INTERNATIONAL JOURNAL of RENEWABLE ENERGY RESEARCH

Anand Kumar Pandey et al., Vol.9, No.4, December, 2019

Multi-Objective Optimal Location and Sizing of

Hybrid Photovoltaic System in Distribution Systems
Using Crow Search Algorithm
Anand Kumar Pandey*‡, Sheeraz Kirmani*

*Department of Electrical Engineering, Faculty of Engineering & Technology, Jamia Millia Islamia, New Delhi-110025, India
([email protected], [email protected])

‡
Corresponding Author; Anand Kumar Pandey, Jamia Millia Islamia, New Delhi-110025, India, Tel: +91 7838434833,
[email protected]

Received: 10.10.2019 Accepted:20.11.2019

Abstract- The optimal size and location of Distributed generation (DG) in a distribution system can minimize the power loss
and improve the voltage profile of the distribution system. But the random placement of DG will not solve the problem of
power loss and voltage profile. Therefore optimization algorithms are used to find the optimal size and location of DG to
minimize the power loss and to improve the voltage profile. This research paper introduces a new metaheuristic algorithm
which is the Crow search algorithm (CSA) to calculate the optimal size and location of multiple solar photovoltaic (PV) units
for reducing the power loss and improving the voltage profile. This paper also discusses the complete mathematical modeling
of probabilistic PV generation, time-varying load modeling, loss calculation, voltage stability, and optimal PV size calculation.
The method proposed in this paper is tested on IEEE 30 and 57 bus test systems and compared with other existing methods like
Genetic algorithm (GA), Particle swarm optimization (PSO) and Hybrid GA-PSO method. CSA has only 2 settings parameters
while GA and PSO have 6 and 3 settings parameters. Because of this, computation time and complexity of the algorithm are
less in comparison to other methods and it gives a better result. Therefore by using the CSA method, the optimal size and
location of the DG can be found which gives less power loss and improved voltage profile, less computational time and easy to
implement compared to other existing methods proposed here.
Keywords Solar PV, Distributed Generation, Crow search algorithm, Genetic Algorithm, Particle swarm optimization, hybrid
GA-PSO method and optimal size.

1. Introduction generations (DG) or embedded generation is that a source

which is connected near to the load [2]. To minimize the
Power system performance in terms of the power losses and power loss and improve the voltage profile placement of DG
voltage profile can be improved by the use of renewable in distribution systems plays a major role. A random
energy sources such as solar, wind, biogas, geothermal etc. placement of DG will lead to more power losses and poor
As the data compiled by ministry of power around 22 percent voltage profile. Therefore DGs are placed at optimal location
of the total electricity produced in India got lost during of optimum size so that power loss can be reduced and
distribution. There are many states in India which have as voltage profile can be improved[3]. In literature different
high as 40 percent transmission and distribution (T&D) methods has been suggested to find the optimal size and
losses. Odisha has 38.2 percent T&D loss in May 2018 location of DG in Distribution system. These methods can be
which is worst among all the state second state is Madhya classified as classical methods, artificial intelligent methods
Pradesh which has 37.7 percent T&D loss which is followed and hybrid intelligence methods. Comparison of each
by West Bengal having 31.44 percent T&D loss [1]. Table1 method is mentioned and given in [4]. From the paper it is
below gives the T & D losses in terms of total electricity clear that each method has some advantages and
generated in different states of India. From the table 1 it is disadvantages both. In [3, 5] analytical methods are proposed
obvious that the T&D losses plays a major role in Power to find the best size and location of DGs. In [6-8] Genetic
system performance. One of the solutions to tackle this algorithm (GA) approach and method is explained to find the
problem is to use distributed generators at load center. The optimal size and location of DG. Paper [9-12] discusses
definition of the Renewable energy resources or Distributed particle swarm optimization (PSO) method to find the
INTERNATIONAL JOURNAL of RENEWABLE ENERGY RESEARCH
Anand Kumar Pandey et al., Vol.9, No.4, December, 2019

Table 1: State wise T&D losses

S.N. State T & D loss S.N. State T & D loss
(in percent) (in percent)
1 Himachal Pradesh 9.54 10 Chhattisgarh 20.84
2 Uttranchal 25.02 11 West Bengal 31.44
3 Haryana 23.7 12 Maharashtra 20.1
4 Rajasthan 25.1 13 Odisha 38.2
5 Uttar pradesh 29.6 14 Telangana 15.9
6 Bihar 31.8 15 Karnatka 15.4
7 Assam 22.1 16 Andhra Pradesh 21.8
8 Gujarat 15.4 17 Kerala 15.26
9 Madhya Pradesh 37.7 18 Tamil Nadu 14

optimal size and location of DG. An Artificial bee colony power can be extracted. The power value can vary with
(ABC) method is explained in [13] to find the size and respect to time due to its normal and abnormal condition
location of DG. In[14] ant lion optimization method is of weather pattern, changes of climate condition. The
used to discuss the problem of DG location and sizing. In probabilistic nature of solar irradiance can be described
the literature different optimization methods has been using the beta probability distribution function as[15].
discussed to find the optimal size and location of DG. In
[15] GA and PSO combined method for the improvement &(()*)
of result has been proposed to find the location and size of (𝑅(-. )(1 − 𝑅*-. ) 0 ≤ 𝑅 ≤ 1, 𝑎, 𝑏 ≥ 0
𝐵𝐷𝐹 = %&(()&(*) (1)
DG. Mixed integer conic programming method is 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
presented in [16] to find the type, location and sizing of Where BDF, is the beta distribution function; a,b are the
battery and DG. In [17] author has included uncertainties parameters of beta distribution function; R denotes the
of DG to find the optimal location and size of DG. In [25- random variable of solar irradiance.
30] different authors have used different methods and B(.)B)
mathematical functions for planning of distribution 𝑏 = (1 − 𝜇)[ − 1] (2)
CD
systems. Methods which comes under classical approach B×*
has disadvantage of large computation time, slow 𝑎 = .-B (3)
convergence and can consider single objective while
The standard deviation (𝜎) and the mean (µ) values are
methods which comes under artificial intelligence category
.given in table 4.
has advantage of efficient performance, needs fewer
iterations and can handle complex problems but The maximum power, fill factor of the panel, voltage,
disadvantage of many setting parameters [4]. Parameter current, and temperature of the panels are derived by
setting is one of the draw backs of optimization algorithms following equations[15]
since it is time consuming work. Algorithms which have
fewer parameters to adjust are easier to implement. In Maximum power:
Crow search algorithm (CSA) only two parameters flight 𝑃(𝑅) = 𝑁 × 𝑓𝑓 × 𝑉L × 𝐼L (4)
length and awareness probability need to be tuned while in
GA six parameters like selection method, cross over The output power at different irradiance P(R) , number of
method, cross over probability, mutation method, mutation modules are (N), fill factor is (ff), voltage and current
probability and replacement method need to be tuned in obtained from the PV is 𝑉L and 𝐼L respectively.
PSO algorithm four parameters inertia weight, maximum Fill factor:
value of velocity, individual learning factor and social PQR× SQR
learning factor need to be adjusted. This paper suggests a 𝑓𝑓 = PTU× SVU
(5)
new metaheuristic method for optimal solar PV placement
which has only 2 setting parameters. The remaining 𝑉WX and 𝐼WX is the voltage and current at maximum power
sections are as follows: Section 2 discusses about solar point; 𝑉YZ and 𝐼[Z is the Open circuit voltage and current.
PV, load and energy storage modelling Section 3 explain
problem formulation section 4 application of Crow search Voltage:
algorithm to find the optimal size of solar PV, Section 5 𝑉L = 𝑉YZ − 𝐶] × 𝑇_Z`aa (6)
discusses about the result and discussion and in the last
Section 6 we discuss about the conclusion. Current:

2. Solar PV, Load and Energy storage modelling: 𝐼L = 𝑅[𝐼[Z + 𝐶c × 𝑇_Z`aa ] (7)
In (6 and 7) the open circuit voltage is (𝑉YZ ); Short circuit
2.1 Solar PV modelling current (𝐼[Z ); voltage and current temperature coefficient
Output of the PV module depends on the temperature and (𝐶] , 𝐶c ).The cell temperature (𝑇_Z`aa ).
solar irradiation; from the panels the maximum amount of

1682
INTERNATIONAL JOURNAL of RENEWABLE ENERGY RESEARCH
Anand Kumar Pandey et al., Vol.9, No.4, December, 2019
v
Cell temperature:
𝜖(𝑃] + 𝐸[ )m = † 𝑃m (𝜏)𝑑𝜏
d -fg
𝑇_Z`aa = 𝑇( + 𝑅( Teg.i ) (8) g
⇒ ∑f‰
Š‹. 𝑃m (𝜏)∆𝑡 (14)
The normal operating temperature (𝑇_Z`aa ), the ambient
temperature (𝑇( ) is in (8). Here, in a 24-h day cycle when the PV output is zero or
small battery supply energy to the load and for PV output
more than load demand battery stores energy. The power
Output power: factors of PV-BES unit are selected for each 1-h period.
. This approach can reduce energy loss, which is related to
𝑃(𝜏) = ∫g 𝑃(𝑅) × 𝐵𝐷𝐹𝑑𝑟 (9) active and reactive power loss index. In this paper PV and
2.2 Load Modelling: battery units are placed at the same bus to avoid energy
In this paper commercial load is assumed. The load under losses during charging of the battery.
study is assumed to follow 24-h daily commercial load 3.2 Active power loss index (APLI) and Reactive power
curve[16]. This is voltage dependant and represented by loss index (RPLI)[15]
active and reactive power injection. 3.2.1 Active Power loss index (APLI)
Active power injection: In a radial distribution system having n branches, the total
oX active power loss (Ploss) can be given as:
𝑃m (𝜏) = 𝑃n (𝜏) × 𝑉m (𝜏) (10)
Reactive power injection:
oq sŒD )•ŒD
𝑄m (𝜏) = 𝑄n (𝜏) × 𝑉m (𝜏) (11) 𝑃aY[[ = ∑m•‹. D 𝑅• (15)
ŽPŒŽ
In above equation (10, 11), the active and reactive loads
are arranged as 𝑃n (𝜏) and 𝑄n (𝜏) ; voltage at kth bus is Total active power loss (Ploss(Pv+Es)) in the system
oX
𝑉m (𝜏), 𝑄n (𝜏);𝑃m (𝜏)is the active power injections and because of the DG and BES both can be given as :
𝑄m (𝜏) is the reactive power injections at bus k. np=1.51
s•D -fsŒs•
and nq=3.40 are respectively commercial load voltage 𝑃aY[[ (𝑃] + 𝐸[ ) = ∑m•‹. D 𝑅• +
ŽPŒŽ
exponents. D D
(• s• -f•Œ(• s•
2.3 Energy storage modelling: ∑m•‹. D 𝑅• + 𝑃aY[[ (16)
ŽPŒ Ž
The battery energy storage (BES) is the battery and it is
modelled to save energy, first calculate the demand power
s“TVV (s”)‚V )
of 24 hrs, and it varies. Beyond the limit of load demand, 𝐴𝑃𝐿𝐼 = s“TVV
(17)
energy is stored in the battery else, it supplies the power to 𝑄„ = 𝑎„ 𝑃„ (18)
the distribution system. The charging and discharging of
Where, 𝑎„ = ±tan (cos-. (𝑝𝑓m ))
battery at bus k in period t can be expressed as [17]
Active power loss index is APLI ; the resistance and the
At Discharging:
s tUu (v)
voltage magnitude of the node is 𝑅• and Ž𝑉• Ž it is
𝜀m (𝜏) = 𝜀(𝜏 − 1) − ∆𝜏, 𝐹𝑜𝑟 𝑃(𝜏) > 0 (12) represented in (15-17).
wt
At charging:
𝜀m (𝜏) = 𝜀(𝜏 − 1) − 𝜂Z 𝑃Z{ (𝜏)∆𝜏, 𝐹𝑜𝑟 𝑃(𝜏) < 0 (13)
The total energy stored in the battery (𝜀m (𝜏)); 𝑃}Uu (𝜏) is
Power at discharging; 𝑃Z{ (𝜏) is at charging; η• , 𝜂Z are the
charging and discharging coefficients of battery; 𝜀m (𝜏) is
total energy stored in the battery.
The upper and lower bounds of the BES unit is taken as PV Inverter
follows[18]:
Wco W(L
𝐸•‚ƒ„ ≤ 𝐸•‚ƒ„ (𝑡) ≤ 𝐸•‚ƒ„
Wco W(L
Where 𝐸•‚ƒ„ and 𝐸•‚ƒ„ represent the minimum and
maximum limit of the energy stored in BES unit. The
minimum and maximum limit are assumed to be 20% and
Load
90% of the total capacity of the BES unit [18].
3. Problem Formulation:

3.1 Conceptual design :

Proposed conceptual model of a stand-alone roof top PV-
BES system is shown in fig. 1a and b. The idea is to BES Inverter
minimize the active and reactive power losses for each
load level at the minimum with a combination of PV-BES
units. This modelling can be accomplished due to the PV
and energy storage, and it can generate hybrid energy with
variation of time and it is given by, Fig. 1(a). Model of a PV-BES system connected to
a load

1683
INTERNATIONAL JOURNAL of RENEWABLE ENERGY RESEARCH
Anand Kumar Pandey et al., Vol.9, No.4, December, 2019
¢ ¢
min(MOI)=min(¤£ P¦¤§ + ¨D Q¦¤§ ) (25)
¥ ¥

Where P¦ and P¦¤§ indicate the active power loss of the

system without and with DG system. Q¦ and Q¦¤§ indicate
the reactive power loss of the system without and with DG
system.
Under the constraints of:
𝑉 Wco ≤ 𝑉m (𝜏) ≤ 𝑉 W(L (26)
DG -®¯ ≤ DG ≤ DG -°± (27)
Vmin and Vmax are minimum and maximum voltage limit of
the bus. DGmin and DGmax are the minimum and maximum
size of the DG.
3.5 Energy loss and Voltage stability
Fig. 1(b). Charging and discharging characteristics Energy loss:
of BES and PV output. Total energy loss in a year over the time period of 24 h
can be obtained as:
3.2.2 Reactive power loss index 𝐸aY[[ = 365 ∑f‰Š‹. 𝑃aY[[ (𝜏) ∆𝜏 (28)
In a radial distribution system having n branches, the total
reactive power loss (Qloss ) can be given as : Where Ploss(t) is active power loss at any time t.
sŒD )•ŒD Voltage stability:
𝑄aY[[ = ∑m•‹. D 𝑋• (19) Active power demand
ŽPŒ Ž
𝑃µ`W(oµ = 𝜆𝑃g (29)
Total reactive power loss (Qloss(Pv+Es)) in the system
Reactive power demand
because of the DG and BES both can be given as :
s•D -fsŒs•
𝑄µ`W(oµ = 𝜆𝑄g (30)
𝑄aY[[ (𝑃] + 𝐸[ ) = ∑m•‹. D 𝑋• + Po and Qo are initial active and reactive power demand
ŽPŒ Ž
D D respectively.
∑ m (• s• -f•Œ(• s•
•‹. D 𝑋• + 𝑄aY[[ (20) 4. Proposed Methodology:
ŽPŒ Ž
•“TVV (s”)‚V )
𝑅𝑃𝐿𝐼 = (21) The output of the solar system depends on the solar
•“TVV
The reactive power loss index is RPLI; 𝑄aY[[ is the reactive irradiance and the temperature. The battery is required to
power loss store the energy, and it can supply the energy when the PV
system is in abnormal condition. The major problem in the
3.3 Multi-objective index (MOI) distribution systems are charging, discharging, PV size,
and DG location. The capacity of the PV panels are varies
Multi-objective index is given as: in accordance with the voltage, and power. We can
𝑀𝑂𝐼 = 𝜎. 𝐴𝑃𝐿𝐼 + 𝜎f 𝑅𝑃𝐿𝐼 (22) estimate how much amount of power can be generated
Where ∑fc-. 𝜎c = 1 from the single panel, and it will be equal in all kind of
The (MOI) is the multi-objective index. It is the addition panels. First, discuss the optimization algorithms to reduce
of active power loss index (APLI) and reactive power loss the optimal size, and the battery is designed to save the
index (RPLI). σ1 and σ2 are impact indices which values power as well as discharge the power, for these reasons the
lies between 0 to 1. In this paper the value of σ1 and σ2 are algorithms are established. The system is analysed in IEEE
taken as 0.7 and 0.3 respectively[19]. Finding the suitable 30, and IEEE 57 bus system.
values of the weight depends on the experience and
Crow Search Algorithm (CSA) [20]
concerns of the engineers. Now the average multiobjective
index (AMOI) over the total time (T=24 h) can be One of the meta-heuristic algorithms is a crow search
calculated as: algorithm, which can be used to find the optimal sizing
. v
𝐴𝑀𝑂𝐼 = ∫g 𝑀𝑂𝐼(𝜏)𝑑𝜏 (23) and location of DG system. The inspiration of the
v
⇒ algorithm is based on the intelligence of the crows. The
.
⇒ f‰ ∑f‰ lifestyle of the crows depend on the following behaviors,
Š‹. 𝑀𝑂𝐼(𝜏)𝑑𝜏 (24)
Minimum value of the AMOI over the total time indicated • Self-awareness
gives the best PV and BES allocation for energy loss • Recognition faces
reduction and voltage stability improvement.
• Using tools
3.4 Objective Function:
• Warn and flock
• Sophisticated communication ways
The objective function is to minimize the active and • Recalling the food’s hidden place
reactive power loss of the system using mutiobjective
index (MOI) i.e. Objective function The brain bodies ratios of the crows are slightly differ
from the humans, while comparing with other birds; the

1684
INTERNATIONAL JOURNAL of RENEWABLE ENERGY RESEARCH
Anand Kumar Pandey et al., Vol.9, No.4, December, 2019
crows are the most intelligent birds in nature. The Step4: Now generate the new position for each crow using
evolutionary process among the crows are defined as eq. (34)
hiding of food and recovering the extra food. The position
of the crow and the number of iteration is considered in Step5: Check the feasibility of each position, if the new
equation (31). position of the crow is feasible the crow updates its
position otherwise stays in the same position.
𝑍(𝑙. 𝑚) =
𝑖 = 1,2, . . , 𝑁 Step6: Evaluate the fitness function of new positions.
〈𝑧. (𝑙, 𝑚), 𝑧 f (𝑙, 𝑚), … … , 𝑧 o (𝑙, 𝑚)〉;𝑓𝑜𝑟 ¾
𝑘 = 1,2, … . , 𝑚𝑎𝑥
Step7: Update memory .If the fitness function value of the
(31) new position of the crow is better than the fitness function
value of the memorized position, the crow updates its
The crow has the best remembering capability, which can memory by the new position.
easily get the best visited location and it is represented by,
Step8: Check termination criterion. Steps 4-7 are repeated
𝑅(𝑙, 𝑚) = 〈𝑟. (𝑙, 𝑚), 𝑟 f (𝑙, 𝑚), … … , 𝑟 o (𝑙, 𝑚)〉 (32) until itermaxis reached. When the termination criterion is
met, the best position of the memory in terms of the
The behaviour of the crows are pursuit, and evasion which objective function value is reported as the solution of the
is defined by, optimization problem.
Pursuit: Table 2: Nomenclature used in Proposed Methodolog
The crows are moving like one by one to knows about the 𝑍(𝑙, 𝑚) Position of the crow
hidden place, moving of first crow never seen the
N Number of individual crows
following crow, unfortunately the follower can take food
k Maximum number of iteration
Evasion: 𝑅(𝑙, 𝑚) Visited location
𝑔(𝑙, 𝑚) Flight length
When the crow knows about the follower, it just take the
decision to move randomly, for that calculate the 𝑅(𝑙, 𝑚) Best position
probability (ap) of random generation. The uniformly f Random variable
distributed values () are taken as 0 and 1. The major
calculations are awareness probability, and taking a
random value. The algorithm related with the works are given in the
pseudo code and it is by,
𝑖𝑓 𝑑a ≥ 𝑎𝑝 = 1 𝑎𝑛𝑑 𝑑a ≤ 𝑎𝑝 = 0
Table 3: Algorithm of CSO
The probability of the random value is greater than
awareness probability, then the condition is given by, Step 1: Initialize Decision variables (load flow)
Step 2: Population size (24hrs)
𝑍(𝑙, 𝑚 + 1) = [𝑧(𝑙, 𝑚) + 𝑑a × 𝑓Â (𝑙, 𝑚) × Ã𝑅(𝑙, 𝑚) − Step 3: Initialize current and voltage temperature
𝑍(𝑙, 𝑚)Ä] (33) coefficients (equation6&7)
Step 4: Initialize the size of PV
The position can be evaluated by, Step 5: Set upper and lower bounds
Step 6:Evaluate the objective function (Min (PV Size))
𝐹Ã𝑧(𝑙, 𝑚 + 1)Ä 𝑖𝑓 𝐹(𝑧(𝑙, 𝑚 + 1) < 𝐹(𝑧(𝑙, 𝑚)) Step 7: Find fitness function of all position between the
% (34) upper and lower bounds
(𝑙, 𝑚) 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
Step 8: Initialize maximum number of iteration
In above representation, 𝐹Ã𝑧(𝑙, 𝑚 + 1)Ä term denotes the Step 9:For
t=1:maximum number of iteration
objective function, in which calculate the sizing of PV.
Choose random variables based on population size
Step1: Input the line data, bus data, bus voltage limit and For
adjustable parameters of the algorithm like flock size (N), i=1:N(24hrs)
maximum number of iterations (itermax), flight length (fl) If rand position>step 3
and awareness probability (AP). Generate new position for first state
Else
Step2: Calculate the active power loss and reactive power For
loss using eqn. (15, 19) and initialize the position and j=1:step 1
memory of crows. Generate new position
End
Step3: Find the value of objective function for each crow End
using eqn. (25) and determine the position. End

1685
INTERNATIONAL JOURNAL of RENEWABLE ENERGY RESEARCH
Anand Kumar Pandey et al., Vol.9, No.4, December, 2019

Step 10: Evaluate fitness of new generated coefficients The definition wise the optimal size is the size, which can
Step 11: Repeat step 9, until reach the better results maximize the internal rate of the return investment of the
End PV system. Fist initializing the load flow, and finding the
losses of the system. The values of generated active and
reactive power, as well as the active and reactive load
The analysis and expressions are mathematically derived power are taken for the calculation. Here the IEEE 30 and
and the results are taken which is given in next section. 57 bus test system is carried out for the analysis of the
proposed system. The details about the test systems are
5. Result and Discussion given in []. The output of the solar PV panel, voltage,
losses and the size of the panel are analyzed for
the IEEE 30 and 57 bus test system, which is discussed in similar calculation and graph can be plotted. All these
this section. For the study purpose 4 units of solar PV units are placed at pre-defined locations unit 1and unit 3
system has been taken at different power factor explained are assumed to be placed at bus number 22, 25, and 29.
in Fig.2. From the solar PV modelling discussed in Unit 2 and unit 4 are placed at 20, 29. By considering
problem formulation section, Output power of the unit 1 these locations optimal size of four units has been
Solar PV panel is calculated using eqn. (1) to (9) and calculated.
Table.1-2 and represented in Fig.3. For unit 2, 3 and 4

Table 4: Mean & Standard deviation of solar irradiance

Hour µ(kw/m2) σ (kw/m2) Hour µ(kw/m2) σ (kw/m2)

1 0 0 13 0.657 0.164
2 0 0 14 0.612 0.147
3 0 0 15 0.497 0.143
4 0 0 16 0.349 0.116
5 0 0 17 0.203 0.081
6 0.007 0.021 18 0.068 0.063
7 0.081 0.036 19 0.003 0.012
8 0.237 0.056 20 0 0
9 0.400 0.087 21 0 0
10 0.523 0.127 22 0 0
11 0.632 0.156 23 0 0
12 0.663 0.162 24 0 0

Table 5: PV module characteristics

PV module characteristics Value

Nominal Cell operating temperature, NOT (0C) 43
Current at maximum power point, IMPP (A) 7.76
Voltage at maximum power point, VMPP (V) 28.36
Short circuit current, Isc (A) 8.38
Open circuit voltage, Voc (V) 36.96
Current temperature coefficients, Ki (A/oC) 0.00545
Voltage temperature coefficients, Kv (V/oC) 0.1278

1686
INTERNATIONAL JOURNAL of RENEWABLE ENERGY RESEARCH
Anand Kumar Pandey et al., Vol.9, No.4, December, 2019
The output power obtained from the unit-1 in each interval
of time is plotted on semi-log graph in Fig.1

Now, by using the crow search optimization algorithm

which is proposed in this article and explained in proposed
methodology section we can find the optimal size of Solar
PV panel. This is given in Table 6. Particle swarm optimization (PSO) method and Hybrid
PSO-GA method and same is plotted in Figure 5. From the
Result of IEEE- 30 bus test system: figure it is clear that Crow search optimization gives better
result compare to other methods. The optimal size of each
Now if we compare the size of solar PV before and after method like HPSO-GA, PSO, GA, and CSA are 2.9e05
the optimization obtained from the proposed method and
plotted in Fig. 3. We comment that size of the solar PV is
reduced after the optimization. In this study four different
solar units have been taken for study on the basis of
different power factor. Figure 4 represent power loss of
the system. The method proposed here is compared with
other intelligent algorithm like Genetic algorithm (GA),

Table 6: Optimal Size obtained from CSA

Hour(hr) Size(MW) Hour(hr) Size(MW)
1 0 13 3.9559
2 0 14 3.0040
3 0 15 2.4303
4 0 16 2.2826
5 0 17 2.0616
6 1.5844 18 1.7363
7 1.8474 19 1.6667
8 2.2219 20 0
9 2.3757 21 0
10 2.4679 22 0
11 2.9078 23 0
12 4.4290 24 0

1687
INTERNATIONAL JOURNAL of RENEWABLE ENERGY RESEARCH
Anand Kumar Pandey et al., Vol.9, No.4, December, 2019

Figure 2: Power Factor of Different Units

Figure 3: PV size before and after optimization of IEEE 30 bus system

1688
INTERNATIONAL JOURNAL of RENEWABLE ENERGY RESEARCH
Anand Kumar Pandey et al., Vol.9, No.4, December, 2019
Figure 4: Power loss of IEEE 30 bus

Figure 5: Comparison of PV size with existing algorithms for IEEE 30 bus

Result of IEEE- 57 bus test system: other intelligent algorithm like Genetic algorithm (GA),
Particle swarm optimization (PSO) method and Hybrid
2.0e05 W, 1.87e05 W and 4.57e04 W respectively. Now PSO-GA method and same is plotted in Figure 5. From the
if we compare the size of solar PV before and after the figure it is clear that Crow search optimization gives better
optimization obtained from the proposed method and result in compare to other methods. The optimal size of
plotted in Fig. 6. We comment that size of the solar PV is each method like HPSO-GA, PSO, GA, and CSA are
reduced after the optimization. In this study four different 2.9e05 W, 2.87e05 W, 5.61e04 W and 4.80e04 W
solar units have been taken for study on the basis of respectively
different power factor. Figure 4 represent power loss of
the system. The method proposed here is compared with

Table 7: Optimal Size obtained from CSA

Hour(hr) Size(mw) Hour(hr) Size(mw)
1 0 13 3.9559
2 0 14 3.0040.7
3 0 15 2.4303.5
4 0 16 2.2826
5 0 17 2.0616.5
6 1.5844.9 18 1.7363.2
7 1.8474.5 19 1.6667.3
8 2.2219.8 20 0
9 2.3757.8 21 0
10 2.4679.9 22 0
11 2.9078.7 23 0

1689
INTERNATIONAL JOURNAL of RENEWABLE ENERGY RESEARCH
Anand Kumar Pandey et al., Vol.9, No.4, December, 2019

12 4.4290.4 24 0

Figure 6: PV size before and after optimization of IEEE 30 bus system

Figure 7: Power loss of IEEE- 67bus

1690
INTERNATIONAL JOURNAL of RENEWABLE ENERGY RESEARCH
Anand Kumar Pandey et al., Vol.9, No.4, December, 2019

Figure 8: Comparison of PV size with existing algorithms for IEEE 30 bus

Table 8: Optimal Size of DG

S.No Test Algorithms Optimal size (Mw) Total size

system (Mw)
Unit 1 Unit2 Unit 3 Unit 4
HPSO-GA 7.39e04 8.33e04 7.38e4 7.39e 2.9e05
PSO 7.17e04 7.18e04 7.23e04 7.24e04 2.0e05
1
IEEE 30 GA 4.66e04 4.69e04 4.70e04 4.71e04 1.87e05
CS 1.27e04 0.99e4 1.11e04 1.19e04 4.57e04
HPSO-GA 7.41e04 7.38e04 7.38e04 7.37e04 2.95e05
PSO 7.17e04 7.22e04 7.13e04 7.24e04 2.87e04
2 IEEE57
GA 1.07e04 1.21e04 1.62e04 1.70e04 5.61e04
CS 1.20e04 1.11e04 0.977e03 1.51e04 4.80e04

6. Conclusion

Thus the proposed system has successfully implemented in generators in radial systems. Electric Power
MATLAB environment, under this, the mathematical Systems Research 2009. 79(6): p. 912-918.
modelling is derived and the constraints of voltage and [6] El-Ela, A.A., S.M. Allam, and M. Shatla,
energy are being satisfied. The power losses of active and Maximal optimal benefits of distributed
reactive are gown through the Newton Raphson load flow generation using genetic algorithms. Electric
analysis. The crow search optimization algorithm has been Power Systems Research, 2010. 80(7): p. 869-
developed to overcome the issue of optimal sizing. The 877.
energy storage modelling is done to save the power; all [7] Evangelopoulos, V.A., P.S. Georgilakis, and
these calculations are based on the time interval, which is Distribution, Optimal distributed generation
to be 24 hrs. The results are made among the particle placement under uncertainties based on point
swarm optimization algorithm (PSO), and genetic estimate method embedded genetic algorithm.
algorithm (GA), and hybrid particle swarm genetic IET Generation, Transmission, 2014. 8(3): p.
algorithm (HPSOGA) in which the voltage, loss, loading 389-400.
parameter, and the size are estimated. The optimal size of [8] Pandey, A.K. and S. Kirmani. Implementation of
the HPSOGA, PSO, GA, and CS for IEEE 30 bus and 67 genetic algorithm to find the optimal timing of
bus test system is given in table 8. From the obtained overcurrent relays. in Power Electronics and
values, the proposed system can enable the efficient results Motion Control Conference (PEMC), 2016 IEEE
rather than the existing methods. International. 2016. Verna, Bulgaria: IEEE.
[9] Aman, M., et al., Optimal placement and sizing of
a DG based on a new power stability index and
References line losses. International Journal of Electrical
Power, 2012. 43(1): p. 1296-1304.
[1] Power, M.o., Ministry of Power, India, Annual [10] Kansal, S., et al., Optimal placement of different
Report. April 2018. type of DG sources in distribution networks.
[2] Pepermans, G., et al., Distributed generation: International Journal of Electrical Power, 2013.
definition, benefits and issues. Energy policy, 53: p. 752-760.
2005. 33(6): p. 787-798. [11] Devi, S., M. Geethanjali, and E. Systems,
[3] Acharya, N., P. Mahat, and N. Mithulananthan,
Optimal location and sizing determination of
An analytical approach for DG allocation in
Distributed Generation and DSTATCOM using
primary distribution network. International Particle Swarm Optimization algorithm.
Journal of Electrical Power Energy Systems,
International Journal of Electrical Power, 2014.
2006. 28(10): p. 669-678.
62: p. 562-570.
[4] HA, M.P., P.D. Huy, and V.K. [12] Kansal, S., V. Kumar, and B. Tyagi, Hybrid
Ramachandaramurthy, A review of the optimal approach for optimal placement of multiple DGs
allocation of distributed generation: Objectives, of multiple types in distribution networks.
constraints, methods, and algorithms. Renewable International Journal of Electrical Power Energy
Sustainable Energy Reviews 2017. 75: p. 293- Systems, 2016. 75: p. 226-235.
312.
[13] Abu-Mouti, F.S. and M. El-Hawary, Optimal
[5] Gözel, T. and M.H. Hocaoglu, An analytical
distributed generation allocation and sizing in
method for the sizing and siting of distributed distribution systems via artificial bee colony

1691
INTERNATIONAL JOURNAL of RENEWABLE ENERGY RESEARCH
Anand Kumar Pandey et al., Vol.9, No.4, December, 2019
algorithm. IEEE transactions on power delivery, [24] Kothari, I.J. and Nagrath, D.P. (2007) Modern
2011. 26(4): p. 2090-2101. Power System Analysis. 3rd Edition, New York.
[14] Ali, E., S.A. Elazim, and A. Abdelaziz, Optimal [25] A. K. Pandey and S. Kirmani, "Efficient
allocation and sizing of renewable distributed Approach for DG Placement and Size in Medium
generation using ant lion optimization algorithm. Voltage Distribution Systems," 2018 2nd IEEE
Electrical Engineering, 2018. 100(1): p. 99-109. International Conference on Power Electronics,
[15] Moradi, Mohammad Hasan, and M. Abedini. "A Intelligent Control and Energy Systems
combination of genetic algorithm and particle (ICPEICES), Delhi, India, 2018, pp. 127-131
swarm optimization for optimal DG location and [26] Essallah, Sirine, Adel Bouallegue, and A. K.
sizing in distribution systems." International Khedher. "Optimal sizing and placement of DG
Journal of Electrical Power & Energy units in radial distribution system." International
Systems 34.1 (2012): 66-74. Journal of Renewable Energy Research
[16] Home-Ortiz, Juan M., et al. "Optimal location- (IJRER) 8.1 (2018): 166-177.
allocation of storage devices and renewable-based [27] D. Motyka, M. Kajanová and P. Braciník, "The
DG in distribution systems." Electric Power Impact of Embedded Generation on Distribution
Systems Research 172 (2019): 11-21. Grid Operation," 2018 7th International
[17] HassanzadehFard, Hamid, and Alireza Jalilian. Conference on Renewable Energy Research and
"Optimal sizing and location of renewable energy Applications (ICRERA), Paris, 2018, pp. 360-364.
based DG units in distribution systems [28] T. Sakagami, Y. Shimizu and H. Kitano,
considering load growth." International Journal "Exchangeable batteries for micro EVs and
of Electrical Power & Energy Systems 101 renewable energy," 2017 IEEE 6th International
(2018): 356-370 Conference on Renewable Energy Research and
[18] Hung, D.Q., N. Mithulananthan, and R. Bansal, Applications (ICRERA), San Diego, CA, 2017,
Integration of PV and BES units in commercial pp. 701-705.
distribution systems considering energy loss and [29] P. Mazidi, G. N. Baltas, M. Eliassi and P.
voltage stability. Applied Energy, 2014. 113: p. Rodríguez, "A Model for Flexibility Analysis of
1162-1170. RESS with Electric Energy Storage and
[19] Lopez, E., et al., Online reconfiguration Reserve," 2018 7th International Conference on
considering variability demand: Applications to Renewable Energy Research and Applications
real networks. IEEE Transactions on Power (ICRERA), Paris, 2018, pp. 1004-1009.
systems, 2004. 19(1): p. 549-553. [30] M. Saleh, Y. Esa, N. Onuorah and A. A.
[20] Chen, S., H.B. Gooi, and M. Wang, Sizing of Mohamed, "Optimal microgrids placement in
energy storage for microgrids. IEEE Transactions electric distribution systems using complex
on Smart Grid, 2012. 3(1): p. 142-151. network framework," 2017 IEEE 6th
[21] Gabash, A. and P. Li, Active-reactive optimal International Conference on Renewable Energy
power flow in distribution networks with Research and Applications (ICRERA), San Diego,
embedded generation and battery storage. IEEE CA, 2017, pp. 1036-1040.
Transactions on Power Systems, 2012. 27(4): p. [31] Y. Tominaga, M. Tanaka, H. Eto, Y. Mizuno, N.
2026-2035. Matsui and F. Kurokawa, "Design Optimization
[22] Ochoa, L.F., A. Padilha-Feltrin, and G.P. of Renewable Energy System Using EMO," 2018
Harrison, Evaluating distributed generation International Conference on Smart Grid
impacts with a multiobjective index. IEEE (icSmartGrid), Nagasaki, Japan, 2018, pp. 258-
Transactions on Power Delivery, 2006. 21(3): p. 263.
1452-1458.
[23] Askarzadeh, A., A novel metaheuristic method
for solving constrained engineering optimization
problems: crow search algorithm. Computers
Structures, 2016. 169: p. 1-12.

1692
INTERNATIONAL JOURNAL of RENEWABLE ENERGY RESEARCH
Anand Kumar Pandey et al., Vol.9, No.4, December, 2019

1693