ACKNOWLEDGEMENT

We are grateful to the Department of Electrical Engineering, for giving us the

opportunity to execute this project, which is an integral part of the curriculum in B. Tech
programme at IIT(ISM)Dhanbad.

We would also like to take this opportunity to express heartfelt gratitude for our project
guide Mr. B. Krishna Naick, who provided us with valuable inputs at the critical stages this
project execution.

We would like to acknowledge the support of every individual who assisted us in making
this project a success and we would like to thank Prof. Sukanta Das, HOD Electrical
Department for his guidance, support and direction without which project would not have
taken shape.

Finally, we would like to thank our family and friends for their unwavering support,
encouragement, and understanding throughout our academic journey. Their support has been
a constant source of motivation for us.

Thank you all for your contributions, without which this project would not have been
possible.

1|Page
 ABSTRACT

Solar photovoltaic (PV) modules are a popular and sustainable source of renewable
energy. Accurate estimation of their parameters is crucial for optimizing their performance
and ensuring their efficient integration into the power grid.

This final year project report presents a comprehensive overview of the estimation
methods for determining the parameters of solar PV modules using optimization techniques.
The optimization algorithm used in this report are war strategy optimization algorithm, grey
wolf optimization algorithm, ant colony optimization algorithm and partial swarm
optimization algorithm.

The advantages and limitations of each optimization algorithm are analysed, and their
mathematical modelling, pseudocode, flow chart to is discussed. Additionally, the report
explores the challenges and uncertainties associated with parameter estimation, such as
environmental conditions, manufacturing tolerances, and degradation effects. Furthermore,
the report highlights the comparison between the algorithms. Finally, the report concludes
with recommendations for future research directions and practical implications for improving
the accuracy and reliability of PV panel parameter estimation.

The findings of this report contribute to the understanding of the current state of the art in
the estimation of solar PV module parameters and provide insights for researchers, engineers,
and policymakers working in the field of renewable energy.

2|Page
 TABLE OF CONTENTS

CERTIFICATES
ACKNOWLEDGMENT………………………………………………………………………1
ABSTRACT…………………………………………………………………………………...2
TABLE OF CONTENTS……………………………………………………………………...3
ABBREVIATION AND SYMBOLS …………………………………………………………5
LIST OF FIGURES AND GRAPHS………………………………………………………….6
LIST OF TABLES…………………………………………………………………………….8
Chapter 1: SOLAR PV………………………………………………………………………...9
1.1 Introduction……………………………………………………………………………9
1.2 Modeling of Solar PV system…………………………………………………………9
1.2.1 Single-diode model…………………………………………………………....9
1.2.2 Graphs……………………………………………………………………..…11

Chapter 2: WAR STRATEGY OPTIMIZATION………………………………………..….14

2.1 Introduction………………………………………………………………………….14
2.2 Steps Involved in War Strategy……………………………………………………...14
2.2.1 Randomized attack……………………………………………………………14
2.2.2 Attack plan……………………………………………………………………14
2.2.3 Drum signals……………………………………………………………...…..15
2.2.4 Defensive plan………………………………………………………………..15
2.3 Attack strategy in WSO…………………………………………………………….15
2.4 Salient features of the proposed algorithm…………………………………………18
2.5 Results and Discussion……………………………………………………………..18
Chapter 3: GREY WOLF OPTIMIZATION………………………………………………...21
3.1 Introduction………………………………………………………………………...21
3.2 Different Types of Optimization Algorithm……………………………………….21
3.2.1 Deterministic optimization algorithms……………………………………….21
3.2.2 Heuristics and Meta-Heuristics……………………………………………....22
[Link] Classification of Metaheuristic Algorithms……………………….....22
3.3 GWO algorithm……………………………………………………………………23
3.4 Proper balance between search and attack………………………………………...25

3|Page
 3.5 Range of parameters for SDM…………………………………………………….27
3.6 Application of GWO ……………………………………………………………...30
3.7 Conclusions………………………………………………………………………..31
Chapter 4: PARTIAL SWARM OPTIMIZATION………………………………………….32
4.1 Abstract…………………………………………………………………………...32
4.2 Introduction……………………………………………………………………….32
4.3 Partial Swarm Optimization Technique…………………………………………..32
4.4 Implementation of PSO Algorithm……………………………………………….33
4.5 Simulation Results………………………………………………………………...36
4.6 Conclusion………………………………………………………………………...40
Chapter 5: ANT COLONY OPTIMIZATION…………………………………………….....41
5.1 Introduction……………………………………………………………………….41
5.2 Biological Inspiration………………………………………………………….….42
5.3 Algorithmic Design……………………………………………………………….44
5.4 Pseudocode………………………………………………………………………..45
5.5 Results…………………………………………………………………………….47
5.6 Applications………………………………………………………………………48
Chapter 6: COMPARISON AND RESULTS………………………………………………..49
REFERENCES……………………………………………………………………………….51

4|Page
 ABBREVIATIONS AND SYMBOLS

Symbols Abbreviations

Ish Shunt resistance current

IL PV cell output current
Iph Photocurrent
Id Diode current
Isd Diode reverse saturation current,
q Electron charge,
k Boltzmann constant,
VL PV cell output voltage,
RS Series resistance,
n Diode ideality factor,
T Temperature in kelvin, and
Rsh Shunt resistance
TC PV cell temperature degree celcius
WSO War Strategy Optimization
GWO Grey Wolf Optimization
PSO Partial Swarm Optimization
ACO Ant Colony Optimization
d-axis Direct axis
q-axis Quadrature axis

5|Page
 LIST OF FIGURES AND GRAPHS

Chapter – 1:

Figure 1.1: Equivalent circuit of single-diode model……………………………………10

Figure 1.2: Voltage Vs Current…………………………………………………………..11

Figure 1.3: Voltage Vs Power……………………………………………………………12

Figure 1.4: Voltage Vs Current…………………………………………………………..12

Figure 1.5: Voltage Vs Power……………………………………………………………13

Chapter – 2:

Figure 2.1: Attack strategy in WSO………………………………………………………15

Figure 2.2: Flowchart of war strategy optimization algorithm…………………………..17

Figure 2.3: Iteration Vs Best score obtained so far ……………………………………...19

Figure 2.4: Voltage Vs Power……………………………………………………………19

Figure 2.5: Voltage Vs Current…………………………………………………………..20

Chapter – 3:

Figure 3.1: Flowchart of GWO…………………………………………………………..26

Figure 3.2: Experimental and estimated data comparison using GWO for single diode
model (i) characteristics curve of I-V (ii) characteristics curve of P-V……………………..29

Figure 3.3: Iteration Vs Best Score obtained so far………………………………………29

Chapter – 4:

Figure 4.1: Flowchart of PSO……………………………………………………………..35

Figure 4.2: Schematic diagram of proposed system………………………………………40

Figure 4.3: Simulation results of Voltage, Current and power tracked using PSO
algorithm……………………………………………………………………...46

6|Page
Chapter 5:

Figure 5.1: Flow Chart for Ant Colony Optimisation. Iteration Vs Best Score obtained so
far………………………………………………………………………………………….47
Figure 5.2: Voltage Vs Power………………………………………………………….47

Figure 5.3: Voltage Vs Current…………………………………………………………47

7|Page
 LIST OF TABLES

Table 2.1: Bounds for the parameters for single-diode

model……………………………….18

Table 3.5: Range of parameters for SDM……………………………………………………27

Table 3.6: Value of absolute error and estimated current obtained by GWO for single
diode system………………………………………………………………………………….28

Table 4.1: PSO parameters……………………………………………………………………

36

Table 4.2: Comparison of converter outputs with different insolation levels…………………

37

Table 6: Comparison Table…………………………………………………………………..50

8|Page
 Chapter 1: Solar PV

1.1 INTRODUCTION

The use of renewable energy sources is increasing in the recent past due to a multitude of
factors associated with climate change and the energy crisis. Solar power systems are widely
available in large-scale photovoltaic (PV) power plants for electricity generation.

In particular, PV systems are usually installed in exposed areas, and these systems are
prone to deteriorated circumstances where severe weather conditions, including rainstorms
and gale, occur. In order to address this issue, a more accurate data-driven model is needed
to identify the essential properties of PV systems in the solar industry.

The detailed analysis of solar model parameter extraction is beneficial in PV power plant
performance evaluations, efficiency calculations as well as maximum power point tracking
(MPPT), and optimal energy management of the PV system. In general, the detailed
modelling of solar PV systems consists of two stages: the first is the development of a
mathematical model and the second is the parameters identification.

The single-diode model (SDM) is used in our report. However, the unspecified
parameters primarily affect the actual performance of the PV models, probably in the form
of unstable and error-prone, if these models are subject to contingent aging of the
equipment. Therefore, prior precise estimation of the PV cell parameters is a key simulation
task. In addition, the installation and optimization of PV systems should also be more
accurate. However, the PV model is presumed as a nonlinear system with non-convex
relation.

1.2 MODELING OF SOLAR PV SYSTEM

This part of report focuses on mathematical models of the solar PV cells and modules.

9|Page
1.2.1. Single-diode model:

Figure 1.1: Equivalent circuit of single-diode model

This is the most widely used mathematical model, From the equivalent circuit visualized
in Figure 1.1, the load current is given by the 1st equation:

I L =I ph−I sd−I sh ………………………………... (1.1)

where IL is the PV cell output current,

Iph is the photocurrent,
Id is the diode current, and
Ish is the shunt resistance current.

The expressions of the currents Id and Ish are given by 1.2 and 1.3 equation:

(e( )−1 )
q V L + I L RS }
{
nkT …………………………… (1.2)
I d=I Sd

10 | P a g e
 V L + I L RS
I Sh= ……………………………
R sh
(1.3)

Where Isd is the diode reverse saturation current,

q is the electron charge,
k is the Boltzmann constant,
VL is the PV cell output voltage,
RS is the series resistance,
n is the diode ideality factor,
T is the temperature in kelvin, and
Rsh is the shunt resistance.

Replacing the current expressions Isd and Id in the 1.1 equation gives PV cell output
current as follows:

(e( )−1)− V
q {V L + I L RS }
+ I L RS
I L =I ph−I sd
nkT L
………………. (1.4)
R sh

1.2.2. Graphs

Figure 1.2 contain the variation of voltage vs current curve with different temperature at
constant irradiance PV module.

11 | P a g e
 Figure 1.2: Voltage Vs Current

Figure 1.3 contain the variation of Voltage Vs Power curve with different temperature at
constant irradiance for PV module.

Figure 1.3: Voltage Vs Power

Figure 1.4 shows the variation of Voltage Vs Current Curve with different irradiance
condition at a constant temperature.

12 | P a g e
 Figure 1.4: Voltage Vs Current

Figure 1.5 shows the variation of Voltage Vs Power curve with different irradiance
condition at a constant Temperature.

Figure 1.5: Voltage Vs Power.

13 | P a g e
 Chapter 2: War Strategy Optimization

2.1 Introduction

War strategy has been an essential component of warfare since ancient times. From the
earliest kingdoms to modern armies, every military force has relied on an effective strategy to
achieve victory on the battlefield. In ancient times, kingdoms-maintained armies for
protection against attacks from other dynasties. These armies consisted of various forces,
such as infantry, chariots, elephants, and more. During times of war, each kingdom planned a
strategy called "Vyuha" to attack the opposing army. The Emperor and the Commander-in-
Chief coordinated the army troops in a specific pattern, allowing their army to meet desired
targets.

The strategy was built based on factors such as targets, risks, challenges, and
opportunities, and was a continuous dynamic process that required coordination and
adaptation. Soldiers were trained to follow the strategy based on the sounds of drums or other
musical instruments, and the positions of the army commander and the king were represented
with flags, making them visible to all soldiers.

14 | P a g e
 The ultimate goal was for the king to conquer the opposing leader, while the army soldiers
aimed to attack the opposing team and advance in rank. Effective war strategy optimization
was essential in ancient times, just as it is today.

2.2 Steps involved in the war strategy are discussed as follows

2.2.1 | Randomized attack

The army distributes its troops evenly and randomly to launch an attack against the
opposing army. The army chief, who possesses the strongest attacking force, is the highest-
ranking army personnel. The king serves as the leader of the various army chiefs.

2.2.2 | Attack plan

The primary goal of this strategy is to attack the enemy. The king leads and directs the
army troops. Soldiers identify the weak points of the opposition (promising search space) and
continuously attack. The king and commander ride in separate chariots adorned with flags.
Soldiers move around dynamically, adjusting their positions based on the king and
commander's location. If a soldier improves their attacking force, their rank increases. If their
new position is unsuitable for combat, the soldier returns to their previous position. The army
moves in all directions and takes large steps to change their position at the war's outset.

2.2.3 | Drum signals

The king gives orders to adjust the strategy based on the situation. A group of soldiers then
beats the drums rhythmically. Based on the rhythm, soldiers adopt a new strategy and adjust
their positions accordingly.

2.2.4 | Defensive plan

The primary objective of this strategy is to protect the king. The commander or army chief
takes the lead and forms a chain around the king using the army troops. Every soldier adjusts
their position based on the nearby soldier and king position. The army tries to explore a large
area of the war field (search space) during the war. To confuse the opposition army, the army
dynamically changes its strategy from time to time.

15 | P a g e
 Figure 2.1: Attack strategy in WSO

2.3 Attack strategy in WSO

Mathematical modelling of the war strategy:

We have created two war strategies. In the first strategy, each soldier updates their position
based on the king and commander's positions. Figure 3 illustrates this attacking model. The
king positions themselves in a suitable spot to launch a powerful attack against the enemy,
and the soldier with the highest attack force assumes the role of the king. At the outset of the
war, all soldiers have the same rank and weight. A soldier's rank improves if they successfully
execute the strategy. However, as the war progresses, all soldiers' ranks and weights update
based on the success of implementing the strategy. As the war approaches its final stage, the
king, army head, and soldiers remain close together as they approach the target.

Xi (t+1) = Xi(t) + 2 x ρ x (C – K) + rand x (Wi x K – Xi(t)) ……….. (2.1)

where Xt (t+1) is a new position, Xi(t) is the previous position, C is the position of the
commander, K is the position of the King, and Wi is the weight. If the attack force (fitness) in
the new position (Fn) is less than that of the previous position Fp, the soldier takes the
previous position.
Xi (t+1) =(Xt(t+1)) x (Fn ≥ Fp) + (Xt(t)) x (Fn < Fp) ……………… (2.2)

16 | P a g e
 If the soldier updates the position successfully, the rank (Ri) of the soldier will be
upgraded.

Ri = (Ri+1) x (Fn ≥ Fp) + (Ri) x ( Fi < Fp) …………………… (2.3)

Based on the rank, the new weight is calculated as:

( )
α
Ri
W i =W i × 1− ………………………………………………. (2.4)
Max iter

Where, α is a tunable parameter. The second strategy position update is based on the
positions of King, army head, and a random soldier. Whereas the ranking and weight
updating remain the same.

Xi (t+1) = Xi (t) + 2 x p x (K-Xrand(t)) + rand x Wi x (C-Xi(t)) ………... (2.5)

17 | P a g e
Figure 2.2 Flowchart of war strategy optimization algorithm

18 | P a g e
2.4 Salient features of the proposed algorithm

[Link] weights in this algorithm are adaptive and constantly changing with each iteration,
unlike the majority of algorithms.

[Link] weight of each soldier in the algorithm is determined by their rank and is unique to
them.

3.A soldier's weight in the algorithm is adjusted based on their ability to improve their
fitness during the update step.

[Link] weights in the algorithm exhibit a non-linear pattern, where significant changes
occur in the early iterations and smaller changes occur in later iterations, ultimately leading
to faster attainment of the global optimum value.

[Link] algorithm being proposed possesses attributes of both exploration and exploitation.

[Link] position updating process in the algorithm occurs in two distinct stages, thereby
enhancing its ability to explore the global optimum solution.

[Link] proposed algorithm is uncomplicated and necessitates less computational

resources.

2.5 Results and Discussion

Single-diode model:

Table 2.1:

Bounds for the parameters for single-diode model:

Parameter Lower bound Upper bound

Iph (A) 0 1
Isd (μA) 0 1
Rsh (Ω) 0 100
Rs (Ω) 0 0.5
n 1 2

19 | P a g e
Figure 2.3: Iteration Vs Best score obtained so far

Figure 2.4: Voltage Vs Power

20 | P a g e
Figure 2.5: Voltage Vs Current

21 | P a g e
 Chapter 3: Grey Wolf Optimization

3.1 Introduction

Optimization is a process through which we can make things fast and reliable and save the
time and cost of any application. In this modern world, almost everywhere we need
optimization as money, time and resources are limited.

Optimization is essentially everywhere, from engineering design to economics and from

holiday planning to Internet routing.

Generally, optimization can be written as:

optimise f1 (x), ... fi (x), …… fN (x)

where x= (x1, x2....,xd )
subject to,

hj (x) =0, gk(x) =0,

where j= (1, 2...... J) and k= (1,2,3……K)

where f1, …, fN are the objectives, while hj and gk are the equality and inequality
constraints, respectively. When N=1, it is called single-objective optimization. And when
N≥2, it becomes a multi-objective optimization problem.

3.2 Different Types of Optimization Algorithm

1. Deterministic optimization algorithms

2. Heuristic and meta-heuristic

3.2.1 Deterministic optimization algorithms:

These algorithms take advantage of the analytical properties of the problem to generate a
solution or sequence of points which converges to a globally optimally solution.

22 | P a g e
 For ex: linear programming, nonlinear programming, and mixed-integer nonlinear
programming, etc.

3.2.2 Heuristics and Meta-Heuristics:

A metaheuristic is a higher-level procedure or heuristic which aims to find, generate, or

select a heuristic (partial search algorithm) that may provide a sufficiently good solution to an
optimization problem.

They are used especially when incomplete or imperfect information is available or when
there is limited computation capacity.

Meta heuristics make relatively few assumptions about the optimization problem being
solved and so may be usable for a variety of problems.

For example: Particle Swarm Optimization (PSO), Ant Colony Optimization, Grey Wolf
Optimization etc.

[Link] Classification of Metaheuristic Algorithms

Metaheuristic algorithm has been classified into four types:

1. Evolutionary
2. Physics-based
3. Swarm-Intelligence based
4. Humans Social Behaviour based

We will provide examples of each category for better understanding.

1. Genetic algorithms, Differential evolution, Genetic programming are examples of

evolutionary based algorithms.
2. Gravitational local search, charged system search, black hole algorithms are examples
of physics-based algorithms.
3. Particle-swarm optimization, ant-colony optimization, Grey Wolf Optimization are
examples of swarm intelligence-based algorithms.

23 | P a g e
3.3 GWO algorithm

It is a population-based meta-heuristics algorithm that works on leadership hierarchy and

hunting mechanism of grey wolves as they do naturally to catch their prey . It was given by
Seyedali Mirjalilii.

 Grey wolves are at the top of the food chain and considered as apex predators.

 Basically, grey wolves live in groups, each having 5-12 members and every member
has its own social dominance.

 Alpha α wolf is considered the dominant wolf in the pack and his/her orders should be
followed by the pack members.

 Beta β are subordinate wolves, which help the alpha in decision-making and are
considered as the best candidate to be the alpha.

 Delta δ wolves have to submit to the alpha and beta, but they dominate the omega.
There are different categories of delta-like Scouts, Sentinels, Elders, Hunters, Caretakers
etc.

 Omega ω wolves are considered as the scapegoat in the pack, are the least important
individuals in the pack and are only allowed to eat at last.

24 | P a g e
 Grey Wolf Hunting follows a complete process that includes tracking, pursuing, and
getting close to the prey. The target is pursued, surrounded, and harassed until it stops
moving. Finally, Attack the prey directly.

To create GWO, the social structure and hunting habits of grey wolves are mathematically
modeled.

The following equations were developed (Mirjalili et al. 2014) to mimic this:

X(t+1) =X(t)−A.D ……………………………… (3.1)

Here, X(t+1) denotes the wolf's new location. X(t) represents the wolf's current position, A
is a coefficient matrix, and D is a vector that is dependent on the prey's location, Xp, and is
computed as follows:

D=|CXp(t)X(t)| …………................................ (3.2)

where C=2 r2 and r2 is a random vector with a range of [0,1]. Note that because vectors
are utilised in these equations, they can be applied to any number of dimensions.

A=2a⋅r1−a ……………………………… (3.3)

25 | P a g e
 Here, 'r1' is produced at random in the range [0,1] for the vector 'a', which progressively
decreases from 2 to 0.

Alpha, beta, and delta are seen to be the three best solutions in GWO since they are aware
of the proper location and have strength within the pack. In order to update their position, the
other wolves should strive to do the following.

X(t+1) =13X1+13X2+13X3 ……………………... (3.4)

where X1, X2, and X3 are determined using Eq.

X1=Xα(t)−A1⋅Dα;
X2=Xβ(t)−A2⋅Dβ;
X3=Xδ(t)−A3⋅Dδ

The formulas for Dα, Dβ, and Dδ are as follows:

Dα=|C1⋅Xα−X| , Dβ=|C2⋅Xβ−X| , Dδ=|C3⋅Xδ−X|

3.4 Proper balance between search and attack

Here, the parameters 'a' and 'A' play a crucial role in determining how to go about finding
the overall best answer via searching and attacking. As the predator approaches the prey, the
vector 'a' value should linearly decline from 2 to 0, and the parameter 'A''s range should shift
in the range [ 2, 2] owing to random components in the parameter.

A >1 or A < -1 is used for search (also known as exploration), while -1<A<1 is used for
attack (also known as exploitation). Thus, exploration and exploitation are the two crucial
activities that determine the global optimum. The solution will stall close to the local
optimum rather than the global optimum if there is an improper balance between the two.
This implies that an extra delay and increase in convergence time will result from more
exploration.

26 | P a g e
 The solution will become trapped in local optima, which again does not result in an
effective performance, if the exploration is constrained in "limited search space" or the
exploitation is limited. Therefore, a suitable strategy for striking a balance between
exploration and exploitation should exist.

In comparison to other optimisation techniques like PSO, ACO, GA, and others, GWO
suggests faster convergence and more qualitative outcomes.

Figure 3.1: Flowchart of GWO algorithm

27 | P a g e
 3.5 Range of parameters for SDM:

Parameter Lower bound for SDM Upper bound for SDM

Ip (A) 0 1

Isd (µA) 0.01 0.5

Rs (Ω) 1E-03 0.5

Rsh (Ω) 0 100

a 1 2

Table 3.6: Value of absolute error and estimated current obtained by GWO for
single diode system.

Observation VL(V) IL(A) IL estimater(A) IAE

I 0.2057 0.764 0.7619 0.0020

II 0.1291 0.762 0.7609 0.0010

III 0.0588 0.7605 0.7601 0.0003

IV 0.0057 0.7605 0.7593 0.0011

V 0.0646 0.76 0.7586 0.0013

VI 0.1185 0.759 0.7579 0.0010

VII 0.1678 0.757 0.7573 0.0003

VIII 0.2132 0.757 0.7566 0.0003

IX 0.2545 0.7555 0.7558 0.0003

X 0.2924 0.754 0.7545 0.0005

28 | P a g e
 XI 0.3269 0.7505 0.7523 0.0018

XII 0.3585 0.7465 0.7481 0.0016

XIII 0.3873 0.7385 0.7406 0.0021

XIV 0.4137 0.728 0.7275 0.0004

XV 0.4373 0.7065 0.7066 0.0001

XVI 0.459 0.6755 0.6746 0.0008

XVII 0.4784 0.632 0.6299 0.0020

XVIII 0.469 0.573 0.5712 0.0017

XIX 0.5119 0.499 0.4993 0.0003

XX 0.5265 0.413 0.4139 0.0009

XXI 0.5398 0.3165 0.3279 0.0014

XXII 0.5521 0.212 0.2130 0.0010

XXIII 0.5633 0.1035 0.1026 0.0008

XXIV 0.5736 -0.01 0.0096 0.0003

XXV 0.5833 -0.123 0.1284 0.0054

XXVI 0.59 -0.21 0.2137 0.0037

Sum of IAE 0.0436

29 | P a g e
 Figure 3.2: Experimental and estimated data comparison using GWO for single
diode model (i) characteristics curve of I-V (ii) characteristics curve of P-V.

Figure 3.3: Iteration Vs Best Score obtained so far

30 | P a g e
3.6 Application of GWO

In resolving numerous issues from the real world, GWO exhibits great performance and
competitive outcomes. As a result of its many benefits, it has been applied to solve both
limited and unconstrained multi-objective technical, medical, and environmental challenges.
The parts that follow provide a thorough discussion.

1. In computer-assisted instruction
In order to locate the nodal sites with the greatest number of nearby members Fouad et
al. (2015) proposed a sink node localization method based on the grey wolf optimisation
strategy. There was formed an objective function. These nodes have large total residual
energies. The performance of a GWO-based technique was assessed in terms of
convergence time and energy cost under various scenarios with various network
capacities. GWO findings surpassed several other outcomes.

2. With binary puzzles

A removal method for noisy features integrating multi-focus images with a mix of
decision information of optimised individual attributes was proposed by Sujatha and
Punith Vathani (2018).

Binary GWO and mutant GWO were the two methodologies suggested by Manikandan et
al. (2016) for gene selection from microarray data.

A noncontinuous, nonlinear, and nonconvex problem is a transfer of economic power.

Jayabarathi et al. (2016) employed crossover and mutation with GWO to solve four economic
dispatch issues, and the findings outperformed those of the other metaheuristics. In terms of
convergence time and energy cost, the GWO-based strategy fared better than several other
methods.

1. In issues that are dynamic and limited and have restrictions

All Pareto optimal solutions are gathered in the archive in the non-dominated sorting
GWO presented by Jangir and Jangir (2018) until the evolution of the last iteration gathers all

31 | P a g e
Pareto optimal solutions in the archive. The algorithm's effectiveness is demonstrated by its
rapid execution and extensive coverage.

By adopting the cluster-first route-second strategy, the overall cost or distance covered by
the vehicles can be kept to a minimum. This approach was suggested by Korayem et al.
(2015). This combines the original GWO with the K-means algorithm to create a capacitated
version of K-GWO, which performs effectively.

The GWO approach was put forth by Teeparthi and Kumar (2016) to address the dynamic
security-constrained optimal power flow (DSCOPE) problem.

3.7 Conclusions:

The algorithm's inspiration came from a natural phenomenon, which was described. The
real mathematical model came next, and the effectiveness in terms of exploration and
exploitation was then looked into.
With their application taking into account parameter tuning, various methods for feature
selection and classification, and subsequently the discussion of hybridised forms, many GWO
forms were described in depth. The applications in several industries, such as engineering,
medicine, power distribution, reliability optimisation, etc., were discussed.
There is no algorithm that can optimise all real-world issues for a global optimum
solution, according to the "No Free Lunch Theorem" (Wolpert and Macready 1997), hence
the following areas can be researched in the future:
 In dynamic multi-objective optimisation problems to update the position of the non-
dominated solutions.

 To accommodate the differences in the restrictions, goal functions, and outcomes.

 To determine whether there may be a variation in the number of wolves in each

repetition.

 It is possible to simplify or eliminate all of the related parameters.

It is possible to suggest a binary and systematic discrete GWO that is more effective.

32 | P a g e
 Chapter 4: Particle Swarm Optimization

4.1 Abstract
The multi-junction solar cell system has been used with the Particle Swarm Optimization
(PSO) technique. The solar panels are constructed from a variety of materials and provide a
steady output through a boost converter. The primary goal of PSO is to determine the boost
converter's duty cycle in order to maintain constant output voltage regardless of the amount
of power generated by solar panels. The proposed method has been thoroughly simulated in
Matlab/Simulink. The simulation's outcome demonstrates that this design can be successfully
implemented in real-world uses.

4.2 INTRODUCTION

 Inspired by social interaction and dynamic movements found in nature's insects, birds,
and fish.
 Employs a swarm of agents that moves around the search space in search of the best
answer.

33 | P a g e
  Each particle in the search space modifies its "flying" in response to both its own
flying experience and that of other particles.
 Each particle adjusts its travelling speed dynamically corresponding to the flying
experiences of itself and its colleagues
 Each particle modifies its position according to:
 its current position
 its current velocity
 the distance between its current position and pbest
 the distance between its current position and gbest

4.3 PARTICLE SWARM OPTIMIZATION TECHNIQUE

Kennedy and Eberhart first proposed the PSO swarm intelligence optimization algorithm
in 1995, and it has since being developed quickly in recent years. It is portrayed in
accordance with how flocks of birds behave. Its advantages include easy implementation and
quick convergence, and it can be effectively used to find the global best solution in a
nonlinear, discontinuous, nondifferentiable curve [6]. A large number of particles are used in
this algorithm [9] in an n-dimensional space. Each particle maintains its initial velocity, ai,
and position, yi, which are both randomly distributed. A particle's position is determined by
its best position to date (pbesti), which is also, the best position of all particles to date.
Position and velocity of particles are updated using the following equation:

34 | P a g e
35 | P a g e
Figure 4.1: Flowchart of PSO algorithm

36 | P a g e
4.4 Implementation of the PSO Algorithm

1. When the parameters and particles are initialized, fix the velocity and position limit.
2. The position and velocity of each particle are initialized at random.
3. The value of Pbest has been determined for each particle. When the particle's best
value has been reached.
4. Gbest value has been set when the best value of the particle has been reached.
5. Based on the Gbest, the velocity and position values have automatically been updated.
6. Keep going back to steps 3 and 4 until you find the best answer.
7. Based on Gbest, the final iteration yielded the optimized value.

The table below displays the parameters' minimum and maximum values for each iteration of
the PSO [5] algorithm, which changes the particle's velocity and updates its position. [Link]
proposed system calls for four MPPT trackers and four PV modules.

Table 4.1: PSO PARAMETERS

[Link]. Parameters Value

1 No of particles 20

2 Min duty cycle 0.01

3 Max duty cycle 0.9

4 Sampling Time 0.1s

5 Max iteration 30

6 wmax 1

7 wmin 0.3

8 C1 min 0.1

9 C1 max 1.05

10 C2 min 0.1

11 C2 max 1.05

37 | P a g e
 In this method, only one MPPT tracker is used to cut costs and computation time. By
monitoring the global maximum point and updating the position and velocity on a constant
basis, the PSO algorithm is used to update the duty cycle of DC-DC Converters.

TABLE 4.2. COMPARISON OF CONVERTER OUTPUTS WITH DIFFERENT

INSOLATION LEVELS:

As depicted in Figure 4.2, boost converters are linked in series to share a single load. The
PSO in this system's design generates duty cycles dc (1), dc (2), dc (3), and dc (4) separately
for each Boost converter to produce a predetermined output.

38 | P a g e
4.5 SIMULATION RESULTS

A DC-DC Boost converter with a switching frequency of 5 KHz makes up the model. The
converter can operate with voltage ranges between 0 and 200V at each of its inputs, per its
specifications.

The proposed system makes use of solar panels made of various materials. The four panels
in this example are each connected to a separate Boost converter and are made of four
different materials, including GaInAsp, GaIn, GaInAs, and GaAsT.

39 | P a g e
 Figure 4.3 displays the simulation outcomes for the PV system using the PSO
optimization technique. he results of the simulation show that the PSO-based MPPT
algorithm can quickly and accurately determine the maximum power of each module, and the
system has successfully sensed the maximum power output.

It should be noted that the temperature was the same for all four materials, and the results
are shown in the table. The insolation value changed from 800 to 1000 W/m2. Regardless of
the radiation, the proposed PSO tracked the maximum point, which is displayed in Table II,
and it provides the converters with a corresponding duty cycle that is not possible with other
MPPT techniques. When compared to other techniques, the PSO method has the following
advantages: Its features include a straightforward structure, simple execution, quick
convergence to the ideal solution, and rapid tracking.

Table II compares converter outputs using the PSO technique for various insolations.

40 | P a g e
 Figure 4 3. Simulation results of Voltage, Current and power tracked using PSO
algorithm.

4.6 CONCLUSION

This project presents a Particle Swarm Optimization technique with an effective duty cycle
initialization. The proposed method was evaluated in a constant, partially-shaded
environment. The results of the simulation led to the following conclusion. It was discovered
that the PSO method had good tracking power even in varying atmospheric conditions and
partially shaded areas.

41 | P a g e
 Chapter 5: Ant colony optimization (ACO)

A Probabilistic Technique for Finding Optimal Paths

5.1 Introduction
ACO, which stands for ant colony optimization, is a type of metaheuristic that utilizes a
population-based approach to discover approximate solutions for complex optimization
problems.

Over time, new ways to solve difficult problems have been developed based on
evolutionary algorithms. These methods fall under a category called "metaheuristics," which
comes from two Greek words that mean "to find" and "one level above." Some examples of
metaheuristics include Particle Swarm Optimization (PSO) and Ant Colony Optimization
(ACO). They're based on the idea of "swarm intelligence," which means designing computer
systems that work together like social insects or animal groups, such as ants, bees, birds, or
fish. The goal is to make these systems smart enough to tackle complex challenges.

Ant colony optimization (ACO) is a method created by Marco Dorigo in the 1990s for
solving optimization problems. ACO uses artificial ants, which are software agents that look
for the best solutions to the problem at hand. To do this, ACO turns the optimization problem
into a graph with weights on each edge. The ants then move along the graph, creating
potential solutions along the way. The process of constructing solutions is based on chance
and is influenced by a "pheromone model," which assigns values to certain parts of the graph
that the ants can change as they move.

42 | P a g e
5.2 Biological Inspiration
Imagine there's a colony of ants looking for food, and there are two paths they can take to
get there. At first, there is no smell on the ground that would guide the ants to the food, so the
chances of them choosing each path are equal at 50%. Now let's say two ants start walking
towards the food, and they both randomly choose different paths since there's an equal chance
of choosing either one.

The two paths the ants can take have different lengths. If an ant takes the shorter path, it
will get to the food faster than the ant that takes the longer path.

Once an ant finds food, it takes some of it and starts heading back to the colony. As it
goes, it leaves behind a trail of pheromones on the ground to help other ants find their way to
the food source. The ant that took the shorter path to get to the food will get back to the
colony sooner than the ant that took the longer path, thanks to the pheromones marking the
shorter route.

43 | P a g e
 When a third ant goes out to look for food, it will choose the path with the shortest
distance based on the pheromones left behind by the previous ants. Since the shorter path has
more pheromones on the ground, the third ant will follow the path with more pheromones
rather than the longer one.

After the ant that took the longer path returns to the colony, more ants start following the
path with the stronger pheromone trail. When another ant sets out to find food, it sees that
both paths have a similar pheromone level, so it randomly chooses one of the paths. In this
case, let's say it chooses the path that was previously less traveled, as shown in the picture
below.

As the ants continue to follow the path with the stronger pheromone trail, the shorter path
will gradually become the preferred route. After repeating this process many times, the
shorter path will have a higher pheromone level than the longer path, making it more likely
for ants to choose this path. Eventually, all ants will follow the shorter path to the food
source.

44 | P a g e
5.3 Algorithmic Design
Regarding the behavior of the ants mentioned earlier, we can create an algorithmic design
for it. To keep things simple, we'll focus on a single food source and a single ant colony, with
only two possible paths for the ants to take. We can represent the situation using weighted
graphs, where the ant colony and food source are the nodes, the paths are the edges, and the
pheromone values are the weights assigned to the edges.

Consider a graph G = (V, E) with vertices V and edges E. Let Vs and Vd denote the source
vertex (representing an ant colony) and destination vertex (representing a food source),
respectively. The graph has two edges, E1 and E2, each assigned a length L1 and L2,
respectively. The edges are also associated with pheromone values R1 and R2, indicating
their strength.

In this context, the initial probability of an ant selecting a path between E1 and E2 can be
expressed as follows:

……………………………………
Ri
Pi= ; i=1 , 2
R 1+ R 2
(5.1)

If the pheromone value R1 of edge E1 is greater than the pheromone value R2 of edge E2,
then the probability of an ant choosing E1 is higher and vice versa.

When an ant travels back through the shortest path (Ei), the pheromone value of the
corresponding path is updated. This update is based on the length of the path as well as the
rate at which the pheromone evaporates.

The update process can be broken down into a series of steps as follows:

1. In accordance with path length –

………………………………………………
K
Ri ← Ri +

(5.2)
Li

45 | P a g e
 The update equation shown above applies to paths with indices i = 1, and
i = 2. The parameter K is a key variable in this model, and the magnitude of the
pheromone update depends on the length of the path. In general, shorter paths receive
a higher amount of pheromone.

2. In accordance with the evaporation rate of pheromones –

…………………………...
(5.3)
Ri ←(1−v )∗Ri

The parameter 'v' is a value between 0 and 1, which controls the rate of pheromone
evaporation. The update equation shown above applies to paths with indices i = 1 and
i = 2.
In each iteration of the algorithm, all ants start at the source vertex (ant colony) Vs.
Then, the ants move from Vs to Vd (food source) using the process described in step 1.
Finally, all ants return to the source vertex and reinforce their chosen path based on
the update equation described in step 2.

5.4 Pseudocode:

Procedure AntColonyOptimization:
Initialize necessary parameters and pheromone trials;
while not termination do:
Generate ant population;
Calculate fitness values associated with each ant;
Find best solution through selection methods;
Update pheromone trial;
end while
end procedure

46 | P a g e
 The step-wise procedures described above provide the means to perform fitness
calculations and update pheromone values in the pseudocode. This establishes the use of Ant
Colony Optimization (ACO) technique. ACO can be applied to various optimization
problems, including the well-known Travelling Salesman Problem (TSP).

Figure 5.1: Flow Chart for Ant Colony Optimisation.

47 | P a g e
5.5 Results

These are the graphs that we got from the Ant Colony Optimization:

Figure 5.2: Iteration Vs Best Cost

Figure 5.3: Voltage Vs Current

48 | P a g e
 Figure 5.4: Voltage Vs Power

5.6 Applications

ACO was initially applied to solve NP-hard combinatorial optimization problems, and
most of the ACO research still focuses on this area. A comprehensive overview of these
applications can be found in the publication by Dorigo and Stützle (2004). Another early
application of ACO was in routing for telecommunication networks, with AntNet (Di Caro &
Dorigo 1998) being a successful example of ACO algorithm in this domain.

Currently, researchers are focused on both the theoretical foundations of ACO algorithms
and their application to new and challenging problems. The development of theoretical
foundations started with Gutjahr, who was the first to prove convergence in probability of an
ACO algorithm (Gutjahr 2000). A review of theoretical results available for ACO can be
found in the publication by Dorigo and Blum (2005). In terms of applications, current
research is exploring the use of ACO for solving dynamic, multi-objective, stochastic,
continuous, and mixed-variable optimization problems. There is also growing interest in
developing parallel implementations capable of taking advantage of new parallel hardware.
The proceedings of the ANTS conference and the Swarm Intelligence journal contain many
papers reporting on current ACO research.

49 | P a g e
 Chapter 6: Comparison and Results

This part of report would take you through the comparison between various optimization
techniques used in this project to estimate the parameter of solar PV module.

War strategy optimization, grey wolf optimization, ant colony optimization, and particle
swarm optimization are all nature-inspired algorithms used in optimization problems. Here is
a comparison of these algorithms:

1. Inspiration: War strategy optimization is inspired by the strategies used in military

warfare, grey wolf optimization is inspired by the social behaviour of grey wolves, ant
colony optimization is inspired by the foraging behaviour of ants, and particle swarm
optimization is inspired by the flocking behaviour of birds.

2. Solution space: War strategy optimization is suitable for discrete optimization

problems, grey wolf optimization is well-suited for continuous optimization problems, ant
colony optimization is suitable for combinatorial optimization problems, and particle
swarm optimization is effective for both continuous and discrete optimization problems.

3. Search strategy: War strategy optimization divides the search space into sub-regions
and optimizes each sub-region, grey wolf optimization uses a pack-based search strategy,
ant colony optimization uses pheromone trails to guide the search, and particle swarm
optimization uses a swarm-based search strategy.

4. Parameter settings: In war strategy optimization and grey wolf optimization, the
parameters are usually fixed, while in ant colony optimization and particle swarm
optimization, some parameters are adaptive and can be adjusted during the optimization
process.

50 | P a g e
5. Performance: All four algorithms have been shown to perform well in various
optimization problems, but their performance may vary depending on the problem at
hand.

Table 6:

Algorithms Ip(A) Isd(µA) Rs(Ω) Rsh(Ω) n RMSE

WSO 0.76077 0.323 0.036377 53.718 1.4811 9.86E- 04
5
GWO 0.7597 0.499 0.0342 83.013 1.5483 9.41E- 04
1
PSO 0.7607 0.400 0.0354 59.012 1.5033 1.38E- 03
ACO 0.7599 0.454 0.0344 55.567 1.5245 9.11E- 04

51 | P a g e
 REFERENCE

[1] M. K. Alam, F. H. Khan, and A. S. Imtiaz, "An efficient power electronics solution for
lateral multi-junction solar cell systems," [Link] IECON, pp.4373-4378, 2011.

[2] High Efficiency Mechanically Stacked Multi-Junction Solar Cells for Concentrator
Photovoltaics, L. Zhao, Ph.D. dissertation, Katholieke University, Leuven Faculty of
Engineering, Arenberg Steel, Belgium, March 2011.

[3] Parallel-connected solar PV system to address partial and rapidly fluctuating shadow
conditions, IEEE Trans. Ind. Electron., vol. 56, no. 5, May 2009, pp. 1548-1556. L. Gao, R.
A. Dougal, S. Liu, and A. P. Iotova.

[4] Solar Energy, vol. 69(Suppl.), n0.1-6, pp. 217-227, 2000. M. Meinhardt, G. Cramer, B.
Burger, and P. Zacharias. "Multi-string converter with reduced specific costs and enhanced
functionality."

[5] Maximum power point tracking of multiple photovoltaic arrays: A PSO approach,
IEEE Trans. Aerosp. Electron. Syst., vol. 47, no. 1, pp. 367-380, Jan. 2011. M. Miyatake, M.
Veerachary, F. Toriumi, N. Fujii, and H. Ko.

[6] Jacob Raglend, Harsha Vardhanreddy, Sarat Kumar Sahoo, W Razia Sultana,
Prabhakar Karthikeyan, and Rajasekhar Reddy International Conference on Artificial
Intelligence and Evolutionary Algorithms in Engineering Systems, "Elimination of
Harmonics in Multilevel Inverter Using Particle Swarm Algorithm as an Optimization
Technique," pp. 265-274, 2014.

[7] Design of PSO- Fuzzy MPPT Controller for Photovoltaic Application, International
Conference on Artificial Intelligence and Evolutionary Algorithms in Engineering Systems,

52 | P a g e
2013, pp. 1339–1348. Prakash J, Sarat Kumar Sahoo, S. Prabhakar Karthikeyan, and Jacob
Raglend.

[8] Raja Das, Sarat Kumar Sahoo, Samhita R., Sugumar S., and Sumathi A International
Conference on Science, Engineering, and Technology, "A Novel Maximum Power Point
Tracking Method for PV Systems," 2014.

[9] Lian Lian Jiang, Douglas L. Maskell, Jagdish C. Patra, L.L. Jiang Et Al., "A Novel
Ant Colony Optimization-Based Maximum Power Point Tracking for Photovoltaic Systems
Under Partially Shaded Conditions", Energy And Buildings, Vol. 58, Pp. 227-236, 2013.

[10] R Sridhar, D Pradeep Vishnuram, A Himabindu, Divya, "Ant Colony Optimization

Based Maximum Power Point Tracking (MPPT) For Partially Shaded Standalone PV
System", international Conference on Innovations in Power and Advanced Computing
Technologies [i-PACT2017]

[11] Srisailam Sreedhar, Devadi Jagadeesh,” A Review On Optimization Algorithms For

MPPT In Solar PV System Under Partially Shaded Conditions”, iOSR Journal of Electrical
and Electronics Engineering (IOSR-JEEE) e-ISSN: 2278-1676, P-ISSN: 2320-3331PP 23-
32

[12] Eshram, T., Chapman, P.L.: ‘Comparison of photovoltaic array maximum power
point tracking techniques’, IEEE Trans. Energy Convers., 2007, 22, (2), pp. 439–450

[13] Subudhi, B., Pradhan, R.: ‘A comparative study on maximum power point tracking
techniques for photovoltaic power systems’, IEEE Trans. Sust. Energy, 2013, 4, (1), pp. 89–
98

53 | P a g e
 [14] Zakzouk, N.E., Elsaharty, M.A., Abdel Salam, A.K., et al.: ‘Improved performance
low-cost incremental conductance PV MPPT technique’, IET Renew. Power Gener., 2016,
10, (4), pp. 561–574

[15] I. Alaya, C. Solnon, K. Ghedira, Ant algorithm for the "multi-dimensional knapsack
problem, Proceedings of Inter-national Conference on Bioinspired Optimization Methods,
their Applications (BIOMA 2004)", 2004, p. 63-72.

[16] Sundareswaran, K., Peddapadi, S., Palani, S.: ‘MPPT of PV systems under partial
shaded conditions through a colony of flashing fireflies’, IEEE Trans. Energy Convers.,
2014, 29, (2), pp. 463–472

[17] Dorigo, M., Maniezza, V., Colorni, A.: ‘Ant system: optimization by a colony of
cooperation agents’, IEEE Trans. Syst. Man Cybern. B, Cybern., 1996, 26, (1), pp. 29–41

54 | P a g e