TRIBHUVAN UNIVERSITY

INSTITUTE OF ENGINEERING

PASCHIMANCHAL CAMPUS, POKHARA

THESIS NO:079/MSDGE/018

Optimal Location of Electric Vehicle Charging Station on Khaireni Feeder-

Lekhnath, Pokhara using PSO Algorithm

by

Shreedhar Dangi

A THESIS

SUBMITTED TO DEPARTMENT OF ELECTRICAL ENGINEERING

IN PARTIAL FULFILLMENT OF THE REQUIREMENT FOR THE
DEGREE OF MASTER OF SCIENCE IN ELECTRICAL ENGINEERING
IN DISTRIBUTION GENERATION

DEPARTMENT OF ELECTRICAL ENGINEERING

POKHARA, NEPAL

April, 2025
Optimal Location of Electric Vehicle Charging Station on Khaireni Feeder-
Lekhnath, Pokhara using PSO Algorithm

by

Shreedhar Dangi

(PAS079MSDGE018)

Thesis Supervisor

Sandeep Dhami
Assistant Professor

Paschimanchal Campus, IOE, TU

A thesis submitted to the Department of Electrical Engineering in partial fulfilment

of the requirements for Degree of Masters of Science in Electrical Engineering in

Distributed Generation

Submitted to:

Department of Electrical Engineering

Institute of Engineering, Paschimanchal Campus

Tribhuvan University

Pokhara, Nepal

April, 2025

ii
 COPYRIGHT

The author has agreed that the library, Department of Electrical Engineering,
Paschimanchal Campus, Institute of Engineering may make this thesis freely available
for inspection. Moreover, the author has agreed that permission for extensive copying
of this thesis for scholarly purpose may be granted by the professor(s) who supervised
the work recorded herein or, in their absence, by the Head of the Department wherein
the thesis was done. It is understood that the recognition will be given to the author of
this thesis and to the Department of Electrical Engineering, Paschimanchal Campus,
and Institute of Engineering in any use of the material of the thesis. Copying or
publication or the other use of this research for financial gain without approval of the
Department of Electrical Engineering, Paschimanchal Campus, Institute of Engineering
and author’s written permission is prohibited.

Request for permission to copy or to make any other use of this thesis in whole or in
part should be addressed to:

Head

Department of Electrical Engineering

Paschimanchal Campus, Institute of Engineering

Lamachour, Pokhara

Nepal

iii
 TRIBHUVAN UNIVERSITY
INSTITUTE OF ENGINEERING
PASCHIMANCHAL CAMPUS
DEPARTMENT OF ELECTRICAL ENGINEERING
The undersigned certify that they have read, and recommended to the Institute of
Engineering for acceptance, a thesis entitled " Optimal Location of Electric Vehicle
Charging Station on Khaireni Feeder-Lekhnath, Pokhara using PSO Algorithm"
submitted by Shreedhar Dangi in partial fulfillment of the requirements for the degree
of Master of Science in Distributed Generation Engineering.

________________________________________________

Supervisor,
Asst. Prof. Er. Sandeep Dhami
Co-ordinator, MSC in Distributed Generation Engineering
Department of Electrical Engineering
Paschimanchal Campus, Institute of Engineering

________________________________________________
External Examiner,

________________________________________________
Committee Chairperson,

Date: April, 2025

iv
 ACKNOWLEDGEMENT

I acknowledge my deep gratitude to my thesis supervisor Asst. Prof. Sandeep Dhami, MSc.
Coordinator of Distributed Generation Paschimanchal Campus for the insightful lessons,
guidance and inspiration without whom I may not have accomplished this study. His effort
has greatly benefited from his encouragement, suggestions, and observations.

Additionally, I would like to express my gratitude to the Paschimanchal Campus faculty

members Associate Prof. Bhrigu Raj Bhattarai, Asst Prof. Suraj Shrestha and Asst Prof.
Menaka Karki of the Department of Electrical Engineering for their insightful comments
and suggestions during the project's many phases.

I would like to extend my gratefulness to all members of Lekhnath Distribution Center and
Pokhara Grid Substation, NEA for providing support with the data and elaboration required
for the modeling of the system and their guidance in the process of the thesis works.

I extend my extreme thanks to Er. Basant Raj Tiwari and Er. Tilak Giri who provided
constant support and guidelines throughout the completion of thesis work. At last, I am very
much thankful to my family and MA.VI center school, Dang family members for their love,
support in completion of my degree.

v
 ABSTRACT

In recent years the concern towards environment quality protection has become a
burning topic and researchers are working tremendously on protection of environment.
Several concepts have been proposed justifying the acceptance of EV as a prime
solution and consequently the world has started switching towards the use of EV. As
to charge these EVs, installation of EVCS should be done technically and economically
feasible. Mostly the EV charging station placed at radial distribution network are
installed without prior detailed system analysis. This thesis work primarily focuses on
finding the optimal location based on minimizing the active power loss along with
favoring the weighted zones in the feeder, complying with different constraints like
voltage regulation, line loading and distance between two EVCS. The optimal
placement problem of EVCS is optimized by using Particle Swarm Optimization
Algorithm. The optimization is performed for IEEE 34 bus system and real radial feeder
called Khaireni Feeder. At present year i.e. 2081 B.S the optimal location for single
EVCS is on Bus 2 and between two EVCS is Bus 2 and Bus 34 at a distance of 7km.
The result of this study shows that, with the installation of one EVCS to dual EVCS,
the bus’s voltage profile decreases as well as active power loss increases from 39.96kW
to 48.59kW for single and dual EVCS installation respectively.

Load forecasting is also performed during 2086 B.S and 2091 B.S for Khaireni feeder.
Various metrics like voltage profile, are assessed and compared for single and dual
EVCS placement for 2081 B.S and 2086 B.S. From the results obtained single and Dual
EVCS placement are viable during 2081 B.S, while only single EVCS is feasible during
2086 B. S.

vi
 TABLE OF CONTENTS
COPYRIGHT................................................................................................................iii
ACKNOWLEDGEMENT ............................................................................................. v
ABSTRACT.................................................................................................................. vi
LIST OF FIGURES ...................................................................................................... ix
LIST OF TABLES ......................................................................................................... x
LIST OF ABBREVIATIONS ....................................................................................... xi
CHAPTER ONE INTRODUCTION ............................................................................. 1
1.1 Background ........................................................................................................ 1
1.2 Problem Statement ............................................................................................. 2
1.3 Objectives .......................................................................................................... 3
1.3.1 Main Objective....................................................................................... 3
1.3.2 Specific Objectives ................................................................................ 3
1.4 Scope and Limitations........................................................................................ 4
1.5 Report Organization ........................................................................................... 4
CHAPTER TWO LITERATURE REVIEW.................................................................. 5
2.1 Optimization Method ......................................................................................... 7
2.2 Classification of EVCS ...................................................................................... 8
2.2.1 Level 1 Slow Charger ............................................................................ 8
2.2.2 Level 2 Fast Charger .............................................................................. 8
2.2.3 Level 3 Rapid Charger ........................................................................... 8
2.3 Types of Distribution System ............................................................................. 8
2.4 Electrical Vehicle Charging Station in RDS ...................................................... 9
2.5 Status of EVCS Installation in Nepal............................................................... 10
2.6 Load Flow Analysis in Radial Distribution System ......................................... 11
2.7 IEEE 34-Test Bus System ................................................................................ 13
2.8 Load Forecasting .............................................................................................. 14
2.8.1 Models of Load forecasting ................................................................. 15
2.8.2 Exponential Decay Saturation Method ................................................ 16
CHAPTER THREE METHODOLOGY ..................................................................... 18
3.1 System Under Study ........................................................................................ 19
3.2 Optimization using PSO Algorithm ................................................................. 21
3.3 Problem Formulation ....................................................................................... 22

vii
 3.3.1 Objective Function ............................................................................... 22
3.3.2 Constraints ........................................................................................... 24
3.3.3 Voltage Deviation Index (VDI) ............................................................ 25
3.3.4 Load Forecasting Using Exponential Decay Saturation Model ........... 25
CHAPTER FOUR RESULT AND DISCUSSION ...................................................... 27
4.1 Analysis in IEEE 34 Test bus system............................................................... 27
4.1.1 With Single and Two EVCS................................................................. 27
4.2 Analysis in Khaireni Feeder............................................................................. 29
4.3 Load Forecasting of Khaireni Feeder .............................................................. 31
4.3.1 Voltage profile and Active Power Loss for future years ...................... 32
CHAPTER FIVE CONCLUSIONS AND RECMMENDATIONS............................. 36
5.1 Conclusions ...................................................................................................... 36
5.2 Recommendations ............................................................................................ 36
APPENDIX-A.............................................................................................................. 41
APPENDIX-B .............................................................................................................. 52

viii
 LIST OF FIGURES

Figure 2.1:Backward/Forward Sweep Flow Algorithm ............................................... 12

Figure 2.2:IEEE 34 Test Bus System ........................................................................... 14
Figure 2.3:Three stages of electricity demand growth................................................. 16
Figure 3.1: Overall Flowchart of proposed work......................................................... 19
Figure 3.2:SLD of Khaireni Feeder ............................................................................. 20
Figure 3.3: Flowchart of Particle Swarm Optimization ............................................... 22
Figure 4.1:Voltage profile of IEEE 34 Test Bus system in different cases .................. 27
Figure 4.2: APL of IEEE 34 Test Bus system in different cases .................................. 28
Figure 4.3: Voltage profile comparison of Khaireni feeder in different cases ............. 29
Figure 4.4: APL comparison of Khaireni feeder in various scenarios ......................... 29
Figure 4.5:Active power loss at each Branch during 2081 .......................................... 30
Figure 4.6: Convergence curve for optimization of two EVCS placement ................. 30
Figure 4.7: Load Growth pattern of Khaireni Feeder for past years ............................ 31
Figure 4.8: Load growth for each type of load during forecasted years ...................... 32
Figure 4.9: Voltage profile with single EVCS of Khaireni Feeder at forecasted years 32
Figure 4.10: APL with single EVCS of Khaireni Feeder at forecasted years .............. 33
Figure 4.11:Active power loss at each Branch during 2086 ........................................ 33
Figure 4.12: Voltage profile of Khaireni Feeder with dual EVCS at forecasted years 34
Figure 4.13: APL with dual EVCS of Khaireni Feeder at forecasted years ................. 34
Figure 4.14: Cumulative VDI for various scenarios .................................................... 35

ix
 LIST OF TABLES

Table 3.1:PSO Parameters for Optimization ................................................................ 21

Table 3.2:Zones and their weight values ...................................................................... 24
Table 4.1: Growth Rates for Different Load Types ...................................................... 31

x
 LIST OF ABBREVIATIONS

AC Alternating Current

APL Active Power Loss

DC Direct Current

FCS Fast Charging Station

EV Electric Vehicle

EVCS Electric Vehicle Charging Station

EVSE Electric Vehicle Supply Equipment

GA Genetic Algorithm

GHG Green House Gas

GIS Geographic Information system

IEEE Institute of Electrical and Electronics Engineers

LV Low Voltage

PSO Particle Swarm Optimization

RDS Rural Distribution System

SVC Static VAR Compensator

XLPE Cross-linked Polyethylene

xi
 CHAPTER ONE : INTRODUCTION
1.1 Background

In recent years the because of air pollution and climate change due to use of gas-driven
vehicle the world has to switch towards using Electrical Vehicles. The IEA has
presented in its report that globally use of EV stock will exceed 300 million by 2030
[1]. The availability of EVCS is a mandatory for the success of EV technology. Many
researchers work was done and the researchers come to a conclusion that the placement
of EVCS in random fashion in distribution network will result in very high demand of
electricity which will negatively impact grid performance and degrades the voltage
regulation, causes a lot of modification in load demand pattern resulting in transformer
overloading and increase power loss, line loading in the system [2]. Also, feeder
capacities to transfer load and reverse capacity of distribution grid substation will be
decreased due to increase in system demand because of plugging EV charger.
Therefore, to increase the penetration and obtain more technical and financial benefit
determining the best location for EVCS in a distribution network is of prime
importance. Here the technical parameters include voltage, frequency, harmonics,
power quality [3]. Proper planning should be done for placement of EVCS in right
location so that negative impact on electrical parameters can be reduced. The optimal
location of EVCS is greatly affected by parameters like road network, land availability,
number of EV users. Because of this, these parameters should also be realized during
study.

Research has been done on possible effects of the loads from charging stations may
have on the distribution network. When the various EV penetration scenarios were
examined on the LV distribution network, it was discovered that the placement of
numerous charging stations negatively impacted the node voltage profile and that the
high EV charging loads negatively impacted the voltage profile of the weaker buses [4].
Determining the best location for EV charging stations inside a distribution network
thus becomes crucial. The Khaireni Feeder of the Lekhnath Power Grid is examined in
this thesis, and the best place for EVCS in a feeder is assessed using an effective Particle
Swarm Optimization Algorithm (PSO) technique.

1
The installation of EV charging stations at different radial distribution system buses
will result in a very high load demand that is dependent on location, time, charging
interval, and the unpredictability of active and reactive power, which will negatively
impact grid operation. When EV charging stations are strategically placed, the electric
distribution system's voltage stability, dependability, and other operational factors will
improve. In order to increase the penetration of electric vehicles in the current radial
distribution system, charging infrastructure installation must be done optimally.

Over time, the electrical demand for charging has evolved. The batteries took a while
to charge at first. These days, faster charging techniques are the outcome of
advancements in solid state technologies. Because of their high current requirements,
these chargers are frequently referred to as fast chargers. A developing field of study
these days is how to offer high-quality electricity for the charging station without
sacrificing the current supply infrastructure for other facilities. The purpose of this
study is to offer a solution for positioning these rapid chargers in a way that allows the
utility to supply a high-quality power source.

1.2 Problem Statement

Nepal's power distribution infrastructure predominantly relies on radial distribution

systems due to their cost-effectiveness, straightforward planning requirements, and
operational simplicity. This design features a single power source with feeders
extending outward like branches of a tree. While economically advantageous for initial
implementation, these systems face significant technical limitations, particularly
regarding voltage profile management. As electricity travels along lengthy feeders,
voltage progressively decreases due to impedance in the lines, resulting in substandard
voltage levels for consumers located farthest from the supply point. As more and more
Electric Vehicle Charging Stations (EVCS) are installed nationwide, this inherent flaw
has become a bigger issue. The rapid, uncoordinated installation of these charging
facilities is exacerbating existing network vulnerabilities, intensifying voltage
degradation, elevating power losses, increasing line loading beyond design capacities,
and introducing harmful harmonic distortions into the system. Without strategic
intervention, these compounding issues threaten both power quality and system
reliability for all consumers connected to these networks.

2
The strategic placement of EVCS represents a critical opportunity to mitigate these
adverse effects while supporting Nepal's transition toward electrified transportation. By
conducting comprehensive allocation analysis, distribution system operators can
identify optimal locations for charging infrastructure that minimize negative grid
impacts while maximizing service availability. This approach requires sophisticated
modeling that accounts for multiple electrical parameters including voltage profiles,
power flow dynamics, thermal limitations of conductors, and power quality metrics.
The goal is to determine installation points that balance the competing needs of EV
users and grid stability without violating operational constraints of the distribution
system. Properly executed, strategic EVCS placement can transform these charging
facilities from potential liabilities into grid assets that support voltage profiles through
appropriate reactive power management, reduce overall system losses, and improve
power quality. This research direction offers Nepal a pathway to accommodate growing
electricity demand from the transportation sector while enhancing the performance of
its existing distribution infrastructure, ultimately supporting broader goals of energy
transition with minimal capital investment in complete system redesign.

1.3 Objectives

1.3.1 Main Objective

The main objective is to find optimal location of EVCS in Khaireni Feeder,Lekhnath

Pokhara using Particle Swarm Optimization (PSO) Algorithm

1.3.2 Specific Objectives

• Perform load flow analysis and find optimal location for single and dual EVCS
in a 34-test bus system.
• Model Khaireni Feeder, perform load flow analysis and find optimal location
for single and dual EVCS in Khaireni feeder.
• Perform load forecasting for 5 years and 10 years in real feeder.
• Find optimal location of EVCS in real feeder after load forecasting using
Particle Swarm Optimization Algorithm and observe voltage profile, power loss
and distance between two EVCS after EVCS penetration in a real feeder.

3
1.4 Scope and Limitations

The study looks at how the distribution network is affected by the placement of EVCSs
both now and after five and ten years with modelling and application of a PSO
algorithm that optimizes the placement of EVCS with the proposed constraints within
the limit.

The limitations of the study include:

• The size of the EVCS is taken as a standard rated size available rather than its
sizing optimization is done.
• The optimal location is targeted to preferred zones.
• Load forecasting is only based on trend of demand growth rather parameters
like temperature, season, population and other parameters could have been
considered.

1.5 Report Organization

The first chapter deals with a brief introduction of the thesis background, problem
statement, objectives, scope and limitation and report organization.

In the second chapter, the detail of review of different literatures are presented. The
various literature related to the thesis works are presented

The third chapter provides description of the methodology used in thesis work in brief.
This includes system under study, Grid parameters calculation method, problem
formulation.

In the fourth chapter, the expected results and discussion are presented.

The fifth chapter presents conclusion of the results and future recommendations for this
study.

4
 CHAPTER TWO : LITERATURE REVIEW

An extensive review of the literature on the impact of EV charging stations on power

distribution feeders served as the foundation for the investigation for this thesis project.
This involved a comprehensive review of research on load profiles, voltage stability,
power quality, as well as an analysis of the relationships between these elements and
the incorporation of EV charging station infrastructure. The basis for gaining a thorough
comprehension of the topic area was laid by this review.

Placing charging stations at weak spots in the distribution network can negatively
impact the voltage stability and reduce the reliability of the system. Additionally, these
stations should be situated close to areas with high charging demand to ensure
practicality. Chicken Swarm Optimization (CSO) is a recent, bio-inspired optimization
method that simulates the social behavior of chickens within a group. Ant Colony
Optimization (ACO) is another swarm-based algorithm that imitates how ants follow
pheromone trails to find optimal paths. Teaching-Learning-Based Optimization
(TLBO) draws inspiration from the educational process, where the most optimal
solution acts as the “teacher” to guide the rest of the population. Lightning Search
Algorithm (LSA), on the other hand, is grounded in physics and replicates the behavior
of lightning during atmospheric discharge events.[5].

An optimal configuration can significantly enhance the profitability of a charging

station. To tackle the optimization challenge, three metaheuristic algorithms are
employed: Particle Swarm Optimization (PSO), Salp Swarm Algorithm (SSA), and
Arithmetic Optimization Algorithm (AOA). Therefore, it is essential to deploy EV fast-
charging stations (EVFCS) in accessible public areas—like parking lots—where
electric vehicles can be charged in under 20 minutes [6]. The Genetic Algorithm (GA)
is a nature-inspired, population-based optimization method that searches for the global
optimum through processes of selection, crossover, and mutation. Compared to other
techniques, GA is recognized for its strong performance and robustness in finding
optimal solutions [7].

Transformer overloading, significant variations in load demand, and electrical network

disruptions involving voltage, frequency, power, and power quality can all result from
the EVCS's disorganized infrastructure development. One of the main objectives of
making electric vehicles more user-friendly is to reduce the charging time. Fast DC

5
charging presents an intriguing prospect in this regard. It makes it possible to shorten
charging times to between ten and twenty minutes. DC Level 1 200/450 V, up to 36 kW
(80 A); DC Level 2 200/450 V, up to 90 kW (200 A); and DC Level 3 200/600 V DC
(proposed) up to 240 kW (400 A) are the three levels of fast DC charging as defined by
the SAE J1772 standard. Off-board electric vehicle supply equipment (EVSE) is used
at all levels.

For the simultaneous placement of EV charging stations and shunt capacitors, S. Muthu
Kannan proposes a mathematical model with three objective functions: maximization
of coverage, minimizing of loss, and node voltage deviations subject to limitations. The
position and rating of shunt capacitors and charging stations serve as the control
variables for optimization. [8].

A multi-objective optimization technique is discussed by Venkata K. Babu Poonam and

K. Swar Nasri in order to achieve simultaneous EVCS & DG placement and size. The
objective of the challenge is to maximize the electrical distribution system's real power
losses, Average Voltage Deviation Index (AVDI), and Voltage Stability Index (VSI).
The conventional IEEE 33-bus and 69-bus test systems were used for simulation
research. To minimize the system objectives, the Teaching-Learning Based
Optimization (TLBO) and Harries Hawk Optimization (HHO) algorithms were chosen

[9].

Srinivas introduces a heuristic method known as Particle Swarm Optimization (PSO)

to enhance the performance of the IEEE 33-bus radial distribution system integrated
with electric vehicle charging stations. The main goal is to determine the most suitable
locations for installing EV charging stations within the existing radial distribution
network, with a focus on minimizing active power losses and maintaining appropriate
voltage levels across the system's buses [10].

Samarendra Pratap Singh considers the two parameters, first is transportation node and
available capacity of substation for determining the geographical position and second
is size of the EVCS. This paper analyses and validated the optimal allocation of place
for EVCS in the specific region of Ayodhya City with help of traffic and power
constraints. Datasets like substations and traffic data of the Ayodhya city were collected,
analyzed and used for validation of proposed methodology. K-means, Particle Swarm

6
Optimization (PSO) and Genetic Algorithm (GA) were used to optimize the proposed
EVCS sites. Results of both optimization technics were compared [11].

It was believed that EVs should be able to go to the closest charging station using only
5% of their remaining charge in order to make it easier for them to access EVCSs in
any area of the city. Since EVs can drive roughly 10 km on a 5% battery charge, the
distance between any location and the closest charging station shouldn't be more than
10 km. [12].

Ming Dong presents a hybrid modeling approach that leverages sequence prediction
techniques for load forecasting on distribution feeders. This method effectively
combines top-down, bottom-up, and sequential patterns embedded in multi-year
datasets. The study explores two advanced sequence prediction architectures—Long
Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) networks—which
address the common issues of vanishing and exploding gradients found in standard
recurrent neural networks [13].

In a separate study, Tai-Hua Yangrui Sun, Qiming Wei, and Yuqi Gao detail the
Saturated Demand Forecast model by segmenting the progression of electricity demand
into three distinct phases. They introduce an enhanced self-adaptive logistic model and
validate its accuracy and reliability by applying it to power demand data from East
China, laying the groundwork for further research on demand saturation forecasting
[14].

2.1 Optimization Method

As an optimization tool, the PSO offers a population-based search process where in

particles, or people, alter their position (state) over time. Particles move through a
multidimensional search space in a PSO system. Each particle in flight modifies its
position based on its own experience (Pbest) and that of an adjacent particle (Gbest), using
the best position that both the particle and its neighbor have experienced. Each particle
moves in an N-dimensional search space with the position and velocity of a particle
could be updated to find optimal location [5].

The velocity and position of each particle in that space is achieved using the following
Equations 2.1-2.2. [15]

𝑣𝑖𝑘+1 = 𝑤 𝑘 . 𝑣𝑖𝑘 + 𝑐1 × 𝑟1 × (𝑃𝑏𝑒𝑠𝑡𝑘 − 𝑥𝑖𝑘) + 𝑐2 × 𝑟2 × (𝐺𝑏𝑒𝑠𝑡𝑘 − 𝑥𝑖𝑘) (2.1)

7
 𝑥𝑖𝑘+1 = 𝑥𝑖𝑘+𝑣𝑖𝑘+1 (2.2)

where 𝑣𝑖𝑘 is the component in dimension d of the ith particle velocity in iteration k, 𝑥𝑖𝑘i
is the component in dimension d of the ith particle position in iteration k, c1 and c2 are
constant weight factors, Pbest is the best position achieved so far by particle i, Gbest is the
best position found by the neighbors of particle i, r1 and r2 are random factors that lies
between 0 and 1 interval, and w is inertia weight.

2.2 Classification of EVCS

Electric Vehicle Charging Stations is categorized on the basis of their charging speed
and capabilities. Level 1, Level 2 and Level 3 are three levels of EV charging stations
among which Level 1 and 2 are AC type chargers while Level 3 is DC charger.

2.2.1 Level 1 Slow Charger

These types of chargers are mostly found in household where the vehicle owners can
charge the EV’s easily and comfortably. It uses a standard household outlet (120 volts
in the US) with power output of less than 6 kW thus, it is the least efficient and takes
the longest time to charge vehicle completely.

2.2.2 Level 2 Fast Charger

It is mostly found on work places and public locations where it can charge the vehicle
a bit faster compered to Level 1 charging. It uses either 230 Volt 1-phase supply or 400
volt 3-phase supply to power the battery with power output less than 22 kW.

2.2.3 Level 3 Rapid Charger

It is the fastest type of charging based on high-voltage DC current that can fed more
than a 100 kW of power directly to the EV battery. They are generally designed to
charge the vehicle up to 80% within short period of time thus, reduces the charging
time drastically as compared to Level 1 and Level 2 charging.

2.3 Types of Distribution System

Distribution systems are usually divided into feeders, distributors, and service mains.
Based on distribution voltage levels, it can be divided into primary and secondary
distribution systems. A primary distribution system has distribution voltage levels of
11 kV, 6.6 kV, or 3.3 kV, while a secondary distribution system has 400 V or 230 V. It

8
can be further divided into three categories based on the connecting scheme:
interconnected, ring main, and radial systems.

A radial distribution system is a type of electrical power distribution network where

power flows in a single direction from the substation to the end consumers. It is the
most commonly used distribution system, especially in rural and suburban areas, due
to its simplicity and cost-effectiveness. However, poor dependability is a major
disadvantage because, in the absence of a backup power source, a single main feeder
failure might result in a total power loss for all downstream users.

As the ring main system's load is supplied by two parallel paths from the substation,
electricity can be readily delivered to the load even in the event of a fault in one of the
paths. If there is an outage in any path, the busbar must support twice as much load. It
can sustain the voltage level at the receiving end and, as a result, improve voltage
regulation due to its increased reliability and power availability due to the parallel
routes. In contrast to RDS, it is more costly.

A power distribution network with several substations and feeders connected to one
another, offering numerous routes for the flow of electricity, is known as an
interconnected distribution system. An interconnected system improves dependability,
flexibility, and efficiency by guaranteeing that energy can be diverted in the event of
failures or maintenance. The main drawback is its complicity in design and operation
as well as costlier.

In the scenario of Nepal, most of the distribution system are in radial mode of
connection. With the growing demand, load connected to the RDS are facing voltage
fluctuation and low reliability. Nepal Electricity Authority (NEA), only concerned
authority for distribution system are implementing various projects like Capacity
upgrading & expansion, installation of auto re-closer & the smart load break switch,
undergrounding of the distribution system.

2.4 Electrical Vehicle Charging Station in RDS

Slow and quick charging are two ways that electric vehicle customers can meet their
demands. As it takes longer to recharge the battery, slow charging is often
recommended for residential use. As a result, it enables the distribution system operator
(DSO) to organize and control how the charging system operates. Fast charging station
(FCS) installation is on the rise, nevertheless, and this is impressive because it enables

9
a superior charging experience that is comparable to combustion car refilling times.
Compared to slow charging, FCS poses additional risks to the grid, such as voltage
fluctuations, imbalances, distortions of the harmonic current and voltage, etc.

As per Nepal Electricity Grid Code 2080 introduced by Electricity Regulatory

Commission, distribution voltage level should be maintained within ±5% of the rated
voltage level in normal operation while the frequency should be within ±2.5% of the
rated frequency. Thus, to maintain voltage, frequency and harmonics level within the
specified limit some of the mitigating measures need to incorporate with the charging
stations.

2.5 Status of EVCS Installation in Nepal

There is a great need for efficient vehicle charging infrastructure as a result of the
growing global usage of electric vehicles (EVs). Not only government organizations
but private builders and infrastructure developers are also working to satisfy the energy
demands of the expanding need. In addition, governments worldwide are collaborating
with oil and gas companies to strategically design and build charging station
infrastructure. A number of nations are putting in EVCS close together to prevent EV
users from being deprived of charging stations in an effort to promote the usage of EVs
and reach zero carbon emissions. Although, most of the EV users get their vehicle
charged through home charging station, publicly accessible charging stations becoming
more and more essential to offer the same degree of accessibility and convenience as
refueling traditional automobiles.

The charging station of power rating less than 22 kW that can be considered as slow
charger were installed more than 6,00,000 in public places throughout the world in
2022, among which China shares the greatest portion. In addition to that, the number
of public fast chargers whose power rating greater than 22 kW increased by 3,30,000
globally in 2022. Public charging station are used to provide charging solutions to those
consumers who do not have consistent access to private charging facilities. [16]

In 2023, 51 advanced fast charging stations for electric vehicles were formally opened
by Nepal Electricity Authority (NEA) in various parts of the nation. Among them, 26
are made to charge large automobiles while remaining 25 are made to charge both small
and large automobiles. Apart from that, Sajha Yatayat has installed fast charging station
of capacity around 1.4 MW to cater the charging solution to approximately 40 EVs at

10
a time, especially large buses and micros. More than that, various private organizations
like Hyundai, Kia, BYD, MG Motor, Tata Motors, Yatri and so on are focused on setting
the charging solutions for both four wheelers and two wheelers.

Approximately 200 charging stations have been deployed countrywide by private

enterprises as of December 2023. The NEA has set an ambitious goal to establish 500
more charging stations nationwide within the current fiscal year, acknowledging the
significance of strong infrastructure in promoting EV adoption. Plans are currently in
place to release tenders for this expansion

2.6 Load Flow Analysis in Radial Distribution System

Load Flow Analysis is conducted to get power system steady state condition. It is
typically carried out for system enhancement, appropriate planning, and long-term
system operation. The Newton-Raphson Method, Backward/Forward Sweep Method,
Fast Decoupled Method, Gauss-Sidel Method, Continuation Method, Artificial
Intelligence Method, and other algorithms can be used with software like NEPLAN,
CYMDIST, MATLAB, DigSilent, ETAP, and so forth.

Backward/Forward Sweep method is an iterative method in which two computational

steps are carried out at each iteration. It consists of two steps: the backward sweep and
the forward sweep. By applying Kirchhoff's Voltage Law and Kirchhoff's Current Law,
Voltage and Currents are calculated from the End node to the Source node in a
backward sweep. While in forward sweep, the source node serves as the beginning
point for calculating the downstream. Node-branch aligned data serves as the
algorithm's input value. Basic data needed for the load flow are active and reactive
powers with the classification of sending and receiving nodes [17]. Because of the fast
convergence capability and appropriate for RDS, this method has been adopted in this
research work.

This algorithm consists of backward sweep and forward sweep. The load flow of the
radial feeder starts with input of resistance and reactance of branches as well as active
and reactive power demand of each bus. Then end nodes of the feeder are determined
using breadth first search method. Now, initialize the voltage of each bus to 1 p.u. and
iteration count K=1. Then, we calculate the load current with the initialized bus voltage
and branch current using backward sweep.

11
 Figure 2.1:Backward/Forward Sweep Flow Algorithm

∗
𝑆
𝐼𝑖𝑘 = ( 𝑘𝑖 ) - 𝑦𝑖 × 𝑉𝑖𝑘−1 (2.3)
𝑉𝑖

𝐽𝑙𝑘 = -𝐼𝑙𝑟 + ∑ 𝐽𝑙𝑟 (2.4)

where, i = 1,2,3……………. n (number of nodes)

Si = power output at node i

Vi = voltage at node i

yi = shunt admittance at node i

Ilr = current injection of node lr calculated from step 1

∑ Jlr = current in branches originated from node lr

lr = 1,2, 3………………..., b (number of branches)

The, voltage at each bus is calculated with the branch current using forward sweep
method.

𝑉𝑙𝑟𝑘 = 𝑉𝑙𝑠𝑘 - 𝑍𝑙 × 𝐽𝑙𝑘 (2.5)

where, ls and lr represent sending and receiving end of branch l

Zl = series impedance of branch l

12
This continuous iterative approach will verify convergence requirements, such as the
voltage differential between two subsequent iterations at each bus. Each bus's voltage
and branch current are saved if the convergence conditions fall within the designated
tolerance limit. Transmission power loss at each branch and total losses are computed
using these preserved values. The total losses are the summation of the branch losses
while the branch loss is calculated using following formula:

𝑃𝑖2 +𝑄𝑖2
𝑃𝑙 = ∑ × 𝑅𝑖 (2.6)
𝑉𝑖2

𝑃𝑖2 +𝑄𝑖2
𝑄𝑙 = ∑ × 𝑋𝑖 (2.7)
𝑉𝑖2

where, Pi and Q i are the total active and reactive power injected through ith node

R i = resistance of ith branch

Xi = reactance of ith branch

The total active and reactive power loss of the system is given by

TPL = ∑𝑛−1
𝑙=1 𝑃𝑙 (2.8)

2.7 IEEE 34-Test Bus System

The current IEEE 34 bus system was used as a test case to determine the best position
for EVCS in a distribution network. The IEEE 34 bus data was used for the radial
distribution feeder because the IEEE Distribution Analysis Subcommittee contains data
for many test instances. A primary utility substation is connected to a number of fixed
and distributed loads in the original system, which is 60Hz, 24.9kV, and 100 MVA.
Constant current, constant impedance, and constant power models (three phase and
single phase) are all included in the load type. The geometric data is used to determine
the line impedances, which are then provided as configurations with information on the
impedance and capacitance matrices in ohms/km and ohms/km. The complete setup
and the model specifics are depicted in Figure 2.2.

From the IEEE 34 Distribution feeder committee the information about line
impedances, load data and branch length are obtained and tabulated in Appendix-A

13
 Figure 2.2:IEEE 34 Test Bus System

2.8 Load Forecasting

Forecasting is a statement of what will happen if certain conditions or trends continues.

In case of electric load forecasting, trend in general have two meanings; trend of load
characteristics and trend of load growth. There are various factors on which load
forecasting depends like Historical Data, Geographical Factors, Land Use, City &
Industrial plan, Alternative Energy sources etc. During forecasting various factors like
electricity consumption, time factors, weather factors, possible customer’s classes need
to be considered. On the basis of Time Horizon load forecasting is classified into
following categories:

• Ultra/very short-term load forecasting

This type of forecasting ranges from a few minutes to an hour ahead and is used for
real-time control, volt-var control, frequency control etc.

• Short term load forecasting

Short term load forecasting is done over an interval ranging from an hour to week.
This type of forecasting is important for different functions as unit commitment,
economic dispatch, energy transfer scheduling, and real- time control.

• Medium term load forecasting

It is used for load forecasting ranging from 1 month to 5 years and sometimes 10 or
more years. Medium term load forecasting is used by the utilities to purchase
enough fuel and for the calculation of various electricity tariffs.
14
 • Long term load forecasting

This type of load forecasting is used by planning engineers and economists to plan
for the future expansion of the system covering 5 to 20 years. Long term forecasting
of distribution feeder is very important as it is used for input to evaluate power
delivery capacity at normal operation and restoration ability during system
contingencies for few years later. [19]

Generally, load forecasting is done for capacity planning, network planning,

Generation and transmission capital investment, financial forecasting, optimum
supply schedule etc.

2.8.1 Models of Load forecasting

Various models have been proposed for forecasting of load. Some of them are
explained as below:

1. Time -Series Forecasting Methods

a) Statistical Methods
• Box-Jenkins basic models (AR, MA, ARMA, ARIMA, ARMAX,
and ARIMAX)
• Kalman Filtering Algorithms in the State space
• Grey models
• Exponential Smoothing
b) Machine Learning Models
• Artificial Neural Network (ANN)
• Support Vector Machines (WNN)
• Fuzzy Logic
• Wavelet Neural Network (WNN)
c) Hybrid Models
• Combination of both statistical and machine learning model
2. Spatial Load Forecasting
• Small Area Load Forecasting
• Non-Analytic
• Trending

15
2.8.2 Exponential Decay Saturation Method

Saturated power demand is important indication showing the trend of power

development. Various factors like industrial structure adjustment, energy consumption
control, technological upgrading, resource environmental constraints will enable
sustained growth in economy when economy grows to certain stage [20]- [21].
Additionally, as economic expansion slows, the growth rate of power consumption will
reduce, stop, or even reverse, indicating a saturated situation. [22]. Many research has
done research on various forecasting models and methods like autoregressive integrated
moving average (ARIMA), regression analysis method [23], neural network methods
[24]. Theses method and models are better for power demand forecasting where there
is high speed growth rate but as demand growth rate decreases or slows down there will
be a lot of forecasting error. Saturated power demand model is used where power
demand is stable.

Figure 2.3:Three stages of electricity demand growth

The fundamental principle of the exponential decay saturation model is that growth
rates tend to decrease over time as systems mature. This is based on observations across
many utility systems where initial rapid growth eventually moderates due to market
saturation, efficiency improvements, and physical limitations.

Mathematical Formulation: The basic mathematical formulation of the exponential

decay saturation model is defined in Equation 2.9- Equation 2.10

gt =g0×αt (2.9)

16
 Where:

• gt is the growth rate at time period t

• g0 is the initial growth rate

• α is the decay factor (typically between 0.90 and 0.98)

• t is the time period (usually in years)

For load forecasting, we apply this decaying growth rate to calculate future loads [25].

𝐿𝑡 = 𝐿0 × ∏𝑡𝑖=1 (1 + 𝑔0 × 𝛼 𝑖 ) (2.10)

Where,

• 𝐿𝑡 =forecasted load at time t

• 𝐿0= initial/base load at time t=0
• ∏𝑡𝑖=1 =product from i=1 to t (multiply each time step’s adjustment
factor cumulatively)

17
 CHAPTER THREE : METHODOLOGY

Khaireni Feeder of Lekhnath DCS Pokhara is used for analysis of optimal location of
EVCS placement. In analysis purpose, ETAP software is used for feeder modelling and
MATLAB is used for optimization code. The PSO algorithm is used to determine the
best place for EVCS in the suggested feeder. The Exponential Decay Saturation model
is used in MATLAB to forecast the load. In order to optimize the single and two EVCS
in a feeder, additional analysis is carried out by allocating appropriate line, load, and
length data along with the necessary weight value. The Particle Swarm Optimization
(PSO) Algorithm is used to solve the optimal EV charging station placement problem

Initially, the research has been conducted on the Radial Distribution System of Standard
IEEE 34 test bus system on which analysis of technical aspects like active power loss,
voltage profile is performed. The Standard IEEE 34 test bus system's findings are
validated using previously published research papers, and then same methodology is
applied for Khaireni Radial Feeder, Lekhnath Pokhara. The voltage profile and active
power loss of the feeder before and after placement of single and two EVCS has been
compared and then optimization process has been implemented for the optimal location
of EVCS in the system so as to reduce active power loss. The objective function for the
optimization is the summation of power loss reduction and zone component under
multiple equality and non-equality constraints that helps to determine the candidate bus
for optimal location of EVCS in the system. The constraints used for optimization in
this thesis works are voltage regulation, line loading and distance between two
Electrical Vehicle Charging Station should be between 7 to 11 Km. Then after, load
forecasting of real feeder for 5 years ,10 years is done and voltage profile, power loss
and optimal location of EVCS is calculated for these scenarios. Finally, voltage profile
and active power loss of the feeder before and after placement of single and two
Electrical vehicle Charging Station is done after EVCS penetration in different
scenarios and is compared with present load flow of real feeder. The required line data,
load data and branch data of IEEE 34 test bus system and proposed real Khaireni Feeder
is mentioned in Appendix A. The detail explanation is done in flowchart shown in
Figure 3.1.

18
 Figure 3.1: Overall Flowchart of proposed work

3.1 System Under Study

The distribution system selected for EV impact research is the feeder of the Pokhara
Grid Substation located in Lekhnath, Pokhara, Nepal as shown in Figure 3.2. The
Khaireni distribution feeder primarily uses mostly XLPE Covered Cable of 100mm 2
and Rabbit conductor placed in horizontal and triangular fashion. The distribution
transformer’s position and line length are extracted from the GIS route map using Arch
map software and site visit. The feeder is around 18km long overall and has a radial
length. There are around 6500 consumers and some industrial loads in this feeder. There
are 45 Distribution Transformers in Khaireni Feeder. In the system, the transformers
are regarded as the buses or load points. 11 kV lines are used as distribution lines, with
transformers acting as load points or buses and grid substations supplying the line as
sources. The three different zones are created based on distribution of load with their
weight value based on Population densities, EV users, and land availability.
19
Figure 3.2:SLD of Khaireni Feeder

20
3.2 Optimization using PSO Algorithm

Initially, a group of particles is randomly initialized in the search space. Each particle
makes use of its memory and flies through the search space for obtaining a better
position than its current one. In its memory, a particle memorizes the best experience
found by itself (Pbest) as well as the group's best experience (Gbest). The velocity and
position of each particle in that space is achieved using the equations 2.1 & 2.2. In this
thesis work parameters in Table 3.1 are used for optimization.

Table 3.1:PSO Parameters for Optimization

Parameters Description Value

Matrix Maximum number of iterations 100

n-pop Size of swarm 30

w Inertia weight 0.73

C1 Cognitive acceleration coefficients 2.05

C2 Social acceleration coefficient 2.05

The procedure for finding the optimal solution using PSO is described below:

1. A population will be randomly generated in the search space.

2. The initial velocity of each particle is randomly generated.
3. Objective function value (OF) for each particle will be calculated.
4. The initial position of each particle i.e. location of EVCS will be selected as its
Pbest, and the position of the particle with minimum objective function among the
population will be choose as Gbest.
5. Particles will move to new positions based on Equation 2.1 – Equation 2.2 and
Equation 3.1.
6. If a particle exceeds the allowed range given by the constraints of objective
function, it will be replaced by its previous position.
7. Objective function value (OF) for each particle will be calculated based on the
updated value of the particle position.
8. Pbest and Gbest will be updated.

21
 Figure 3.3: Flowchart of Particle Swarm Optimization
9. The stopping criterion will be checked. If it satisfies, the algorithm will be
terminated and position of the particle stored in Gbest will be selected as the optimal
location of EVCS. Otherwise, Steps 5 to 8 will be repeated.

3.3 Problem Formulation

3.3.1 Objective Function

The addition of EVCS to RDS degrades the voltage profile and raises the system’s
power losses and line loading. A number of factors are taken into account while
designing an EVCS’s optimal location like EVCS placement area distribution, feeder
capacity, channel capacity, power loss, and voltage regulation etc. In this work multi
objective functions i.e. Active loss and zone component are taken as objective function,
Fobj.

Min {𝑂𝐹} = w1*(APL) + w2*zone component (3.1)

22
Where w1 and w2 are the user defined weights,

w1=0.8 and w2=0.2

Active Power loss Component (APL)

In this thesis work, minimizing the total active power loss will be used as the objective
function, Fobj

The total APL after EVCS placement (PTotalLoss) is obtained by referring to Equation 3.2.
𝑁br
𝑃TotalLoss = ∑𝑖=1 𝑃Loss ,𝑖 (3.2)

where Nbr represents all of the branches.

The total active power generation PTotalGen was obtained from the slack capacity.
Therefore, the total active power loss is obtained by subtracting the total active power
generation from the total load PTotalLoad on the distribution network [2]. This is in
accordance with Eq 3.3-3.4.

𝑃TotalLoss = 𝑃TotalGen − 𝑃TotalLoad (3.3)

𝑁
𝐸𝑉𝐶𝑠
𝑃TotalLoad = 𝑃Loadbase + ∑𝑏=1 𝑃𝐸𝑉𝐶𝑠,𝑏 (3.4)

Zone Component

In this thesis work zone component is used as another objective function for optimal
location of EVCS. The zone component is categorized into three sections: industrial,
urban and semi urban zone based on distribution of load pattern. The zone with load
more than 70kW as a lump load is considered as an industrial zone. The Urban zone is
defined as the area feeding 30-70 kW of distributed load. Similarly, the semi-urban zone
is defined as the area feeding less than 30kW of distributed load.

Zone components are defined based on the types of loads that the buses mostly have.
Buses with high number of domestic loads are designated as urban zone, while buses
with high industrial loads are designated as industrial zone. Similarly buses with
commercial loads are designated as semi-urban zone. Classification as such is based on
the thought that urban zone should be given highest priority for EVCS installation as it
serves a greater number of individual electrical consumers. Similarly, the semi-urban
zone serves lesser individual consumers and accordingly, the priority is set lower than
the urban zone. Finally, the industrial zone is least favored for EVCS installation as it
serves bulk load rather than distributed individual consumers. The weight value for each

23
zone is assigned based on the population available there. The weights are assigned as
defined in Table 3.2

Table 3.2: Zones and their weight values

Zone Types Zone Weights

Urban 0.1

Semi-Urban 0.2

Industrial 0.7

3.3.2 Constraints

Constraints used in this thesis work for particle swarm optimization is briefly
described below.

3.3.2.1 Inequality Constraints

a) Voltage Regulation Constraints

In order to comply with the grid voltage regulation standard, voltage magnitude of
proposed distribution bus network should satisfy the constraints mentioned in
Equation 3.5 before and after EVCS placement [2]

𝑉reg 𝑚𝑖𝑛 ≤ 𝑉bus ≤ 𝑉reg 𝑚𝑎𝑥 (3.5)

where ,

Vreg min=0.95

Vreg max=1.05

b) Line Loading Constraints

In a network, the feeder runs normally when the load is no greater than 80 %.
Therefore, the difference between 80 % of the generation power and the base load
𝑇𝑜𝑡𝑎𝑙
is the maximum capacity of the EVCS 𝑃𝐸𝑉𝐶𝑆 [2].
Total
𝑃𝐸𝑉𝐶𝑠 ≤ 0.8𝑃TotalGen − 𝑃Loadbase (3.6)

24
 c) Distance Constraints

Distance constraint is introduced to the fulfill the requirement of the installation of

two consecutive EVCS to be in the range of designated distance as mentioned in
Equation 3.7. For optimal location of two EVCS in the proposed feeder the distance
component is used as second objective function. The distance between two EVCS
should be between 7 to 11 Km. The distance travelled by EVs with 5% battery
charge is about 10 km, so the distance between any point and the nearest charging
station should not exceed 10km [12]. The Dmin and Dmax is the minimum and
maximum distance from first optimal location to second optimal location for EVCS
placement.

Dmin ≤DEVCS≤D max (3.7)

where, Dmin=7km, Dmax=11km

3.3.3 Voltage Deviation Index (VDI)

Voltage Deviation Index is a measure used to quantify the quality of voltage profiles in
a power system [26].

VDI = √[(1/𝑛) × Σ(𝑉𝑖 − 𝑉𝑟𝑒𝑓)2 ] (3.8)

where, n is the total number of buses, Vref is the nominal voltage and Vi is the actual
voltage at bus ‘i’.

A higher VDI indicates significant voltage deviations, which may lead to power quality
issues. A lower VDI suggests stable and well-regulated voltage. If VDI approaches zero,
the system voltage is very close to nominal.

3.3.4 Load Forecasting Using Exponential Decay Saturation Model

In our load forecasting analysis, we employed the exponential decay saturation model
to project future electricity demand. This approach accounts for the natural tendency of
growth rates to diminish over time as systems mature.

We began by analyzing historical load data from 2076-2081, which revealed distinct
growth patterns across domestic, commercial, and industrial sectors. Each sector
showed evidence of growth moderation, particularly in the later years. To capture this
saturation effect mathematically, we implemented an exponential decay model with the
formula in Equation 3.9

25
 gt =g0×αt (3.9)

𝐿𝑡 = 𝐿0 × ∏𝑡𝑖=1 (1 + 𝑔0 × 𝛼 𝑖 ) (3.10)

To get the value of 𝛼 following mathematical computation is done:

• 𝑔0 = 14.57%
• Then minimization is performed for
𝐽(𝛼) = ∑4𝑡=1 (𝑔𝑡 − 14.57 ⋅ 𝛼 𝑡 )2 (3.11)

In expanded form:

𝐽(𝛼) = (11.21 − 14.57𝛼)2 + (7.70 − 14.57𝛼 2 )2 + (5.67 − 14.57𝛼3 )2 +

(0.98 − 14.57𝛼 4 )2 (3.12)
𝑑𝐽
Then optimal 𝛼 is found by taking the derivative , setting it to 0. Through this
𝑑𝛼
mathematical optimization, we determined that a decay factor of 0.95 provided the best
fit to our historical data. This value minimized prediction errors while maintaining
realistic long-term behavior. The forecasting process involved:

• Calculating weighted growth rates from recent historical data

• Applying the exponential decay function to project diminishing future growth
rates
• Computing future loads by applying these decaying growth rates to current
consumption levels using Equation 3.10.

26
 CHAPTER FOUR : RESULT AND DISCUSSION
In this thesis work, at first the required data for IEEE- 34 is acquired and load flow is
done. Then optimization of EVCS placement is done using PSO algorithm. Comparison
is made for optimal placement of single and dual EVCS placement. After that, the
required line and load data for Khaireni Feeder is also acquired and optimization of
EVCS placement is performed. Comparison is made for optimal placement of single
and dual EVCS placement in the present year i.e. 2081 B.S. Afterwards, load
forecasting is done for the Khaireni feeder based on the trend of load demand of past 6
years, for three types of loads i.e. domestic, industrial and commercial for next 5 and
10 years. Comparison is made for optimal placement of single and dual EVCS
placement in Khaireni feeder is performed as well.

4.1 Analysis in IEEE 34 Test bus system

The backward/forward sweep algorithm has been used in MATLAB to analyze load
flow. The model validation of IEEE 34 test bus system is carried out by evaluating the
voltage regulation requirement. The voltage profile of the system is depicted in Figure
4 and meets the voltage regulation standard. Analysis shows that at Vbase= 24.9kV total
active power loss is 40.15 kW and a minimum voltage of 0.989081p.u. at Bus 27 which
is identical to the literature [18].

4.1.1 With Single and Two EVCS

In this case PSO optimization is done with line data, load data and branch data of 34
test bus system and optimal location of single and two EVCS is obtained with EVCS
of capacity 142kW where with single EVCS optimal location is found at Bus 13 which

Figure 4.1:Voltage profile of IEEE 34 Test Bus system in different cases

27
degrades the voltage of Bus 27 to 0.98902 p.u. Similarly, optimal location of two EVCS
is found to be at Bus 2 and 26 at a distance of 7.05 km which further degrades the
voltage of Bus 27 to 0.988352 p.u which is shown in Figure 4.1

Figure 4.2 depicts the Active Power Loss of IEEE 34 Test Bus system at various
scenarios. This shows that the APL at Base case is 40.15 kW that increases to 40.75 kW
which is 1.49% with single EVCS penetration. Similarly, during two EVCS penetration
the APL increase to 43.77 kW which is 9.016% of base case APL. These results justified
that as the system load increases active power losses increases.

Figure 4.2: APL of IEEE 34 Test Bus system in different cases

28
4.2 Analysis in Khaireni Feeder

Figure 4.3 shows the voltage profile of khaireni feeder at various cases. At base case
load flow, the minimum voltage is found at Bus 50 which is 0.9673 p.u. With addition
of single EVCS at optimal location i.e at Bus 2, the least voltage magnitude of the
system is 0.9669 pu which is at Bus 50. Similarly, for two EVCS placement at optimal
locations i.e at Bus 2 and Bus 34 at a distance of 7 km the least voltage magnitude of
the system is 0.9634 pu which is at Bus 50.

1.01

0.99
voltage (p.u)

0.98

0.97

0.96

0.95

0.94 Base Case Single EVCS Dual EVCS

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51

Figure 4.3: Voltage profile comparison of Khaireni feeder in different cases

Figure 4.4 shows that the Active Power loss increases from 38.79kW at base case to
39.96 kW which is about 3.01 % increment at single EVCS penetration and then to
48.59 kW at dual EVCS placement which is about 25.26% increment to that of base
case. This result is justified because as the system load increases, the active power
losses are bound to increase.

Figure 4.4: APL comparison of Khaireni feeder in various scenarios

29
 Figure 4.5:Active power loss at each Branch during 2081
The convergence curve for optimization of two EVCS placement is shown in Figure
4.6.The first figure shows the normalized fitness curve while the second shows the
actual fitness convergence

Figure 4.6: Convergence curve for optimization of two EVCS placement

30
4.3 Load Forecasting of Khaireni Feeder

A straight forward exponential decay saturation model is used in the load forecasting
analysis because upon inspecting the trend of load demand for past six years, the load
demand tends to saturate as the year passes by. So, this model applies a consistent
percentage reduction to the growth rate each year to simulate the natural tapering of
growth that occurs in maturing systems. The load demand trend for each type of loads,
i.e. domestic, commercial, industrial ranging from 2076 B.S to 2081 B.S was acquired
from Lekhnath DCS. The load growth pattern of overall system in the past six years is
used. The growth rate was calculated for 2086 B.S and 2091 B.S as defined in

Table 4.1. The future load demand at each bus of the feeder is calculated by simply
scaling the present load demand by respective year’s growth. The load forecasting was
done by writing code in MATLAB based on exponential decay saturation model.

Table 4.1: Growth Rates for Different Load Types

Load Type Growth Rate: 2086 B. S Growth Rate: 2091 B. S

Domestic 30.46% 62.85%

Commercial 40.36% 82.82%

Industrial -0.10% -0.18%

Figure 4.7: Load Growth pattern of Khaireni Feeder for past years

31
 Figure 4.8: Load growth for each type of load during forecasted years

4.3.1 Voltage profile and Active Power Loss for future years

1. For Single EVCS

After load forecasting of khaireni feeder using exponential decay saturation method,
the optimal location for single EVCS for next 5 year i.e 2086 B.S is obtained which
came out as Bus 2 but optimal location for next 10 years is not obtained because the
voltage regulation doesn’t meet the regulation standard i.e. voltage regulation
constraint. The voltage profile of real feeder after single EVCS placement in Khaireni
Feeder during 2086 B.S is shown in with comparison to single EVCS placement in
2081 B.S.

Figure 4.9: Voltage profile with single EVCS of Khaireni Feeder at forecasted years

Similarly, the APL of Khaireni feeder with single EVCS at different forecasted years is
depicted in Figure 4.10. Figure 4.10 shows that with optimal placement of single EVCS

32
the APL increases from 38.79 kW to 62.759 kW at 2086 B.S. The branches losses of
Khaireni feeder for 2086 B.S. is shown in Figure 4.11.

Figure 4.10: APL with single EVCS of Khaireni Feeder at forecasted years

Figure 4.11:Active power loss at each Branch during 2086

33
2. For Dual EVCS

Furthermore, the optimal location of dual EVCS is evaluated and came out at Bus 2 and
Bus 38 at a distance of 9.32 km. Voltage profile of 2086 B.S is compared with dual
EVCS placement during 2081 B.S as shown in Figure 4.12.Figure 4.13 shows the
comparison of Active power loss for two EVCS placement at 2081 B.S and 2086 B.S.
which shows that the APL increases from 48.59 kW at 2081 B.S to 74.93 kW at 2086
B.S.

Figure 4.12: Voltage profile of Khaireni Feeder with dual EVCS at forecasted years

Figure 4.13: APL with dual EVCS of Khaireni Feeder at forecasted years

Though the optimal location is found complying with the technical constraints, one
of the optimal locations is changed to Bus 38 from Bus 32 during 2081 B.S. This

34
 makes the optimal location practically infeasible as the EVCS structures are fixed
installation and aren’t movable. As the dual EVCS placement becomes practically
infeasible during 2086 B.S the optimal placement of EVCS after next 5 years i.e.
2091 B.S is also obviously infeasible.

3. Voltage Deviation Index

Figure 4.14: Cumulative VDI for various scenarios

In this thesis work Cumulative VDI is calculated at each optimal scenario as
described in Figure 4.14. As seen in the figure single EVCS at 2081 B.S has a lowest
VDI, which is also an expected outcome. The lower VDI indicates a better voltage
profile across the system. Similarly, scenario single EVCS at 2086 B.S has the
highest VDI indicating the poor voltage profile across the system.

35
 CHAPTER FIVE : CONCLUSIONS AND RECMMENDATIONS

5.1 Conclusions

As specified by the IEEE standard in radial distribution system increasing the load
degrades the voltage profile of the buses and also increases the active power loss in the
system. Also, as the placement of EVCS is done in random fashion in radial distribution
network it will result in further negative impact on grid parameters due to a lot of
modification in load demand pattern as a result of which there will be degradation of
voltage regulation, increase in power loss, increase in line loading e.tc. Therefore, it is
important to find optimal location of EVCS placement so that there will be minimum
negative impact to system parameters.

Initially for the work validation the load flow is performed on IEEE 34 test bus system.
Thereafter integration optimization is done with single and dual EVCS integration using
PSO algorithm. In this case optimal location for single and dual EVCS is obtained. The
result shows that with the placement of EVCS voltage profile degrades but remains
within the regulation limit and power loss increases with the increase in number of
EVCS.

After validation of work in IEEE 34 test bus system, optimization is performed in

Khaireni feeder of Lekhnath DCS, Pokhara within 11 KV distribution line using PSO
algorithm at present scenario and after 5 years and 10 years with forecasted load data
and load growth rate pattern. It is concluded that the network is within regulation limit
even after placement of single and two EVCs in feeder for single EVCS and two EVCS
placement for present scenarios i.e. 2081 B.S. Here the optimal location for single
EVCS placement is obtained at Bus 2. For two EVCS placement optimal location is
obtained at Bus 2 and Bus 34. During the forecasted year 2086 B.S, only single EVCS
placement is viable. For rest of other scenarios; two EVCS placement during 2086 B.S
and single and Dual EVCS placement during 2091B.S, reallocation must be analyzed
accordingly. The VDI comparison is also done in this thesis work which shows that the
VDI is least during single EVCS installation at 2081 B.S, while it is highest during
single EVCS installation at 2086 B.S.

5.2 Recommendations

The recommendations for this thesis work are mentioned as below:

36
• Load forecasting model can be further improved by including several technical
and practical parameters.
• Economic analysis can be done for the optimal placement of EVCS in the
feeder.
• As the dual EVCS placement in 2086 B.S and both single and dual EVCS
placement in 2091B.S were infeasible, this problem can be mitigated by
penetration of DGs in the feeder, use of hybrid compensators like PV-
DSTATCOM, SVCs etc.

37
 REFERENCES

[1] Y. L. a. A. Jenn, "Impact of electric vehicle charging demand on power

distribution grid congestion," Proceedings of the National Academy of Sciences,
vol. 121, p. 18, 2024.

[2] S. D. a. A. M. a. R. D. C. Nugraha, "Hybrid Genetic Algorithm-Modified Salp

Swarm Algorithm to Determine the Optimal Location of Electric Vehicle
Charging Station in Distribution Networks," IEEE Access, vol. 12, p. 3, 2024.

[3] M. a. Y. M. a. I. H. a. R. M. a. A. N. a. K. M. A. a. W. M. a. S. A. M. Altaf, "PSO-

based optimal placement of electric vehicle charging stations in a distribution
network in smart grid environment incorporating backward forward sweep
method," IET Renewable Power Generation, vol. 18, p. 15, 2024.

[4] A. A. a. K. Strunz., "Modeling of an electric vehicle charging station for fast dc

charging.," in IEEE International Electric Vehicle Conference, 2012.

[5] S. D. a. X.-Z. G. a. K. T. a. K. K. a. P. Mahanta, "Nature-Inspired Optimization

Algorithms Applied for Solving Charging Station Placement Problem : Overview
and Comparison," Archives of Computational Methods in Engineering, vol. 28,
no. 2021, pp. 91-106, 2021.

[6] P. a. P. a. S. a. K. Antarasee, "Optimal Design of Electrical Vechile Fast-Charging

Station's Structure Using Metaheuristic Algorithms," Sustainability, vol. 15, no.
2022, p. 771, 2022.

[7] A. Amin, W. Tareen, M. Usman, H. Ali, I. Bari, B. Horan, S. Mekhilef, M. Asif,

S. Ahmed and Mahmood, " A Review of Optimal Charging Strategy for Electric
Vehicles under Dynamic Pricing Schemes in the Distribution Charging Network,"
Sustainability, p. 12, 2020.

[8] S. a. K. D. Muthukannan, "{Multiobjective Planning Strategy for the Placement

of Electric-Vehicle Charging Stations Using Hybrid Optimization Algorithm,"
IEEE Access, vol. 10, no. 2022, pp. 48088-48101, 2022.

[9] V. P. a. K. Swarnasri, "Multi-Objective Optimal Allocation of Electric Vehicle

Charging Stations and Distributed Generators in Radial Distribution Systems
using Metaheuristic Optimization Algorithms," Engineering, Technology \&
Applied Science Research, 2020.

[10] D. a. R. R. M. Srinivas, "Optimal Placement of Electric Vehicle Charging Station

by Considering Dynamic Loads in Radial Distribution System," in 2022

38
 International Conference on Automation, Computing and Renewable Systems
(ICACRS), 2022.

[11] P. T. a. S. N. S. S. P. Singh, "Optimal Allocation of EVS Charging Stations with

Help of Traffic and Power Constraints using PSO: A Case Study of Ayodhya
City," in 2024 4th International Conference on Advance Computing and
Innovative Technologies in Engineering (ICACITE), Greater Noida,India, 2024.

[12] S. F. a. N. T. a. G. M. a. C. H. Keleshteri, "New optimal planning strategy for

plug‐in electric vehicles charging stations in a coupled power and transportation
network," The Journal of Engineering, vol. 2023, no. 3, 2023.

[13] M. a. S. J. a. S. Q. Dong, "Multi-year long-term load forecast for area distribution

feeders based on selective sequence learning," Energy, vol. 206, no. 2020, p.
118209, 2020.

[14] T.-H. a. S. R. a. W. Q. a. G. Y. Yang, "Saturated Demand Forecast of Regional

Power Grid Based on Amended Self-Adaptive Logistic Model: A Case Study of
East China," IEEE Access, vol. 9, no. 2021, 2021.

[15] A. a. A. a. K. Maleki, "Scrutiny of multifarious particle swarm otimization for

finding the optimal size of a PV/wind/battery hybrid system," Renewable Energy,
vol. 80, no. 2015, pp. 552-563, 2015.

[16] N. Shrestha, "Financial Analysis of Utility Scale Photovoltaic System with

Battery Energy Storage System in Nepal," IOEGC, 2023.

[17] T. C. S. &. M. T. Deosaria, "Load flow analysis using Forward and Backward
sweep, and minimising power losses using Genetic Algorithm," International
Journal of Advances in Engineering and Management (IJAEM), vol. 4, no. 2022,
pp. 763-772, 2022.

[18] S. P. Ashok, "Modeling and Protection Scheme for IEEE 34 Radial Distribution
Feeder with and without Distributed Generation," IEEE Access, p. 394, 2014.

[19] M. D. a. L. Grumbach, "A Hybrid Distribution Feeder Long-Term Load

Forecasting Method Based on Sequence Prediction," IEEE, p. 12, 2019.

[20] G. W. A. S. G. a. J. R. M. I. M.Elders, " Identification of long-term scenarios of

electricity network development," in Proc. Int Conf.Future Power Syst,
Amsterdam,Netherland, 2005.

[21] X. Xin, "Survey of saturated load analysis technology for urban power," Electr.
Power Automat. Equip, 2014.

39
[22] S. F. a. R. J. Hyndman, "Short-term load forecasting based on a semi-parametric
additive model," IEEE, vol. 27, no. Feb, 2012, pp. 134-141, 2012.

[23] A. A. M. I. K. a. T. S. Ö. M. Kankal, " Modeling and forecasting of Turkeys energy

consumption using socio-economic and demograhic variables," vol. 88, pp. 1927-
1939, May 2011.

[24] C. S. Q. Z. a. X. Z. L. Lan, "Short-term load forecasting method based on RBF

neuralNetwork and an s system," in Proc. CSEE, 2005.

[25] H. L. Willis, "Spatial Electric Load Forecasting: Second," Newyork, 2022.

[26] M. C. a. D. Das, "Voltage stability analysis of radial distribution networks,"

International Journal of Electrical Power & Energy Systems, vol. 23, pp. 129-
135, 2001.

40
 APPENDIX-A

1. Parameters of IEEE 34 Test Bus System (VBase=24.9KV, SBase=100MVA)

From To Length (km) R (Ω) X (Ω) P (kW) Q (kVAR)

1 2 0.6 0.117 0.048 230 142.5
2 3 0.55 0.1073 0.044 0 0
3 4 0.55 0.1645 0.0457 230 142.5
4 5 0.5 0.1495 0.0415 230 142.5
5 6 0.5 0.1495 0.0415 0 0
6 7 0.6 0.3144 0.054 0 0
7 8 0.4 0.2096 0.036 230 142.5
8 9 0.6 0.3144 0.054 230 142.5
9 10 0.4 0.2096 0.036 0 0
10 11 0.25 0.131 0.0225 230 142.5
11 12 0.2 0.1048 0.018 137 84
3 13 0.3 0.1572 0.027 72 45
13 14 0.4 0.2096 0.036 72 45
14 15 0.2 0.1048 0.018 72 45
15 16 0.1 0.0524 0.009 13.5 7.5
6 17 0.6 0.1794 0.0498 230 142.5
17 18 0.55 0.1645 0.0457 230 142.5
18 19 0.55 0.2079 0.0473 230 142.5
19 20 0.5 0.189 0.043 230 142.5
20 21 0.5 0.189 0.043 230 142.5
21 22 0.5 0.262 0.045 230 142.5
22 23 0.5 0.262 0.045 230 142.5
23 24 0.6 0.3144 0.054 230 142.5
24 25 0.4 0.2096 0.036 230 142.5
25 26 0.25 0.131 0.0225 230 142.5
26 27 0.2 0.1048 0.018 137 85
7 28 0.3 0.1572 0.027 75 48
28 29 0.3 0.1572 0.027 75 48
29 30 0.3 0.1572 0.027 75 48
10 31 0.3 0.1572 0.027 57 34.5
31 32 0.4 0.2096 0.036 57 34.5
32 33 0.3 0.1572 0.027 57 34.5
33 34 0.2 0.1048 0.018 57 34.5

41
 2. Parameters of Khaireni Feeder

Length Conductor PL QL
Bus ID (km) Type R (Ω) X (Ω) (kW) (kVAR)
Substation 0 0
Hotel Ravi Mahal 0.75 XLPE 0.2437 0.096 31.1 18.6
Gachyafaat 0.12 Rabbit 0.1033 0.034 19.5 11.6
DadaNak I 0.4 XLPE 0.13 0.051 38.9 23.3
DadaNak II 0.2 XLPE 0.065 0.025 38.9 23.3
ERMC 0.22 XLPE 0.0715 0.028 19.5 11.7
NTC Office 0.35 XLPE 0.1137 0.044 38.9 23.3
TalChowk 0.18 XLPE 0.0585 0.023 38.9 23.3
Talchowk Height 0.2 XLPE 0.065 0.025 19.5 11.7
BhatBhateni 0.1 XLPE 0.0325 0.012 38.9 23.3
Sujal Foods 0.25 XLPE 0.0812 0.032 155.4 93.9
Naya Talchowk 0.3 XLPE 0.0975 0.038 38.9 23.3
Naya Talchowk II 0.2 XLPE 0.065 0.025 0 0
Thulakhor 0.2 Rabbit 0.1722 0.058 19.5 11.6
Thulakhor Khani
Pani 0.1 Rabbit 0.0861 0.029 19.5 11.6
Adarsha Chowk 0.1 XLPE 0.0325 0.012 0 0
Bhandari Rice Mill 0.2 Rabbit 0.1722 0.058 19.5 11.6
Hatchery 0.08 Rabbit 0.0688 0.023 19.5 11.6
Gumbaz 0.6 XLPE 0.195 0.076 38.9 23.3
Gumbaz Ncell 0.1 Rabbit 0.0861 0.029 9.68 5.74
BhandariDhik 0.05 XLPE 0.0162 0.006 38.9 23.3
PU Gate 0.8 XLPE 0.26 0.102 38.9 23.3
Naba Durga 0.4 XLPE 0.13 0.051 19.5 11.7
PowerHouse 0.8 XLPE 0.26 0.102 0 0
Power House I 0.1 Rabbit 0.0861 0.029 19.5 11.6
Power House II 0.1 Rabbit 0.0861 0.029 19.5 11.6
Ghotghote 0.01 Rabbit 0.0086 0.002 19.5 11.6
Oil Corporation 0.2 Rabbit 0.1722 0.058 28.9 17.1
Karyashiddi 0.2 Rabbit 0.1722 0.058 96.8 57.4
Seti Hydro 0.5 Rabbit 0.4305 0.145 96.8 57.4
Bhatyako Chauki 0.8 XLPE 0.26 0.102 38.9 23.3
Gagangauda I 0.5 XLPE 0.1625 0.064 38.9 23.3
Grihyalaxmi 0.25 XLPE 0.0812 0.032 29.1 17.4
GaganGauda II 0.3 XLPE 0.0975 0.038 38.9 23.3
Tallo Gagangauda 0.5 XLPE 0.1625 0.064 38.9 23.3
Chaplang 0.5 XLPE 0.1625 0.064 0 0
Chhaplyang 0.4 Rabbit 0.3444 0.116 19.5 11.6
Apukaseri 0.8 Rabbit 0.6888 0.232 96.8 57.4
Lameahal I 0.5 XLPE 0.1625 0.064 19.5 11.7
Lameahal II 0.1 XLPE 0.0325 0.012 38.9 23.3
Majuwa 0.6 XLPE 0.195 0.076 38.9 23.3
Majuwa Pani Tanki 0.25 XLPE 0.0812 0.032 19.5 11.7

42
Bio Gas 0.2 Rabbit 0.1722 0.058 19.5 11.7
Eaklyakhet 0.5 Rabbit 0.4305 0.145 19.5 11.7
Kotre Pool 0.7 XLPE 0.2275 0.089 0 0
Seti Hydro 1 Rabbit 0.861 0.291 38.7 23
Tallo Pudi 0.8 Rabbit 0.6888 0.102 19.5 11.7
Seti Hydro II 0.4 Rabbit 0.3444 0.116 38.7 23
Upallo Pudi 0.1 Rabbit 0.0861 0.029 19.5 11.7
Upallo Pudi II 0.5 Rabbit 0.4305 0.145 19.5 11.6
Kotre 0.06 XLPE 0.0195 0.007 19.5 11.7
Feeder Length 17.57

3. Load Factor of Khaireni Feeder

Date Maximum Load Energy Days Load Factor

2080 Poush 256 756.05 29 0.2227
2080 Magh 344 607.32 29 0.1331
2080 Falgun 266 679.71 30 0.1863
2080 Chaitra 80 409.37 30 0.3730
2081 Baishak 212 646.9 31 0.2153
2081 Jestha 212 473.59 32 0.1527
2081 Asar 320 832.43 31 0.1835
2081 Shrawan 69 468.75 32 0.4643
2081 Bhadra 248 855.08 31 0.2432
2081 Ashoj 150 1110.09 30 0.5395
2081 Kartik 129 894.65 30 0.5056
2081 Mangshir 187 1098.87 30 0.4284
Total Load Factor 0.30

43
4. SLD of IEEE 34 Test Bus System with optimal placement of Two EVCS

44
5. SLD of Khaireni Feeder with optimal placement of Two EVCS

45
 6. Energy demand for various type of load in the past years

Year Domestic (MWhr) Commercial (MWhr) Industrial (MWhr)

2076 25932379 5907757 10266602.46
2077 29710306 8250152.57 10358239.22
2078 33041892 8848689.33 11304698.84
2079 35584749 9049204.47 10836216.41
2080 37600820.3 9871596.42 10276331.61
2081 37969034 9998046.48 10494682.34

7. Past Growth Rate of Khaireni Feeder

Historical Domestic Commercial Industrial Total Growth

Year Load Load Load Rate
2076-2077 14.57% 39.65% 0.89% 14.75%
2077-2078 11.21% 7.25% 9.14% 10.09%
2078-2079 7.70% 2.27% -4.14% 4.28%
2079-2080 5.67% 9.09% -5.17% 4.11%
2080-2081 0.98% 1.28% 2.12% 1.23%

8. Forecasted Load Data of Khaireni Feeder

2086 B. S 2091 B. S
SN
P (kW) Q (kVAR) P (kW) Q (kVAR)
1 0 0 0 0
2 40.629 24.29904 50.0244 29.9181
3 27.3702 16.28176 35.6499 21.20712
4 50.819 30.43912 62.5707 37.47805
5 50.819 30.43912 62.5707 37.47805
6 27.3702 16.42212 35.6499 21.38994
7 50.819 30.43912 62.5707 37.47805
8 50.819 30.43912 62.5707 37.47805
9 27.3702 16.42212 35.6499 21.38994
10 50.819 30.43912 62.5707 37.47805
11 155.245 93.8061 155.12 93.73098
12 50.819 30.43912 62.5707 37.47805
13 0 0 0 0
14 27.3702 16.28176 35.6499 21.20712
15 27.3702 16.28176 35.6499 21.20712
16 0 0 0 0
17 27.3702 16.28176 35.6499 21.20712
18 27.3702 16.28176 35.6499 21.20712
19 50.819 30.43912 62.5707 37.47805
20 13.5868 8.056664 17.697 10.493868

46
21 50.819 30.43912 62.5707 37.47805
22 50.819 30.43912 62.5707 37.47805
23 27.3702 16.42212 35.6499 21.38994
24 0 0 0 0
25 27.3702 16.28176 35.6499 21.20712
26 27.3702 16.28176 35.6499 21.20712
27 27.3702 16.28176 35.6499 21.20712
28 40.564 24.00156 52.835 31.26222
29 96.7032 57.3426 96.6258 57.29668
30 96.7032 57.3426 96.6258 57.29668
31 50.819 30.43912 62.5707 37.47805
32 50.819 30.43912 62.5707 37.47805
33 40.8448 24.42264 53.2006 31.81068
34 50.819 30.43912 62.5707 37.47805
35 50.819 30.43912 62.5707 37.47805
36 0 0 0 0
37 27.3702 16.28176 35.6499 21.20712
38 96.7032 57.3426 96.6258 57.29668
39 27.3702 16.42212 35.6499 21.38994
40 50.819 30.43912 62.5707 37.47805
41 50.819 30.43912 62.5707 37.47805
42 27.3702 16.42212 35.6499 21.38994
43 27.3702 16.42212 35.6499 21.38994
44 27.3702 16.42212 35.6499 21.38994
45 0 0 0 0
46 50.5577 30.0472 62.249 36.9955
47 27.3702 16.42212 35.6499 21.38994
48 50.5577 30.0472 62.249 36.9955
49 27.3702 16.42212 35.6499 21.38994
50 27.3702 16.28176 35.6499 21.20712
51 27.3702 16.42212 35.6499 21.38994

47
 9. Voltage Profile of 34 Test Bus system

Bus No Base Case Single EVCS Dual EVCS

1 1 1 1
2 0.998894 0.9988639 0.998833
3 0.99793 0.99787209 0.997841
4 0.99662 0.99656191 0.996489
5 0.995495 0.99543668 0.995326
6 0.994435 0.99437684 0.994228
7 0.993712 0.99365335 0.993504
8 0.993314 0.99325604 0.993107
9 0.992848 0.99279003 0.992641
10 0.992624 0.99256602 0.992417
11 0.992538 0.99247961 0.992331
12 0.992512 0.99245382 0.992305
13 0.997865 0.9977696 0.997776
14 0.997806 0.99771056 0.997717
15 0.99779 0.99769455 0.997701
16 0.997789 0.9976933 0.9977
17 0.993601 0.99354243 0.993348
18 0.992908 0.99284927 0.992613
19 0.992143 0.99208518 0.991797
20 0.99153 0.99147123 0.991136
21 0.990996 0.99093801 0.990556
22 0.990388 0.99032977 0.989884
23 0.989889 0.98983016 0.98932
24 0.989419 0.98936104 0.988774
25 0.989194 0.98913526 0.988497
26 0.989107 0.98904851 0.988378
27 0.989081 0.9890226 0.988352
28 0.993648 0.99358962 0.993441
29 0.993605 0.99354713 0.993398
30 0.993584 0.99352589 0.993377
31 0.99256 0.99250172 0.992353
32 0.992496 0.99243741 0.992288
33 0.992464 0.99240526 0.992256
34 0.992453 0.99239454 0.992245

48
10. Voltage profile of Khaireni feeder at 2081 B.S

Bus No Base Case Single EVCS Dual EVCS

1 1 1 1
2 0.9959 0.9956 0.995276939
3 0.9959 0.9956 0.995256845
4 0.9938 0.9935 0.992996983
5 0.9928 0.9925 0.991883053
6 0.9917 0.9914 0.990686399
7 0.99 0.9897 0.988805568
8 0.9891 0.9888 0.987861818
9 0.9882 0.9879 0.986839371
10 0.9877 0.9874 0.986334717
11 0.9866 0.9863 0.985105834
12 0.9855 0.9851 0.983788572
13 0.9847 0.9844 0.982936666
14 0.9846 0.9843 0.982868842
15 0.9846 0.9843 0.982851886
16 0.9843 0.984 0.982523873
17 0.9843 0.9839 0.982456021
18 0.9842 0.9839 0.98244245
19 0.9822 0.9819 0.980126143
20 0.9822 0.9819 0.980117706
21 0.982 0.9817 0.97993456
22 0.9794 0.9791 0.976974745
23 0.9782 0.9779 0.975547765
24 0.9757 0.9754 0.972747013
25 0.9757 0.9754 0.972712746
26 0.9757 0.9754 0.972695613
27 0.9757 0.9754 0.972725744
28 0.9753 0.975 0.972334632
29 0.975 0.9747 0.971994284
30 0.9746 0.9742 0.97156876
31 0.9741 0.9737 0.970712689
32 0.9731 0.9728 0.96950778
33 0.9726 0.9723 0.968938631
34 0.9721 0.9718 0.968285557
35 0.9713 0.971 0.967484013
36 0.9706 0.9702 0.966749221
37 0.9702 0.9698 0.966337721
38 0.9695 0.9692 0.965652707
39 0.97 0.9697 0.966213962
40 0.9699 0.9696 0.966113612
41 0.9694 0.9691 0.965591718
42 0.9692 0.9689 0.965407697
43 0.9692 0.9688 0.965338539
44 0.9691 0.9688 0.965252087
49
 45 0.9689 0.9685 0.965033306
46 0.9677 0.9673 0.963827584
47 0.9675 0.9672 0.963702122
48 0.9674 0.9671 0.963551824
49 0.9674 0.967 0.963517205
50 0.9673 0.9669 0.963430715
51 0.9689 0.9685 0.96502928

11. Forecasted voltage Profile of Khaireni feeder

SN With Single EVCS With Dual EVCS

2086 2091 2086 2091
1 1 1 1 1
2 0.9946 0.9935 0.99423 0.9931
3 0.9945 0.9935 0.99407 0.9931
4 0.9919 0.9902 0.99141 0.9896
5 0.9907 0.9885 0.99004 0.9879
6 0.9893 0.9869 0.98857 0.986
7 0.9871 0.9843 0.98626 0.9831
8 0.9861 0.983 0.9851 0.9817
9 0.9849 0.9816 0.98385 0.9802
10 0.9843 0.9809 0.98323 0.9794
11 0.983 0.9792 0.98173 0.9777
12 0.9815 0.9774 0.98009 0.9757
13 0.9805 0.9762 0.97903 0.9744
14 0.9804 0.9761 0.97893 0.9743
15 0.9804 0.9761 0.97891 0.9742
16 0.98 0.9757 0.97852 0.9738
17 0.9799 0.9755 0.97842 0.9737
18 0.9799 0.9755 0.9784 0.9736
19 0.9773 0.9724 0.97556 0.9703
20 0.9773 0.9724 0.97555 0.9702
21 0.9771 0.9721 0.97533 0.97
22 0.9739 0.9682 0.9717 0.9656
23 0.9723 0.9663 0.96995 0.9636
24 0.9693 0.9626 0.96654 0.9595
25 0.9692 0.9626 0.96649 0.9595
26 0.9692 0.9625 0.96647 0.9594
27 0.9692 0.9626 0.96651 0.9595
28 0.9688 0.9622 0.9661 0.9591
29 0.9685 0.9618 0.96576 0.9587
30 0.9681 0.9614 0.96533 0.9583
31 0.9671 0.9599 0.96399 0.9564
32 0.9658 0.9583 0.96249 0.9546

50
33 0.9652 0.9576 0.96178 0.9538
34 0.9645 0.9568 0.96097 0.9528
35 0.9635 0.9555 0.95971 0.9513
36 0.9626 0.9543 0.95854 0.9499
37 0.9621 0.9539 0.95763 0.9494
38 0.9614 0.9532 0.95601 0.9487
39 0.9618 0.9534 0.95781 0.9488
40 0.9617 0.9532 0.95767 0.9486
41 0.961 0.9523 0.95696 0.9477
42 0.9607 0.952 0.9567 0.9473
43 0.9606 0.9519 0.95661 0.9472
44 0.9605 0.9517 0.95648 0.947
45 0.9602 0.9513 0.95619 0.9467
46 0.9586 0.9493 0.95455 0.9446
47 0.9584 0.949 0.95437 0.9443
48 0.9582 0.9488 0.95417 0.9441
49 0.9582 0.9487 0.95412 0.944
50 0.958 0.9486 0.954 0.9439
51 0.9602 0.9513 0.95619 0.9467

51
 APPENDIX-B

1. MATLAB Code for Optimal Location of Dual EVCS

% Function to calculate base power loss - extracted for reuse

function base_PL = calculate_base_PL(loaddata, linedata, currentsRef)
[base_PL, ~, ~, ~] = calculate_loadflow(loaddata, linedata, currentsRef);
end

% Function to create zone weights - extracted for reuse

function zone_weights = create_zone_weights(loaddata)
n_buses = size(loaddata, 1);

% Identify zones
urban_zones = [];
rural_zones = [];
for i = 1:n_buses
if loaddata(i,2)>=30 && loaddata(i,2)<70
urban_zones = [urban_zones, i];
elseif loaddata(i,2)>=70
rural_zones = [rural_zones, i]
end
end

semi_urban_zones = setdiff(2:n_buses, [urban_zones, rural_zones]);

% Create zone weights

zone_weights = ones(n_buses, 1) * 0.7;
zone_weights(semi_urban_zones) = 0.2;
zone_weights(urban_zones) = 0.1;
end

function [best_positions, best_fitness, convergence_data, power_loss_curve] =

optimize_dual_evcs_placement(loaddata, linedata, branch_distances, evcs_power,
n_particles, max_iterations, currentsRef, test_iter)

% PSO parameters
w = 0.729;
c1 = 2.05;
c2 = 2.05;

% Problem dimensions
n_buses = size(loaddata, 1);
urban_zones = []; % High load density areas
rural_zones = []; % Medium load density areas
for i= 1:n_buses
if loaddata(i,2)>=30 && loaddata(i,2)<70
urban_zones=[urban_zones,i];
elseif loaddata(i,2)>=70
rural_zones=[rural_zones,i];

52
 end
end

semi_urban_zones = setdiff(2:n_buses, [urban_zones, rural_zones]);

% Create zone weights

zone_weights = ones(n_buses, 1) * 0.7;
zone_weights(semi_urban_zones) = 0.2;
zone_weights(urban_zones) = 0.1;

% Calculate distance matrix using branch distances

distance_matrix = create_distance_matrix(branch_distances, n_buses);

% Initialize particles for two locations

positions = zeros(n_particles, 2);
for i = 1:n_particles
while true
pos1 = randi([2, n_buses]);
pos2 = randi([2, n_buses]);
if pos1 ~= pos2
dist = distance_matrix(pos1, pos2);
if dist >= 7 && dist <= 11 % Distance constraint in km
positions(i,:) = [pos1, pos2];
break;
end
end
end
end

velocities = zeros(n_particles, 2);

% Initialize best positions and fitness

personal_best_pos = positions;
personal_best_fit = inf(n_particles, 1);
global_best_pos = zeros(1, 2);
global_best_fit = inf;

% Initialize arrays to store convergence data at each iteration

convergence_data = zeros(max_iterations, 1); % Weighted fitness values
power_loss_curve = zeros(max_iterations, 1); % Actual power losses in kW
best_power_loss = inf;

% Main PSO loop

for iter = 1:max_iterations
% Evaluate each particle
for i = 1:n_particles
[fitness, is_feasible, metrics] = evaluate_dual_fitness(positions(i,:), loaddata,
linedata, ...
evcs_power, zone_weights, distance_matrix,
base_PL, currentsRef);

53
 % Update personal best if feasible and better
if is_feasible && fitness < personal_best_fit(i)
personal_best_fit(i) = fitness;
personal_best_pos(i,:) = positions(i,:);

% Update global best

if fitness < global_best_fit
global_best_fit = fitness;
global_best_pos = positions(i,:);
best_power_loss = metrics.power_loss;
print_dual_solution_metrics(positions(i,:), metrics, distance_matrix);
end
end
end

% Store best values for this iteration

convergence_data(iter) = global_best_fit;
power_loss_curve(iter) = best_power_loss;

% Update velocities and positions

for i = 1:n_particles
for d = 1:2
r1 = rand();
r2 = rand();
velocities(i,d) = w*velocities(i,d) + ...
c1*r1*(personal_best_pos(i,d) - positions(i,d)) + ...
c2*r2*(global_best_pos(d) - positions(i,d));

new_pos = round(positions(i,d) + velocities(i,d));

positions(i,d) = max(2, min(new_pos, n_buses));
end

% Check and correct distance constraint

if ~check_distance_constraint(positions(i,:), distance_matrix)
positions(i,:) = personal_best_pos(i,:);
end
end

% Display progress
if global_best_fit ~= inf
dist = distance_matrix(global_best_pos(1), global_best_pos(2));
fprintf('Iteration %d: Best Loss = %.4f kW, Locations = [%d, %d], Distance =
%.2f km\n', ...
iter, best_power_loss, global_best_pos(1), global_best_pos(2), dist);
else
fprintf('Iteration %d: No feasible solution found yet\n', iter);
end
end

54
 best_positions = global_best_pos;
best_fitness = global_best_fit;

% Handle the case where no feasible solution was found during any iteration
if isinf(global_best_fit)
convergence_data = convergence_data * 0; % Set all values to 0
power_loss_curve = power_loss_curve * 0; % Set all values to 0
else
% Remove any inf values from early iterations (if any)
for i = 1:max_iterations
if isinf(convergence_data(i))
if i > 1
convergence_data(i) = convergence_data(i-1);
power_loss_curve(i) = power_loss_curve(i-1);
else
convergence_data(i) = best_fitness;
power_loss_curve(i) = best_power_loss;
end
end
end
end

% Print final results

if best_fitness ~= inf
fprintf('\nFinal Results for Test %d:\n', test_iter);
fprintf('Optimal EVCS Locations: Bus %d and Bus %d\n', best_positions(1),
best_positions(2));
fprintf('Distance between locations: %.2f km\n',
distance_matrix(best_positions(1), best_positions(2)));
fprintf('Final System Losses: %.4f kW\n', best_power_loss);

% Calculate final state with EVCS at optimal locations

mod_loaddata = loaddata;
mod_loaddata(best_positions(1), 2) = mod_loaddata(best_positions(1), 2) +
evcs_power;
mod_loaddata(best_positions(1), 3) = mod_loaddata(best_positions(1), 3) +
evcs_power*0.3;
mod_loaddata(best_positions(2), 2) = mod_loaddata(best_positions(2), 2) +
evcs_power;
mod_loaddata(best_positions(2), 3) = mod_loaddata(best_positions(2), 3) +
evcs_power*0.3;

[final_PL, final_V, final_loading] = calculate_loadflow(mod_loaddata, linedata,

currentsRef);

% Save results to Excel file with test iteration number

save_results_to_excel(best_positions, distance_matrix, final_V, final_PL,
final_loading, test_iter);
else
fprintf('\nNo feasible solution found in Test %d.\n', test_iter);

55
 end
end

function [fitness, is_feasible, metrics] = evaluate_dual_fitness(positions, loaddata,

linedata, ...
evcs_power, zone_weights, distance_matrix,
base_PL, currentsRef)
% Create modified load data with both EVCS
mod_loaddata = loaddata;
for i = 1:2
mod_loaddata(positions(i), 2) = mod_loaddata(positions(i), 2) + evcs_power;
mod_loaddata(positions(i), 3) = mod_loaddata(positions(i), 3) +
evcs_power*0.3;
end

% Run load flow

[PL, V, loading] = calculate_loadflow(mod_loaddata, linedata, currentsRef);

% Store metrics
metrics = struct();
metrics.min_voltage = min(V);
metrics.max_voltage = max(V);
metrics.max_loading = max(loading);
metrics.distance = distance_matrix(positions(1), positions(2));

% Check constraints
voltage_violated = any(V < 0.95 | V > 1.05);
loading_violated = any(loading > 180);
distance_feasible = (metrics.distance >= 7) && (metrics.distance <= 11);

is_feasible = ~voltage_violated && ~loading_violated && distance_feasible;

if is_feasible
% Calculate actual power loss in kW for tracking
metrics.power_loss = PL;

% Normalized power loss component (dimensionless)

loss_component = (PL - base_PL)/base_PL;

% Zone component calculation

% Higher weights (urban zones) should give lower component value
zone1 = zone_weights(positions(1));
zone2 = zone_weights(positions(2));
zone_component = 1 - (zone1 + zone2)/2; % normalized to [0,1]

% Combined weighted fitness without distance component

w1 = 0.8; % Power loss weight
w2 = 0.2; % Zone weight

fitness = w1*loss_component + w2*zone_component

56
 % Store component values for reporting
metrics.loss_component = loss_component;
metrics.zone_component = zone_component;
metrics.zone1_weight = zone1;
metrics.zone2_weight = zone2;
else
fitness = inf;
end
end

function distance_matrix = create_distance_matrix(branch_distances, n_buses)

% Initialize distance matrix
distance_matrix = inf(n_buses);

% Set diagonal elements to zero

for i = 1:n_buses
distance_matrix(i,i) = 0;
end

% Fill direct distances from branch data

for i = 1:size(branch_distances,1)
from_bus = branch_distances(i,1);
to_bus = branch_distances(i,2);
distance = branch_distances(i,3);
distance_matrix(from_bus,to_bus) = distance;
distance_matrix(to_bus,from_bus) = distance;
end

% Floyd-Warshall algorithm for shortest paths

for k = 1:n_buses
for i = 1:n_buses
for j = 1:n_buses
if distance_matrix(i,j) > distance_matrix(i,k) + distance_matrix(k,j)
distance_matrix(i,j) = distance_matrix(i,k) + distance_matrix(k,j);
end
end
end
end
end

function feasible = check_distance_constraint(positions, distance_matrix)

dist = distance_matrix(positions(1), positions(2));
feasible = (dist >= 7) && (dist <= 11);
end

param_values = {
test_iter;
optimal_buses(1);
optimal_buses(2);

57
 distance_matrix(optimal_buses(1), optimal_buses(2));
total_losses;
min(bus_voltages);
max(bus_voltages);
max(line_loading);
};

summary_table = table(param_names, param_values, 'VariableNames',

summary_vars);

% Create bus voltage data as a table

bus_numbers = (1:length(bus_voltages))';
voltages = bus_voltages(:);

% Create a logical array for EVCS placement

is_evcs = false(length(bus_voltages), 1);
is_evcs(optimal_buses(1)) = true;
is_evcs(optimal_buses(2)) = true;

% Convert logical to string 'Yes'/'No' for better readability

evcs_placement = cell(length(bus_voltages), 1);
evcs_placement(is_evcs) = {'Yes'};
evcs_placement(~is_evcs) = {'No'};

bus_table = table(bus_numbers, voltages, evcs_placement, ...

'VariableNames', {'Bus_Number', 'Voltage_pu', 'EVCS_Placement'});

% Try to write tables to Excel file

try
% Write summary to Sheet 1
writetable(summary_table, filename, 'Sheet', 'Summary', 'WriteVariableNames',
true);

% Write detailed bus voltages to Sheet 2

writetable(bus_table, filename, 'Sheet', 'Bus_Voltages', 'WriteVariableNames',
true);

fprintf('Results for Test %d saved to %s\n', test_iter, filename);

catch ME

fprintf('Error saving results for Test %d: %s\n', test_iter, ME.message);

end
end
end

58
4/19/25, 10:36 AM Gmail - [IOEGC16] Editor Decision

Shreedhar dangi <dangishreedhar239@gmail.com>

[IOEGC16] Editor Decision

Suwarna Lingden <conference-noreply@ioe.edu.np> Sun, Apr 6, 2025 at 1:47 AM
To: Shreedhar Dangi <dangishreedhar239@gmail.com>, Sandeep Dhami
<sandeep.dhami@ioepas.edu.p>, Basant Raj Tiwari <rajtiwarib42@gmail.com>, Milan Rimal
<rimalmilan@gmail.com>

Shreedhar Dangi, Sandeep Dhami, Basant Raj Tiwari, Milan Rimal:

We are pleased to inform you that your manuscript titled "Optimal Location of Electric
Vehicle Charging Station on Khaireni Feeder-Lekhnath, Pokhara using PSO Algorithm"
submitted to 16th IOE Graduate Conference is Accepted for presentation in the
Conference as well as inclusion in the Peer-Reviewed Proceedings. Please note that
inclusion in hard copy proceedings is contingent upon your timely response to further
edits, if any, during the publication process.

With Warm Regards,

IOEGC-16 Editorial Team

https://mail.google.com/mail/u/0/?ik=02edf9f491&view=pt&search=all&permmsgid=msg-f:1828594213620181099&simpl=msg-f:18285942136201… 1/1
 IOE Graduate Conference
[Placeholder for
Publication
Information]

Optimal Location of Electric Vehicle Charging Station on Khaireni

Feeder-Lekhnath, Pokhara using PSO Algorithm
Shreedhar Dangia , Sandeep Dhami b , Basant Raj Tiwari c , Milan Rimal d
a Department of Electrical Engineering, Pashchimanchal Campus, Institute of Engineering, Tribhuwan University, Nepal
b Department of Electrical Engineering, Pashchimanchal Campus, Institute of Engineering, Tribhuwan University, Nepal
b Department of Electrical Engineering, Pashchimanchal Campus, Institute of Engineering, Tribhuwan University, Nepal
c Electrical Engineer

! a dangishreedhar239@gmail.com, b sandeep.dhami@ioepas.edu.np, c rajtiwarib42@gmail.com, c rimalmilan@gmail.com

Abstract
In recent years the concern towards environment quality protection has become a burning topic and researchers are working
tremendously on protection of environment. Several concepts have been proposed justifying the acceptance of EV as a prime
solution and consequently the world has started switching towards the use of EV. As to charge these EVs, installation of EVCS
should be done technically and economically feasible. Mostly the EV charging station placed at radial distribution network are
installed without prior detailed system analysis. This paper primarily focuses on finding the optimal location based on minimizing the
active power loss along with favoring the weighted zones in the feeder, complying with different constraints like voltage regulation,
network loading capacity,and distance between two EVCS. The optimal placement problem of EVCS is optimized by using Particle
Swarm Optimization Algorithm. The optimization is performed for IEEE 34 bus system and real radial feeder called Khaireni Feeder.
Load forecasting is also performed during 2086 B.S and 2091 B.S for latter feeder. Various metrics like voltage profile, APL and
THD are assessed and compared for single and dual EVCS placement for 2081 B.S and 2086 B.S. From the results obtained single
and Dual EVCS placement are viable during 2081 B.S, while only single EVCS is feasible during 2086 B.S.
Keywords
Optimal location ,Particle Swarm Optimization, Active Power Loss,Voltage Sensitivity Factor

1. Introduction considered. Similarly, as EVCS are comprised of non-linear

loads, harmonic assessment is also required to further justify
In today’s world the air pollution and climate change due to the installation. So, THD analysis is also done in this paper at
use of gas-driven vehicle has drag people to switch towards different optimal locations at various scenarios.
using Electrical Vehicles. The IEA has presented its report that
globally use of EV stock will surpass upto 300 million by
2030[1]. The availability of EVCS is a mandatory for the 2. Literature Review
success of EV technology. Many Researcher work was done
and the resarchers come to a conclusion that the placement of The investigation for this paper started with a thorough
EVCS in random fashion in distribution network will result in analysis of the body of research on the subject of EV charging
very high demand of electricity which will negatively impact stations’ effects on power distribution feeders. This involved a
grid performance and degrades the voltage regulation,causes comprehensive review of research on load profiles, voltage
a lot of modification in load demand pattern resulting in stability, power quality, as well as an analysis of the
transformer overloading and increase power loss,line loading relationships between these elements and the incorporation
in the system[2].Also feeder capacity to transfer load and of EV charging station infrastructure . The basis for gaining a
reverse capacity of distribution grid substation can be thorough comprehension of the topic area was laid by this
decreased due to increase in system demand because of review. Placement of charging stations at the weak points of
plugging EV charger. Therefore to increase the penetration the distribution network will cause degradation of voltage
and obtain more technical and financial benefit determining profile, reliability indices. Moreover, the locations of charging
the best location for EVCS in a distribution network is prime stations must not be too far away from the point where
important [3]. Proper planning should be done for placement charging demand arises. CSO is novel swarm
of EVCS in right position so that negative impact on electrical intelligence-based bio-inspired algorithm mimicking the
network can be reduced. The optimal location of EVCS is also behavior of chickens in a swarm. ACO is a bio-inspired swarm
greatly affected by parameters like road network, land intelligence-based algorithm that mimics the trailing behavior
availability, number of EV users. Also as the EVCS placement of ants. TLBO is a nature-inspired algorithm mimicking the
is a fixed installation so its placement should also justify the teaching and learning process. The individual of the
installation in the near future when the load demand population having the best fitness value is assigned as teacher.
increases and the network topology changes. So, in this paper, LSA is a physics-based algorithm that mimics the
the viability of placement of EVCS is also assessed by load phenomenon of lightning [4].
forecasting the load demand in the radial feeder being
Venkata K. Babu Poonam and K. Swar Nasri addresses a multi

Pages: 1 – 8
Optimal Location of Electric Vehicle Charging Station on Khaireni Feeder-Lekhnath, Pokhara using PSO Algorithm

objective optimization technique to obtain simultaneous 4. Problem Formulation

EVCS and DG placement and sizing. The problem is
formulated to optimize real power losses, Average Voltage 4.1 Objective Function
Deviation Index (AVDI), and Voltage Stability Index (VSI) of
The addition of EVCS to RDS degrades the voltage profile and
the electrical distribution system. Simulation studies were
raises the system’s power losses and line loading. A number
performed on the standard IEEE 33-bus and 69-bus test
of factors are taken into account while designing an EVCS’s
systems. Harries Hawk Optimization (HHO) and
optimal location like EVCS placement area distribution, feeder
Teaching-Learning Based Optimization (TLBO) algorithms
capacity, channel capacity, power loss, and voltage regulation
were selected to minimize the system objectives [5]. Srinivas
etc. In this work, the active power loss and zone component is
proposes a heuristic algorithm called Particle Swarm
realized as objective function as defined in Equation 1 [11].
Optimization (PSO) to optimize the IEEE 33 bus radial
distribution system with electric vehicle charging stations. min {OF } = w 1 · (AP L) + w 2 · zone component (1)
The prime objective is to place the EVCS at optimal location in
existing radial distribution network by considering the real where w1 and w2 are the user defined weights,
(active) power losses and also the voltage at the buses of the w1= 0.8 and w2= 0.2
system [6]. Zone component are defined based on the types of load that
In order to facilitate the accessibility of EV’s to EVCS in any part the buses mostly have. Buses with high number of domestic
of the city, it assumed that EV should be able to drive with 5 loads are designated as urban zone, while buses with high
percent of their remaining charge to reach the nearest charging industrial loads are designated as industrial zone. Similarly
station. Therefore, the distance traveled by EV with 5 percent buses with commercial loads are designated as semi-urban
battery charge is about 10 km, so the distance between any zone. Classification as such is based on the thought that urban
point and the nearest charging station should not exceed 10km zone should be given highest priority for EVCS installation as
[7]. it serves more number of individual electrical consumers.
Similarly, the semi-urban zone serves lesser individual
Ming Dong proposes a hybrid modeling method using consumers and accordingly, the priority is set lower than the
sequence prediction for load forecasting of distribution feeder. urban zone. Finally the industrial zone are least favored for
The proposed method seamlessly integrates top-down, EVCS installation as it serve bulk load rather than distributed
bottom-up and sequential information hidden in multi-year individual consumers. The weights are assigned as defined in
data. Two advanced sequence prediction models Long Table 1.
Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU)
Table 1: Zone types and their respective weights
networks are investigated in the paper [8]. Tai-Hua Yangrui
Sun , Qiming Wei, And Yuqi Gao explains Saturated Demand Zone Type Weights
Forecast model that divides the growth process of power Urban 0.1
demand into three stages, proposes a modified self-adaptive Semi-Urban 0.2
Logistic model, and takes the power demand in East China as Industrial 0.7
an example to verify the accuracy and practicability of the
model, and carries out the saturation prediction research [9].
Mohamad Faizal Ishak and Jabbar Al-Fattah Yahaya identify 4.2 Constraints
and investigate the effects of harmonics when EV is connected There are two types of constraints: Equality constraints and
to the grid system and identify the most suitable method to Inequality constraints
mitigate the harmonics that occur due to the EV charging
station [10]. 4.2.1 Equality Constraints

The following operating conditions must be fulfilled during

3. Challenges of EVCS placement the optimization process [11].

Radial distribution system is widely preferred in Nepal due to P TotalLoss = P TotalGen → P TotalLoad (2)
its low initial cost and simple in planning and operation.
However, in lengthy feeder, voltage towards the farthest point
N!
EV C s
from the supply end suffers from low voltage due to voltage
P TotalLoad = P Loadbase + P EV C s,b (3)
drop along the length of the feeder. Meanwhile, installation of b=1
the EVCS is increasing annually in the same distribution
system with minimal or no improvement resulting more
voltage drop , power losses, line loading and THD in the 4.2.2 Inequality Constraints
system. In order to some how minimize the negative impact of
haphazard EVCS installation on the system, proper allocation a) Voltage Regulation constraints:
becomes crucial. Therefore, this study intended to focus on In order to keep the proper stable voltage magnitude of
enhancing performance of the EV charging station with the proposed distribution bus network, the absolute voltage value
perfect location in radial distribution system without violating at all nodes of the distribution system should meet the defined
the electrical parameters. constraints before and after EVCS placement [11].

Vreg mi n ↑ Vbus ↑ Vreg max (4)

2
 IOE Graduate Conference

Table 2: EVCs parameters[12] feasible in coming years. Load Forecasting is done using
Exponential decay Saturation Model for next 5 and 10 years.
Parameter Value Unit
Number of EVCs 2 unit
No of Level 2 Charger 2 unit 5.1 Voltage Sensitivity Factor (VSF)
Capacity of level 2 charger 22 kw
VSF is defined as the ratio of voltage change (dV) to the change
Level 3 Charger 4 unit
in active load(dP). It is the measure sensitivity of the system
Capacity of level 3 charger 60 kw voltage with stepwise loading increment[13]. Mathematically,
Total Capacity of EVCs 284 kw it is expressed as
where, " "
" dV "
Vreg min=0.95 V SF = "" " ↓P < P max (7)
dP "
Vreg max=1.05
High value of VSF indicates lesser voltage stability, that means
even with small changes on loading behavior, there is
b)Line Loading constraints:
significant change in voltage drop .
In a network, the feeder runs normally when the load is no
greater than 80 % .Therefore, the difference between 80 %
of the generation power and the base load is the maximum
capacity of the EVCS P t ot al EVCS [11].

Total
P EV C s ↑ 0.8P TotalGen → P Loadbase (5)

c) Distance constraints:
It is the constraints for optimal location of two EVCS in a
system.The D mi n and D max is the minimum and maximum
distance from first optimal location to second optimal
location for EVCS placement [7].

D mi n ↑ D EVCS ↑ D max (6)

where,
D mi n = 7 km
D max = 11 km

5. Methodology

Before implementing the placement problem in real Feeder ,

the placement assessment is done in benchmark system
called IEEE 34 Bus system. To perform placement
optimization, the respective line parameters and load
parameters are acquired. PSO algorithm is used for finding
optimal location of EVCS in IEEE 34 Bus radial system.
Optimization of EVCS placement is done considering two
scenarios i.e. Single EVCS placement and Two EVCS
placement. Similarly the EVCS placement is done in Khaireni
Feeder fed from the Lekhnath Substation, Lekhnath-Pokhara,
Nepal. To perform placement optimization, the respective line
parameters and load parameters are acquired. A Khaireni Figure 1: Flowchart of Optimal Location of EVCs using PSO
feeder model is modeled in ETAP as well, to perform Algorithm
Harmonic Analysis before and after EVCS placement. PSO
algorithm is used for finding optimal location of EVCS for both
scenarios as discussed earlier in IEEE 34 Bus system. Once the 5.2 Optimization Method
optimal locations are found for each scenarios, THD analysis
As an optimization tool, the PSO offers a population-based
is performed in ETAP by placing the EVCS in optimal locations.
search process wherein particles, or people, alter their
Further to perform the comparison the placement with worst
position (state) over time. Particles move through a
case scenario the EVCS placement is also done at most
multidimensional search space in a PSO system. Each particle
sensitive bus.
in flight modifies its position based on its own experience
Furthermore load forecasting is also done to forecast the load (P best ) and that of an adjacent particle (G best ) , using the best
demand in next five and ten years and assess whether the position that both the particle and its neighbor have
placement of EVCS at earlier optimal locations will be still experienced. Each particle moves in an N-dimensional search

3
Optimal Location of Electric Vehicle Charging Station on Khaireni Feeder-Lekhnath, Pokhara using PSO Algorithm

space with the position and velocity of a particle could be

updated to find optimal location.
Table 3: Parameters of PSO[14]

Parameter Description Value

matrix Maximum number of iterations 100
n-pop Size of swarm 30
w Inertia weight 0.73
c1 Cognitive acceleration coefficients 2.05
c2 Social acceleration coefficients 2.05

5.3 Load Forecasting Model

The fundamental principle of the exponential decay
saturation model is that growth rates tend to decrease over
time as systems mature. This is based on observations across
many utility systems where initial rapid growth eventually
moderates due to market saturation, efficiency improvements,
and physical limitations.
Mathematical Formulation The basic mathematical
formulation of the exponential decay saturation model is
defined in Equation in 8.

g t = g 0 ↔ ωt (8)

Where:

• g t is the growth rate at time period t

• g 0 is the initial growth rate

• ω is the decay factor (typically between 0.90 and 0.98)

• t is the time period (usually in years)

For load forecasting, we apply this decaying growth rate to

calculate future loads [15]:

t
#
L t = L0 ↔ (1 + g 0 ↔ ωi ) (9)
i =1

Where:

• L t is the load at time t

• L 0 is the initial load

$
• represents the product of all terms from i = 1 to t

6. System Under Study

The distribution system selected for EV impact research is the

feeder of the Pokhara Grid Substation located in
Lekhnath,Pokhara, Nepal as shown in Figure 2. The Khaireni
distribution feeder primarily uses mostly XLPE Covered Cable
of 100mm 2 and Rabbit conductor placed in horizontal Figure 2: Single Line Diagram of Khaireni Feeder
fashion. The distribution transformer’s position and line

4
 IOE Graduate Conference

length are extracted from the GIS route map using Arch map
software and site visit. The feeder is around 18 km long overall
and has a radial length. There are 45 distribution transformers
in Khaireni Feeder. In the feeder network, the transformers
are regarded as the buses or load points. 11 kV lines are used
as distribution lines, with transformers acting as load points
or buses and grid substations supplying the line as sources.
The three different zones are created based on distribution of
load with their weight value based on population densities, EV
Penetration,and land availability.
The EV load can be represented with different load model like
constant current, constant Power,constant Voltage,constant
impedance etc. For easy analysis in optimal location of EVCS
a constant power EVCS of capacity 142 kW which consists of Figure 4: APL of IEEE-34 Test Bus system in different cases
two 60kW level 3 charger and one 22kW level 2 charger is used.
EVCS placement at optimal locations i.e at Bus 2 and Bus 34 at
a distance of 7 km, the least voltage magnitude of the system
is 0.9634 pu which is at Bus 50. Again when a single EVCS
7. Simulation Results
placement is done at most sensitive bus i.e. Bus 50, the least
voltage magnitude of the system is 0.9608 pu.
7.1 IEEE 34-Test Bus System
The model validation of IEEE 34 test bus system is carried out
by evaluating the voltage regulation requirement. The voltage
profile of the system is depicted in Figure 4 and meets the
voltage regulation standard. The voltage profile of IEEE 34 test
bus system in Figure 3 shows that at base case, Bus 27 has
minimum voltage of 0.989081p.u. In this test bus system, the
optimal location of single EVCS is at Bus 13 which degrades
the voltage of Bus 27 to 0.988 p.u. Similarly, optimal location
of two EVCS is found to be at Bus 2 and 26 at a distance of 7.05
km which further degrades the voltage of Bus 27 to 0.986 p.u.

Figure 5: Voltage profile comparison of Khaireni feeder in

different cases

Figure 3: Voltage profile of IEEE-34 Test Bus system in

different cases

Figure 4 depicts the Active Power Loss of IEEE 34 Test Bus

system at various scenarios. This shows that the APL at Base
case is 40.15 kW which increases to 40.75 kW with single EVCS
penetration. Similary during two EVCS penetration the APL
increase to 43.77 kW which is justified as the system active
Figure 6: APL comparison of Khaireni feeder in various
power losses increases as the load increases.
scenarios

7.2 Khaireni Feeder Figure 6 shows that the Active Power loss increases from
38.79kW at base case to 39.96 kW at single EVCS penetration,
Figure 5 shows the voltage profile of khaireni feeder at various
48.59 kW at dual EVCS placement and finally 47.85 kW at
cases. At base case load flow, the minimum voltage is found
single EVCS placement at most sensitive bus. This result is
at Bus 50 which is 0.9673 p.u. With addition of single EVCS at
justified because as the system load increases, the active
optimal location i.e at Bus 2, the least voltage magnitude of the
power losses are bound to increase.
system is 0.9669 pu which is at Bus 50. Similarly, for two

5
Optimal Location of Electric Vehicle Charging Station on Khaireni Feeder-Lekhnath, Pokhara using PSO Algorithm

7.3 THD Analysis of Khaireni Feeder

The THD analysis of Khaireni feeder at different cases are
done in ETAP software. At first, the system’s THD is modeled
using various harmoinc sources defined in Typical-IEEE 6
pulse 1, Typical-IEEE 12 pulse 1, Typical-IEEE 6 pulse
2,Typical-IEEE Fluorescent, Rockwell 6 pulse VFD libraries. To
model the THD introduced by Charging system used in EVCS, Figure 11: Voltage Waveform with two EVCS installation
Typical Locomotive Current Source based library is used in
ETAP. The system waveform of voltage is shown in Figure 7.
Figure 8 depicts presence of frequencies other than
fundamental frequency in the system.

Figure 12: Harmonics content with two EVCS installation

Similarly, THD analysis is also done for EVCS placement at

Figure 7: System Waveform without EVCS installation most sensitive bus in the Khaireni Feeder. During the
placement of EVCS the THD at Bus 50 is found to be 2.58% .
The waveform of the voltage and frequency content at Bus 50
is shown in Figure 13 and Figure 14 respectively.

Figure 8: Harmonics presence in system without EVCS

installation

Figure 13: Voltage Waveform at sensitive bus EVCS

THD analysis is done for single EVCS placement in the Khaireni installation
Feeder in the optimal location. During the placement of single
EVCS at optimal location i.e. Bus 2 the THD at Bus 2 is found to
be 2.078%Ṫhe waveform of the voltage and frequency content
at the Bus 2 is shown in Figure 9 and Figure 10 respectively.

Figure 14: Harmonics content at sensitive bus EVCS

installation

Figure 9: Voltage Waveform with single EVCS installation

7.4 Load Forecasting of Khaireni Feder
A straightforward exponential decay saturation model is used
in the load forecasting analysis. The load demand trend for
each type of loads, i.e. domestic, commercial, industrial
ranging from 2076 B.S to 2081 B.S was acquired from Lekhnath
DCS. The load growth pattern of overall system in the past six
years is depicted in Figure 15.The growth rate was calculated
Figure 10: Harmonics content with single EVCS installation for 2086 B.S and 2091 B.S as defined in Table 4. The future load
demand at each buses of the feeder is calculated by simply
THD analysis is done for two EVCS placement in the Khaireni scaling the present load demand by respective year’s growth
Feeder at the optimal locations. During the placement of two rate.
EVCS at optimal location i.e. Bus 2 and Bus 34, the THD at Bus Table 4: Growth Rates for Different Load Types
2 and Bus 34 is found to be 2.34% and 2.56% respectively. The
waveform of the voltage and frequency content at the Bus 2 Load Type Growth Rate: 2086 B.S Growth Rate: 2091 B.S
and Bus 34 is shown in Figure 11 and Figure 12 respectively. Domestic 30.64% 62.85%
Commercial 40.36% 82.82%
Industrial -0.10% -0.18%

6
 IOE Graduate Conference

B.S and 2086 B.S. Though the optimal location is found

complying with the technical constraints, one of the optimal
location is changed to Bus 38 from Bus 32 during 2081 B.S.
This makes the optimal location practically infeasible as the
EVCS structures are fixed installation and aren’t movable.

Figure 15: Load Growth of Khaireni Feeder

7.4.1 Voltage profile and APL for future years

After load forecasting of khaireni feeder using exponential
decay saturation method, the optimal location for single EVCS
for next 5 year i.e 2086 B.S is obtained which came out as Bus 2
but optimal location for next 10 years is not obtained because Figure 18: Active Power loss with single EVCS of Khaireni
the voltage regulation doesn’t meet the regulation standard i.e. Feeder at forecasted years
voltage regulation constraint. The voltage profile of real feeder
after single EVCS placement in Khareni Feeder during 2086
B.S is shown in Figure 17 with comparison to single EVCS
placement in 2081 B.S.

Figure 16: Forecasted load of Khaireni Feeder

. Figure 19: Voltage profile of Khaireni Feeder with dual EVCS

at forecasted years

Figure 17: Voltage profile of Khaireni Feeder with single EVCS

at forecasted years
Figure 20: APL with dual EVCS of Khaireni Feeder at
Similarly, the APL of kahireni feeder with single EVCS at forecasted years
different forecasted years is depicted in Figure 18. Figure 18
shows that with optimal placement of single EVCS the APL
increases from 38.79 kW to 62.759 kW at 2086 B.S. 8. Conclusion
Furthermore, the optimal location of dual EVCS is evaluated The optimal location of EVCS placement is performed in
and came out at Bus 2 and Bus 38 at a distance of 9.32 km. Khaireni feeder of Lekhnath DCS,Pokhara within 11 KV
Voltage profile of 2086 B.S is compared with dual EVCS distribution line using PSO algorithm. It is concluded that the
placement during 2081 B.S as shown in Figure 19. Figure 20 network is within regulation limit even after placement of
shows the comparison of APL for two EVCS placement at 2081 single and two EVCs in feeder for single EVCS and two EVCS

7
Optimal Location of Electric Vehicle Charging Station on Khaireni Feeder-Lekhnath, Pokhara using PSO Algorithm

placement for 2081 B.S. Here the optimal location for single systems using metaheuristic optimization algorithms.
EVCS placement is obtained at Bus 2. For two EVCS placement Engineering, Technology & Applied Science Research, 2020.
optimal location is obtained at Bus 2 and Bus 34. VSF analysis [6] Dandu Srinivas and M. Ramasekhara Reddy. Optimal
shows that the most sensitive bus is Bus 50 which is also placement of electric vehicle charging station by
within the voltage regulation and line loading limit after considering dynamic loads in radial distribution system.
placement of single EVCS. Simlarly THD anlaysis for the single In 2022 International Conference on Automation,
Computing and Renewable Systems (ICACRS), pages
and dual EVCS placement during 2081 B.S complies within the
212–217, 2022.
THD regulation standard. During the forecasted year 2086 B.S,
[7] S. F. Keleshteri, T. Niknam, M. Ghiasi, and H. Chabok. New
only single EVCS placement is viable. For rest of other
optimal planning strategy for plug-in electric vehicles
scenarios; two EVCS placement during 2086 B.S and single charging stations in a coupled power and transportation
and Dual EVCS placement during 2091 B.S, reallocation must network. The Journal of Engineering, 2023, 2023.
be analyzed accordingly. [8] Ming Dong, Jian Shi, and Qingxin Shi. Multi-year long-
term load forecast for area distribution feeders based on
selective sequence learning. Energy, 206:118209, 06 2020.
References [9] Tai-Hua Yang, Rui Sun, Qiming Wei, and Yuqi Gao.
Saturated demand forecast of regional power grid based
[1] Yanning Li and Alan Jenn. Impact of electric on amended self-adaptive logistic model: A case study of
vehicle charging demand on power distribution grid east china. IEEE Access, 9:1190–1196, 2021.
congestion. Proceedings of the National Academy of [10] Mohamad Faizal Ishak and Jabbar Al-Fattah Yahaya.
Sciences, 121(18):e2317599121, 2024. Harmonic analysis and mitigation of mass adoption for
[2] Syechu Dwitya Nugraha, Mochamad Ashari, and electric vehicle charging station (100kw) in distribution
Dedet Candra Riawan. Hybrid genetic algorithm- system. 1:57–61.
modified salp swarm algorithm to determine the optimal [11] Syechu Dwitya Nugraha, Mochamad Ashari, and
location of electric vehicle charging station in distribution Dedet Candra Riawan. Hybrid genetic algorithm-
networks. IEEE Access, 12:132332–132343, 2024. modified salp swarm algorithm to determine the optimal
[3] Mishal Altaf, Muhammad Yousif, Haris Ijaz, Mahnoor location of electric vehicle charging station in distribution
Rashid, Nasir Abbas, Muhammad Adnan Khan, networks. IEEE Access, 12:132332–132343, 2024.
Muhammad Waseem, and Ahmed Mohammed Saleh. [12] Arnaldo Arancibia and Kai Strunz. Modeling of an electric
Pso-based optimal placement of electric vehicle vehicle charging station for fast dc charging. 2012 IEEE
charging stations in a distribution network in smart International Electric Vehicle Conference, pages 1–6, 2012.
grid environment incorporating backward forward [13] Sanchari Deb, Kari Tammi, Karuna Kalita, and
sweep method. IET Renewable Power Generation, Pinakeshwar Mahanta. Impact of electric vehicle
18(15):3173–3187, 2024. charging station load on distribution network. Energies,
[4] Sanchari Deb, Xiao-Zhi Gao, Kari Tammi, Karuna Kalita, 11(1), 2018.
and Pinakeswar Mahanta. Nature-inspired optimization [14] Adil Amin, Wajahat Ullah Khan Tareen, Muhammad
algorithms applied for solving charging station placement Usman, Haider Ali, Inam Bari, Ben Horan, Saad Mekhilef,
problem : Overview and comparison. Archives of Muhammad Asif, Saeed Ahmed, and Anzar Mahmood. A
Computational Methods in Engineering, 28(1):91–106, review of optimal charging strategy for electric vehicles
January 2021. under dynamic pricing schemes in the distribution
[5] Venkata Ponnam and Katragadda Swarnasri. Multi- charging network. Sustainability, 12(23), 2020.
objective optimal allocation of electric vehicle charging [15] H. Lee Willis. Spatial Electric Load Forecasting: Second
stations and distributed generators in radial distribution Edition. Marcel Dekker, New York, 2002.

8
 Page 1 of 51 - Cover Page Submission ID trn:oid:::3117:450461350

Shreedhar Dangi
Optimal Location of Electric Vehicle Charging Station on
Khaireni Feeder-Lekhnath, Pokhara using PSO
Tribhuvan University

Document Details

Submission ID
trn:oid:::3117:450461350 43 Pages

Submission Date 9,694 Words

Apr 19, 2025, 10:26 AM GMT+5:45

54,244 Characters

Download Date
Apr 19, 2025, 10:30 AM GMT+5:45

File Name
again.pdf

File Size
1.3 MB

Page 1 of 51 - Cover Page Submission ID trn:oid:::3117:450461350

 Page 2 of 51 - Integrity Overview Submission ID trn:oid:::3117:450461350

16% Overall Similarity

The combined total of all matches, including overlapping sources, for each database.

Match Groups Top Sources

117Not Cited or Quoted 15% 13% Internet sources
Matches with neither in-text citation nor quotation marks
13% Publications
0 Missing Quotations 0% 0% Submitted works (Student Papers)
Matches that are still very similar to source material

4 Missing Citation 1%
Matches that have quotation marks, but no in-text citation

0 Cited and Quoted 0%

Matches with in-text citation present, but no quotation marks

Integrity Flags
0 Integrity Flags for Review
Our system's algorithms look deeply at a document for any inconsistencies that
No suspicious text manipulations found. would set it apart from a normal submission. If we notice something strange, we flag
it for you to review.

A Flag is not necessarily an indicator of a problem. However, we'd recommend you

focus your attention there for further review.

Page 2 of 51 - Integrity Overview Submission ID trn:oid:::3117:450461350

 Page 3 of 51 - Integrity Overview Submission ID trn:oid:::3117:450461350

Match Groups Top Sources

117Not Cited or Quoted 15% 13% Internet sources
Matches with neither in-text citation nor quotation marks
13% Publications
0 Missing Quotations 0% 0% Submitted works (Student Papers)
Matches that are still very similar to source material

4 Missing Citation 1%
Matches that have quotation marks, but no in-text citation

0 Cited and Quoted 0%

Matches with in-text citation present, but no quotation marks

Top Sources
The sources with the highest number of matches within the submission. Overlapping sources will not be displayed.

1 Internet

elibrary.tucl.edu.np <1%

2 Internet

pubs.iscience.in <1%

3 Internet

www.sid.ir <1%

4 Publication

Maleki, Akbar, Mehran Ameri, and Farshid Keynia. "Scrutiny of multifarious partic… <1%

5 Internet

www.mdpi.com <1%

6 Internet

docplayer.net <1%

7 Publication

Bidhan Pokhrel, Nawraj Bhattarai, Ramhari Poudyal, Rupesh Gautam, Khem N. P… <1%

8 Publication

Ritu Kumari, Amit Kumar. "Vehicle-to-Grid Technology for Microgrid Frequency R… <1%

9 Publication

Sobhan Farjam Keleshteri, Taher Niknam, Mohammad Ghiasi, Hossein Chabok. "… <1%

10 Publication

Saman Kanzi. "Multi-Objective Optimisation Technique for Optimum Allocation of… <1%

Page 3 of 51 - Integrity Overview Submission ID trn:oid:::3117:450461350