Content uploaded by Ahmad Jaffar Khan
Author content
All content in this area was uploaded by Ahmad Jaffar Khan on Jul 05, 2021
Content may be subject to copyright.
A Hybrid Bacterial Foraging Tabu Search Heuristic
Optimization for Demand Side Management in
Smart Grid
Ahmad Jaffar Khan1, Nadeem Javaid1,∗, Zafar Iqbal2, Naveed Anwar3, Abdul Saboor1, Inzimam-ul-Haq4and
Umar Qasim5
1COMSATS Institute of Information Technology, Islamabad 44000, Pakistan
2PMAS Arid Agriculture University, Rawalpindi 46000, Pakistan
3University of Wah, Wah Cantt 47040, Pakistan
4COMSATS Institute of Information Technology, Wah Cantt 47040, Pakistan
5Cameron Library University of Alberta Edmonton Canada
∗Corresponding author: www.njavaid.com, nadeemjavaidqau@gmail.com
Abstract—With the advent of Smart Grid (SG), it provides
the consumers with the opportunity to schedule their power
consumption load efficiently in such a way that it reduces their
energy cost while also minimizing their Peak to Average Ratio
(PAR) in the process. We in this paper target the appliances
to schedule in such a way that it increases User Comfort
(UC) and decreases electricity consumption load which benefits
both consumer and utility. In this paper, we proposed hybrid
of Bacterial Forging Algorithm (BFA) and Tabu Search (TS)
Algorithm using different Operational time Interval (OTI) to
schedule appliances while balancing User Comfort which is the
main objective of the Demand Side Management (DSM). This
paper tries to reduce both waiting time and electricity cost
simultaneously in the new hybrid Bacterial Foraging Tabu Search
(BFTS) technique. Real time pricing (RTP) scheme was used to
get the total cost of electricity consumed. We compared the results
of proposed hybrid scheme with Bacterial Forging (BFA) and
Tabu Search (TS) Algorithm using different Operational time
Interval (OTI). The result shows effectiveness of using hybrid
Bacterial Foraging Tabu Search (BFTS) technique for Demand
Side Management (DSM).
Index Terms—Smart grid, Demand side management, Real
time pricing, Heuristic techniques
I. INTRODUCTION
In smart grid (SG) efficient electricity consumption using
modern technologies, is currently one of the significant target
area for research [1]. Advance metering groundwork and
demand management policies are important steps which help
us in development of distributed network for efficient energy
consumption.
Demand Side Management (DSM) programs enable con-
sumer to efficiently control their electricity consumption loads
by shifting their load from on-peak hours to off-peak hours
to avail various economic incentives [2]. DSM programs
developed in the early days where based on unidirectional
communication from the utility to consumer like direct load
control (DLC) [3]. The DLC is most successful for consumer
that use high electric load appliances while it had almost no
effect for consumers using many low electric load appliances.
Demand response (DR) programs has two major categories; 1)
Incentives based (DLC based) 2) dynamic pricing (time-of-use
pricing (TOU) and real-time pricing (RTP) schemes). In the
above pricing schemes, consumers are encouraged to control
their electricity load based on economic incentives voluntarily
to avoid peak load conditions [4]. Considerable number of
techniques have been worked on solve different forms of DMS
as explained in [5] [6].
In literature the DSM is mainly discussed as a way of
reducing Peak to Average Ratio (PAR) and electricity cost.
Different techniques have been proposed such as mixed-integer
linear programming [5], game theory [7], backtracking algo-
rithm [8] and evolutionary algorithm (EA) [6] for scheduling
DSM schemes but EA gives way impressive results as its
adjustable nature allow us to increase UC by changing load
patterns based on consumer lifestyle. In the [9] EA gives
much better results in comparison to simulated annealing as
well as greedy search algorithm technique while EA such as
Bacterial Forging Algorithm (BFA) looks for optimal schedule
for consumer appliances by using steps like chemotaxis,
reproduction and elimination-dispersal. In BFA suboptimal
schedules are replaced by the more optimal during elimination-
dispersal step. The time required to find an optimal schedule
depends on number of population steps. In our proposed BFA-
based scheduling we have limit the population steps to thirty
as increasing steps doesnt increases the optimality of results.
In this article we will use hybrid Bacterial Foraging Tabu
Search (BFTS) technique made from Tabu Search-based (TS)
and BFA-based scheduling approach to schedule consumer ap-
pliances in SG. The both schedules consumer appliances based
on RTP information received from utility. The consumption of
electricity is reduced in a specific time slot by shifting loads as
compared to reducing the load. Energy management controller
unit (EMCU) can only schedule appliances that are Elastic in
nature in the home unit. We have considered only single home
use case for our simulation with different OTIs.
Remaining article is organized as follows. Related work
is presented in section II. Problem statement is described in
section III. System model in section IV. The simulation results
and their discussion in section V. The conclusion in section
VI.
II. REL ATED W OR K
Lots of work is being done by researchers nowadays in the
field of SG, Scheduling appliances on the basis of various
schemes.Some articles related to SG are discussed below:
K Muralitharan [10] put forward an optimization method
implemented using Multiobjective EA to optimize energy
consumption costs used by the consumer which helps in load
balancing so amount of time delay that occurs on execution
of an appliance is minimize. They used Time of Use (TOU)
pricing model in their paper. The results given in [10] reveals
that a consumer can use power under the threshold level to
bypass paying additional in term of utility bill thus allowing
consuming more power on lower costs. while [11] Awais pro-
posed how genetic algorithm can be used to reduce electricity
cost by shifting several electric appliances from on-peak hours
to off-peak which according to results helps in reducing cost
and PAR.
Study [12] proposed a Genetic DSM model to lower PAR
and waiting time of home appliances for home consumers. In
the paper they used genetic DSM with 20 different consumer
who were using different types of appliances with each of
them having different characteristics from each other. This
model used RTP scheme in its simulations. The results of
the simulation presented in G-DSM model showed lowering
of individual and 20 consumer costs by 39.39% ,45.85%
respectively and PAR by 17.17%, 45.24% for individual and
20 consumers respectively.
In paper [13] model is proposed for lowering the electricity
cost using dynamic pricing scheme to prevent peak hours. This
model comprises of smart meter, control node wireless sensor
networks and in-home display. Data flow between utility
and consumer smart meter is controlled by Advance meter
infrastructure and smart meter controls both master controller
and Advance meter infrastructure. The master controller coor-
dinates and controls both the controllable and uncontrollable
appliances schedules and the most efficient schedule is sent to
each control node through wireless sensor networks while in
home display starts the complete process.
Game theory based on dynamic pricing with respect to
residential and commercial sector of Singapore electric market
was studied [13]. Half hourly RTP, TOU and Day-Night (DN)
pricing schemes are assessed and compared with each other
to find which of the three pricing strategies help in reduction
of peak load reduction. Layered model [15] for DSM which
comprises of power generator, demand response aggregator
and consumers was proposed. Consumer discomfort level are
taking in to consideration in demand side management and this
model leads us to multiobjective problem thus allowing the
multiobjective algorithm to be used to find Pareto efficiency
helping us to find the fair solution.
Worked [16] on the economic effects of DMS in the smart
grid where electricity generation cost is one of the parameters.
They used IEEE 24 bus RTS with optimal power flow to get
the cost of power generation of 24 hours. They used day ahead
pricing scheme and implemented load data DSM to be able to
lower the power generation costs and in the end comparing
the cost of total power generation. They used differential
evolution to explain the optimal power flow (DC-OPF) in each
hour while finding a recent load curve. Another study [17] in
which authors proposed a model for distributed control system
based on Organization Multiagent System (MAS). The main
objective of this model was to address peak load problem
using Cyber Physical systems. An Artificial Immune Network
algorithm was implemented by the control decision process of
the system to regulate the energy consumption of all devices
connected to smart grid. In the first step ASPECS (Agent-
Oriented Software Process for Engineering Complex Systems)
methodology was used to develop the proposed models which
during simulation lowered energy consumption by 20%. They
used day ahead prices scheme in their simulations.
Proposed [18] a model on how to intelligently coordinate the
plug-in electric vehicles (PEV) demand of energy consumption
in a distribution system. They proposed a system based on
giving each PEVs a score using fuzzy expert system depending
upon PEV attributes like state of charge, battery capacity,
charger max rating and departure time of the vehicle and
the PEV efficiently charged to increase satisfaction of users
without breaking the grid constraints. The results of the
simulation indicate that PEVs which are more critical, having
shorter parking time and high charge time required benefit the
most among thus increase in user satisfaction.
Proposed [19] a hybrid approach of BFA and genetic
algorithm (GA) for cost minimization and load management
by shifting all the load form on-peak to off-peak hours.
They used RTP scheme in their hybrid model whose results
after simulations verify that the model reduces total cost
of electricity and peak average ratio. Another similar study
[20] proposed GA and Earthworm Optimization Algorithm
based schemes for minimizing PAR, and energy cost and
maximizing UC by scheduling appliances in home energy
management using Real time pricing scheme. Author [21]
gave comparative performances analysis of different heuristic
algorithms which include harmony search algorithm, enhanced
differential evolution and hybrid of both harmony search
differential evolution. They used multiple knapsack theory
in their problem that the highest capacity of the consumer
electric consumption must be in range compared to his electric
bill which he can bear. They used RTP, day ahead and TOU
pricing schemes in their simulations. The table.1 below gives
us a comparison between different techniques along with their
achievements and Drawbacks.
III. PROB LE M STATEMENT
DSM programs are developed to solve energy shortage
problems, generated due to increase in consumption of energy.
Load management based on DR is most used DSM strategy,
which causes load to be shifted from on-peak towards off-peak
hours [2], thus reducing energy consumption cost in process.
But this causes peaks to generate at off-peak hours, while also
increasing appliances operational waiting time. Many systems
like [10]- [21] have been developed with aims to minimize
cost, PAR, and User comfort while using OTI of 1 hour. In
which trade off occurs between cost and PAR, and cost and
user comfort, which are not fully considered by most of the
researchers. As authors in [11], [12], [14], [16], [17], [19], [20]
and [21] never considered comfort. Few researcher ignored
PAR in [10], [11], [14], [16] and [18] due to which a burden
was created on utilities due generation of peaks on off peak
hours.
Cost minimization and shifting load to off peak hours does
not makes a system efficient, many other factors are involved.
User comfort in relation to waiting time of appliance is one
of the factors involved. Operation of the appliances need to
be flexible so that their operation can be rescheduled in such
a way that it neither generates peaks nor it effects cost and
PAR too much.
Fig. 1: Proposed HAN architecture
IV. SYSTEM MODEL
In a standard Home area network (HAN) as shown in Fig 1,
appliance load is generally divided in to two categories man-
ageable and non-manageable loads. Most of the past papers on
energy management work is being mainly done to minimize
manageable loads as it consumes the most energy and are
easier to predict in HAN. We can further sub categorized
manageable load in to three different categories.
A. Brustload shift-able
Appliances in these categories are those that can be delayed
and can be interrupted once they have started execution until
the scheduler allows them to run again. E.g., vacuum cleaner,
water heater, etc.
B. Brustload non-interruptible
Appliances in these categories are those which can be
delayed but once started the scheduler wont be able to stop
them in the middle of the execution until they complete their
required execution time. E.g., washing machine, dish washer,
oven, etc.
C. Base load
Appliances are considered as base loads those that cant be
delayed and interrupted in any case. E.g., AC, refrigerator, etc.
Non-manageable loads are of those that are interactive in
nature which allows for little scheduling window. Moreover,
energy consumption of these appliances is insignificant. E.g.
TV, lights, PC, phones, etc.
D. Detail of different home appliances
In our simulations we studied an average size modern home
having following major appliances: vacuum cleaner, water
heater, water pump, Refrigerator, air conditioner (AC), Dish
washer, washing machine, Cloth dryer, oven morning, over
usage in evening, electric vehicle. We understand that each
appliance has a defined power consumption and time cycle
required to complete its execution and we took these values
from [22] for our simulations shown in Table II. We have used
RTP to determine electricity cost in our simulations. Now we
will describe appliances in our home separately.
1) Vacuum cleaner:The vacuum cleaner is used for 360
minutes a day while its power load is 0.7kW. Vacuum cleaner
is categorized in the Brustload shift-able category.
2) Water heater:The water heater remains on for 720
minutes during a day and its power load is 5kW. It is
categorized in the Brustload shift-able category.
3) Water pump:The water heater remains on for 720
minutes during a day and its power load is 5kW. It is
categorized in the Brustload shift-able category.
4) Dishwasher:Dishwasher has an operation time of 105
minutes during a day and it has power load of 1.35kW.
Dishwasher is categorized in the Brustload non-interruptible
category.
5) Washing Machine and Cloth Dryer:Washing machine
and Cloth dryer work in sync to each other. The washing
machine takes 45 minutes to complete its operation while 0.8
kW of power load. After 15 minutes execution of washing
machine dryer starts and takes 60 minutes to complete while
power load of 2.4 kW during process. Washing machine
and cloth dryer are both categorized in the Brustload non-
interruptible category.
6) Oven Usage in Morning:Oven is one of those appli-
ances in the house hold that is used more than once a day. As
it is used for cooking breakfast in the morning and making
lunch/dinner in the evening. So, we can treat oven used in
the morning and evening as separate appliances. Oven used in
the morning is only for 30 minutes having a power load of
1.2kW. Oven is categorized in the Brustload non-interruptible
category.
7) Oven Usage in Evening:Oven used in the evening is
for longer time and more than one burner is used. The oven is
usage runs for 90 minutes in the evening while having power
load of 1.2kW.
8) Electrical Vehicle:Electrical vehicles nowadays are
gaining acceptance as alternative to petroleum based vehicles
among public. The batteries of Electric vehicles are mostly
charged using home electricity and it takes 150 minutes
to fully charge its batteries while having a power load of
3kW. Electrical vehicle is categorized in the Brustload non-
interruptible category.
TABLE I: Summary of Related Work
Technique(s) Targeted Area Objective(s) Limitation(s)
Multiobjective Optimization Technique
[10]
Residential area Cost minimization and UC
maximization
Compromising on PAR
An Efficient Genetic Algorithm [11] Residential commercial and
industrial area
Cost minimization Compromising on PAR and UC
A generic demand-side management
model [12]
Residential area PAR, Waiting time and Cost
minimization
Compromising on UC
Optimal Residential Appliance
Scheduling [13]
Home energy management Cost and UC minimization Initial Setup cost is ignored
Game-Theory based dynamic pricing
strategies [14]
Residential and commercial
area
Cost minimization Ignoring UC and PAR
A Layered Approach [15] DSM UC and energy efficiency Ignoring Cost
Optimal Day-Ahead Scheduling in SG
[16]
DSM Cost reduction Ignoring UC and PAR
Multiagent model for Distributed Peak
Shaving System [17]
DSM PAR reduction Ignoring UC and Cost
Online Intelligent Demand Manage-
ment of Plug-In Electric Vehicles in
Future Smart Parking Lots [18]
Distribution systems UC maximization Ignoring Cost and PAR
Demand Side Management using Hy-
brid Bacterial Foraging and Genetic Al-
gorithm Optimization Techniques [19]
DSM Cost and PAR minimization Ignoring UC
Genetic Algorithm and Earthworm Op-
timization Algorithm for Energy Man-
agement in SG [20]
DSM Cost and PAR minimization Compromising on UC
Towards optimization of metaheuristic
algorithms for IoT with HAS, EDE and
HSDE [21]
DSM Cost and PAR minimization Compromising on UC
TABLE II: Appliances data used in our Simulations
Group Appliance Power
Rating
(kWh)
Daily
Usage
(hrs)
Water Heater 5 12
Interruptible
Load
Vacuum
Cleaner
0.7 6
Water Pump 1 8
Base Load Refrigerator 1.2 18
AC 2.5 15
Dish washer 1.35 1.45
Washing ma-
chine
0.8 .45
Cloth Dryer 2.4 1
Non-
Interruptible
Oven Morn-
ing
1.2 .30
Load Vacuum
cleaner
1.2 1.45
Electrical car 3 2.30
9) Refrigerator:Refrigerator is one of those appliances
that work for all day long with only going to off state when
the inside temperature is lower than set one and turn on when
the inside temperature is higher than the set one. we have
used 18 hours as the estimated time of refrigerator runs, and
its power load is 1.2kW. Refrigerator is categorized in the
Baseload category.
10) Air Conditioner:Air conditioner is also an appliance
whose runtime depends on temperature of the rooms. If the
rooms temperature goes lower than the set one it turns off for
a while until temperature the temperature it not high again.
So, we estimated that it runs 15 hours and its power load is
2.5kW. Air conditioner is categorized in the baseload category.
E. Duration of operation
We are using in our simulation time slots of 96, 48, 24. In
the 96 timeslot each slot represents 15 mins and in 48 timeslot
each slot represents 30 mins of the day while in 24 timeslot
each time represents 1 hour of the day.
F. BFA
Animal that are successful in foraging strategies are favored
by nature as compared to those that are poor at it thus are
eliminated. The generations eliminated are replaced by healthy
ones. The main accomplishment of this algorithm is that it
allows the bacteria cell to swarm together towards an optimal
solution. The following three successive steps are performed to
accomplish this. a) Chemotaxis this step is used to know about
life time of bacteria and we find out fitness pi of bacteria by
finding closeness of it to other bacterias new position i after a
tumble and the step size Ci is between [-1,1] in the direction
of the tumble. b) Reproduction where reproduction of next
generation takes place by cells which have done good in their
lifetime. c) Elimination dispersal last step in which poor cells
are eliminated. Algorithm 1 gives us the idea on how the BFA
works.
G. TS
Basically, a neighborhood optimization algorithm [23]. TS
uses neighborhood searching technique to keep searching
repeatedly from a potential solution for an even better solution
in a neighborhood unless a criterion to stop the search is met.
It has following steps: a) Create a neighborhood solution from
potential solution. b) Create Tabu List (TS) which stores all the
forbidden moves. c) Add Aspiration criteria (AC). c) Perform
Algorithm 1 BFA
1: procedure
Initialize parameters;
Initialize the location of the population;
2: for j = 1:Ncdo
3:
4: for j = 1:Npdo
5: calculating initial location
6:
7:
8: for e = 1:(a-1) do
9: Compute fitness: J=Fit
function(i)
10: end for
11: s=0
12:
13: while s < Nsdo
14: swim loop
15: if J(i) ≤Jlast(i) then
16: Evaluating current fitness
17: J(i)=J(i)
18: calculate new position
(i,:)
19:
20: for e = 1:(a-1) do
21: fitness function
Fitfunction(i)
22: end for
23:
24: else
25: s=Ns
26: end while
27: end for
28: Appliances with best fitness are
saved.
29: end for
TS. d) After each loop updates the TL. e) If TL is not updated
then update AC. f) Checks ending criteria to end execution.
Algorithm 2 gives us the idea on how the TS works.
V. SIMULATIONS AND DISCUSSIONS
In the following section we will be discussing the results of
simulations carried out using MATLAB by applying BFA, TS
and our new hybrid BFTS technique. We will evaluate them
based on PAR, Energy consumption, UC, cost of electricity
and feasible region. We used in our simulations 11 appliances
in our home. Each appliance belongs to a specific group and
has specific power rating which is given in table 1 above. We
are using RTP scheme in our simulations to calculate cost of
energy consumption.
A. Energy consumption
Given above Fig. 2a, 2b, 2c shows us the power consump-
tion pattern when using BFA Scheduling, TS, Unscheduled
Algorithm 2 TS
Initialize all parameter
2: x’=best solution among trails
S(x) sample of neighborhood S(x)εN(x)
4: Current solution x0εX
Set tabu list TL=150
6: Set aspiration criteria AC=0
Set iteration counter K = 0
8: Randomly generate initial solution
Randomly generate trail solutions S(x)εN(x)and sort them
in ascending order to obtain SS(X).
10: Let x’ be the best trail solution in SS(X)
ifx0>x
12: X00 =x
else
14: X”=x’
end while k¡= number of iterations
16: For i=1:TL
Perform tabu search
18: If X00 >X0
Update X” in tabu list
20: Else X” not present in tabu list
update X” in aspiration criteria
22: end end end
If stopping criteria is satisfied
24: perform termination
else k=k+1
26: end
and hybrid BFTS under different OTIs. The output shows us
that BFA, TS and hybrid BFTS scheduling is more efficient in
lowering load management as compared to unscheduled power
consumption. When using hybrid BFTS scheduling power
consumption has less peaks as all peaks are scheduled to off
peak hours from on peak and these peaks remain similar in
different OTIs showing the time interval has no effect on Load
peaks. So, we can conclude from the above figures that hybrid
BFTS load management technique is shifting appliances load
to off peak hours.
Energy consumption graph for multiple homes using OTI of
1 hour is shown in Fig. 3a, 3b, 3c for 50, 30, and 20 homes. All
multiple homes energy load graphs show shifting of load by
different scheduling algorithms. Increase in number of homes
only increases load but not scheduling algorithms behavior.
B. Cost of Electricity
Electricity cost which the consumer has to pay when using
BFA, TS, Unscheduled and hybrid BFTS scheduled under
different OTI is shown in Fig. 4. As shown in the Fig. 4
when using OTI of 1 hour electricity cost for BFA scheduled,
TS scheduled, BFTS scheduled and Unscheduled electricity
consumption is $1670.20, $1630.20, $1600.50 and $1920.56
respectively which shows that hybrid BFTS scheduling elec-
tricity costs considerably less as compared to other techniques.
When using OTI of 30 mins and 15 mins Fig. 3 clearly shows
that hybrid BFTS technique still has lower cost as compared
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Time (hours)
0
5
10
15
Load(kWh)
Unscheduled
BFA
TS
BFTS
(a)
0 5 10 15 20 25
Time (hours)
0
50
100
150
200
250
300
Load(kWh)
Unscheduled
BFA
TS
BFTS
(b)
1 2 3 4 5 6 7 8 910 11121314151617181920 21222324252627282930 313233343536373839 40414243444546474849 50515253545556575859 60616263646566676869 70717273747576777879 80818283848586878889 90919293949697
Time (Minutes)
0
0.5
1
1.5
2
2.5
3
3.5
Load(kW)
Unscheduled
BFA
TS
BFTS
(c)
Fig. 2: Energy consumption with different techniques under
(a) 1-hour OTI; (b) 30-minutes OTI; (c) 15-minutes OTI;
to BFA and TS techniques thus proving that different OTI
changes have little to no effect on cost of electricity during our
simulations. So we can conclude our hybrid BFTS scheduled
reduces electricity cost more as compared to TS and BFA
scheduled.
BFTS scheduling algorithm’s cost of energy consumption
by multiple homes using OTI of 1 hour is still less as compared
to other scheduling algorithms as shown in Fig. 5.
0 5 10 15 20 25
Time (hours)
0
100
200
300
400
500
600
700
800
Load(kWh)
Unscheduled
BFA
TS
BFTS
(a)
0 5 10 15 20 25
Time (hours)
0
50
100
150
200
250
300
350
400
450
Load(kWh)
Unscheduled
BFA
TS
BFTS
(b)
0 5 10 15 20 25
Time (hours)
0
50
100
150
200
250
300
Load(kWh)
Unscheduled
BFA
TS
BFTS
(c)
Fig. 3: Energy consumption by multiple homes with various
techniques using 1-hour OTI (a) 50 homes; (b) 30 homes; (c)
20 homes;
OTI 15 OTI 30 OTI 60
0
500
1000
1500
2000
Total Cost (Cents)
Unschedule
BFA
TS
BFTS
Fig. 4: Total cost of electricity with different techniques
under various OTIs
20 Homes 30 Homes 50 Homes
0
2
4
6
8
10
Total Cost (Cents)
104Unschedule
BFA
TS
BFTS
Fig. 5: Total cost of electricity consumed by multiple homes
with different techniques
OTI 15 mins OTI 30 mins OTI 60 mins
0
5
10
15
20
25
Appliance waiting time (minutes/hours)
BFA
TS
BFTS
Fig. 6: Waiting time with different techniques under various
OTIs
20 Homes 30 Homes 50 Homes
0
1
2
3
4
5
6
Appliance waiting time (hours)
BFA
TS
BFTS
Fig. 7: Waiting time when energy consumed by multiple
homes with different techniques
C. UC
We can relate UC to both the cost of electricity consumed
and amount of time waited before using it. But we are using
UC in relation to that amount of time a user must wait before
he can use an appliance. In the Fig. 6 waiting time of OTI 30,
15 is in minutes while hours in OTI 60. We can clearly see
that hybrid BFTS scheduled has less waiting time as compared
to BFA and TS scheduled when using different OTI so we can
conclude that hybrid BFTS scheduled gives maximum UC.
BFTS scheduling algorithm’s waiting time when scheduling
multiple homes using OTI of 1 hour is still less as compared to
other scheduling algorithms as shown in Fig. 7. Thus we can
conclude that increase in load won’t effect BFTS effectiveness.
OTI 15 OTI 30 OTI 60
0
1
2
3
4
5
6
Peak Average Ratio
Unschedule
BFA
TS
BFTS
Fig. 8: PAR with different techniques under various OTIs
20 Homes 30 Homes 50 Homes
0
0.5
1
1.5
2
2.5
3
3.5
Peak Average Ratio
Unschedule
BFA
TS
BFTS
Fig. 9: PAR when energy consumed by multiple homes with
different techniques
D. PAR
DSM in SG is beneficial to both the consumers and utility.
Lower Peak Average Ratio not only helps utility in retaining
stability in electric load but also sometimes helps consumers
in lowering costs in their electricity bill. The above Fig. 8
clearly gives us an idea of performance of BFA, TS and hybrid
BFTS technique in terms of PAR as compared to Unscheduled
electric consumption. From above figures it’s clear that hybrid
BFTS technique has higher PAR as compared to TS scheduled
thus we can conclude that it doesn’t minimizes PAR.
BFTS scheduling algorithm’s PAR when scheduling multi-
ple homes using OTI of 1 hour is still more as compared to
TS scheduling algorithms as shown in Fig. 9. Thus we can
conclude that increase in load won’t help in lowering BFTS
PAR.
E. Feasible Region
Feasible region is the area that covers the set of some points
which contain all the solutions possible with respect to our
fitness function. Fig. 10a, 10b, 10c shows feasible region for
energy consumption and electricity cost for 1 hour, 30 mins
and 15 mins OTIs respectively. A set of points P1 (1.2000,
32.8200), P2 (1.2000, 9.7200), P3 (11.1000, 89.9100), P4
(11.1000, 303.5850) gives us the bounded region as shown
in Fig. 6a which gives us the cost region for OTI of 1 hour.
P1 (1.2000, 32.8200) represents when cost is maximum and
load is minimum and P2 (1.2000, 9.7200) represents when cost
and load both are minimum. While P3, P4 gives us the points
when load is maximum but prices are minimum and maximum
0 2 4 6 8 10 12
Power consumption (kWh)
0
100
200
300
400
Cost per hour ($)
P2(1.2000, 9.7200)
P1(1.2000, 32.8200) P5(6.56, 180)
P4(11.1000 , 303.5850)
P6(11.100, 180 )
P3(11.1000, 89.9100)
(a)
01234567
Power consumption (kW)
0
100
200
Cost($)
P2(0, 0)
P1(0, 0)
P5(3.29, 90.0015)
P4(6.8750 , 188.0313)
P6(6.8750, 90.0015 )
P3(6.8750, 55.6875)
(b)
0 0.5 1 1.5 2 2.5 3 3.5
Power consumption (kW)
0
50
100
Cost($)
P1(0, 0)
P2(0, 0)
P6(3.4375, 71.1100)
P4(3.4375 , 94.0156)
P5(2.60, 71.1100)
P3(3.4375, 27.8438)
(c)
Fig. 10: Feasible Region in (a) 1-hour OTI; (b) 30-minutes
OTI; (c) 15-minutes OTI;
respectively. We get feasible region made of points P1, P2, P3,
P4, P5, P6 after setting some constraints in which P6 (11.100,
180) shows us that schedule load must never increase more
then 11.100 kWh at any time slot. While P5(6.56, 180) shows
us that at any time slot price more then $180 must not be
scheduled.
Similarly, a set of points P1 (0, 0), P2 (0, 0), P3 (6.8750,
55.6875), P4 (6.8750, 188.0313) gives us the bounded region
as shown in Fig. 10b which gives us the cost region for
OTI of 30 mins. P1 (0, 0) P2 (0, 0) represents when cost is
maximum, load is minimum and cost, load both are minimum
respectively. While P3, P4 gives us the points when load is
maximum but prices are minimum and maximum respectively.
We get feasible region made of points P1, P2, P3, P4, P5, P6
after setting some constraints in which P6 (6.8750, 90.0015)
shows us that schedule load must never increase more then
6.8750 kW at any time slot. While P5(3.29, 90.0015) shows
us that at any time slot price more then $90.0015 must not be
scheduled. Feasible region of 15 mins OTI is shown in Fig.
10c which shows that an appliance must not be scheduled if
price is higher then $71.1100 at any given time slot or if the
load is higher then 3.4375 kW.
VI. CON CL US IO N AN D FUTURE WORK
This paper presented an hybrid BFTS scheduled to man-
age consumers energy consumption. We reduce waiting time
and cost of electricity consumption using our hybrid BFTS
technique. We evaluated our proposed hybrid BFTS technique
by using BFA and TS technique. The simulation results of
hybrid BFTS-based scheduled clearly shows that the proposed
technique is better as compared to BFA, TS and Unscheduled
electricity consumption. The electricity cost and waiting time
is minimized thus increasing UC. We must understand that
we can’t reduce cost, waiting time and PAR simultaneously.
There needs to be a trade-off between either cost, PAR and
waiting time which can’t be lowered together. Hybrid gives
lower waiting time and cost with respect to other techniques.
We used RTP pricing scheme to calculate the cost of electricity
consumption after scheduling our appliances. In future work,
we would work on BFA and Flower Pollination Algorithm
hybrid scheduling algorithm.
REFERENCES
[1] Davoli, F., Repetto, M., Tornelli, C., Proserpio, G. and Cucchietti,
F., 2012. Boosting energy efficiency through smart grids. International
Telecommunication Union (ITU).
[2] Gellings, C.W., 1985. The concept of demand-side management for
electric utilities. Proceedings of the IEEE, 73(10), pp.1468-1470.
[3] Costanzo, G.T., Zhu, G., Anjos, M.F. and Savard, G., 2012. A system
architecture for autonomous demand side load management in smart
buildings. IEEE Transactions on Smart Grid, 3(4), pp.2157-2165.
[4] Baharlouei, Z., Hashemi, M., Narimani, H. and Mohsenian-Rad, H.,
2013. Achieving optimality and fairness in autonomous demand response:
Benchmarks and billing mechanisms. IEEE Transactions on Smart Grid,
4(2), pp.968-975.
[5] Mohsenian-Rad, A.H., Wong, V.W., Jatskevich, J., Schober, R. and Leon-
Garcia, A., 2010. Autonomous demand-side management based on game-
theoretic energy consumption scheduling for the future smart grid. IEEE
transactions on Smart Grid, 1(3), pp.320-331.
[6] Miao, H., Huang, X. and Chen, G., 2012. A genetic evolutionary task
scheduling method for energy efficiency in smart homes. International
Review of Electrical Engineering (IREE), 7(5), pp.5897-5904.
[7] Samadi, P., Mohsenian-Rad, H., Schober, R. and Wong, V.W., 2012.
Advanced demand side management for the future smart grid using
mechanism design. IEEE Transactions on Smart Grid, 3(3), pp.1170-
1180.
[8] Mohsenian-Rad, A.H., Wong, V.W., Jatskevich, J. and Schober, R., 2010,
January. Optimal and autonomous incentive-based energy consumption
scheduling algorithm for smart grid. In Innovative Smart Grid Technolo-
gies (ISGT), 2010 (pp. 1-6). IEEE.
[9] Khan, M.A., Javaid, N., Mahmood, A., Khan, Z.A. and Alrajeh, N., 2015.
A generic demandside management model for smart grid. International
Journal of Energy Research, 39(7), pp.954-964.
[10] Muralitharan, K., Rathinasamy Sakthivel, and Yan Shi. ”Multiobjective
optimization technique for demand side management with load balancing
approach in smart grid.” Neurocomputing 177 (2016): 110-119.
[11] Awais, M., Javaid, N., Shaheen, N., Iqbal, Z., Rehman, G., Muhammad,
K. and Ahmad, I., 2015, September. An efficient genetic algorithm based
demand side management scheme for smart grid. In Network-Based
Information Systems (NBiS), 2015 18th International Conference on (pp.
351-356). IEEE.
[12] Khan, M.A., Javaid, N., Mahmood, A., Khan, Z.A. and Alrajeh, N.,
2015. A generic demandside management model for smart grid. Interna-
tional Journal of Energy Research, 39(7), pp.954-964.
[13] Shirazi, E. and Jadid, S., 2015. Optimal residential appliance scheduling
under dynamic pricing scheme via HEMDAS. Energy and Buildings, 93,
pp.40-49.
[14] Srinivasan, D., Rajgarhia, S., Radhakrishnan, B.M., Sharma, A. and
Khincha, H.P., 2017. Game-Theory based dynamic pricing strategies for
demand side management in smart grids. Energy, 126, pp.132-143.
[15] Li, D., Sun, H. and Chiu, W.Y., 2016, October. A layered approach for
enabling demand side management in smart grid. In Control, Automation
and Information Sciences (ICCAIS), 2016 International Conference on
(pp. 54-59). IEEE.
[16] Verma, P.P., Soumya, P. and Swamp, K.S., 2016, March. Optimal day-
ahead scheduling in smart grid with Demand Side Management. In Power
Systems (ICPS), 2016 IEEE 6th International Conference on (pp. 1-6).
IEEE.
[17] Lizondo, D., Araujo, P., Will, A. and Rodriguez, S., 2017, April. Mul-
tiagent Model for Distributed Peak Shaving System with Demand-Side
Management Approach. In Robotic Computing (IRC), IEEE International
Conference on (pp. 352-357). IEEE.
[18] Akhavan-Rezai, E., Shaaban, M.F., El-Saadany, E.F. and Karray, F.,
2016. Online intelligent demand management of plug-in electric vehicles
in future smart parking lots. IEEE Systems Journal, 10(2), pp.483-494.
[19] Khalid, A., Javaid, N., Mateen, A., Khalid, B., Khan, Z.A. and Qasim,
U., 2016, July. Demand Side Management using Hybrid Bacterial Forag-
ing and Genetic Algorithm Optimization Techniques. In Complex, Intel-
ligent, and Software Intensive Systems (CISIS), 2016 10th International
Conference on (pp. 494-502). IEEE.
[20] Khan, A., Mushtaq, N., Faraz, S.H., Khan, O.A., Sarwar, M.A. and
Javaid, N., 2017, November. Genetic Algorithm and Earthworm Opti-
mization Algorithm for Energy Management in Smart Grid. In Interna-
tional Conference on P2P, Parallel, Grid, Cloud and Internet Computing
(pp. 447-459). Springer, Cham.
[21] Kazmi, S., Javaid, N., Mughal, M.J., Akbar, M., Ahmed, S.H. and
Alrajeh, N., 2017. Towards optimization of metaheuristic algorithms for
IoT enabled smart homes targeting balanced demand and supply of
energy. IEEE Access.
[22] Qayyum, F.A., Naeem, M., Khwaja, A.S., Anpalagan, A., Guan, L. and
Venkatesh, B., 2015. Appliance scheduling optimization in smart home
networks. IEEE Access, 3, pp.2176-2190.
[23] Shafiq, S., Fatima, I., Abid, S., Asif, S., Ansar, S., Abideen, Z.U. and
Javaid, N., 2017, November. Optimization of Home Energy Management
System Through Application of Tabu Search. In International Conference
on P2P, Parallel, Grid, Cloud and Internet Computing (pp. 37-49).
Springer, Cham.
