Wireless Pers Commun (2017) 93:481–502

DOI 10.1007/s11277-017-3959-z

Genetic Algorithm Based Demand Side Management

for Smart Grid

C. Bharathi1 • D. Rekha1 • V. Vijayakumar1

Published online: 21 January 2017

Ó Springer Science+Business Media New York 2017

Abstract Electricity usage at electricity rush hour (peak hour) may vary from each and
every service area such as industrial area, commercial area and residential area. Equalizing
the power consumption in industry may lead to the utilization of power in other service
areas in an efficient way. Although industries have comparably lesser number of power
consuming device types than other service areas the power consumption is quite high. To
meet the demands rising in the industry, shiftable loads (devices) can be redistributed
equally to all the working time slots based on the average power utilization. It can be done
in a flexible manner by shaping the loads using Demand Side Management (DSM) tech-
nique in Smart Grid. The main objective is to minimize the power utilization during the
electricity rush hour by effectively distributing the power available during off-peak hour.
Evolutionary algorithm can be well adapted to problems where optimization is the core
criteria. Any maximization or minimization problem can be solved efficiently using evo-
lutionary algorithm. Hence, to obtain the optimized fitness function of load redistribution
in industry Genetic Algorithm in Demand Side Management (GA-DSM) is chosen and it
has benefited with an overall reduction of 21.91% which is very remarkable. In addition to
this the evaluation of the fitness function using GA-DSM is carried out in other two
industrial dataset models (steel plant and wind power plant) which is unavailable so far in
the literature.

Keywords Smart Grid  Demand Side Management  Load redistribution  Flexible load
shaping  Genetic Algorithm

& C. Bharathi
[email protected]
D. Rekha
[email protected]
V. Vijayakumar
[email protected]
1
School of Computing Science and Engineering, VIT University, Chennai, Tamil Nadu, India

123
482 C. Bharathi et al.

1 Introduction

Effective power consumption and distribution are major issues everywhere but power
blackout is the major issue in India. Power blackout is mainly because of improper usage
or wastage of power. This issue is prevalent in the traditional power system which can be
overcome by using Smart Grid. Smart Grid can be used to utilize the power in an efficient
way by distributing electricity equally even during the electricity rush hour [1].
Smart Grid is a digital power grid hosted in real time environment which is integrated with
Information and Communication Technology in a network. In this network, there is a two-
way communication between the user and the power distributor and vice versa. Varying the
price tariff of electricity between the peak and off-peak hours is one of the main advantages in
Smart Grid. Power distributors charge less amount during off-peak hours to make the users to
utilize electricity during this particular time period [2–5]. During the electricity rush hour,
there is a rise in power demand. To handle this situation, extra burden prevails to increase the
power productivity. Demand Side Management (DSM) proves to be more beneficial in
reducing the power consumption rather than pressurizing on additional power generation
[6–8]. DSM aims to reduce or shift the load from peak hours to off-peak hours [9].
In DSM, load can be projected as a curve (load curve). DSM has six main objective load
curve representation which is shown in Fig. 1 [10–12].
a. Peak clipping: Reduction of loads from the electricity rush hour.
b. Valley filling: Construction of loads in off-peak hours.
c. Load shifting: Shifting loads from electricity rush hour to off-peak hours.
d. Load reduction: Conservation of power by reducing the complete load for a long term
(in hours or in days).
e. Load growth: Increase in the consumption of power with certain limit.
f. Flexible load shaping: Redistribution of load to various time slots for On-Demand
Service.
The management of power utilization during peak hour can be viewed as a mini-
mization problem for which flexible load shaping is used. The advantages of flexible load
shaping are given as follows:
1. No need of clipping loads in electricity rush hour.
2. No need of building the entire load on off-peak hours.

Fig. 1 Load curve representation in DSM

123
Genetic Algorithm Based Demand Side Management for Smart Grid 483

3. No need to mutate or switch off the entire load in particular time step and run it in off-
peak hours.
4. Conservation of power or growth of the load can be avoided by redistributing the load.
The reason for high power consumption in industry can be summed up by two factors, one
is the longer usage period and the other is the power consumption of the individual device.
In this paper, loads are categorized into two types based on their timeslots. If a load has
free working time slots, it is categorized under shiftable load, and then it can be shifted to
that time slot on demand. If a load is running in all working time slots, it is categorized
under non-shiftable load; it cannot be subjected to the technique of Genetic Algorithm in
Demand Side Management (GA-DSM).
This paper is organized as follows: Sect. 2 discusses the related works from which the
way to accomplish the objective function is mentioned. Section 3 introduces the Genetic
Algorithm and its contribution in DSM of Smart Grid. In Sect. 4, details on simulation
conducted to evaluate the variation in fitness functions with GA-DSM and DSM without
GA are given. In Sect. 5, conclusion and future work are provided.

2 Related Works

Integrating high scale renewable energy sources, establishing contacts among the different
users-and-utilities and providing different services to meet the demands of end users are
highly tough in the sector of power management. One plausible solution available to these
problems is to use Smart Grid [2].
A suitable functional grid of electricity is deployed to reduce huge investments in the
electricity grid by creating, transmitting and consuming the electricity on demand. This
clearly points that Smart Grid is successful in reducing cost and efficiently utilizing power.
Smart meters, distributed generation, distributed storage and DSM are new technologies to
attain a sustainable, highly efficient and reliable power supply. Major issue commonly
faced while developing the Smart Grid is with the enhancement of DSM [5, 6].
Eco-friendly, economical generation, efficient power consumption along with effective
DSM in future Smart Grid Networks aid to address the challenges of Information and
Communication Technology (ICT) integration [2, 4, 5, 7]. In German power system, the
restructuring and modernization has been offered only through ICT, which further
emphasizes the importance of ICT.
DSM provides best solution for the challenge of transforming huge volume of
unstructured data into structured data or information to the power system [10]. It also
defines the information of construction system and the application of business in five
different layers: collection of data, transmission of data, analysis of data, integration of
information and projection of information [5].
The two main programs in DSM are energy efficiency programs and demand response
(DR) programs. The main aim of DR is to reduce the power consumption in electricity
during rush hours by shifting demand to off-peak hours [13]. To reduce the peak demand,
utilities make a transformation in load pattern of the end users by means of shiftable loads.
A Genetic Algorithm (GA) based solution is proposed to solve the objective function
which contains discontinuous functions [14]. It is highly complicated to solve that
objective function with traditional optimization techniques. Load shifting may cause
inconvenience. The objective is to minimize inconvenience and also to minimize the
generation cost. Based on the polynomial function and genetic algorithm, the end user can
reduce significant amount of electricity bill.

123
484 C. Bharathi et al.

Load shifting is done using heuristic based evolutionary algorithm to minimize the bill of the
end user and it should be inversely proportional to the electricity price [15]. Simulations have
been done on data sets of all service areas where different device types have also been handled to
solve the minimization problem mathematically. The problem faced by the industry is because
of higher power consuming device types. Reduction and management of this load may lead the
users to utilize the power efficiently in residential and commercial areas.
The load curve pattern based on the power usage data monitored from the smart meters is
discussed in [16]. Artificial neural network (ANN) is trained with these data sets to classify
the upcoming new load patterns. To classify the load pattern a classification tool is imple-
mented. For each type of load, they find out the appropriate strategy of DSM by following
these steps: data fetching, creating the load patterns curve, choosing the appropriate strategy
of DSM, training ANN, simulation and network validation. Based on all the above strategies,
flexible load shaping seems to be a good solution to the power management problems.
Currently all the DSM research works are mainly focusing only on the residential area
[17–21]. But in developing countries like India and China, the global electricity demand in
industrialization is getting increased. Industrial area acquires one-third of the global power
demand and it have high energy need. Power consumption by industrial sectors having few
distinctive characteristics such as they are high level consumers of electricity [22, 23].
By comparing residential area, lots of issues are there in industrial area like scheduling
of loads for the whole process of production [24]. Average peak loads can be around
hundreds of MW and also attain hundreds of millions of kWh as annual consumption. This
sector is frequently connected to the high level voltage grids. Industrial sectors are not
much involved in defining, standardizing, and other Smart Grid related research activities
than other sectors such as residential and commercial sector.
DSM is an opening for these industrial sectors to rise their profits [25]. For example, it
can move the activities of energy intensive to off-peak hours by electricity bill reduction
and also by doing the flexible operational method based on their demand.

3 Algorithm Description

Evolutionary algorithm can be applied to a wide spread domain due to their random nature and
ease in adaptability [26, 27]. It finds the best possible solution even in places where compu-
tationally intensive tasks are involved. It can outdo other techniques in the aspect of solving the
difficult problems which cannot be even attempted to solve using other techniques.
For adopting any optimized problem into evolutionary algorithm there are two prime
steps [28]:
1. Formulating an encoding method for the given scenario.
2. Iteratively going through various operations of evolutionary algorithms.

3.1 Genetic Algorithm

Genetic Algorithm is a popular category of the generous division of evolutionary algorithm. It

is a global search algorithm, based on the genetic mechanism with genes called chromosomes,
used mainly to provide solution to the optimization problems using binary digits [29].
The three main steps in Genetic Algorithm are initialization of population, evaluation of
fitness function and generation of new population. Chromosomes are randomly generated
which produces the original population with constant number of individuals which could

123
Genetic Algorithm Based Demand Side Management for Smart Grid 485

be a feasible solution to the problem. Using the fitness function, each individual is eval-
uated to provide a near optimal solution [30].
The reproduction process contains three major tasks such as selection, crossover and
mutation [31]. Parents are chosen from the fitness level and they are subjected for the
generation of new population. Crossover is the process of generating new offspring from
parents. Here the bits are randomly recombined to produce the offspring. There are many
types in crossover like single point crossover, multi point crossover, uniform crossover,
arithmetic crossover, etc. Mutation is an operator used in Genetic Algorithm to alter one or
more values in the generation of new population from its initial state. The entire process is
repeated for n times to obtain the best optimal solution for the new population.

3.2 Genetic Algorithm in Demand Side Management (GA-DSM)

Genetic Algorithm gives the best solution based on the objective function of load shaping
in DSM using fitness function.
Xt X L
Minimize lk  pi ð1Þ
i¼1 k¼1

where t is the number of time step, L is the number of load types (device types), l is the total
number of devices in each load type and p is the power of each load in kilowatt hour (kWh).
1
Fitness function ¼ 1 þ Pt PL ð2Þ
i¼1 k¼1 lk  pi

Equation (1) and (2) represent the objective function and fitness function of the Genetic
Algorithm respectively.

Algorithm 1: Genetic Algorithm for flexible Load shaping in Demand Side Management (GA-
DSM)
Take load types ‘L’, total number of matrices ‘n’, number of devices ‘l’, current matrix ‘m’ and power
consumption ‘p’ from the table of load data
while 1< L < n do
while 0 < l < m do
(1) Calculate the total ‘p’ consumption of ‘l’ no. of devices
(2) Compute the fitness function
(a) Fitness function – without GA-DSM
(b) Fitness function – with GA-DSM
i. Population initialization
ii. best optimal fitness function selection based on the minimal ‘p’ during peak
hour ‘ph’.
iii. Crossover is carried out on ‘l’ based on the minimal ‘p’ for time slot ‘t’.
<crossover>
iv. After analyzing the behavior of the new population, mutation is done on ‘l’
based on ‘t’ during ‘ph’. <mutation>
v. The fitness function matrix of the ‘l’ with minimal ‘p’ during ‘ph’ is taken as
the best fitness function for each ‘L’. <optimal solution>
end while
Same process is repeated for each ‘L’ and the best optimal fitness function is chosen.
end while

123
486 C. Bharathi et al.

Based on Algorithm 1, data set of each industrial load types is considered to be a

population and fitness function (2) is calculated. After that the chromosome with higher
fitness is subjected to the proposed genetic algorithm. The ‘n’ random chromosomes are
generated with the constant number of individuals of original population. Crossover is
done with the selective parent load types by redistributing the shiftable loads to various
available time slots. Fitness function is evaluated for each offspring. Mutation is carried
out by considering the electricity rush hours and off-peak hours. The solution for this
minimization problem is obtained after comparing the average of fitness values with GA-
DSM and DSM without GA.
In this work, the encoding of power consumption and device populations has been
encoded using numerical assumptions from data generated in industrial environments.
After looking through all these facts, this optimization problem of power utilization fits
appropriately into the GA-DSM based evolutionary algorithm.

4 Simulation and Results

Simulations were carried out using MATLAB computing environment by giving the input
of various load types (device types) from five different industries. In order to analyze the
performance of our GA-DSM a comparison was made with the heuristic based evolu-
tionary algorithm [15] which is based on ‘‘load shifting’’. The proposed GA-DSM algo-
rithm, on other hand is based on ‘‘flexible load shaping’’. It was found that this GA-DSM
algorithm performed much better than the heuristic based evolutionary algorithm. To
increase the reliability of the comparison results, data in Table 1 is taken from [15] itself. It
represents the various load types in 1st Industry with six time steps. It contains six load
types of both shiftable and non-shiftable loads. The arc furnace and induction motors fall
under the category of non-shiftable load. So it cannot be subjected to GA-DSM. Other
shiftable loads are considered as population and are subjected to the Genetic Algorithm.
Fitness functions and its average are calculated with GA-DSM and DSM without GA
respectively. Based on that, a graph is generated which is shown in Fig. 3.
From the graph in Fig. 2, let us consider that the first three time steps (1st, 2nd and 3rd
hour) are the electricity rush hours for that industry which is represented in the line of
average of fitness function of DSM without GA. Next three time steps are the off-peak

Table 1 Data of 1st industrial area devices [15]

S. no Load types Power consumption of load in kWh Number of
devices
1st hour 2nd hour 3rd hour 4th hour 5th hour 6th hour

1 Water heater 12.5 12.5 12.5 12.5 – – 39

2 Welding machine 25 25 25 25 25 – 35
3 Fan/AC 30 30 30 30 30 – 16
4 Arc furnace 50 50 50 50 50 50 8
5 Induction motor 100 100 100 100 100 100 5
6 DC motor 150 150 150 – – – 6
Total – – – – – – – 109

123
Genetic Algorithm Based Demand Side Management for Smart Grid 487

Fig. 2 Comparison of average of fitness functions—1st industry

Fig. 3 Percentage of power utilization difference—1st industry

hours, represented in the line of average of fitness function with GA-DSM. By applying
Algorithm 1, many of the loads are shifted to the off-peak hour to get the optimized
solution.
The difference between power reduction before and after the application of proposed
GA-DSM is around 23.84%. The heuristic based evolutionary algorithm [15] gives a peak
demand power reduction of 14.2% as a difference between with GA-DSM and DSM
without GA for the same data. It is evident that the proposed GA-DSM algorithm benefits
with more than thrice the times of power reduction. Considering the power reduction in

123
488 C. Bharathi et al.

Table 2 Rush hour demand reduction of 1st industry

Percentage reduction % HDSM [15] GA-DSM

14.2 23.84

Table 3 Data of 2nd industrial area devices

S. no Load types Power consumption of load in kWh Number of
devices
1st hour 2nd 3rd 4th 5th 6th 7th 8th
hour hour hour hour hour hour hour

1 Heat pump heat 10 10 10 – – – – – 14

strips
2 Electric furnace 10.5 10.5 25 10.5 10.5 – – – 20
3 Heat pump w/o heat – – – 9.77 9.77 9.77 9.77 9.77 18
strips 5.0 ton
4 Central AC 3 – 3 – 3 – 3 – 4
5 Electric water heater 500 500 500 500 – – – – 45
6 Freezer (manual 32 32 32 32 32 32 32 32 32
defrost)
Total – – – – – – – 133

Fig. 4 Comparison of average of fitness functions—2nd industry

detail, a power reduction of 48% is found in the 1st hour. Similarly the reduction of power
utilization is 48% in 2nd hour, reduction of power utilization is 48% in 3rd hour, increase
of power utilization is 22.5% in 4th hour, and increase of power utilization is 32% in 5th
hour and increase of power utilization is 75% in 6th hour which is shown in Fig. 3.

123
Genetic Algorithm Based Demand Side Management for Smart Grid 489

Fig. 5 Percentage of power utilization difference—2nd industry

Table 4 Data of 3rd industrial area devices

S. no Load types Power consumption of load in kWh Number of
devices
1st 2nd 3rd 4th 5th 6th 7th
hour hour hour hour hour hour hour

1 Comfort Central AC 2 ton 1450 – – – – – 1450 7

2 Comfort Central AC 4 ton 2750 2750 2750 – – – – 3
3 Dehumidifier 252 252 252 – 252 – 252 33
4 Fan 30 30 30 – 30 – 76 76
5 Water heater 310 310 310 – 310 – 310 50
6 Pool pump 375 375 375 – 375 – 375 50
7 Sprinkler system 28 28 28 – 28 – 28 36
8 Broiler 7 7 7 7 7 7 7 24
9 Dishwasher 30 30 30 30 30 30 30 9
Total – – – – – – – – 288

From Table 2 it is evident that the performance of GA-DSM algorithm differs on a huge
margin from the heuristic based evolutionary algorithm (HDSM) [15]. Hence the rest of the
simulations are carried out only with GA-DSM.
Table 3 indicates the data of 2nd industrial area (steel plant) Devices. Figure 4 shows
the comparison graph of average fitness function for 2nd industry with 8 time slots. 1st to
4th hour are the electricity rush hours and 5th to 8th hour are the off-peak hours in DSM
without GA. After applying GA-DSM the power utilization reduction during 1st hour is
35.5%, the power utilization reduction during 2nd hour is 37%, the power utilization
reduction during 3rd hour is 37.7%, the power utilization reduction during 4th hour is

123
490 C. Bharathi et al.

Fig. 6 Comparison of average of fitness functions—3rd industry

Fig. 7 Percentage of power utilization difference—3rd industry

38.5%, increase of power utilization during 5th hour is 66%, increase of power utilization
during 6th hour is 50.6%, increase of power utilization during 7th hour is 44.9% and
increase of power utilization during 8th hour is 37.5% is represented in Fig. 5. From this
the minimization of power utilization during the electricity rush hours is achieved by
17.49%.
Table 4 indicates the data of 3rd industrial area devices. Figure 6 shows the comparison
graph of average fitness function for 3rd industry with 7 time slots. 1st, 2nd, 3rd and 5th
hours are the electricity rush hours and 4th, 6th and 7th hour are the off-peak hours in DSM

123
Genetic Algorithm Based Demand Side Management for Smart Grid 491

Table 5 Data of 4th industrial area devices

S. no Load types Power consumption of load in kWh Number of
devices
1st 2nd 3rd 4th 5th 6th
hour hour hour hour hour hour

1 Elec. Furnace (cold climate) 26.5 26.5 26.5 – – 26.5 50

2 Elec. furnace (warm climate) 7.941 7.941 7.941 – – 7.941 75
Electric space heater (high) 1.44 1.44 1.44 – – 1.44 45
3 Gas furnace (for the blower) 7.5 7.5 7.5 – 7.5 7.5 90
4 Electric space heater (medium) 0.9 0.9 – – – 0.9 36
5 Electric space heater (low) 0.6 0.6 0.6 – – 0.6 88
6 Frost freezer 205 205 205 – 205 205 56
7 Central air conditioner (2.5 tons) 3.5 3.5 – 3.5 3.5 3.5 6
Total – – – – – – – 446

Fig. 8 Comparison of average of fitness functions—4th industry

without GA. After applying GA-DSM the power utilization reduction during 1st hour is
2.5%, the power utilization reduction during 2nd hour is 1.8%, the power utilization
reduction during 3rd hour is 0.8%, the power utilization is increased during 4th hour by
6.3%, the power utilization reduction during 5th hour is 0.35%, increase of power uti-
lization during 6th hour is 15.6%, increase of power utilization during 7th hour is 9.9% is
represented in Fig. 7. From this the minimization of power utilization during the electricity
rush hours is achieved by 18.55%.
Table 5 indicates the data of 4th industrial area devices. Figure 8 shows the comparison
graph of average fitness function for 4th industry with 6 time slots. 1st to 3rd hour are the
electricity rush hours and 4th to 6th hour are the off-peak hours in DSM without GA. After
applying GA-DSM the power utilization reduction during 1st hour is 17.8%, the power

123
492 C. Bharathi et al.

Fig. 9 Percentage of power utilization difference—4th industry

utilization reduction during 2nd hour is 17.8%, the power utilization reduction during 3rd
hour is 17%, the power utilization is increased during 4th hour by 20.5%, increase of power
utilization during 5th hour is 61.5%, increase of power utilization during 6th hour is 31%,
is represented in Fig. 9. From this the minimization of power utilization during the elec-
tricity rush hours is achieved by 14.54%.
Table 6 indicates the data of 5th industrial area (wind power plant) devices. Figure 10
shows the graph of comparison of average fitness function for 5th industry with 10 time
slots. 1st to 6th hour and 8th are the electricity rush hours and 7th hour, 9th and 10th hour
are the off-peak hours in DSM without GA. After applying GA-DSM the power utilization
reduction during 1st hour is 19.5%, the power utilization reduction during 2nd hour is 20%,
the power utilization reduction during 3rd hour is 20.3%, the power utilization reduction
during 4th hour is 20.3%, the power utilization reduction during 5th hour is 19%, the power
utilization reduction during 6th hour is 11.5%, increase of power utilization during 7th
hour is 15%, increase of power utilization during 8th hour is 10%, increase of power
utilization during 9th hour is 27% and increase of power utilization during 10th hour is
32% is represented in Fig. 11. From this the minimization of power utilization during the
electricity rush hours is achieved by 35.15%.
Table 7 indicates the demand reduction of all the industries during rush hour. The load
shaping is done in the industrial area with few power consuming load types using GA-
DSM. The goal is achieved by the overall reduction of power utilization of all the five
industries is 21.91% during the electricity rush hours.
The same GA-DSM has been repeated for residential and commercial sectors rush hour
reduction and the results are discussed below.
Table 8 indicates the data of 1st Residential Area Devices. Figures 12 and 13 shows the
comparison of average of fitness functions and percentage of power utilization difference
of the same area respectively. From this, the minimization of power utilization during the
electricity rush hours is achieved by 23.81%.
Table 9 indicates the data of 2nd Residential Area Devices. Figures 14 and 15 shows
the comparison of average of fitness functions and percentage of power utilization

123
 Table 6 Data of 5th industrial area devices
S. no Load types Power consumption of load in kWh Number of
devices
1st 2nd 3rd 4th 5th 6th 7th 8th 9th 10th
hour hour hour hour hour hour hour hour hour hour

1 Induced draft 3000 3000 3000 3000 3000 3000 3000 – – – 5

2 Boiler feed pump 5000 5000 5000 5000 5000 – – – – – 10
3 Forced draft 5000 5000 5000 5000 5000 5000 – – – – 10
4 Electrostatic precipitator 120 120 120 – – – 120 120 120 – 6
transformer
5 Condensate extraction pump 600 600 600 600 600 600 600 600 – – 10
Genetic Algorithm Based Demand Side Management for Smart Grid

6 Circulating water pump 6000 6000 6000 6000 6000 6000 6000 6000 – – 10
Total – – – – – – – – – – – 51
493

123
494 C. Bharathi et al.

Fig. 10 Comparison of average of Fitness Functions—5th industry

Fig. 11 Percentage of power utilization difference—5th industry

difference of the same area respectively. From this, the minimization of power utilization
during the electricity rush hours is achieved by 21.15%.
Table 10 indicates the data of 1st Commercial Area Devices. Figures 16 and 17 shows
the comparison of average of fitness functions and percentage of power utilization dif-
ference of the same area respectively. From this the minimization of power utilization
during the electricity rush hours is achieved by 19.29%.

123
Genetic Algorithm Based Demand Side Management for Smart Grid 495

Table 7 Rush hour demand

Industries Percentage reduction
reduction of all the five industries
1st Industry 23.84
2nd Industry 17.49
3rd Industry 18.55
4th Industry 14.54
5th Industry 35.15

Table 8 Data of 1st residential area devices

S. no Load types Power consumption of load in kWh Number of
devices
1st 2nd 3rd 4th 5th 6th
hour hour hour hour hour hour

1 Central AC 5 5 5 5 5 5 60
2 Well pump 8 8 8 – – 8 430
3 Hair dryer 2.4 2.4 2.4 2.4 – – 158
4 Dish washer 1.2 1.2 1.2 1.2 – – 290
5 Vacuum 1.2 1.2 1.2 – – – 248
cleaner
6 Laptop 0.75 0.75 0.75 0.75 – – 236
7 Oven 2.25 2.25 2.25 2.25 – 2.25 125
Total – – – – – – – 1547

Fig. 12 Comparison of average of fitness functions—1st residential area

Table 11 indicates the data of 2nd Commercial Area Devices. Figures 18 and 19 shows
the comparison of average of fitness functions and percentage of power utilization dif-
ference of the same area respectively. From this the minimization of power utilization
during the electricity rush hours is achieved by 26.29%.

123
496 C. Bharathi et al.

Fig. 13 Percentage of power utilization difference—1st residential area

Table 9 Data of 2nd residential area devices [15]

Load types Power consumption of load in kWh Number of devices

1st hour 2nd hour 3rd hour

Dryer 1.2 – – 189

Dry washer 201.6 – – 288
Washing machine 134 107.2 – 268
Oven 362.7 – – 279
Iron 340 – – 340
Vacuum cleaner 63.2 – – 158
Fan 57.6 57.6 57.6 288
Kettle 812 – – 406
Toaster 43.2 – – 48
Rice-cooker 50.15 – – 59
Hair dryer 87 – – 58
Blender 19.8 – – 66
Frying pan 10.1 – – 101
Coffee maker 44.8 – – 56
– – – – 2604

From Table 12 it is evident that the performance of GA-DSM algorithm differs on a

huge margin from the heuristic based evolutionary algorithm (HDSM) [15].
The considerable reduction in rush hour using the GA-DSM algorithm is shown in
Figs. 13 and 17. Hence the algorithm performs well for all areas such as industrial, resi-
dential and commercial areas.

123
Genetic Algorithm Based Demand Side Management for Smart Grid 497

Fig. 14 Comparison of average of fitness functions—2nd residential area

Fig. 15 Percentage of power utilization difference—2nd residential area

5 Conclusion

The main objective of load shaping is the redistribution of load in flexible manner is
achieved for various industries using GA-DSM. The power utilization for various data sets
in two different conditions (DSM without GA and with GA-DSM) is represented using
various graphs. The overall reduction of power utilization of all the five industries is
21.91% during the electricity rush hours as discussed in Sect. 4 and is given in Table 7.
This shows the achievement of the proposed GA-DSM using flexible load shaping. The
absence of work on ‘‘flexible load shaping’’ in the literature provoked us to explore the use
of GA-DSM to resolve the issues of power demand reduction. The novelty of the work

123
498 C. Bharathi et al.

Table 10 Data of 1st commercial area devices

S. no Load types Power consumption of load in kWh Number
of devices
1st 2nd 3rd 4th 5th 6th 7th 8th 9th 10th
hour hour hour hour hour hour hour hour hour hour

1 Broiler 88 88 – 88 88 88 – – 88 – 89
2 Dishwasher 32 32 32 – 32 32 – – 32 – 110
3 Roaster 55 55 55 – 55 55 – – 55 – 93
4 Oven (self 65 65 65 65 65 65 – – 65 – 78
cleaner)
5 Coffee 2 2 2 2 2 2 2 2 2 2 60
maker
6 Hot plate 4 4 4 4 4 4 4 4 4 4 13
7 Oven 26 26 26 – 26 26 – – 26 – 128
8 Bottle – 7 7 – 7 7 – – 7 – 52
warmer
9 Trash 29 29 29 – 29 29 – – 29 29 79
compactor
Total – – – – – – – – – – – 702

Fig. 16 Comparison of average of fitness functions—1st commercial area

revolves around the fact that we have tried to resolve the power demand during rush hour
with an available yet unexplored solution. The results are encouraging and further
exploration can be useful if the flexible load shaping objective is maintained.
Smart Grid applications involve some computational requirements that can be met by
utilizing the Cloud Computing. Scheduling mechanism for timely demand response,
security framework for Smart Grid, data analysis in Smart grid are the major issues exist
for the purpose of using Cloud Computing to service the information management in the
Smart Grid. Any two of these issues will be analyzed and solved in the future.

123
Genetic Algorithm Based Demand Side Management for Smart Grid 499

Fig. 17 Percentage of power utilization difference—1st commercial area

Table 11 Data of 2nd commercial area devices [15]

S. no Load types Power consumption of load in kWh Number of
devices
1st hour 2nd hour 3rd hour

1 Water dispenser 300 – – 156

2 Dryer 409.5 – – 117
3 Kettle 369 307.5 – 123
4 Oven 385 – – 77
5 Coffee maker 198 198 – 99
6 Fan/AC 325.5 279 – 93
7 Conditioner 224 196 168 56
8 Lights 174 152.25 130.5 87
Total – – – – 808

Fig. 18 Comparison of average of fitness functions—2nd commercial area

123
500 C. Bharathi et al.

Fig. 19 Percentage of power utilization difference—2nd commercial area

Table 12 Rush hour demand

Type of area Percentage reduction %
reduction of 2nd residential and
2nd commercial area HDSM [15] GA-DSM

Residential 18.3 21.15

Commercial 18.3 26.29

References
1. Yi, W., & Dong, W. (2015). Modeling and simulation of discharging characteristics of external melt
ice-on coil storage system. International Journal of Smart Home, 9(2), 179–192.
2. Basso, G., Gaud, N., Gechter, F., Hilarie, V., & Lauri, F. (2013). A framework for qualifying and
evaluating Smart Grids approaches: Focus on multi-agent technologies. Smart Grid and Renewable
Energy, 4, 333–347.
3. Li, Q., & Zhou, M. (2011). The future-oriented grid-Smart Grid. Journal of Computers, 6(1), 98–105.
4. Wissner, M. (2011). The Smart Grid–A saucerful of secrets? Applied Energy, 88(7), 2509–2518.
5. Wu, Y., Chen, J., & Liu, L. (2011). Construction of China’s Smart Grid information system analysis.
Renewable and Sustainable Energy Reviews, 15(9), 4236–4241.
6. Bakker, V., Bosman, M. G. C., Molderink, A., Hurink, J. L., & Smit, G. J. M. (2010). Demand side load
management using a three step optimization methodology. In Smart Grid Communications (Smart-
GridComm), First IEEE international conference, pp. 431–436.
7. Mahmood, A., Ullah, M. N., Razzaq, S., Basit, A., Mustafa, U., Naeem, M., et al. (2014). A new
scheme for demand side management in future Smart Grid networks. Procedia Computer Science, 32,
477–484.
8. Gellings, C. W. (1985). The concept of demand-side management for electric utilities. Proceedings of
the IEEE, 73(10), 1468–1470.
9. Sithara, N. N., & Saminathan, V. (2014). Optimal scheduling of ICT for demand side management in
smart grid. International Journal of Advanced Research in Electrical, Electronics and Instrumentation
Engineering, 3(4), 9187–9193.

123
Genetic Algorithm Based Demand Side Management for Smart Grid 501

10. Gellings, C. W., & Chamberlin, J. H. (1987). Demand-side management: Concepts and methods.
Lilburn: Fairmont Press.
11. Maharjan, I. K. (2010). Demand side management: Load management, load profiling, load shifting,
residential and industrial consumer, energy audit, reliability, urban, semi-urban and rural setting.
Saarbrücken: LAP Lambert Academic Publication.
12. Kothari, D. P., & Nagrath, I. J. (2011). Modern power system analysis. New York: Tata McGraw-Hill
Education.
13. Mohsenian-Rad, A. H., Wong, V. W., Jatskevich, J., Schober, R., & Leon-Garcia, A. (2010). Auton-
omous demand-side management based on game-theoretic energy consumption scheduling for the
future smart grid. IEEE Transactions on Smart Grid, 1(3), 320–331.
14. Jayadev, V., & Swarup, K. S. (2013, October). Optimization of microgrid with demand side manage-
ment using Genetic Algorithm. In IET conference on power in unity: A whole system approach, pp. 1–6.
15. Logenthiran, T., Srinivasan, D., & Shun, T. Z. (2012). Demand side management in smart grid using
heuristic optimization. IEEE Transactions on Smart Grid, 3(3), 1244–1252.
16. Macedo, M. N. Q., Galo, J. J. M., De Almeida, L. A. L., & Lima, A. D. C. (2015). Demand side
management using artificial neural networks in a Smart Grid environment. Renewable and Sustainable
Energy Reviews, 41, 128–133.
17. Tsagarakis, G., Thomson, R. C., Collin, A. J., Harrison, G. P., Kiprakis, A. E., & McLaughlin, S.
(2016). Assessment of the cost and environmental impact of residential demand-side management.
IEEE Transactions on Industry Applications, 52(3), 2486–2495.
18. Adika, C. O., & Wang, L. (2014). Demand-side bidding strategy for residential energy management in a
smart grid environment. IEEE Transactions on Smart Grid, 5(4), 1724–1733.
19. Liu, Y., Yuen, C., Huang, S., Hassan, N. U., Wang, X., & Xie, S. (2014). Peak-to-average ratio
constrained demand-side management with consumer’s preference in residential Smart Grid. IEEE
Journal of Selected Topics in Signal Processing, 8(6), 1084–1097.
20. Yao, E., Samadi, P., Wong, V. W., & Schober, R. (2016). Residential demand side management under
high penetration of rooftop photovoltaic units. IEEE Transactions on Smart Grid, 7(3), 1597–1608.
21. Affonso, C. M., & da Silva, R. V. (2015). Demand side management of a residential system using
simulated annealing. IEEE Latin America Transactions, 13(5), 1355–1360.
22. Agha, M. H., Thery, R., Hetreux, G., Hait, A., & Le Lann, J. M. (2010). Integrated production and
utility system approach for optimizing industrial unit operations. Energy, 35(2), 611–627.
23. Samad, T., & Kiliccote, S. (2012). Smart grid technologies and applications for the industrial sector.
Computers & Chemical Engineering, 47, 76–84.
24. Reka, S. S., & Ramesh, V. (2016). Industrial demand side response modelling in smart grid using
stochastic optimisation considering refinery process. Energy and Buildings, 127, 84–94.
25. Zhang, X., Hug, G., Kolter, Z., & Harjunkoski, I. (2015). Industrial demand response by steel plants
with spinning reserve provision. In North American Power Symposium (NAPS), pp. 1–6. IEEE.
26. Michalewicz, Z. (2013). Genetic algorithms ? data structures = evolution programs. Berlin: Springer
Science & Business Media.
27. Holland, J. H. (1992). Adaptation in natural and artificial systems: An introductory analysis with
applications to biology, control, and artificial intelligence. Cambridge: MIT press.
28. Goldberg, D. E. (2006). Genetic algorithms. London: Pearson Education India.
29. Xu, J., Tan, S., & Panda, S. K. (2011). Optimization of economic load dispatch for a microgrid using
evolutionary computation. IECON 2011-37th annual conference on IEEE industrial electronics society.
IEEE, pp. 3192–3197.
30. Lee, K. Y., & El-Sharkawi, M. A. (Eds.). (2008). Modern heuristic optimization techniques: Theory and
applications to power systems (Vol. 39). New York: Wiley.
31. Yadav, P. K., & Prajapati, N. L. (2012). An overview of genetic algorithm and modeling. International
Journal of Scientific and Research Publications, 2(9), 1–4.

123
502 C. Bharathi et al.

C. Bharathi received the B.E. degree for Computer Science and

Engineering from Anna University, Trichy in 2011 and the M.Tech
degree specialising in Multimedia Technologies from Anna University,
Chennai in 2013. From 2009 to 2013 she has worked on various
projects in Cloud Computing and Network Securities. She has also
published papers in notable journals during this time. Presently, she
pursues her interest in Smart Grid for her Ph.D. degree from VIT
University, Chennai.

Dr. D. Rekha is currently working as Senior Assistant Professor in

VIT University. She has more than 11 years of experience which
includes 7 years in teaching and 4 years in Research. She is also a
Division Chair of Network and Security Research Group, her area of
research includes Wireless Sensor Networks, Multihop Networks,
evolutionary algorithms, Mobile Cloud Computing and Internet of
things. She published many papers in reputed journals. She served as
General and Technical Program Chair of numerous conferences.

Dr. V. Vijayakumar is currently working as Professor in VIT

University; He has more than 15 years of experience which includes
9 years in teaching and 6 years in Industry. He is also a Division Chair
of Cloud Computing Research Group; His area of research includes
Grid Computing, Cloud computing, Big Data, Web semantics and also
involved in the domain like Bio-medical applications-Mammogram,
Autism, Immune system and other areas like key management, security
issues in cloud and grid computing. During his research he is also
collaborated with University of Singapore, Nan yang Technological
University and Deakin University. He is the reviewer for International
Journal of supercomputing -Springer and Guest Editor for international
journal of Big Data intelligence and International journal of high
performance communication networks.

123