart grid environment. HEMS is respon-
sible for two-way communication between smart meter and user demand. HEMS
helps in giving efficient demand response (DR) which is also known as DSM [2].
HEMS performs scheduling of smart appliances to balance load for operational time
according to user demand. This scheduling helps in reducing PAR in any hour to
minimize electricity cost.
Demand response model consists of aggregator, which is responsible for com-
munication between house hold appliances for their scheduling and running time
for predefined time duration [1]. DSM became successful for multiple homes with
contribution of generators, retailers, large users and aggregator. Demand from mul-
tiple user including smart home, smart business, and industrial area is controlled by
control center. Power is generated from wind turbine, conventional, hydroelectric,
nuclear, and solar panels [2].
In [3], the grid power system with renewable energy resources is modeled to
generate extra energy. The system consists of batteries, multiple users, smart appli-
ances, and energy providers work efficiently under power generation from renew-
able resources. In this paper [4], a system model is proposed to examine system
performance, energy and power consumption calculations, performance parameter
optimization and energy management. The system consists of smart meter, HEMS
controller, and appliances. Local area network (LAN) is used to share appliances
control. Appliances are categorized in schedulable (flexible) and real time devices
(less flexible).
DSM in smart grid based on regularization, bi-directional framework, and new-
ton method is applied in this paper [5]. The regularization helps to improve in-
terrupts in appliance scheduling and minimizing PAR by reducing duration. Bi-
directional framework helps in bidirectional communication among agents and
HEMS. The newton method helps in fast convergence and this increase user comfort
in terms of waiting time. The HEMS system model in smart grid is shown in fig 1.
In this paper, section 2 reflects the related work in smart grid. Section 3 illustrates
system model of this research work. In addition, meta-heuristic techniques including
HSA, EDE, BFA, and GA are highlighted in section 4. Moreover, simulation results
and discussions are presented in section 5. Finally, complete work is concluded in
section 6.

Demand side management using meta-heuristic techniques and ToU in smart grid 3
Fig. 1 Home Energy Management System in Smart Grid
2 Related Work
In this section, the previous work done in smart grid is described. Energy optimiza-
tion is focus in smart grid to control energy consumption. Energy maximization,
load balancing, controlling power consumption, using renewable energy sources,
and minimizing PAR is achieved by different work in smart grid.
In [6], authors focused on load scheduling and energy consumption using renew-
able energy resources. Appliances are categorized as interruptible and fixed appli-
ances. In this proposed system, each user can sell excessive power generated by
renewable energy sources. To model the connection between user and generated
by renewable energy resources, a game theoretical approach adopted by authors.
The proposed system has reduced energy consumption from power grid, and mini-
mized electricity cost. Load balancing and PAR problems are addressed by gener-
ating power from renewable energy resources. However, cost of renewable energy
resources is neglected in this proposed system.
In this paper [7], the authors proposed a decentralized system to establish a con-
nection between demand response and user. The proposed system is used to manage
load to avoid peak in on-peak hours. Balancing load in on-peak hour cause reduction
in electricity cost. However, the waiting time of user in increased. In this proposed
system, when the HLM does not find acceptable load profile then complete load
shifted to peak hours. Convergence is very slow and it required many iterations to
finally receive balanced load profile.

4 Sajeeha Ansar et al.
Efficient cost reduction for residential load scheduling in smart grid is proposed
in [8]. The proposed load scheduling algorithm works for cost reduction in DSM
system. Day ahead bidding and RTP mechanisms are used in this proposed sched-
uler. The distributed energy resources (DERs) are used in proposed load scheduling
algorithms; this cause high computational cost and increase user waiting time. The
service charges are also considered for better results in cost efficiency. The cost of
implementing DERs is neglected in this proposed system.
In [9], the proposed DSM technique deals with the load management in resi-
dential area for single and multiple homes. The authors focused on maximizing user
comfort, minimizing electricity cost and PAR. The proposed system is a hybrid tech-
nique of genetic algorithm and wind driven algorithms. In this proposed system, the
scheduler shift load from on-peak hours to off-peak hours for interruptible appli-
ances to reduce energy consumption in peak hour. User comfort in terms of waiting
time of appliances is neglected in proposed hybrid technique during load balancing.
Energy consumption scheduling mechanism by load balancing for residential
area in smart grid is proposed in [10]. In this proposed technique, author achieved
balanced load for each hour using proposed scheduling mechanism. The proposed
system schedule appliances to minimize energy consumption in on peak hour, max-
imizing user comfort by scheduling power and operational time. Load balancing
achieves minimum power consumption, electricity cost and PAR in peak hours.
In [11], the authors focused on reducing electricity cost and PAR by schedul-
ing power usage in smart homes. The authors proposed energy management system
(EMS) and scheduling method for proposed EMS. In this proposed system, authors
combined RTP and inclining block rate (IBR) pricing schemes. Hybrid pricing sig-
nals performed better to achieve reduction in electricity cost and PAR. The authors
focused on optimizing power consumption, however proposed system implemented
by strong assumptions of same power consumption in each hour.
In this paper [12], the authors proposed a system to schedule appliances in smart
homes. The focus of the work is to minimize electricity cost by load balancing in
peak hour for interruptible appliances. Ahead of time pricing signal is used in this
proposed model. User comfort is compromised in terms of waiting time to run re-
quired appliance. The proposed system is based on wireless connection between
smart meter, smart appliances, and system model. However, unavailability of inter-
net cause rise in PAR and load unbalancing cause maximizing electricity cost.
3 System Model
The proposed system model is composed of 12 appliances for single smart home
in smart grid. However, the authors in [9], have used same appliances classification
and power ratings for multiple homes using RTP pricing signal. Appliances are cat-
egorized as: shift able and non-shift able appliances as shown in fig.2. In our system
model, we have used ToU pricing signals for electricity cost calculation. The sched-
uler performs appliance scheduling for 24 hours according to TOU pricing signal as

Demand side management using meta-heuristic techniques and ToU in smart grid 5
Fig. 2 System Model
1 2 3 4 5 6 7 8 9 101112131415161718192021222324
Time (hours)
8
10
12
14
16
18
Price (cent)
Price Signal
Fig. 3 TOU Pricing signal
shown in fig. 3. The scheduler performs load balancing in on-peak hours and off-
peak hours to minimize PAR in any hour. Smart meter is installed in smart home for
bi-directional communication between utility and HEMS. Appliances classification,
life time and power ratings are shown in table 1. We have used equation 1 to calcu-
late electricity cost for 24 hours, equation 2 is used to calculate load and as shown
in equation 3 PAR is calculated using this equation. These equations are used by the
authors in [13] for electricity cost, load, and PAR calculations.
Electricity-Cost =
24
∑
hour=1
Ehour
Rate ×PApp
Rate (1)
Load =PApp
Rate ×App (2)
PAR =max(load2)
ˆa(load2)(3)
4 Meta-heuristic Techniques
In this section, meta-heuristic techniques have been discussed to perform appliances
scheduling by HEMS for DSM. 12 appliances of single home are scheduled using
these techniques for HEMS performance evaluation.

6 Sajeeha Ansar et al.
Table 1 Appliances and Power Ratings
Appliance Categories Appliance Name Power Rating (kwh) Life Time (hours) Deferrable Load
Elastic and shift able appliances Space Heater 1 9 Yes
Space Heater 1 9
Heat Pump 0.11 4
Portable Heater 1.00 5
Water Heater 4.50 8
Clothes Washer 0.51 9
Clothes Dryer 5.00 5
Dishwasher 1.20 11
First-Refrigerator 0.50 24
Fixed load appliances Fan 0.5 11 No
Furnace Fan 0.38 8
Central AC 2.80 12
Room AC 0.90 5
4.1 EDE
It consists of multiple efficient features over other optimization techniques. In EDE,
main steps are: mutation, crossover, and selection phase. Initially, random popu-
lation is generated using upper and lower bounds for random function. In mutation
phase, randomly three target vectors are selected from initially created memory. One
mutant vector is formed by taking difference to any two previously selected target
vector and adding the results in third target vector as shown in equation 4 given
below.
Vj,i,G+1=xbest,G+F(xr1,G) + F(xr2,G−xr3,G)(4)
In crossover phase, a random value is generated and compared with crossover rates.
However, these crossover rates are predefined in EDE. If the random number is less
than crossover rate then information is taken from selected mutant vector. Mean-
while, if random value is greater than the crossover rate then target vector becomes
trial vector of this optimization technique. In EDE, it generates 5 trial vectors based
on 5 distinct crossover rates as shown in equation 5 to equation 9. After getting five
trial vectors, fitness of these five trial vectors is calculated and the vector having
minimum value is selected for final trial vector.
U1j,i,G+1=(Vj,i,G+1,if(randb(j)) ≤0.3=Irand
xj,i,G,if(randb(j)) >0.36=Irand (5)
U2j,i,G+1=(Vj,i,G+1,if(randb(j)) ≤0.6=Irand
xj,i,G,if(randb(j)) >0.66=Irand (6)
U3j,i,G+1=(Vj,i,G+1,if(randb(j)) ≤0.9=Irand
xj,i,G,if(randb(j)) >0.96=Irand (7)

Demand side management using meta-heuristic techniques and ToU in smart grid 7
U4j,i,G+1=(randb(j)) ×xj,i,G(8)
U5j,i,G+1=(randb(j)).vj,i,G+ (1−(randb(j)))×xj,i,G(9)
In selection phase, the selected trial vector is compared with target vector and the
vector having minimum fitness value is selected for next generation. The authors in
[14] have used equation 5 to equation 9 for EDE algorithm 1. EDE Parameters and
Values are shown in table 3.
4.2 HSA
The steps involved in this evolutionary algorithm are: random initial population,
harmony improvising process, memory consideration, and pitch adjustment for new
generation. Initially, harmony memory is created randomly using random function
by specifying upper and lower number range. After completing first step of ini-
tial random memory creation using equation 10, harmony improvising process gets
started. HSA Parameters and Values are shown in table 4.
xi,j=l j +rand() ×(Uj−lj)(10)
In harmony improvising step, generation of new harmony is based on harmony
memory consideration rate (HMCR) and pitch adjustment ratio. In this step, a ran-
dom number is generated and compared with HMCR. If the generated value is less
than HMCR then the existing harmony memory contributes in selecting new har-
mony. If the randomly generated number is greater than HMCR then a new random
value is generated to create new harmony using equation 11 as given below.
Vi,j=(xrandb(j),i f (rand()) <HMCR,
lj+(rand()) ×(Uj−lj),else (11)
The harmony selected in memory consideration process further go through the pro-
cess of pitch adjustment ratio. In this step, a random number is generated and if it
is less than pitch adjustment ratio then the existing harmony memory contributes
in selecting new harmony. If the randomly generated number is greater than pitch
adjustment ratio then a new random value is generated to create new harmony using
equation 12.
Vi,j=(Vi,j±(rand()) ×bwj,(rand()) <PAR,
Vi,j,else (12)
After getting a final new vector, compare it with worst harmony value in existing
harmony memory. If new results are better than worst, replace it in previous worst
harmony value in existing harmony memory. HSA complete steps are shown in algo-

8 Sajeeha Ansar et al.
Table 2 EDE Parameters and Values
Parameter Value
Population Size 30
Number of appliances 12
Maximum pitch adjustment rate 0.9
Minimum pitch adjustment rate 0.4
Harmony memory consideration rate 0.9
Maximum bandwidth 1.0
Minimum bandwidth 0.0001
Maximum iteration 100
Lower limit 0.1
Upper limit 0.9
Stopping Criteria Max. iteration
Table 3 HSA Parameters and Values
Parameter Value
Population Size 30
Number of appliances 12
Number of target vectors 3
Number of mutant vector 1
Number of crossover rates equations 5
Number of trail vectors 5
Maximum iteration 100
Lower limit 0.1
Upper limit 0.9
Stopping Criteria Max. iteration
rithm 2. Detail description of all symbols, parameters used in equations, algorithms
are shown in table 5 and 6, respectively.
4.3 BFA
Among nature inspired optimization techniques, BFA is most commonly used opti-
mization technique. BFA technique is based on real bacteria foraging process. In this
optimization algorithm, stochastically and collectively it allows the cell to swarm for
optimal solution. The steps involve in BFA are: chemotaxis step, reproduction and
elimination-dispersal. In chemotaxis step, it is life duration of the bacteria based on
number of chemotactic steps. In reproduction cell, it is basically selection phase of
this algorithm. In this step, bacteria cells performed well over their life duration are
selected for next generation. Elimination dispersal step is based on fitness function
in which previous expired cells are discarded and new population is inserted.
4.4 GA
To find optimal solutions, GA is the most popular optimization technique. The con-
cept behind genetic algorithm is alike chromosomes. The main steps involve in GA
are: selection, crossover, and mutation. In selection phase, initially population is
generated randomly which is basically representation of chromosomes. Then for se-
lection process this generated population is broken down. Crossover phase is then
performed on selected chromosomes from selection phase. In mutation phase, bits
are changed randomly and finally a chromosome is selected based on fitness func-
tion. GA is relatively better algorithm for optimal solution [4], [9]. While, prob-
abilistic nature of GA does not guarantee optimality. GA performs best for larger

Demand side management using meta-heuristic techniques and ToU in smart grid 9
population, while BFA performs best for small population [13]. Execution time of
GA is less as compare to other meta-heuristic techniques [14].
Algorithm 1 HSA
Input:(HMS,NVAR,HMCR,PAmin,PAmax,BW min,BWmax ,maxItr,X l ,
Xu)
1: for Hour =1→24 do
2: if hour <24 then
3: Select electricity cost of next hour
4: else
5: Select electricity cost of current hour
6: end if
7: for j=1→maxItr do
8: Pitch adjustment
9: for p=1→NVAR do
10: Bandwidth adjustment
11: end for
12: for I=1→NVAR do
13: if rand(1)<H MCR then
14: Select new harmony from existing
15: if rand(1)<PA then
16: V[´
i,´
j] + rand()
17: else
18: V[´
i,´
j]−rand()
19: end if
20: else
21: Randomly select new harmony
22: end if
23: end for
24: end for
25: if new <HM(worst)then
26: HM(worst) = new
27: else
28: HM(worst) = HM(worst)
29: end if
30: end for
5 Simulation Results and Discussions
In this section, the simulation results show performance comparison of implemented
meta-heuristic techniques. Moreover, meta-heuristic techniques do not guarantee
optimal solutions [9]. In addition, computational time and optimal solutions are im-
portant parameters in research. The implementation of meta-heuristic techniques is
based on random initial population generation process. Therefore, the confidence
interval is calculated based 10 times average to evaluate performance of HEMS. Fig
3 elucidates the pricing rate for 24 hour. TOU is commonly used tariffs for electric-

10 Sajeeha Ansar et al.
Algorithm 2 EDE
Input:(po psize,NA,maxItr,Xl ,X u)
1: B1=rend perm(popsize)
2: Randomly select 3 target vectors T, T1, T2, T3
3: MutantvectorM =T1+0.5∗(T2−T3)
4: for m=1→maxItr do
5: for n=1→NA do
6: if rand(1)>=0.3then
7: t1 = T else
8: t1 = M
9: end if
10: end for
11: for n=1→NA do
12: if rand(1)>=0.6then
13: t2 = T else
14: t2 = M
15: end if
16: end for
17: for n=1→NA do
18: if rand(1)>=0.9then
19: t3 = T else
20: t3 = M
21: end if
22: end for
23: for n=1→NA do
24: if rand(1)∗t argetvector then
25: t4 = Telse
26: t4 = M
27: end if
28: end for
29: for n=1→NA do
30: if rand(1)∗Mutantvector +1rand (1)∗targetvector then
31: t5 = Telse
32: t5 = M
33: end if
34: end for
35: F1 = eletricitycost * t1
36: F2 = eletricitycost * t2
37: F3 = eletricitycost * t3
38: F4 = eletricitycost * t4
39: F5 = eletricitycost * t5
40: New trial vector = min [F1, F2, F3, F4, F5]
41: if Newtrialvector >T1then
42: New target vector = New trial vector
43: else
44: New target vector = T1
45: end if
46: end for

Demand side management using meta-heuristic techniques and ToU in smart grid 11
Table 4 Detail Description of Symbols
Symbols Description
Ehour
Rate Electric Rate per Hour
PApp
Rate Power Rate of an Appliance
App Appliance
Vj,i,G+1Mutant Vector
xbest,GBest target vector
xr1,GFirst target vector
xr2,GSecond target vector
xr3,GThird target vector
xj,i,GTarget vector
U1j,i,G+1First trial vector
U2j,i,G+1Second trial vector
U3j,i,G+1Third trial vector
U4j,i,G+1Fourth trial vector
U5j,i,G+1Fifth trial vector
xi,jInitial harmony memory
rand() Built in function for random value generation
ljLower limit of rand function
UjUpper limit of rand function
vi,jNew harmony memory
bwjBandwidth
Table 5 Detail Description of Parameters
Parameters Description
HMS Population size HSA
NVAR Number of appliances in EDE
HMCR Harmony memory consideration rate
PAmax Pitch adjustment maximum value
PAmin Pitch adjustment minimum value
BWmin Bandwidth minimum value
BWmax Bandwidth maximum value
maxItr Maximum iteration
Xu Random function upper limit
Xl Random function lower limit
Popsize Population size EDE
NA Number of appliance in EDE
F1 Fitness value 1st trail vector
F2 Fitness value 2nd trail vector
F3 Fitness value 3rd trail vector
F4 Fitness value 4th trail vector
F5 Fitness value 5th trail vector
t1 1st trial vector
t2 2nd trial vector
t3 3rd trial vector
t4 4th trial vector
t5 5th trial vector
HM(worst) worst value from harmony memory
1 2 3 4 5 6 7 8 9 101112131415161718192021222324
Time (hours)
0
20
40
60
80
100
120
140
160
180
Electricity Cost (cent)
Unscheduled
HSA
EDE
GA
BFA
Fig. 4 Electricity Cost
1 2 3 4 5 6 7 8 9 101112131415161718192021222324
Time (hours)
0
2
4
6
8
10
12
Load (kWh)
Unscheduled
HSA
EDE
GA
BFA
Fig. 5 Per hour Load Demand
ity pricing which varies in all countries. Fig 4 clearly demonstrates electricity cost
consumed for 12 appliances during 24 hours. The electricity cost is high in on-peak
hours from 11th hour to 17th hour for unscheduled load. Electricity cost for sched-

12 Sajeeha Ansar et al.
Unscheduled HSA EDE GA BFA
0
500
1000
1500
2000
2500
Total Cost (cent)
Fig. 6 Total Cost
HSA EDE GA BFA
0
1
2
3
4
5
Waiting Time (hours)
Fig. 7 Waiting Time
uled algorithms is less then unscheduled because load is balanced by scheduling
techniques.
The load is completely distributed in off-peak hours and on-peak hours; it cause
electricity cost within range of 40 - 60 cents for off-peak hours. HSA algorithms
performs selection of fitness value less than worse; it enables the scheme to fully
distribute load in 24 hours. EDE algorithm compares electricity cost for two hours
including current and next hour; this helps to schedule high power consumption
appliances accordingly to reduce electricity cost. Load balancing cause balanced
energy consumption and reduction in electricity cost.
Fig 5 represents the load balancing through optimization techniques. Load is
balanced for 24 hours in on-peak hours to off-peak hours. To minimize the cost,
most of the schedulers shift load toward off-peak hours which cause maximization
in PAR. Load unbalancing through schedulers can lead the smart cities to starvation.
However, in our simulation results, it is clear that load is balanced in 24 hours.
Fig 6 shows total cost consumed during 24 hours. The total cost for unscheduled
is higher than scheduled algorithms. The behaviour of EDE is better among all opti-
mization techniques. Moreover, cost of BFA is highest among all implemented tech-
niques. Total cost for all implemented optimization techniques vary from 1500 cents
to 1700 cents. However, there is not a significant difference in total cost consump-
tion. However, these scheduling techniques help in reduction of total cost consumed
in 24 hours as compare to unscheduled case.
Fig 7 illustrates the total waiting time in appliances operational time for unsched-
uled and scheduled algorithms. The EDE optimization algorithm performs best in
terms of less waiting time of appliances through scheduling criteria. However, wait-
ing time for BFA is highest among all scheduled algorithms. Thus, total waiting time
is reduced as compare to unscheduled scenario. Moreover, user comfort in terms
of waiting time is compromised in implemented algorithm. These implemented

Demand side management using meta-heuristic techniques and ToU in smart grid 13
Unscheduled HSA EDE GA BFA
0
0.5
1
1.5
2
2.5
3
3.5
PAR
Fig. 8 PAR
Unscheduled HSA EDE GA BFA
0
20
40
60
80
100
120
140
160
Total Load (kWh)
Fig. 9 Total Load
schemes perform efficient where the user concern is to minimize electricity cost
and power consumption in on-peak hours.
Fig 8 clarify simulation result for PAR of implemented scheduled techniques
with comparison of unscheduled case. The result shows that the PAR is highest
for BFA. Moreover, HSA performs best among scheduling techniques. Appliance
having high power rate consumes high electricity cost in peak hours in unscheduled
scenario. In scheduled process, it balance load to avoid peak in any hours. Therefore,
this effect user comfort in terms of waiting time. However, total load for all opti-
mization techniques remains same as illuminated in fig 9. Total load is operational
time of all smart appliances which is to be completed for whole day. However, these
scheduling techniques help in reduction of total cost in 24 hours. Therefore, total
load for all implemented scheduled algorithms and unscheduled are same. More-
over, the number of appliances for both are same and their power consumption re-
quired for complete day remains same for all techniques. Load is balanced in all
scheduled techniques.
6 Conclusion
Efficient energy consumption, PAR reduction and load balancing for DSM is fo-
cus in smart grid to prevent starvation. In this paper, performance comparison of
meta-heuristic algorithms is evaluated in terms of cost minimization and PAR re-
duction. Efficient energy consumption is achieved through scheduler, which helps
in scheduling smart appliances within smart home. Smart grid helps in reducing
electricity cost by load balancing. The objectives of study are achieved in terms of
minimizing power consumption in peak hours to reduce electricity bills. Secondly,

14 Sajeeha Ansar et al.
balancing load in on-peak hours and off-peak hours to minimize PAR. The utility
comfort is achieved in terms of controlling energy consumption in on-peak hours.
However, user comfort in term of waiting time of appliances. However, there is
trade-off between electricity cost reduction and increase in waiting time.
References
1. Mhanna, Sleiman, Gregor Verbi, and Archie C. Chapman. ”A faithful distributed mechanism
for demand response aggregation.” IEEE Transactions on Smart Grid 7, no. 3 (2016): 1743-
1753.
2. Ye, Feng, Yi Qian, and Rose Qingyang Hu. ”A real-time information based demand-side man-
agement system in smart grid.” IEEE Transactions on Parallel and Distributed Systems 27, no.
2 (2016): 329-339.
3. Nguyen, Hung Khanh, Ju Bin Song, and Zhu Han. ”Distributed demand side management with
energy storage in smart grid.” IEEE Transactions on Parallel and Distributed Systems 26, no.
12 (2015): 3346-3357.
4. Basit, Abdul, Guftaar Ahmad Sardar Sidhu, Anzar Mahmood, and Feifei Gao. ”Efficient and
Autonomous Energy Management Techniques for the Future Smart Homes.” IEEE Transactions
on Smart Grid (2015).
5. Li, Chaojie, Xinghuo Yu, Wenwu Yu, Guo Chen, and Jianhui Wang. ”Efficient computation for
sparse load shifting in demand side management.” IEEE Transactions on Smart Grid 8, no. 1
(2017): 250-261.
6. Samadi, Pedram, Vincent WS Wong, and Robert Schober. ”Load scheduling and power trading
in systems with high penetration of renewable energy resources.” IEEE Transactions on Smart
Grid 7, no. 4 (2016): 1802-1812.
7. Safdarian, Amir, Mahmud Fotuhi-Firuzabad, and Matti Lehtonen. ”Optimal residential load
management in smart grids: A decentralized framework.” IEEE Transactions on Smart Grid 7,
no. 4 (2016): 1836-1845.
8. Ma, Jinghuan, He Henry Chen, Lingyang Song, and Yonghui Li. ”Residential load scheduling
in smart grid: A cost efficiency perspective.” IEEE Transactions on Smart Grid 7, no. 2 (2016):
771-784.
9. Javaid, Nadeem, Sakeena Javaid, Wadood Abdul, Imran Ahmed, Ahmad Almogren, Atif
Alamri, and Iftikhar Azim Niaz. ”A hybrid genetic wind driven heuristic optimization algo-
rithm for demand side management in smart grid.” Energies 10, no. 3 (2017): 319.
10. Zhu, Ziming, Jie Tang, Sangarapillai Lambotharan, Woon Hau Chin, and Zhong Fan. ”An
integer linear programming based optimization for home demand-side management in smart
grid.” In Innovative Smart Grid Technologies (ISGT), 2012 IEEE PES, pp. 1-5. IEEE, 2012.
11. Zhao, Zhuang, Won Cheol Lee, Yoan Shin, and Kyung-Bin Song. ”An optimal power schedul-
ing method for demand response in home energy management system.” IEEE Transactions on
Smart Grid 4, no. 3 (2013): 1391-1400.
12. Ma, Kai, Ting Yao, Jie Yang, and Xinping Guan. ”Residential power scheduling for de-
mand response in smart grid.” International Journal of Electrical Power and Energy Systems
78 (2016): 320-325.
13. Khalid, Adia, Nadeem Javaid, Abdul Mateen, Bilal Khalid, Zahoor Ali Khan, and Umar
Qasim. ”Demand Side Management Using Hybrid Bacterial Foraging and Genetic Algorithm
Optimization Techniques.” In Complex, Intelligent, and Software Intensive Systems (CISIS),
2016 10th International Conference on, pp. 494-502. IEEE, 2016.
14. Zafar, Ayesha, Samia Shah, Rabiya Khalid, Sardar Mehboob Hussain, Hassan Rahim, and
Nadeem Javaid. ”A meta-heuristic home energy management system.” In Advanced Informa-
tion Networking and Applications Workshops (WAINA), 2017 31st International Conference
on, pp. 244-250. IEEE, 2017.