You are on page 1of 8

Future Generation Computer Systems 98 (2019) 708–715

Contents lists available at ScienceDirect

Future Generation Computer Systems


journal homepage: www.elsevier.com/locate/fgcs

Optimal power flow using Moth Swarm Algorithm with Gravitational


Search Algorithm considering wind power

Shilaja C. , Arunprasath T.
Department of EEE, Kalasalingam Academy of Research and Education, Krishnankoil, India

highlights

• Presently, in order to fulfill the power demand and make cost effective power supply, it is forced to use the renewable energy resources as an
alternative energy sources.
• In this paper it is aimed to provide a hybrid approach MSA-GSA by integrating Moth Swarm Algorithm (MSA) and Gravitational Search Algorithm
(GSA) for power systems with Wind energy sources.
• The test cases, with and without wind power are considered for solving the objective functions of deteriorating the Fuel cost for reduced power
loss.
• Finally, the simulation results are tested with IEEE 30-bus, IEEE 57 bus and IEEE 118 bus with and without wind power offers better results when
associated with the existing procedures.

article info a b s t r a c t

Article history: One of the major problems to be solved in power systems is Optimum Power Flow. Presently, in order
Received 31 October 2018 to fulfill the power demand and make cost effective power supply, it is forced to use the renewable
Received in revised form 12 December 2018 energy resources as an alternative energy sources. Various existing research studies utilized different
Accepted 18 December 2018
artificial intelligence and metaheuristic approaches for solving OPF problems in power system. But, in
Available online 5 April 2019
this paper it is aimed to provide a hybrid approach MSA–GSA by integrating Moth Swarm Algorithm
Keywords: (MSA) and Gravitational Search Algorithm (GSA) for power systems with Wind energy sources. To do
Optimal power flow (OPF) this, the WDF (Weibull Distribution Function) is used for demonstrating the alternating nature of wind
Wind power farm. The test cases, with and without wind power are considered for solving the objective functions
Generation cost of deteriorating the Fuel cost for reduced power loss. Finally, the simulation results are tested with
Power loss IEEE 30-bus, IEEE 57 bus and IEEE 118 bus with and without wind power. The proposed MSA–GSA
algorithm offers better results when associated with the existing procedures
© 2019 Elsevier B.V. All rights reserved.

1. Introduction when it is incorporated in remote areas while processing OPF.


The fundamental goal of operating power flow is to manage basic
Most of the renewable energy resources which are intermit- dispatching process, distributing the demands and to minimize
tent like wind energy and solar photovoltaic energy that shows the loss in transmission. Moreover, it develops and impression
gradual penetration which reveals a lot of disadvantages in the in reducing the cost of overall generation including the needs
current operational processed power system that ultimately end and the operation process [1,2]. The electrical energy develops an
up in the limitations of integrating new resources. The intermit- unavoidable space for the users, industrialist who uses the energy
tent technology of wind generation imposes lot of challenges such system and all the stock holders in the usage of energy with help
as risk management, low cost and few transmissions assets when of wind power. During the time when the remote energy systems
they are not completely dispatched. got connected, an effective wind power is generated. When com-
To incorporate wind energy generations to electrical energy pared with renewable power systems, the remote energy systems
systems, OPF (Optimum Power Flow) plays an important issue. have shown great accuracy. The power flow frequency may get
Major problems like planning and operating energy systems arise collapsed in the absence of inter connected power systems. In
earlier days, several researchers discussed on different optimiza-
∗ Corresponding author. tion methods which shows the various problems and how it get
E-mail address: shilaja@klu.ac.in (Shilaja C.). solved with help of OPF solutions [3,4]. To solve OPF problems in

https://doi.org/10.1016/j.future.2018.12.046
0167-739X/© 2019 Elsevier B.V. All rights reserved.
Shilaja C. and Arunprasath T. / Future Generation Computer Systems 98 (2019) 708–715 709

the energy system, generally two types of methods are used such
as intelligent method and traditional method. GSA (gravitational
search algorithm) is one of the latest high-tech heuristic way
of algorithm which works on the basis of mass interactions and
Newton’s law of gravitation. According to various studies GSA is
said to be the most efficient quality in performing and solving
the problems of optimization [5–9]. The tremendous speed in
identifying the solutions is to modify the accuracy of the process
and thus it is the major feature of the GSA [10–12]. Additionally,
GSA is also known to be memory less and it effectively works like
the algorithms which has high memory [11,12]. To solve the OPF
problem, this paper develops a new heuristic optimization called
GSA that is created in nonlinear optimization issues along with
both inequality and equality constraint in the system.
The ultimate uses are minimizing the cost of the fuel like
quantum function cost, valve pointed effect with function cost,
quantum function cost with piecewise, enhanced profile of the
voltage and voltage stable enhancement in ostensible and pos-
sibility situations. IEEE30 bus and 57 bus test systems are used
to assess and sort out the process of this proposed work. To
solve the OPF issues, the resulted simulation demonstration that
are obtained from the proposed work. The final result has been
compared to the reports from the literature.
The proposed paper gives the OPF (optimal power flow) issues
in wind energy system for the network operated situations equal
to the report [13–15] along the profile of the wind speed that
are featured with the help of the functions of Weibull proba-
bility [16]. On the basis relative or probabilistic frequencies in
histograms of predicting mistakes, OPF design that is the reason
for the excess value of adjusting wind intermittent capacity [14].
Hence, this current study takes initiation generator on the basis
of wind energy and thus the proactive energy capacity varies in
wind turbines which are not essential for the OPF formulation.
The proactive energy supplies those are capable of variables in
speed wind turbine doubles the fed induction generator, which
is used for the bus voltage enhancement [15], along the other
working situations predicted as the same. Wind OPF issues are
Fig. 1. Flowchart of Hybrid MSA–GSA algorithm.
applied on the IEEE 30 bus test systems by merging the hybrid
of optimization algorithms. Few researchers have already said
that the implementation of wind OPF in IEEE 39 bus systems
by merging its own adaptive evolution programmers are avail- But, there is limited ability in the heuristic approach to make
able. Though the process [14] and on the basis of cost design solutions in efficient cost [21,22]. Now a day, Unit Commitment
recommended in [9] said to be the most practical one. No other (UC) and transmission-constrained OPF models on the basis of
work can be compared with these literatures that validate the stochastic programming [23–26], chance constrained optimiza-
work of various methods employed. Additionally, the impact of tion [27], interval programming [28,29] and robust optimiza-
the change in the profile of wind speed and wind energy cost co- tion [30–32] methods are proposed to make endogenous in the
efficient of optimum dispatch formulation of energy plants are time of emergency.
not assessed [14,15]. This proposed paper predicts the energy A general disadvantage of programming in stochastic and the
network that is inter related with the wind farm that has variable cause of restrained optimization is that a need to the exact math-
speed wind turbine that is capable to supply proactive energy to ematical designs of generating wind in ambiguity and changeabil-
the grid. ity. Generally, lack of accuracy arises in wind generation using
The remaining part of the paper is sequenced as follows. In single or multivariate Gaussian distribution [33]. Moreover, pro-
Section 2 related work are described of OPF, MSA and GSA. In gramming in stochastic really needs the stage where a large
Section 3 proposed work of this paper were presented clearly. In quantity of scenes that leads to practically in a huge processing
Section 4 obtained results of propped work for this paper were times, also for a huge number in the dual gap [34]. When the
discussed in detailed. Finally, Section 5 concludes the proposed quantity of the scenario is decreased in terms of scenario re-
work of this paper. ducing methods [35], the compulsory advantages obtained along
with the programming stochastic reduces simultaneously [29]
2. Related work and [34]. Unlike other methods, the robust optimization and pro-
gramming in interval designs gives the wind productions in the
The central wind forecast is used to transmit the convenient given range along the predicted area. In comparison to stochastic
generation of deterministic OPF models [17], which means there method, these methods results in an overall conserving solutions
is no variability in account of endogenous and uncertainty in gen- due to the likeness of individual scenarios in the surrounding
eration of wind [18]. To impose further requirements in security, ranges [5,6]. It is understood that no organized techniques to
Heuristic policies and reserve margins which are calculated ex- select the range of budget which is uncertainty, even the conser-
ogenously [19,20] are used to make certain reliability of system. vative of robust optimization is decreased by altering the budget
710 Shilaja C. and Arunprasath T. / Future Generation Computer Systems 98 (2019) 708–715

in uncertainty [7] or utilizing effective uncertainty set [9]. Like- (b) Security constraints:
wise, by taking the ramping realistic scenarios or utilizing various
objective optimization methods, the interval program conserva- ViL,min ≤ ViL ≤ ViL,max ∀i = 1, . . . , NL (11)
tive decreases [8]. [27] formulates and implements a special CC SiL ≤ SiL,max ∀i = 1, . . . , NL (12)
OPF with the help of cutting plane method and demonstrates
the measurable capacity to the greater systems. On the basis of where, Pg denotes the output from power generator at buses
this method, it is also said that visualizing the RCC OPF which Pv excluding slack bus. Vg denotes the generated voltage, TLC
provides certain range in the mean variance of wind prediction denotes the arrangement of the transformer and Qc denotes the
mistakes, but it is negotiable. In this proposed work, a RCC OPF compensation of the shunt-VAR.VL denotes the voltage summary
is formulated to decrease the imprecision of predicting the Gaus- to the load bus, θ denotes the argument voltages in all the buses,
sian supplies to the wind predicting mistakes. This technique is excluding slack bus. Ps denotes the generated active power and
applied in [10] as an open reason for optimization packages that Qs denotes the generated reactive powers. Pmin max min
Gi , PGi , QGi and
develops into a design simple and robust in distribution which max
QGi denotes the minimum and maximum active and reactive
constraints OPF and various other settings of optimization. En- powers at ith generation unit respectively. Vmin max min max
Gi , VGi , QGi , QGi ,
hanced versions of BPA systems are used to illustrate the process Tmin and T max
denotes the lower and upper limits of the voltage
i i
of RCC OPF application on the huge inter linked networks. This amplitude, compensator capacity and tap changer at ith units
study also demonstrates the robust constraints which results in
respectively. VLi,min , VLi,max denotes the minimum maximum load
huge scale interlinked networks in cost conservations and is the
voltage at an ith unit, SLi and SLi,max denotes the apparent power
most efficient technique to make a decision in real time proactive
energy balances, tremendous ramping of generators and overload flow and maximum apparent power flow at an ith unit. Also
in transmissions. there will be a penalty function value can be added with the
objective function value in case of any violation occur at operating
3. Proposed work constraints.

This section elaborates the proposed work of this paper. The 3.2. Moth Swarm Algorithm (MSA)
following below sub-sections are carried out the OPF, MSA, GSA
and Hybrid MSA–GSA algorithm. In order to improve the accuracy in terms of convergence
rate, execution time and CPU utilization a hybrid optimization
3.1. Optimal power flow approach MSA–GSA is designed by integrating GSA into MSA. The
spiral path movement of MSA is replaced by straight line path in
In this section the optimum power flow problem is math- order to reduce the searching time. Similarly, the onlooker swarm
ematically demonstrated. It is assumed that the OPF problem update is eliminated from MSA and all the calculation updating
is a non-linear problem whereas the main motto of the paper are carried out on the normal swarms directly. It helps to update
is to reduce the fuel cost by decreasing the best values of the the information about the swarm directly and it reduces the
control-variables with respect the various constraints such as
CPU utilization and searching time. Hence this hybrid approach
equal and in-equal conditions. The mathematical formation of the
improves the efficacy of the optimization in OPF problem. All the
OPF problem is:
three algorithms MSA, GSA and MSA–GSA are simulated using
Objective Function value (OFV) = min(OF(x, u)) (1) IEEE-30, 57 and 118 bus systems over test cases. In this paper
there are only five cases are taken for simulation and the obtained
results are compared to verify the performance. Table 1 shows
With respect to g (x, u) = 0 and h (x, u) ≤ 0. (2)
the fuel cost comparison among all the three approaches. From
where, OF denotes the objective function, x is a vector of fixed Table 1, it is noticed that the proposed hybrid approach reduced
variables and u is a vector of control variable and F to be reduced. the fuel cost than the other two approaches and it shows its
The variable x is symbolized as: efficacy.
x = [VL , Q, Ps , Qs ] (3) The entire moth is represented as a matrix as:
⎡ ⎤
The variable u is symbolized as: m1,1 ··· m1,d
⎢ . .. .. ⎥
M = ⎣ .. . . ⎦ (13)
u = [Pg , Vg , Tc , Qc ] (4)
mn,1 ··· mn,d
The set of all equality constraints (g) are
NB where the total moths are represented as n and d denote the
variable representing the dimensions. The values of moths are

Vj Cij cos (δi −
[
Pi = 0 = PGi − PDi − Vi
represented as:
j=1
OM1
⎡ ⎤
δj + Bij sin δi − δj
) ( )]
(5)
⎢OM2 ⎥
Qi = 0 = QGi − QDi −
⎣ .. ⎦
OM = ⎢ (14)

NB
∑ .
Vj Cij sin δi − δj + Bij cos δi − δj
[ ( ) ( )]
Vi (6) OMn
j=1
The main component of the MSA is flames and it is repre-
sented as a matrix as:
(a) Inequality Constraints: ⎡ ⎤
F1,1 ··· F1,d
Vmin
Gi ≤ VGi ≤ Vmax
Gi ∀ i = 1, 2, . . . , NPV (7) ⎢ . .. .. ⎥
F = ⎣ .. . . ⎦ (15)
Pmin
Gi ≤ PGi ≤ Pmax
Gi ∀ i = 1, 2, . . . , NPV (8)
Fn,1 ··· Fn,d
Qmin max
Gi ≤ QGi ≤ QGi ∀ i = 1, 2, . . . , NPV (9)
Moth moves toward the best flame is considered as the best
Tmin
i ≤ Ti ≤ Tmax
i ∀ i = 1, 2, . . . , NT (10) position of the moths. In accordance to the flame the position of
Shilaja C. and Arunprasath T. / Future Generation Computer Systems 98 (2019) 708–715 711

the moth is updated using: 3.3. Gravitational Search Algorithm (GSA)

Mi = S(Mi , Fj ) (16) GSA (Gravitational Search Algorithm) is a current heuristic


where the above equation says, ith moth (Mi ), jth flame in the algorithm of optimization which was proposed in the new era
spiral path (S). From this the spiral path of the MSA algorithm is to identify the effective solution in solving the issues in search
calculated using: paces with the help of written rules. This was greatly inspired by
and extracted from the Newton’s law that tells that every single
S Mi , Fj = Di · ebt · cos (2π t) + Fj
( )
(17) particle in the whole of the universe will certainly attract toward
the other particle using a force which is directly related to the
The distance between a moth i and a flame j is Di , b is used product of the weight and inversely proportional to its square of
to shape the logarithmic based spiral path and t is generated distance among them [6].
between [−1, 1] randomly. The distance Di is calculated between GSA is said to be the group of agents that is candidate solutions
ith moth and jth flame using the following formula as: that has weight proportional to the range in fitness process.
⏐ ⏐ When generation mass attracts one another by the gravitational
Di = ⏐Fj − Mi ⏐ (18)
forces among them. The greater the mass, the huge the attraction
More number of iterations reduces the more number of flames force is. Hence, the greatest mass that is by chance closer to the
where leads to obtaining the objective function value effectively universally less attraction to various mass in proportion to the
and it can be calculated using the following formula as: distance. In [5–7], N agents are present in the system. The place
( ) of every agent or mass that is a candidate result for the issue.
N−1
flamenumber = round N − l × (19) Here, N denotes the various dimension of the issue and x denotes
T the place of the agent in that particular dimension. The algorithm
where N is the maximum number of flames, T is the total may start with the random placement of all agents in the correct
number of flames and l is the current repetition number. Also, the place. In all the epochs, it is said that the gravity from the agent
lower/upper boundary limits of the control-variables are given j is at the specific place. It is illustrated as follows. Here, Maj
denotes the active gravitation mass connected to agent j, Mpi
as:
denotes the passive gravitation mass connected to the agent i,
lb = [Pmin , Vmin , Tmin , Qc min ] (20) G(t) denotes the gravitation stability at the time t, e denotes little
constant, while Rij(t) denotes the Euclidian distances among the
and agents i and j. The gravitation constant G and Euclidian distances
ub = [Pmax , Vmax , Tmax , Qc ]. (21) among the agents i and j are measured as follows:
max
G (t) = G0 + exp(− ∝ ×iter/maxiter) (27)
MSA algorithm is used to resolve the OPF problem in various
circumstances such as: Here, ‘a’ denotes the descending coefficient, G0 denotes the
Case-1: Minimizing the generation fuel cost: basic initial gravitation constant, iter indicates current iteration
NG NG and maxiter denotes the total number in iterations. During a
problem pace the dimension d and the overall force which applies
∑ ∑
Fi (PGi ) = ai + bi Pgi + ci P2gi + penalty
( )
c1 = (22)
on the agent i that is estimated as follows:
i=1 i=1

Rij (t ) = Xi (t), Xj (t)2


 
Case-2: Minimizing the total emission: (28)
NG ( Here, r as well as j is a randomly chosen numbers from the
αi + βi Pgi + γi P2gi + µi exp ξi Pgi
∑ ) ( )
c2 = (23) interval [0,1].
+penalty In Accordance to the law of motion, the acceleration of the
i=1
single agent is equivalent to the resulted force and contrary of
Case-3: Improving voltage profile:
its mass, hence the acceleration of all the agents is measured as
npq follows:

c3 = |Vi − 1| + penalty (24) N

i=1 Fdi (t) = randj Fdij (t) (29)
Case-4: Improving Voltage profile including fuel cost minimiza- j=1,j̸ =i
tion: Here d indicates the dimension of the issue, t indicates the
NG npq specific time, and Mi denotes the mass of the object i. the volume
∑ ∑
ai + bi Pgi + ci P2gi + η
( )
c4 = |Vi − 1| + penalty (25) and the place of the agents are measured as follows:
i=1 i=1
fi ti (t) − Worst(t)
mi (t ) = (30)
Case-5: Minimizing the emission and generation fuel cost: best (t ) − Worst(t)
NG NG Here d denotes problem dimension where r and i is a ran-
∑ ∑
ai + bi Pgi + ci P2gi + αi + βi Pgi + γi P2gi
( ) ( )
c5 = domly chosen number in the space [0,1]. It is known from (3.7–
i=1 i=1 3.8). The obtained agent’s velocity can be defined as last velocity’s
+ µi exp ξi Pgi + penalty fraction that is included in the acceleration. In addition to that,
( )
the position of the last is mostly equal to the current with its
Case-6: Minimizing the generation fuel cost at value point effect: present velocity. Fitness assessment is used define the agents. It
NG is understood that the greatest mass agent is the effective agent
when compared with the other. The above equations depict that

ai + bi Pgi + ci P2gi
( )
C6 =
if the agent is heavy, high in attracting force and slow in the
i=1
⏐ movement. The law of gravity is based on the greater attraction
+ ⏐di ∗ sin ei ∗ Pmin
( ( ))⏐
gi − Pgi
⏐ + penalty (26) (3.2), and law of motion is greatly influenced on the slow motion
712 Shilaja C. and Arunprasath T. / Future Generation Computer Systems 98 (2019) 708–715

Fig. 2. General steps for gravitational search algorithm [5]. Fig. 3. Steps for MSA–GSA algorithm.

Table 1
(3.6) [5]. The following equations update the entire agent’s mass. Comparison of power loss in MSA, GSA and MSA–GSA algorithms. (IEEE-
30/57/118 bus).
Here fiti (t) denotes the value of the fitness in agents i in time t,
Method IEEE-30 bus IEEE-57 bus IEEE-118 bus
best (t) represents effective agent in time t, as well worst(t) is the
Ploss (MW) Ploss (MW) GSAPloss (MW)
weaker agent in time t. Worst(t) and best(b) for a minimization
MSA [24] 4.5409 15.0526 139.16
problem of minimization as measured given below: GSA [22] 4.39 15.611 135.69
MSA–GSA 3.09 14.9658 119.500
best (t ) = minj∈{1,...,N } Fitj (t) (31)
Worst (t) and best (b) for the problem maximization is esti-
Table 2
mated as given: Comparison of fuel cost in MSA, GSA and MSA–GSA algorithms. (IEEE-30/57/118
bus).
worst (t ) = minj∈{1,...,N } Fitj (t) (32)
System GSA [22] MSA [24] MSA–GSA
The following equation defines the normalization of estimated IEEE-30 bus 799.08 799.09 799.01
mass. IEEE-57 bus 41 679.94 41 697.193 41 658.58
IEEE-118 bus 129 640.72 129 688.72 129 458.64
mi (t)
Mi (t ) = ∑N (33)
j=1 mj (t)

Generally, in GSA initially all the agents initialize with values gravitational constant, gravitational force as well as the resulted
randomly. Every single agent is meant to be the contender so- forces between each agents are measured by using (3.2) (3.3)
lution. As soon as the agents are been initialized, the position and (3.5). Then, acceleration of the particles can be defined as
and the velocity of the entire agents defines as (3.7 & 3.8). In (3.6). In every single iteration, the accurate solution is updated.
the other hand, the calculation of the gravitational invariables Then measuring the obtained accelerations will be updated the
and the masses are the few more criteria (3.3 & 3.9). At the end, accurate solution, the velocity of the agents must be measured
GSA can be blocked by satisfying the last criterion. The process (4.1). At last, positions of the agents should be updated (4.2).
of the GSA is depicted in Fig. 1. Two major fundamental feature The step to update the velocity and the position is blocked when
is taken into account such as the capacity of the algorithm that meeting the last criteria. The process of PSOGSA is depicted in
search the complete parts of research places and the capacity Fig. 2. To evaluate the work of PSOGSA, the below criterion are
to bring out the better solution. Exploration means researching mentioned:
the complete issue space where as exploitation is congregating
the most accurate solution. The two important features should • PSOGSA, the eminence of the solutions or fitness must be
be held by the population oriented algorithm in order to assure taken considered to update the process.
the accurate solution. P best is applied to exploration capacity • Agents with the better solution attempt to draw the further
and G best is applied for exploitation capacity in PSO. Proper agents that are exploring various parts of research space.
values are chosen randomly for the parameters (Go & a) in GSA to • If all the agents nearing better solution, they tend to move
guarantee the exploration and slower movement in heavy agents leisurely; in such case, it allows them by exploiting the
to guarantee exploitation capacity [5,8]. It gives a comparison universal best.
among GSA and few known heuristic algorithm of optimization • PSOGSA also employs a memory (gbest) by saving the solu-
such as PSO. The obtained results clearly show that GSA consists tion that found; hence it is available all the time.
of many advantages in the optimization field. Hence, GSA has • Every agent can observe best solutions (gbest) and be liable
a difficulty in searching speed in the later iterations [9]. The
to it.
hybrid algorithm called MSA–GSA is implemented to enhance the
• Adjusting c01& c02, abilities of universal probing and lim-
drawbacks.
ited probing must be balanced.
3.4. The hybrid MSA–GSA algorithms
4. Result and discussion
The fundamental plan of PSOGSA is that combining the capa-
bility for communal thought (gbest), PSO with restricted search This section illustrate the validation and performance of Hy-
ability in GSA. To combine the algorithms, (4.1) is presented in brid MSA–GSA.
this way: Tables 1 and 2 show the power loss and fuel cost obtained
PSOGSA, initially, entire agents randomly work. Every agent using the entire three algorithms from the experiment. From the
is referred as the contender solution. Next to the initialization, results, it is noticed that the hybrid MSA–GSA algorithm obtained
Shilaja C. and Arunprasath T. / Future Generation Computer Systems 98 (2019) 708–715 713

Fig. 4. Comparison of Fuel Cost vs. Iterations for IEEE-30 bus system using GSA, MSA, and MSA–GSA..

Fig. 5. Comparison of Fuel Cost vs. Iterations for IEEE-57 bus system using GSA, MSA, and MSA–GSA..

Fig. 6. Comparison of Fuel Cost vs. Iterations for IEEE-118 bus system using GSA, MSA, and MSA–GSA..
714 Shilaja C. and Arunprasath T. / Future Generation Computer Systems 98 (2019) 708–715

Fig. 7. Comparison of Fuel Cost vs. Iterations for IEEE-30 bus system using GSA, MSA, and MSA–GSA..

Table 3
Comparison of convergence time in MSA, GSA and MSA–GSA algorithms (IEEE-30/57/118 bus).
System CPU time in GSA [22] CPU time in MSA [24] CPU time using MSA–GSA
IEEE-30 bus 14.91 10.2716 7.87
IEEE-57 bus 45.87 42.34 39.12
IEEE-118 bus 67.87 63.12 61.32

reduced fuel cost and less power loss comparing with other two Table 4
Power loss comparison among various method for power system with wind
approaches.
system.
The convergence rate of the proposed approach is computed
System Ploss in IEEE-30 Ploss in IEEE-57 Ploss in IEEE-118
in terms of a number of iterations in accordance with optimum
MSA [24] 7.46 20.02 132.45
fuel cost. The comparison is carried out for IEEE-30, IEEE-57 and
GSA [22] 7.81 21.12 139.69
IEEE-118 bus systems and it is showed in Fig. 1, Fig. 2 and in Fig. 3 MSA–GSA 7.39 20.003 110.34
respectively.
From Fig. 6 it is noticed that the proposed hybrid MSA–GSA
approach obtained the optimum results within a less number of Table 5
iterations and it shows the time consumption and efficiency in Fuel cost comparison among various method for power system with wind system.
CPU utilization. Similarly, when the number of bus increases then System Fuel cost ($/h) Fuel cost ($/h) Fuel cost ($/h)
using GSA [22] using MSA [24] using MSA–GSA
the number of iterations also increases. Table 3 shows the time
(in seconds) execution time taken for all the three algorithms in IEEE-30 bus 958.263 1001.0445 945.112
IEEE-57 bus 5184.3948 5076.5839 5001.324
the simulation. From Table 3, it is noticed that the time taken by IEEE-118 bus 184 686.8248 179 145.3604 176 849.32
the proposed hybrid approach is less comparing with the other
approaches. From Figs. 1, 2, and 3, it is proved that the number of
iterations taken by the proposed hybrid approach is less than the Table 6
other approaches comparatively. Hence this paper states that this Average CPU utilization time comparison among various method for power
hybrid MSA–GSA approach is suitable for solving OPF problem system wind system.
with the higher number of buses and various related factors to the Time (S) GSA [22] MSA [24] MSA–GSA
OPF problem. Similarly, the MSA–GSA approach is simulated and CPU Time for IEEE-30 bus 51.4 67.672 41.85
tested with and without integrating wind power systems with the CPU Time for IEEE-57 bus 67.75 68.01 49.85
CPU Time for IEEE-118 bus 87.34 92.55 59.648
regular power systems. To evaluate the efficacy of MSA–GSA the
obtained results in terms of Fuel cost and convergence rate with
the wind and without wind generation is given here.
Similarly, the power system with wind power generation is the complexity in the system is also reduced by adding the best
also experimented and verified through the simulation results. features of the individual approaches.
The simulation results are compared among MSA, GSA and MSA– The convergence rate of the three approaches is obtained from
GSA approach for IEEE-30, 57 and 118 bus systems in terms of the simulation results and shown in Fig. 3.
power loss (Table 4), fuel cost (Table 5), CPU utilization time To verify and evaluate the performance of the hybrid approach
(Table 6) and convergence rate (Fig. 4). There are some deviations the convergence rate of the three approaches are compared and
between the MSA and GSA approaches due to more number of shown in Fig. 7. From Fig. 7, it is absorbed that the hybrid
assignment and comparisons for searching the optimum results. approach performs better than the other approaches in terms
Comparing with the MSA and GSA the hybrid MSA–GSA approach of convergence rate. This paper tested and verifies the obtained
obtained a better result in terms of reduced power loss, fuel cost simulation results after each optimization process in terms of ac-
and less utilization of the CPU. The number of comparisons and curacy comparing with the existing results. The control variables,
Shilaja C. and Arunprasath T. / Future Generation Computer Systems 98 (2019) 708–715 715

constraints, state and dependent variables used in this paper is [19] S. Duman, U. Guvenc, N. Yorukeren, Gravitational search algorithm for
same with the existing studies [22,24] and in addition to that economic dispatch with valve-point effects, Int. Rev. Electr. Eng. 5 (6)
IEEE-118 bus system is also attached. Most of the results are (2010) 2890–2895.
[20] O. Ceylan, A. Ozdemir, H. Dag, Gravitational search algorithm for post-
verified and certain results are given in the paper such as fuel
outage bus voltage magnitude calculations, in: International Universities’
cost, time and convergence rate in terms of iterations. From the Power Engineering Conference, Wales (UK); 31 August–3 September, 2010.
results, it is noticed that the hybrid approach proved as better [21] E. Rashedi, H. Nezamabadi-pour, S. Saryazdi, BGSA: binary gravitational
than the other existing approaches. search algorithm, Nat. Comput. 9 (2010) 727–745.
[22] E. Rashedi, E. Rashedi, H. Nezamabadi-pour, Saryazdi, GSA: a gravitational
5. Conclusion search algorithm, Inform. Sci. 179 (2009) 2232–2248.
[23] C. Li, J. Zhou, Parameters identification of hydraulic turbine governing
system using improved gravitational search algorithm, Energy Convers.
The major goal of the paper is providing a hybrid optimiza-
Manag. 52 (2011) 374–381.
tion method by joining the most significant features of the two [24] J. Hetzer, DC. Yu, K. Bhattarai, An economic dispatch model incorporating
excellent meta heuristic technique to improve the effectiveness wind power, IEEE Trans. Energy Convers. 23 (2008) 603–611.
to solve OPF issues. To complete it MAS and GSA techniques are [25] R.A. Jabr, B.C. Pal, Intermittent wind generation in optimal power flow
chosen and combine its characteristics in a hybrid method known dispatching, IET Gener. Transm. Distrib. 3 (2009) 66.
as MSA–GSA algorithm. This method uses the genuine character [26] S. Mishra, Y. Mishra, S. Vignesh, Security constrained economic dispatch
of both MA and GSA methods for good efficiency. Regarding considering wind energy conversion systems, in: IEEE Power Energy Soc.
Gen. Meet., 2011.
the functional application, all the hybrid techniques decreases
[27] E. Ela, M. O’Malley, Studying the variability and uncertainty impacts of
the CPU use, limits the application time and junction rate with variable generation at multiple timescales, IEEE Trans. Power Syst. 27 (3)
enhanced cost of the fuel of IEEE-57, IEEE-118, and IEEE-30 bus (2012) 1324–1333.
system. This results, clearly says that hybrid techniques are the [28] L. Shi, C. Wang, L. Yao, Y. Ni, M. Bazargan, Optimal power flow solution
best when compared with the other recent approaches. incorporating wind power, IEEE Syst. J. 6 (2012) 233–241.
[29] A. Panda, M. Tripathy, Optimal power flow solution of wind integrated
References power system using modified bacteria foraging algorithm, Int. J. Electr.
Power Energy Syst. 54 (2014) 306–314.
[30] HT. Jadhav, R. Roy, A comprehensive review on the grid integration of
[1] T. Niknam, R. Azizipanah-Abarghooee, R. Sedaghati, A. KavousiFard, An
doubly fed induction generator, Int. J. Electr. Power Energy Syst. 49 (2013)
enhanced hybrid particle swarm optimization and simulated annealing
8–18.
for practical economic dispatch, Energy Educ. Sci. Technol. Part AJ 30 (1)
[31] Y. Li, Q. Wu, M. Li, J. Zhan, Mean–variance model for power sys-
(2012) 553–564.
[2] T. Niknam, F. Golestaneh, Enhanced adaptive particle swarm optimization tem economic dispatch with wind power integrated, Energy 72 (2014)
algorithm for dynamic non-convex economic dispatch, IET, J. Mag. Gener. 510–520.
Transm. Distrib. 6 (5) (2012) 424–435. [32] R. Bacher, H.P. Van Meeteren, Real-time optimal power flow in automatic
[3] J. Aghaei, T. Niknam, R. Azizipanah-Abarghooee, J.M. Sanchez, Scenario generation control, IEEE Trans. Power Syst. 3 (4) (1988) 1518–1529.
based dynamic economic emission dispatch considering load and wind [33] D. Lew, G. Brinkman, E. Ibanez, B. Hodge, J. King, The western wind and
power uncertainties, Int. J. Electr. Power Energy Syst. 47 (2013) 351–367. solar integration study, Nat. Renew. Energy Lab. 2010, NREL/TP-5500.
[4] L.B. Shi, C. Wang, L. Yao, Y. Ni, M. Bazargan, Optimal power flow solution [34] R. Doherty, M. O’Malley, A new approach to quantify reserve demand in
incorporating wind power, IEEE, Gener. Syst. J. 6 (2) (2012) 233–241. systems with significant installed wind capacity, IEEE Trans. Power Syst.
[5] J.M. Morales, A.J. Conejo, J. Perez-Ruiz, Economic valuation of reserves in 20 (2) (2005) 587–595.
power systems with high penetration of wind power, IEEE Trans. Power [35] P.A. Ruiz, C.R. Philbrick, E. Zak, K.W. Cheung, P.W. Sauer, Uncertainty
Syst. 24 (2) (2009) 900–910. management in the unit commitment problem, IEEE Trans. Power Syst.
[6] A. Papavasiliou, S.S. Oren, R.P. O’Neill, Reserve requirements for wind 24 (2) (2009) 642–651.
power integration: A scenario-based stochastic programming framework,
IEEE Trans. Power Syst. 26 (4) (2011) 2197–2206.
[7] J. Wang, M. Shahidehpour, Z. Li, Security-constrained unit commitment Dr C.Shilaja, completed her M.E. in Sathyabama uni-
with volatile wind power generation, IEEE Trans. Power Syst. 23 (3) (2008) versity Chennai. & Ph.D. in Power System at VIT
1319–1327. University Vellore. Now currently doing POST DOCT
[8] S.M. Ryan, R.J.B. Wets, D.L. Woodruff, C. Silva-Monroy, J.-P. Watson, To- FELLOW in Kalasalingam Academy of Research And
ward. scalable, Toward scalable, parallel progressive hedging for stochastic Education Krishnankoil. She has published 12 Inter-
unit commitment, Proc. 2013 IEEE Power and Energy Soc. General Meeting national journals and 3 international conference.The
(PES) 2013 (2013) 1–5. area of interest focused on Optimal Power Flow,Power
[9] Y. Wang, Q. Xia, C. Kang, Unit commitment with volatile node injections System Optimization, Renewable Energy.
by using interval optimization, IEEE Trans. Power Syst. 26 (3) (2011)
1705–1713.
[10] W. Lei, M. Shahidehpour, L. Zuyi, Comparison of scenario-based and
interval optimization approaches to stochastic SCUC, IEEE Trans. Power
Syst. 27 (2) (2012) 913–921.
[11] R.A. Jabr, Adjustable robust OPF with renewable energy sources, IEEE Trans. Dr T.Arunprasath, has completed his Ph.D. at
Power Syst. 28 (4) (2013) 4742–4751. Kalasalingam Academy of Research And Education Kr-
[12] R.A. Jabr, S. Karaki, J.A. Korbane, Robust multi-period OPF with storage and ishnankoil Tamilnadu. Currently Working in Associate
renewables, IEEE Trans. Power Syst. 30 (5) (2015) 2790–2799. professor,Department of EEE,Kalasalingam Academy of
Research And Education . His area of interest focused
[13] A. Lorca, X.A. Sun, Adaptive robust optimization with dynamic uncertainty
on MedicalImageprocessing ,Image Reconstruction,PET
sets for multi-period economic dispatch under significant wind, IEEE Trans.
Imaging .
Power Syst. 30 (4) (2015) 1702–1713.
[14] M. and Lubin, JuMPChance.jl: JuMPChance 0.1.0, 2015 [Online]. Available:
http://dx.doi.org/10.5281/zenodo.13740.
[15] A. Nemirovski, A. Shapiro, Convex approximations.
[16] D. Bertsimas, M. Sim, The price of robustness, Oper. Res. 52 (1) (2004)
35–53.
[17] HR. Hassanzadeh, M. Rouhani, A multi-objective gravitational search al-
gorithm, in: Proceedings of the Communication Systems and Network
Conference, 2010, pp. 7–12.
[18] SR. Balachandar, K. Kannan, A meta-heuristic algorithm for set covering
problem based on gravity, Int. J. Comput. Math. Sci. 4 (2010) 223–228.

You might also like