INTEGRATION OF PHOTOVOLTAIC SYSTEM IN

EXISTING DISTRIBUTION NETWORK IN NEPAL

A DISSERTATION
Submitted in partial fulfillment of the
requirements for the award of the degree
of
MASTER OF TECHNOLOGY
in
ALTERNATE HYDRO ENERGY SYSTEMS

By
MEEN BAHADUR DHAMI

ALTERNATE HYDRO ENERGY CENTRE

INDIAN INSTITUTE OF TECHNOLOGY ROORKEE
ROORKEE-247 667 (INDIA)
MAY, 2017
 CANDIDATE’S DECLARATION

I hereby declare that the work which is presented in this dissertation report, entitled,
“Integration of Photovoltaic System in Existing Distribution Network in Nepal”,
submitted in partial fulfillment of the requirements for the award of the degree of Master
of Technology in “Alternate Hydro Energy Systems” in Alternate Hydro Energy
Centre, Indian Institute of Technology, Roorkee, is an authentic record of my own work
carried out during the period from June 2016 to April 2017 under the supervision of Dr. D.
K. Khatod, Associate Professor, Department of Electrical Engineering, Indian
Institute of Technology Roorkee, Roorkee (India).

I also declare that I have not submitted the matter embodied in this report for award
of any degree.

Date: ..., May, 2017 .........................................

Place: IIT, Roorkee (Meen Bahadur Dhami)

CERTIFICATE

This is to certify that the above statement made by the candidate is correct to the best of
my knowledge.

.........................................

(Dr. D. K. Khatod)
Associate Professor,
Department of Electrical Engineering,
Indian Institute of Technology,

I
 ACKNOWLEDGEMENT

Firstly, I would like to express my sincere gratitude and appreciation to my

respected guide, Dr. D. K. Khatod, Associate Professor of Department of Electrical
Engineering for the continuous support of graduate studies. I am highly indebted for his
guidance, supervision, and help in providing all the necessary information and support
during this study. His creative thinking, motivation and immense knowledge have been
source of inspiration throughout this work.
I would like to express my gratitude to the faculties and staffs of Alternate Hydro
Energy Center for their kind co-operation and encouragement in the present work.
My sincere thanks also go to my friends and colleagues for their unlimited support
and having contributed to a good working environment. My special thanks go to the
Department of Hydrology and Metrology (DHM), Nepal and Nepal Electricity Authority
(NEA) for providing data for this dissertation work.
I express my thanks to ITEC program, Government of India for facilitating the M.
Tech. program. In addition, I would like to express my thanks to Government of Nepal,
who nominated me for this program.

(Meen Bahadur Dhami)

Date: - ... May,
2017

II
 ABSTRACT

Solar photovoltaic (SPV) has become a significant part of power generation in

many countries during recent years. This is a green and clean form of energy and hence is
the most popular alternative source of energy in the world. Through its policy, Government
of Nepal has recently declared to reduce present energy crisis by searching some quick
possible options of production of electricity like SPV. Solar photovoltaic (SPV) as
distributed generation (DG) in existing grid will minimize power loss, improve voltage
profile and power quality in the distribution system. But, inappropriate DG allocation in the
system may react adversely. Therefore, optimal location and size of DG plays important
role in the network.
In this dissertation work, the placement and sizing of PV in distribution network is
determined using optimization. Seasonal average hourly solar potential and load profile for
the four seasons are determined based on available data. Based on capacity factor (CF),
suitable solar PV module is selected.
Multi-objective optimization problem is formulated to minimize energy loss and to
minimize cost of SPV in a radial distribution system. The constraints are the size of SPV,
bus voltage, line current and maximum SPV penetration. The proposed methodology uses
Genetic Algorithm (GA) to optimize the multi-objective function. This technique increases
the diversity of the optimization variables and does not trap in a local minimum. In
addition, a fuzzy set theory is used to select the best compromising solution from the
Pareto-optimal set obtained from GA. To validate this method, an 11 kV, 21 bus radial
distribution system is selected which is situated in Kathmandu, Nepal. The obtained results
illustrate the effectiveness of the proposed method.

III
 TABLE OF CONTENTS

CANDIDATE’S DECLARATION I
CERTIFICATE I
ACKNOWLEDGEMENT II
ABSTRACT III
LIST OF FIGURES VII
LIST OF TABLES VIII
LIST OF ABBREVIATIONS IX

Chapter - 1 INTRODUCTION 1

1.1 General 1
1.2 Distributed Generation 1
1.2.1 Definition 1
1.2.2 Benefits of DG 2
1.2.3 Adverse effects of DG 3
1.3 Solar Photovoltaic 3
1.3.1 Solar photovoltaic cell 4
1.3.2 Types of solar PV system 7
1.3.2.1 Isolated solar PV system 7
1.3.2.2 Grid-connected solar PV system 8
1.3.3 Photovoltaic power output 9
1.4 Distribution system 9
1.5 Dissertation objectives 10
1.6 Organization of report 11

Chapter - 2 LITERATURE REVIEW 12

2.1 General 12
2.2 Review of literature 12
2.2.1 Solar photovoltaic related studies for Nepal 12

IV
2.2.2 Review of solar cell material 13
2.2.3 Review of load flow method for distribution systems 14
2.2.4 Review of impact of photovoltaic on grid 15
2.2.5 Review of optimization for distributed generation sizing and locating problems 16
2.3 Gaps identified 21

Chapter - 3 PROBLEM FORMULATION AND SOLUTION ALGORITHM 22

3.1 General 22
3.2 Problem Statement: 22
3.3 Analysis of distribution system for load flow 23
3.4 Mathematical modeling of distribution line 23
3.4.1 Load flow analysis 26
3.5 Problem formulation 27
3.5.1 Formulation of optimization problem 27
3.5.2 Constraints 29
3.5.2.1 Equality constraints 29
3.5.2.2 Inequality constraints 29
3.5.2.3 Constraints Handling 30
3.6 Genetic Algorithm 31
3.6.1 Introduction: 31
3.6.2 Outline of Genetic Algorithm 31
3.6.3 GA based Solution Methodology 33
3.6.3.1 Coding of the solution 33
3.6.3.2 Objective Functions 34
3.6.3.3 Selecting the best result 34
3.6.3.4 Setting of parameters 35

Chapter - 4 SOLAR ENERGY RESOURCE ASSESSMENT 36

4.1 General 36
4.2 Estimation of solar radiation on tilted surface 36
4.3 Selection of solar PV unit 40
4.3.1 Capacity Factor (CF): 42

V
Chapter - 5 RESULT AND DISCUSSION 44

5.1 General: 44
5.2 Case Study: 44
5.3 Result and Discussion 47
5.3.1 Result of energy loss minimization 48
5.3.2 Result of cost minimization 51
5.3.3 Result of Multi-Objective Optimization 51
5.3.4 Result Analysis 56

Chapter - 6 Conclusion and future work 57

6.1 General 57
6.2 Conclusion 57
6.3 Future Work 58

REFERENCES 59

Appendix: A 64

Base case load flow program in MATLAB 64

Appendix: B 68

Single objective optimization using MATLAB GA Toolbox 68

Appendix: C 73

Multi-objective optimization using MATLAB GA Toolbox 73

VI
 LIST OF FIGURES

Fig. 1.1: Operation of PV cell 4

Fig. 1.2: Mono and poly crystalline silicon PV cell 5
Fig. 1.3A: Solar PV panel 5
Fig. 1.3B: Solar PV Array 5
Fig. 1.4: I-V and P-V characteristics curve of a SPV module 6
Fig. 1.5A: I-V curve for various irradiances at constant temperature 6
Fig. 1.5B: I-V curve at constant irradiance with varying temperature 6
Fig. 1.6: Isolated solar PV system 7
Fig. 1.7: Components of a grid-connected system 9
Fig. 3.1: Single line diagram of a distribution line 23
Fig. 3.2: Distribution grid with solar PV generator 25
Fig. 3.3:Genetic Algorithm Scheme. 32
Fig. 4.1: Seasonal variation of solar radiation in horizontal surface 37
Fig. 4.2: Seasonal global solar radiation in tilted surface. 39
Fig. 4.3:Daily solar PV power output of a 1.0 kWp PV array 42
Fig. 5.1: Distribution system with 21 bus 43
Fig. 5.2: Seasonal load profile for distribution network. 45
Fig. 5.3: Fitness function evaluation for 21 Bus 46
Fig. 5.4: Optimal size and allocation of PV units for 21 Bus 47
Fig. 5.5: Comparison of voltage profile with PV during maximum load condition 48
Fig. 5.6: Multi-Objective Pareto front representation 49
Fig. 5.7: Effect of PV capacity of the energy losses 52
Fig. 5.8: Effect of PV capacity on PV cost 53

VII
 LIST OF TABLES

Table 4.1: seasonal average global solar radiation (Wm-2) on horizontal surface 36
Table 4.2: Specification of different modules 40
Table 4.3: CF of the PV modules available 42
Table 5.1: Line Data and Load Data of the distribution system 44
Table 5.2: Parameters of Genetic Algorithm 46
Table 5.3: Result for PV size and location 47
Table 5.4: Annual total energy loss reduction with PV 48
Table 5.5: PV location and size for cost minimization 49
Table 5.6: Summary of multi-objective optimization results 52
Table 5.7: The best result based on fuzzy theory 53

VIII
 LIST OF ABBREVIATIONS

ACO: Ant Colony Optimization

AI: Artificial Intelligent
BIBC: Bus injection to branch-current
CBF: Cost Benefit Factor
CF: Capacity Factor
DE: Differential Evolution
DER: Distributed Energy Resources
DG: Distributed Generation
DHM: Department of Hydrology and Metrology
EA: Evolutionary Algorithm
EP: Evolutionary Programming
FCM: Fuzzy C Mean
FF: Fill Factor
GA: Genetic Algorithm
GoN: Government of Nepal
HS: Harmony Search
IA: Immune Algorithm
I-V: Current-Voltage
LD: Line Data
LP: Linear Programming
MINLP: Mixed Integer Nonlinear Programming
MPP: Maximum Power Point
MPPT: Maximum Power Point Tracker
NB: Number of Buses
NEA: Nepal Electricity Authority
NLP: Nonlinear Programming
NSGA: Non-dominated Sorting Genetic Algorithm
O&M: Operation and Maintenance

IX
OF: Objective Function
PSO: Particle Swarm Optimization
P-V: Power-Voltage
RES: Renewable Energy Sources
RET: Renewable Energy Technology
SA: Simulated Annealing
SPV: Solar Photovoltaic
STC: Standard Test Condition

X
Chapter - 1 INTRODUCTI

ON

1.1 General

Nowadays, life without electrical energy cannot be imagined. This is the

most essential daily needs for human life. The global energy consumption is increasing
day by day to improve the living standards of human life; however, the conventional
source of energy is on the verge of extinction. Present energy demand of Nepal has
been fulfilled from traditional sources of energy. One of the major sources of electrical
energy in Nepal is to develop hydropower plant, but it has many barriers such as
economic and technical constraints. Therefore, it becomes necessary to look for
alternative sources for electrical energy production such as wind and solar. Through its
policy, Government of Nepal has recently declared to reduce load shedding by
searching quick possible options for production of electrical power. One of the great
opportunities to Nepal is that it has large amount of solar energy incidence on the
territory which can partially fulfill the gap between demand and supply[ CITATION
Adh12 \l 1033 ]. Knowing that fact, Government of Nepal has recently shown the
interest in solar energy system, particularly photovoltaic grid integrated system to
reduce energy shortage in a quick way without polluting the environment. In addition,
that, the present electrical networks are operating in their full capacity and further
loading them is very difficult. To reduce this type of problem distributed generation are
necessary to install at the load center.

1.2 Distributed Generation

1.2.1 Definition
Distributed generation (DG) is small power generating unit that serves
electricity closer to the consumer load. The main sources of distributed generator are
solar photovoltaic, small hydropower, wind energy, biomass, etc.[ CITATION
Ack01 \l 1033 ]. Nowadays, it has gained many popularity in power sector due to its
1
ability in voltage improvement, power loss reduction, low investment cost, and
increased reliability. DG can be used as in isolated way or integrated with grid. In
distribution system, DG may provide benefit to the consumers as well as to utilities.
According to [ CITATION Jha14 \l 1033 ], the DGs can be classified based on source
of energy and size.
Based on energy source employed in power generation, the DG are categorized
as renewable DG and non-renewable DG. The renewable DGs includes small
hydropower, solar power, wind, biomass, geothermal, etc. While non-renewable DGs
are diesel, gas or natural gas which cannot be replenished. Based on capacity of
generation the distributed generation belongs the size smaller than 50MW[CITATION
Shi16 \l 1033 ].
Integration of DG requires to analyze several factors, such as: technology to
be used, numbers and capacity of DG, best location and type of network connection.
The operating characteristics of the network should be evaluated properly and the
impact of DG addition should also be evaluated. Otherwise, improper location and
sizing of DG may lead to higher system loss, increase in voltage and increase in
investment cost which is opposite as desired. Due to this reason, the proper
optimization method capable for the best solution in the given network is useful for the
planner.

1.2.2 Benefits of DG
The role of distributed generation (DG) is increasing day by day in the
emerging modern power systems. The main reasons for usage of DG are as follows:
 Decrease the overall system losses.
 Decrease transmission and distribution related costs.
 Easy operation and maintenance.
 Generally, gestation period is shorter and investment risks are relatively low
for distributed generation (DG) plants.
 Improve both the reliability and efficiency of the system.
 Improve the system voltage and load factor.
 It is easier to find the location for DG installation for small generators.
 Latest technology available for wide range of capacity.
 More flexibility in energy management system.

2
  Renewable energy based DGs are more environmental friendly compared to
coal, oil or gas fired power plants.

1.2.3 Adverse effects of DG

Despite the numerous benefits of DG, there are certain drawbacks also:
 The intermittency of the renewable energy based DG causes the instability
to the system.
 Consumers that uses solar PV connected network are likely to be blackout
in night time because PV does not supply electricity at night time.
 Voltage variations and power fluctuations and hence decreases quality of
power and increases instability in the system.
 Power demand requires both active and reactive power but most of the DGs
like SPV supply active power only.
 Integration of DG changes the direction of power flow; hence potential
effects on the metering and protection systems.
 Inappropriate location and sizing of DG may degrade the quality of the grid.
 Solar and wind based DG units are highly variable due to weather
conditions and hence these types of DG units do not support in base load or
peak load.
 Increases losses and power factor modification.
 SPV like DG increases total harmonic distortion.

1.3 Solar Photovoltaic

Nowadays, solar photovoltaic (SPV) is the mostly used alternative source of

energy in the globe[ CITATION Sha15 \l 1033 ]. There are many areas where plenty of
reliable solar radiation is available but not the electricity. Therefore, utilization of solar
photovoltaic is the good solution to provide electricity for such areas.
Solar photovoltaic (SPV) systems are sustainable environmental friendly
distributed source of energy. There potential can be utilized where the conventional
power grid is absent. Technically, they can be installed near load centers, providing
benefits to the consumers that bulk power generation cannot be installed. In addition,
SPV energy system is fuel free, low maintenance, and noiseless. Its gestation period is
3
very short[ CITATION Ene11 \l 1033 ]. It is highly mobile and portable because of its
light weight. It also presents a high-power capability per unit of weight. The SPV
power output matches very well with peak load demands. However, solar PV power is
highly intermittent in nature due to variations in solar radiation and temperature levels.
Due to this reason, where the solar power is directly connected to the load, the
delivered power may not be reliable.

1.3.1 Solar photovoltaic cell

The sun delivers its energy in the form of light and heat. In solar PV based
generation, sunlight is directly converted into the electricity by using solar cell which
is called photoelectric effect. A solar cell is a diode that converts solar radiation into
electricity without any form for mechanical or thermal interlink. A PV cell is made of
silicon material having similar properties to p-n junction diode[ CITATION Nav \l
1033 ]. Because of the positive charge on one side and negative charge on the other
side, movement of this charge in the material creates an electric field in the material.

Fig. 1.1: Operation of PV cell

When solar radiation strikes on the PV cell, electrons are dislodged from
depletion region. If an electrical conductor attached to the negative and positive
terminals of the cell, forming an electrical circuit, electrons will flow in the form of
electric current that is electricity. The electricity thus generated is in the form of direct
current (DC) as shown in Fig. 1.1[ CITATION Boy09 \l 1033 ]. An inverter, a power
electronic device that transforms dc power generated from PV array into ac power that
is the most useful form of power used in ordinary electrical equipment.
Solar photovoltaic cells are made of thin coating of light-sensitive
semiconductor materials that use photons to dislodge electrons to drive an electric

4
current. There are two types of technology used for PV cell: crystalline silicon, as
shown in Fig. 1.2; and thin film, which is newer and growing technology.

Fig. 1.2: Mono and poly crystalline silicon PV cell

Mono crystalline silicon PV cells are produced by melting high-purity silicon

crystal of diameter 150 mm and thickness 150-200 microns. Poly crystalline silicon
cells are manufactured by pouring pure silicon into a mold and crystalized. Poly
crystalline silicon PV cell are less expensive than the mono crystalline cell
[ CITATION Tan \l 1033 ]. Thin film solar cell is fabricated by forming a stack of thin
continuous layers of silicon wafer barely 0.3 to 2 micrometers thick onto glass or
stainless steel substrates. The cost of raw material used in thin film solar cell is lower
than the capital cost of processing and equipment because the raw material used for
layers are so thin.

5
 Fig. 1.3A: Solar PV panel Fig. 1.3B: Solar PV Array

A PV module is formed by interconnecting individual cell in series and/or

parallel combination as shown in Fig. 1.3a. The combination of PV module is PV array
which is shown in Fig. 1.3b. The output power depends on the way of connecting the
multiple PV modules inside an array. The PV modules are connected in parallel to
increase the current and PV modules are connected in series to increase the voltage. To
increase the output power modules are combined in series and parallel configuration.

Fig. 1.4: I-V and P-V characteristics curve of a SPV module.

The V-I and P-V characteristic curve for a typical value of solar radiation and a
fixed temperature is shown in Fig. 1.4. The curve shows the behavior of the PV system
under all load conditions from open circuit to short circuit. The current-voltage
characteristic curve of a SPV cell/module is non-linear which makes it difficult to
determine the maximum power point[ CITATION Sur14 \l 1033 ]. Therefore, a
maximum power point tracker (MPPT) is required so that the PV array always output
maximum power. This will increase the efficiency and reliability of the PV array.

6
 Fig. 1.5A: I-V curve for various Fig. 1.5B: I-V curve at constant

irradiances at constant temperature. irradiance with varying temperature

PV modules in the field are not working on the standard test condition (STC)
i.e. cell temperature 25oC, solar radiation 1000 W/m2 and spectrum distribution with an
air mass AM=1.5. The above Fig. 1.5A & Fig. 1.5B show the effect of solar radiation
and temperature on solar PV modules. The short circuit current varies proportionally
with the effective solar radiation (H) while the open circuit voltage varies
logarithmically with H for constant temperature. As the cell temperature of PV module
increases, the open circuit voltage falls drastically while the short circuit current has
minimal increment at fixed solar radiation (H).

1.3.2 Types of solar PV system

Based on the end use application, the solar photovoltaic systems are classified
as: isolated and grid connected solar PV systems.

1.3.2.1 Isolated solar PV system

7
 Fig. 1.6: Isolated solar PV system

Isolated solar photovoltaic (SPV) system is sometimes called Off-grid solar PV

systems. The general layout of off-grid solar system is shown in Fig. 1.5
above[ CITATION Tan \l 1033 ]; this type of solar system is usually used where there
is absence of utility grid such as rural areas or off-shore islands. But they may also be
installed in urban areas in case where it is inconvenient or too costly to tap electricity
from the grid. In many countries, traffic lights are powered by stand-alone solar PV
systems[ CITATION Tan \l 1033 ]. The stand-alone solar PV system requires battery
back up to store electricity and can provide electricity to users when there is no
sunshine such as at night or during cloudy weather. The deep cycle batteries are used
to store electricity for off-grid systems that may store more electricity for longer
duration of time such type of batteries are lead-acid, nickel-cadmium or lithium-ion
batteries.
The main components used in off-grid solar PV system are solar array, charge
controller, battery banks, and loads. Charge controller plays major role in this type of
solar system to control the charge of the battery.

1.3.2.2 Grid-connected solar PV system

In this system, the power generated by the PV array can be either used by the
consumers or injected into the utility grid. Grid-connected solar photovoltaic (SPV)
system reduces the dependency on the grid and gives financial benefits to the users by
injecting the generated power into the grid. In a grid-connected PV system, the grid
acts as an unlimited energy storage capacity. This increases the overall reliability and
efficiency of the system. This type of solar system gives benefit to the grid as well as
users. During day-time when there is excess power generated from the solar system,
the power can be delivered to the grid and when there is shortage of power generation
electricity may be taken out from the grid. A typical grid-connected solar system is
shown in Fig. 1.6 below that consists of the following components:
 PV Module/array: this convert sunlight directly to electricity.
 Energy storage (optional)

8
  Inverter: converts the DC current generated by the solar PV modules to
AC current.
 Transformer (for Voltage step/down)
 Utility Grid

Fig. 1.7: Components of a grid-connected system

When
photovoltaic
system is connected to the power system, it is necessary to analyze the
dynamic study of the entire system. Inverter plays an important role in
the design of grid-connected PV system, because it acts as an interface between the PV
system and utility grid. The main function of inverter is that it helps in conditioning the
power output of the PV array and simultaneously provides a control that helps in
allowing the PV generated power to the grid.

1.3.3 Photovoltaic power output

The output power generated from the solar photovoltaic system depends upon
the active area of solar photovoltaic generator and the solar incident radiation on the
solar PV array. The power output, P, for a solar PV module can be determined as
below:
P=ηPV HA

(1.1)

where, η PV is the efficiency of the solar cell and

H is the total solar radiation on PV array and A is the area of the solar module
The efficiency of solar module is given as:

output V × I
efficiency ( η PV )= = (1.2) V and I are the voltage and current
input H × A

produced by the module respectively.

9
1.4 Distribution system

The electric power system consists of generating stations where primary energy
is converted into electric power, transmission lines and distribution systems. A
distribution system is the final stage in power delivery network which carries
electricity from the transmission system to the individual consumers. Industrial and
large commercial users usually get three-phase supply directly from the primary
distribution feeder, as they have their own transformers and in certain cases can
directly utilize the higher voltages. However, for the remaining consumers, who
generally require only single-phase power, power is usually transmitted for the last
half-mile or so over lateral feeders that carry one phase. A distribution transformer,
typically mounted on a pole or located underground near the customer, steps this
voltage down to the secondary distribution level, which is safe.
The power distribution system is generally operated in radial configuration
where there is only one source and power radiates from source to the load, but
integration of solar photovoltaic in the existing system will change the direction of
power flow which makes difficult for planning and operation[ CITATION Ten03 \l
1033 ]. In addition, that, other parameters such as voltage control, power quality,
protection system, faults and system losses also get affected.
Generally, load flow studies are performed to determine the normal steady-state
operation of the power system. A load flow study uses simplified diagram represented
by per unit (pu) system and focuses on various forms of ac power (i.e. voltage, voltage
angle, real power and reactive power). The distribution networks have the following
features:
 Radial or weakly meshed networks
 High R/X ratios
 Multi-phase, unbalanced operation
 Unbalanced distributed load
 Distributed Generation
The distribution system has high R/X ratio and hence higher voltage drop. Due
to this nature of the distribution network the conventional load flow methods are not
satisfactory.

10
1.5 Dissertation objectives

The main objectives of the dissertation report are followings:

1. To identify and select an existing distribution network in a particular site in
Nepal and to investigate the potential of solar energy over the area, by
identifying and analyzing the solar radiation and geographical data.
2. To develop a suitable objective function to minimize the energy loss and
cost of installation of PV system and to describe an optimal planning
method to determine the size and configuration for the PV system using
multi-objective Genetic Algorithm (GA).
3. To analyze the integration effect of solar PV systems on the voltage profile
of the network.

1.6 Organization of report

This dissertation report is organized in a total of seven (7) chapters. These are
introduction, literature review, problem formulation and solution algorithm, solar
energy resource assessment, result and discussion, conclusion and future work, and
references.

Chapter 1 introduces basic information about the distributed generation (DG)

and its impact on distribution network. It also gives an overview of solar photovoltaic
system. This chapter discusses location and sizing as well as contributions and
objectives of dissertation. Chapter 2 discusses literature reviews related to the
integration of distribution generation. This chapter gives some issues related to the
photovoltaic system in Nepal, power flow analysis, different optimization techniques
that were used earlier for cost of DG placement, sitting and sizing, case studies, and
software tools used.

Similarly, chapter 3 gives the problem formulation and solution algorithm.

Chapter 4 gives details about solar energy potential assessment. Chapter 5 gives the
results obtained from the simulations for the case study. Chapter 6 gives the conclusion
and future scope of work.

11
Chapter - 2 LITERAT

URE REVIEW

2.1 General

This chapter reviews literature studies and work done on previous similar
studies for the harnessing and development of solar photovoltaic system. An extensive
literature search was conducted to collect the available information on integration of
photovoltaic in existing grid.

2.2 Review of literature

Some of literature surveys are presented below:

2.2.1 Solar photovoltaic related studies for Nepal

Jha [ CITATION Jha14 \l 1033 ], presented the potential status of renewable
energy based DG in Kathmandu. He identified potential of the renewable energy based
DG like PV, wind, micro-hydro on hourly based potential for different seasons and
succeeded to achieve the optimal location and size of DGs in the distribution network
to minimize energy loss and DG cost. In this study, 12 bus distribution was selected
and obtained results were tested effectively using genetic algorithm tool.
Adhikari et al.[ CITATION Adh13 \l 1033 ], used daily global solar radiation,
sunshine hours and meteorological data for Jumla, Pokhara, Kathmandu, and
Biratnagar to derive the regression constants for the estimation of global solar radiation
for those selected sites. They found that high altitude location (Jumla) has high
potential of global solar radiation and low altitude location (Biratnagar) has least value
of solar radiation.
Bhandari & Stadler [CITATION Bha11 \l 1033 ], suggested that solar PV
system is the convincing alternative energy sources for rural electrification in Nepal.
Because of the promotional program from the government as well as from the private
sector, the PV system expansion would be increased in future until the government
continues provide subsidy for the PV system.

12
 Pokharel [ CITATION Pok03 \l 1033 ], introduced the main characteristics of
alternative energy technologies used in Nepal and discussed the barriers for the
promotion of these technologies. Gurung et al.[ CITATION Gur12 \l 1033 ], reviewed
about the renewable energy situation in Nepal and highlighted the policies and
subsidies for the optimal utilization of the REs in rural and isolated communities.
K.C. et al.[ CITATION KCS11 \l 1033 ], presented the status of renewable
energy technologies (RETs) such as micro-hydro, solar power, wind energy, bio-fuel
etc. and highlighted the opportunities and barriers for the RETs.

2.2.2 Review of solar cell material

Boylastad and Nashelskey [ CITATION Boy09 \l 1033 ], presented the
principle of photovoltaic action. When a photon from Sun’s radiation hits a p-n
junction and an electron is dislodged across the depletion region in the n-side and
accelerated by the field and pushed towards the n-region and hole is pushed towards
the p-region which creates the depletion region. The operation of solar cell depends
upon the action of solar irradiation and the temperature. They discussed the
characteristics of PV cell that the increase in irradiation causes an increase in the
current and a decrease results in decrease of the current. This is because the current
produced is directly proportional to solar radiation. Solar cell temperature has an
impact on the voltage of the cell. As the temperature increases, the voltage reduces.
The overall operation of a solar PV cell is affected by its efficiency, irradiance and
temperature. These parameters cause changes to the power output of the cell and
consequently the maximum power point (MPP) by Suri et al.[ CITATION Sur14 \l
1033 ].
Tan and Seng [ CITATION Tan \l 1033 ], presented about the material used in
solar cell that is silicon and semi-conductive in nature. The materials in use today for
majority of solar cells are mono-crystalline, poly-crystalline and amorphous silicon by
Dubey et al. [ CITATION Dub12 \l 1033 ].
Pihl [ CITATION Phi09 \l 1033 ], gave the importance of the solar power, an
important renewable energy source, which is being increased in the most part of the
world. He also highlighted about the principle of solar thermal power plant technology
and the technology uses for the photovoltaic effect known as solar photovoltaic power.

13
In a solar PV system, a solar cell is used to convert energy in sunlight into a direct
current (dc).

2.2.3 Review of load flow method for distribution

systems
Teng [ CITATION Ten03 \l 1033 ], proposed a direct method for load flow in
unbalanced three-phase distribution load using two derived matrices, the bus-injection
to branch-current (BIBC) matrix and the branch–current to bus-voltage (BCBV) matrix
and multiply those matrices to obtain load flow solutions. He used special topological
nature of the distribution network and he managed to achieve direct solution. The
proposed method is less time-consuming and useful for distribution systems.
Rupa and Ganesh[ CITATION Rup14 \l 1033 ], proposed a backward/forward
sweep method to determine the power flow in distribution network. They determined
power losses for each branch and voltage magnitude of each bus. This method has been
tested on 33 bus radial distribution system using MATLAB.
According to Haque [ CITATION Haq96 \l 1033 ], several methods have been
developed to solve the load flow problem of a high voltage system but these methods
may not applicable for a low voltage distribution network because a distribution
system if fed at one point and the branches of the system have a wide range of R and X
values. Also, the R/X ratios of branches in a distribution system are relatively high
compared to a transmission system. That is why the conventional Newton-Raphson
Method, Gauss Seidel Method, Fast Decoupled load flow method and their
modifications are not suitable for the distribution system.
Meerimatha et al.[ CITATION Mee15 \l 1033 ], recommended a novel load
flow method in distribution system using forward backward matrix method. The
characteristics of distribution network was fully utilized to make direct solution
possible. In forward sweep a voltage drop calculation is done with branch currents.
Then, nodal voltages are updated in a forward sweep starting from source node to the
last node assuming constant power flow in the branches. During Backward sweep the
current or power flow solution is obtained with possible voltage updates.
Bhujel et al.[ CITATION Bhu12 \l 1033 ], proposed a simple model to study
the load flow analysis for a typical radial distributed system including the distributed
generation connected to the buses modeled as PV bus and negative PQ buses. This
method is basically a modified form of backward/forward sweep method for a radial

14
distribution system and distributed generations (DGs) were considered to test the
effectiveness of the proposed model.

2.2.4 Review of impact of photovoltaic on grid

According to Suri et al. [CITATION Sur14 \l 1033 ], it was discovered that
cloud cover has a huge impact on solar PV output and the variability of clouds causes
variable output of the power from the PV generator. Furthermore, Paulescu et al.
[ CITATION Pau13 \l 1033 ], explained the modeling of the systems for better
operation with varying climatic conditions. Therefore, it is very important to consider
the variability of cloud cover as an important parameter on the power production when
studying the effects or impacts of integrating solar PV power to a grid.
According to studies conducted by Widen et al. [ CITATION Wid10 \l 1033 ],
the integration of solar PV has a lot of challenges from the systems modeling and
simulation point of view. The study looked at the stochastic processes study of the
impact of solar PV power voltage levels with extensive use of MATLAB. This study
along with Solanki et al. [ CITATION Sol12 \l 1033 ], presented the effects of
photovoltaic integration on the grid losses. Different penetration levels were used to
determine the impact of increased PV penetration on grid losses. These studies showed
that the system losses would decrease at lower photovoltaic penetration and increase
with increased penetration. This is because reverse power flow increases the feeder line
loading and consequently the losses. In addition, it could interfere with protection
operations in the grid.
Similarly, Shaha et al. [ CITATION Sha15 \l 1033 ], and Schoene et al.
[ CITATION Sch13 \l 1033 ], had presented issues that are related with integration of
photovoltaic to a grid. One of the challenges of photovoltaic integration discussed in
the papers was voltage level and profile. The integration could cause an increase in
voltage level and profile but dependent on the grid configuration and how it is from the
source of main power.
According to Willis [ CITATION Wil04 \l 1033 ], feeders away from the
substation experienced lower voltage than the sending end voltage. Placement of SPV
like DG in this type of distribution system improves the voltage profile along the
feeders is not the ultimate factor in determining the maximum penetration level. Losses

15
and line loading are very important parameters of a power system. Photovoltaic
integration to a grid has potential to alter the two parameters. Transmission lines and
feeders have a limit to which they can be loaded. Additionally, losses can either reduce
or increase the operation costs of a distribution grid.
Enernex [ CITATION Ene11 \l 1033 ], presented the feeder line loading in a
radial grid. With distributed photovoltaic, the line loading reduces with increased
photovoltaic penetration until reverse power flow occurs. Reverse power flow
increases the line loading with additional increase in penetration levels.

2.2.5 Review of optimization for distributed generation

sizing and locating problems
Shivarudraswamy et al.[CITATION Shi16 \l 1033 ], presented a technique for
optimal location and sizing of distributed generation (DG) in distribution systems
considering real time scenario. Using this method, they managed to achieve an
acceptable limit of reliability and voltage profile. This method has been evaluated
using 14 bus distribution network and optimization method has been implemented
using Genetic Algorithm (GA) technique and for time varying loads.
The study by Widen et al.[ CITATION Wid10 \l 1033 ], showed that voltage
level and profile other parameters, for example, line loading and losses, should be
analyzed together with voltage level and profile in determining the capacity of grid.
Helal et al. [ CITATION Hel12 \l 1033 ], presented a methodology for the
optimal distributed generation location and sizing in distribution system using Genetic
Algorithm Technique. In this study, Cost Benefit Factor (CBF) was used to evaluate
the benefits from the added DG units to the system performance. They also discussed
about the Genetic Algorithm(GA) technique for optimization problem and formulated
the aforementioned problem using this technique.
Khatod, et al., [ CITATION Kha13 \l 1033 ], discussed about the evolutionary
programming based method for the optimal placement of solar and wind based DG in
radial distribution system. The load model and DG model were implemented by
probabilistic model. A sensitivity index method was applied to reduce the search space
and quickly determine the location and size of DG. The fitness function was to
minimize energy loss by considering different constraints to the system. In this
method, they developed index method to generate population and this reduces the

16
computational time. This method was tested on 69 bus, 12.66 kV test system and
concluded with the significant amount of loss reduction with voltage improvement.
Banihashemi et al.[CITATION Ban \l 1033 ], suggested a new optimization
method NSGA-II to optimize the voltage profile, loss and investment simultaneously
which facilitates the planner to avoid from the pre-judgment before decision making
and gives the opportunity to choose the best answers. They used non-dominated sorting
(NSGA-II) for the optimization of DG problems. The NSGA-II provides the collection
of optimized solution that determines the best DG units to be installed in which
location.
Sosic and Skokljev [ CITATION Sos13 \l 1033 ], studied the optimal location
and sizing of photovoltaic plant using genetic algorithm with two-fold objective
function. The first objective was to locate the network node and second, to maximize
the sizing of the plant. They implemented genetic algorithm (GA) for optimal location
and optimal sizing of the distributed generation. Genetic algorithm is a search and
optimization technique adopting the natural selection and genetics.
In [ CITATION Sao14 \l 1033 ], Saonerkar and Bagde, presented optimized
DG placement problem in radial distribution system using reconfiguration and
capacitor placement in the system using GA. In this study, genetic algorithm (GA) was
used for finding optimum number of DG units and optimum total power loss of IEEE
33-bus distribution system.
Walla et al.[ CITATION Wal12 \l 1033 ], attempted to establish the impacts on
voltage and determination the maximum distributed photovoltaic generation to the
distribution grids in Sweden. Three grids were considered in the studies and showed
increase in voltage rise at the bus. Of the three grids, two had voltage rise but without
voltage violation and one of the grids experienced overvoltage at some buses.
Hung et al.[ CITATION Hun14 \l 1033 ], implemented biomass and wind DG
as dispatch able and non-dispatchable distributed generation respectively to minimize
energy losses by locating and sizing of DG in a distribution system. They proposed
analytical method to identify power factor and size of DG units and simultaneously
consider time-varying generation and demand also. This method has been applied to
different renewable DG with and without power factor limit to 69 bus IEEE test
distribution system. The result shows that the despatchable or a combination of
dispatchable and non-dispatchable DG sources can minimize annual energy loss
significantly.
17
 Acharya et al.[ CITATION Ach06 \l 1033 ], proposed analytical approach to
find the optimal location and size of DG for power loss minimization. This method was
implemented in three different test systems by varying size and complexity. The results
obtained from the proposed method was compared with the exhaustive load flow
method and that of loss sensitivity method. From the results the loss sensitivity method
is not suitable to the best placement for DG.
Murthy and Kumar,[ CITATION Mur13 \l 1033 ] showed that the sensitivity
based approach for the comparison of optimal DG allocation in a radial distribution
network. They combined the loss sensitivity and index vector method for optimal
allocation of DG units. The results were compared with the novel and voltage
sensitivity index method. This found out that the reactive power should also be injected
to the system to improve voltage profile and optimal size of DG units with lagging
power factor and unity power factor were separately determined. For these power
factors, they calculated the cost of loss savings and cost of power supplied through the
selected IEEE 69-bus system. Based on the obtained results, lagging power factor DGs
were better than unity power factor DGs.
Singh et al.[ CITATION Sin07 \l 1033 ] presented that the assumptions under
constant power load model may lead to an inconsistent and misleading results. The
load model affect the DG size and location in the system. They presented different load
models for different size of DG to minimize active and reactive power loss. 38 bus
system was implemented to show the effect of various load models.
Abdi and Afshar[ CITATION Abd13 \l 1033 ], proposed Monte Carlo
simulation method to determine the result from the improved PSO method for optimal
placement of DG in existing distribution network. The objective functions were to
minimize loss cost, to improve voltage and reliability in the system. Voltage dependent
load model for three consumer areas namely residential, commercial and industrial
were considered in the study. The proposed method was tested on 33 bus IEEE test
system and obtained result was better than PSO.
Jabar and Pal [ CITATION Jab09 \l 1033 ], proposed an ordinary optimization
method for locating and sizing of DG. The best location and size of DG depended on
the compromise between the maximization of DG size and minimization of loss. This
method selects few number of alternatives from the large number of potential DG units
and set objective functions for selected samples using linear programming and simulate
the functions using objective function programming. The method has been
18
implemented in MATLAB and tested for 69 bus distribution system. The obtained
result is fairly well for the DG system because the ordinary optimization is
deterministic and easily implemented in any software.
Ameri et al. [ CITATION Ame15 \l 1033 ], proposed a GA optimization
technique to find optimal sizing and location of multi distributed generations in
electrical networks. The objective function was based on a linearized model to
calculate the real power losses as a function of power generators. The multi DG units
have been used to minimize power loss taking DG capacity, line capacity and voltage
as constraints. This method was tested on 14, 30, and 57 IEEE bus systems to
demonstrate the effectiveness of the method. The results showed that the GA with
combined linear model was very efficient to reduce real power losses by finding the
optimal allocation of DG units.
Arya et al.[ CITATION Ary12 \l 1033 ], highlighted differential evolution (DE)
technique considering voltage stability for the planning of DG. A sub transmission line
was selected to minimize loss by optimal DG placement using incremental voltage (
dV /dP) method. This method was developed for wind based DG and tested on 6 and
30 bus system and compared with the bare bone PSO method. This method was less
time consuming and efficient for DG placement in a distribution system.
Ziari et al. [ CITATION Zia10 \l 1033 ], determined the placement and sizing
of DG in semi-urban 37-bus Roy Billinton Test System using Discrete Particle Swarm
Optimization (DPSO) and Genetic Algorithm (GA). The use of DPSO was the
diversity of the optimizing variables that those variable not to be trapped in local
minimum. Junjie et al.[ CITATION Jun12 \l 1033 ], proposed immune algorithm (IA)
for DG allocation in distribution system. This method is a biometric intelligent
behavior of the biological system. Since load growth in a system is dynamic in nature
and hence dynamic load model is required. The analysis was done considering
environmental cost, DG capacity cost, purchase power cost and loss cost.
Kumar and Gao [ CITATION Kum10 \l 1033 ], presented mixed integer
nonlinear programming (MINP), a mathematical programming with discrete and
continuous variables and nonlinear in objective function and constraints, to determine
the optimal location and size of the DG for hybrid as well as pool market. The
objective was to minimize fuel cost as well as line losses for conventional and DG
sources. Firstly, they selected an optimal DG placement with nodal price and power
losses. Then MINP was implemented for selected site. This showed that the actual
19
price and power loss reduced significantly. It was observed that the nodal price
increases when demand increases but optimal DG location nodal price goes down.
However, more DG penetration in the network has no effect on the nodal price. The
proposed method has been tested on IEEE 24 bus reliability test system.
El-Zonkoly [ CITATION ElZ11 \l 1033 ], utilized multi-objective index based
method to select the optimal location and size of multiple DG units considering non-
unity power factor in distribution system using different load models. This method
suggested that the load model can significantly affect the optimal allocation of DG in
distribution network. The proposed method considered active and reactive power loss,
voltage improvement and also considered short circuit level parameter to represent
protection devices. Integration of DG was proposed on most sensitive buses which
were likely to be collapsed. The proposed method was tested on 30-bus IEEE mesh
system and 38 bus radial system.
Wang and Singh[ CITATION Wan08 \l 1033 ], presented an Ant Colony
Optimization (ACO) algorithm for optimal placement of DG with reclosers in a
distribution network. The system average interruption duration index (SAIDI) and
system average interruption frequency index (SAIFI) being the composite reliability
indices that were minimized by optimal allocation of DG. The proposed method was
compared by taking two practical distribution system using GA technique.
Mohammedi et al.[ CITATION Moh131 \l 1033 ] described a multi-objective
optimization technique to determine the optimal location and size of DG in radial
distribution system. The main objectives were to minimize loss and to improve voltage
stability. NSGA-II was used to obtain the Pareto-optimal set of solutions that form a of
non-dominated set of solutions to select the best compromise solution. A fuzzy-based
technique was employed to select the best solution. The method was tested on 12, 33,
69 and 85 bus systems to validate the effectiveness of proposed method and detailed
performance was analyzed.
Vahidinasab[ CITATION Vah14 \l 1033 ], proposed a multi-objective
optimization and probabilistic design for optimal planning of Distributed Energy
Resources (DER) in competitive markets. The suggested nonlinear programming based
method has attempted to achieve a compromise between cost minimization and
pollution emission minimization considering electrical load and market uncertainties.
A hybrid fuzzy FCM-MCS model was employed for modeling of electricity prices and
a combined roulette-wheel/ Monte-Carlo simulation model was used for generation of
20
load scenario. The objective was to determine type, location and size of wind, PV, fuel
cell, micro turbine, gas turbine, and diesel engine. The test was done on primary
distribution network and provided to select the best Pareto-optimal solution to the
decision maker to its own preference.
Lamsal et al. [ CITATION Lam16 \l 1033 ], presented a new methodology for
locating and sizing of distributed generation to determine loss in distribution system
using B-coefficient matric method. The loss evaluation has been done by using
different combination of distributed generations (DGs). Use of B-coefficient matrix
does not require separate load flow calculation to calculate power loss in the system.
Le and Qiang [ CITATION LeJ16 \l 1033 ], carried out a research on optimal
allocation of DG to maximize the benefits of renewable energy based DG and improve
voltage profile of the system. They proposed PSO method which employed non-linear
decreasing inertia-weight to improve the convergence of the algorithm.

2.3 Gaps identified

From the literature reviews carried out in this chapter, following gaps have
been identified:
1. There are very few and limited studies available for integration of solar
photovoltaic system with existing grid in Nepal.
2. There are limited studies related to dynamic stability analysis due to the
intermittency of solar photovoltaic energy injection to the grid.
3. Many authors have proposed a load flow algorithm for solving balanced
radial distribution network, but practical distribution network is
unbalanced in nature.

21
Chapter - 3 PROBL

EM FORMULATION AND SOLUTION ALGORITHM

3.1 General

This chapter presents the modelling of distribution feeder and the solution
algorithm that has been adopted. The research work began from site selection, data
collection, modeling of distribution system, solar energy resource assessment, problem
formulation, and load flow analysis for base case. Load flow analysis was performed
on the test case, problem formulation was developed to find the optimal location and
size of solar photovoltaic system to reduce the system loss and investment cost. The
problem formulation was single objective and multi-objective to implement with the
use Genetic Algorithm technique. For load flow problem MATLAB software was used
and for optimization problem MATLAB Toolbox was used. For design and economic
analysis MS Excel was used. The detailed method carried out during research work
was as below:

3.2 Problem Statement:

Normal distribution systems are designed to operate unidirectional power flow

that is power flowing from source to the end users. When photovoltaic is connected to
the system the power flow occurs in bidirectional causing change in voltage profile of
the system. Inappropriate location and size of PV may rise in system losses, decline in
power quality and voltage rise in at the end of feeders. In that reason, installation of PV
with optimal site and size is very important.
A multi-objective problem is formulated to reduce the system loss and the cost
associated to the installation cost, and operation and maintenance cost of the SPV. To
select the optimal solution from the multi-objective optimization, a fuzzy-based set
theory is employed such that the best compromise solution is obtained. The maximum
value among the fuzzy membership function is selected for this purpose.

22
3.3 Analysis of distribution system for load flow

The first step to this study is to analyze load flow for the distribution network
with base case and with PV connected. This analysis is done through the MATLAB
software with multiple photovoltaic unit that can be modeled as negative PQ buses.
When the power flow program is run with this software the results obtained are the
voltages, angles, currents, power loss, total annual energy loss, total load of the system
and power supplied by the photovoltaic system. By varying the size of the PV and
location of installation the output results are obtained. Then these results obtained will
be compared with the base case.

3.4 Mathematical modeling of distribution line

Vmδm Vnδn
Ῑij
S Z=Ri+jXi

Pi+jQi
Bus m Bus n

Load(Pj+jQj)

Fig. 3.8: Single line diagram of a distribution line

The three-phase distribution line is assumed to be balanced and represented by

an equivalent single line diagram. Single line to ground conductance and capacitance at
the distribution voltage level is very small and thus is can be neglected. A simple
model of a radial distribution feed is shown in the Fig. 3.1.

23
 Current flowing through the line is given by:

|V m| δm −|V n|δ n (3.1)

I ij =
Ri + j X i
here, Iij= branch current of the line i to j
Vmδm=voltage and angle of bus m

Vnδn=voltage and angle of bus n

Ri= resistance of branch m to n

Xi= reactance of branch m to n

Again,

¿ Pi + jQ i
I ij = (3.2)
Vm
V n=V m−I ij Z i

Pi− jQi
¿ V m−
( V ¿m )
( Ri + j X i )

Pi Ri +Qi X i P X −Q X
(
¿ V m−
Vm ¿
) (
− j i i ¿ i j (3.3)
Vm )
The voltage magnitude and angle at bus j can be written as,

|V n|=√( V 2m −2 ( P i Ri +Qi X i ) + ( P 2i +Q2i )( R 2i + X 2i )/V 2m ) (3.4)

Angle δ n=δ m−tan −1 ¿ ¿

The active and reactive power losses of the line can be expressed by equations
(3.6) and (3.7):
2 (P¿¿ j)2 +Q j2
Ploss (i )=|I ij| Ri= 2
× Ri¿
|v n|
(3.6)

24
 2 (P¿¿ j)2 +Q j2
Q loss (i )=|I ij| Xi= 2
× Xi ¿
|v n|
(3.7)

The active power loss for every hour in a distribution line is the energy loss.

Total annual active energy loss is given by,

4 24 ( N br )
Eloss =∑ ∑ ∑ Ploss ( i) ,t ,s . N s ,(3.8)
s=1 t=1 i=1

Where Ploss(i),t,s is the active power loss of branch ‘i’ at time ‘t’ of the season ‘s’,
Eloss is the total annual active energy loss, N s is the number of days in different season
and N br is the total number of branches.
When solar photovoltaic is added to the feeder at the load bus then it injects
active power to the bus at unity power factor which is given in Fig.3.2. This SPV
generator is treated as negative load (PQ) bus because it will not generate reactive
power and hence no role in voltage control of the bus.

PPV
Vmδn Vnδn PV
Iij
S Z=Ri+jXi

Load (Pj,Qj)

Fig. 3.9: Distribution grid with solar PV generator

Inverter is operating to inject unity power factor by adjusting the fire angle of
the power electronic devices, the active power generated by the PV generator is Ppv and
there is no reactive power because of unity power factor. Then, the resultant active and
reactive power losses become as shown in equations (3.9) and (3.10):

(3.9)
2 2
2 ( P j−P PV )2+Q j2 2 ( P j−P PV ) +Q j
Ploss (i )=|I ij| Ri= 2
× RiQloss (i )=|I ij| X i = 2
× Xi
|V n| |V n| (3.10)
25
 At this stage, the voltage at different buses will be changed due to the change in
active power injection and hence the power loss will be changed. Then the total annual
energy loss is different from the base case which was discussed in above case.

3.4.1 Load flow analysis

The power distribution system is generally operated in radial configuration. The

voltage drop in in the feeder mainly depends upon the loading where the maximum
voltage drop occurs during peak load. For radial distribution system, backward/forward
sweep method is suitable because it converses very fast and less time consuming. In
backward sweep, updated power flow or current in each branch is estimated starting
from the last layer and moving towards the source node. In forward propagation,
voltage drop is calculated with branch currents and these values are used to update bus
voltages in forward direction. During the forward sweep, the power in each branch is
held constant. The load flow algorithm is given as follows[ CITATION Mee15 \l
1033 ]:
1. Read the system data.
2. Form BIBC matrix as given as follows:
I. Line data arranged in matrix form as LD matrix.
II. Initialize BIBC matrix to null matrix.
III. Set Br=size(LD,1).
IV. For i=1 to Br.
V. If LD(i,2) ==1, go to next step, otherwise go to step IX.
VI. Find BIBC (LD(i,3)-1, LD(i,3)-1) =1
VII. Go to step XI.
VIII. Find BIBC (:, LD(i,3)-1) = BIBC (:, LD(i,2)-1).
IX. Find BIBC (LD(i,3)-1, LD(i,3)-1) =1.
X. Increment i by one.
XI. Is i<=Br, (Br = number of branches), then go to step VI. otherwise go
to next step.
3. Then calculate bus injection current vector:
I=conj((P-jQ)/V))

26
 Where, I=bus injection current vector
P=real power injection vector
Q=reactive power injection vector
V=bus voltage vector

4. Calculate branch current matrix as:

Ibr= BIBC*I
5. Form diagonal impedance matrix, ZD
6. Compute bus voltage as:
TRX=BIBCꞌ*ZD*BIBC
T=TRX*I
VB=V1-T
Where VB is the bus voltage vector and V1 is substation voltage vector.
7. Find voltage mismatch dV=max (abs (Vnew - Vold)
8. Repeat steps(iii) to (vii) until the algorithm converges with accuracy tolerance.
9. Calculate real and reactive power loss.

3.5 Problem formulation

This problem is a mixed integer non-linear optimization problem with fitness

function, equality constraints and inequality constraints. The main objective of this
study is to minimize the total energy loss and installation cost of PV system while
satisfying operating constraints. The constraints are power flow, thermal limits of line,
voltage limits and PV power injections limits. In this formulation, PV is modeled as
negative PQ bus that injects power at unity power factor. Only some of the buses are
selected to connected the solar PV which has voltage below 0.95 p.u. levels.

3.5.1 Formulation of optimization problem

In this formation of optimization problem, first single objective problem is
formed to optimize size and location at the same time. There are two variables PV
location and capacity which has to be determined from optimization. Candidate
locations are identified by calculating the normalized voltage. First calculate the

27
normalized voltage |V i|/0.95 for all the buses. Then, those buses having normal
voltage value <1.0 are selected as candidate bus for PV placement.
The total active energy loss for a distribution network is given by,

4 24 ( N br )
Eloss =∑ ∑ ∑ Ploss ( i) ,t ,s . N s ,(3.11)
s=1 t=1 i=1

Where, i – branch number

Nbr – total number of branches
Ploss(i),t,s – real power loss of branch ‘i’ at time ‘t’ of season ‘s’.
Eloss – Total annual energy loss of distribution network.

A multi-objective problem formulation is to determine the best location and

size of PV such that the overall cost of PV will be minimized. The PV cost is
formulated as:
The investment cost (CINSTAL) depends on installed capacity of the SPV and
operation and maintenance cost (CO&M) depends on the number of units generated from
the system. The annual installation cost per kW capacity can be calculated by

installation cost /kW × r ( 1+r )n

C INSTAL = n
(3.12) ‘r’ is discount rate and ‘n’ is the
(1+ r ) −1
number of years for the project life.

The annual operation and maintenance cost per kW can be calculated by

CO&M= annual energy generation × unit running price

(3.13)

The total annual cost (CTOTAL) for solar photovoltaic is given as,
(3.14)
C TOTAL=N PV ( C INSTAL + CO∧ M )

Where NPV is the number of PV modules in one-kilowatt capacity.

Hence, first objective is to minimize annual energy loss.

28
 Minimize, f1(x) =system energy loss

(3.15)

Second objective is to minimize total cost of PV system

Minimize, f2(x) = total cost (3.16)

Multi-objective optimization problem is,

f(x) = [min f1(x), min f2(x)] (3.17)

3.5.2 Constraints
solar PV integration optimization problem are subjected to the following constraints:

3.5.2.1 Equality constraints

The equality constraints are the power flow equation:

N bus
PGi−P Li−V i ∑ V j Y ij cos ( θij + δ j−δ i )=0 (3.18)
j=1

N bus
QGi−Q Li +V i ∑ V j Y ij cos ( θ ij + δ j −δ i ) =0(3.19)
j=1

Where, i=1,2,3 , … , N bus, PGi ,Q Gi is, the active/reactive power generated at the
ith bus; P Li , Q Li is the active (reactive) power demand at the same bus and Y ij ∧θij are
the admittance and angle of branch connecting bus i and j.

3.5.2.2 Inequality constraints

1. Bus voltage constraint:

0.95 pu≤ V i ≤ 1.05 pu, i=2, 3, ..., Nbus (3.20)

2. Generation constraint:

0 ≤ P PVi ≤ Pmax
PVi (3.21)

29
 3. Maximum PV penetration constraint:

Maximum injected PV capacity should be less than 50% of the total

active power load in the distribution system.

N PV Nbus

∑ P PV ≤ 0.5 × ∑ PL (3.22)
i i
i=1 i=2

4. Line thermal rating:

|Sij|≤ Sijmax (3.23)

Where Pmax
PVi is, the upper active power generating unit limits of PV, Sij is the

apparent power line capacity and Smax

ij is the maximum permissible loading capacity of
the line connected between bus i and bus j.

3.5.2.3 Constraints Handling

The equality constraints are the power balance equations. These constraints are
handled by the load flow method. Backward/forward sweep method is suitable to
solve the load flow method for radial distribution system because it is very efficient
method for distribution networks.
The inequality constraints like voltage limit, line loading limits are handled by
introducing the penalty function in the fitness function.
F (x)=f ( x )+ P ( x) (3.24)

Where, F(x) is the new function from the objective function, f(x) and the
penalty function, P(x).
N BUS
P ( x ) =∑ λ v ¿ ¿ ¿
i=1

Where, λ v and λ s are the penalty factors and values for these factors have been

considered 10000 in this report. V lim

i
¿¿
and Slim
i
¿¿
are taken as follows:

max max
V i ,if V i >V i

Vi
lim ¿=
{ min min
¿ V i ,if V i <V i
min
¿ V i ,if V i ≤V i ≤V i
max
¿

(3.26)

30
 max max
S ij , if S ij >Sij
lim ¿=

Sij
{ min
¿ S ij ,if Sij< S ij
min
min

max
¿ Sij ,if S ij ≤ S ij ≤ S ij
¿

(3.27)

3.6 Genetic Algorithm

3.6.1 Introduction:
Genetic algorithm is a method for solving optimization problems that is
inspired by the natural selection of evolution GA starts from a set of random generated
initial population of individual/string[ CITATION Ame15 \l 1033 ]. Once the initial
population is generated, each individual is evaluated by objective function. In each step
three main operators (reproduction, crossover, and mutation) are performed to create
the better new population from the current population. The genetic algorithm applies
selection and reproduction criteria at random from current population to be parents and
uses them to form new generation containing the genes of parents. The solution having
good fitness will undergo reproduction.
GA does not need to select good initial populations and the fitness value come
after the corrected value of each generation[ CITATION Hel12 \l 1033 ]. The solution
is achieved from the best individual of the last populations.
GA uses multiple search points instead of single search point, thereby finding
more hills and valleys, reducing the chance of getting struck in local
minima[ CITATION Ban \l 1033 ]. GA based solution techniques are getting more
popularity due to their ability to find global optima rapidly, robustness in convergence
rate and regardless of nature of complexity. There are major two advantages of using
GA for optimization: It does not involve many mathematical assumptions about
problems and Due to evolutionary nature, it will search for solutions rather than the
specific inner structure of the problems.

3.6.2 Outline of Genetic Algorithm

The steps involved in GA based methodology for optimization are given below
and the flow chart for GA based optimization problem is shown in Fig. 3.3.
1. Generate random population of chromosomes, that is, suitable solutions for the
problem.

31
2. Evaluate the value of objective function (OF) for each chromosome in the
populations.
3. Create a new population by repeating the selection, crossover and mutation
operation.

STAR
T

Randomly generate initial

population

Evaluate objective function

f(x)

Operate
crossover

NO Check
for
constrai
nt?s

YES Update
parameters

Offspring

Mutation

Generate new
population

NO
Check
stopping
criteria?

YES

Results

32
EN
D
 Fig. 3.10:Genetic Algorithm Scheme.

 Selection/reproduction: select two best parents from a population

according to their fitness. This process is based on probabilistic
approach and parents having greater value have higher probability to
reproduce.
 Crossover: with a crossover probability, crossover operator recombines
the relevant information carried out by the parents to form a new
offspring.
 Mutation: with a mutation probability, it alters one or more genes values
in chromosome from its initial population. First, one string is selected
with less probability and random changes a bit of the string to form new
offspring.
4. Use new generated population to evaluate the fitness function and further run
the algorithm.
5. If the end condition is satisfied, or permissible generation is achieved, return
the best chromosome as the solution; otherwise go to step (3).

3.6.3 GA based Solution Methodology

The global optimization toolbox provides genetic algorithm function that search
for the global solutions to optimization problems. The input arguments to the
optimization function are:
 Number of variables for the fitness function.
 Fitness function.
 Options structure.

The following steps are employed for genetic algorithm based optimization process:

3.6.3.1 Coding of the solution

33
 When applying genetic algorithm to determine the location and size of SPV
unit problem, the first step is the coding of the potential solution. In a general way, the
potential solution is a configuration of the network with PV units connected in some
locations. Therefore, the coded variables are the candidate locations and the respective
size of the PV units. Hence, each chromosome is coded by a solution vector whose
dimension is the number of candidate locations and content is the size of installed
units.
To reduce the search space, first identify the candidate buses for PV placement

by determining the normalized voltage magnitude for all buses as:

|V i|
Norm [ i ] = (3.28)
0.95

Buses that have normalized values less than 1.00 are taken as candidate buses

for PV placement and reset the size of PV to zero in other buses.

x i=¿ i=1,2, 3..., Nbus (3.29)

Where, DG pl and DG ¿¿¿ are the placement and size of PV unit.

3.6.3.2 Objective Functions

In global optimization toolbox function, first write a file that computes the

objective functions. The function should accept the length of a solution vector and

return a scalar. For multi-objective optimization function, the function should return a

row vector of objective function values. The objective functions that are discussed in

section 3.5.1 are as follows:

Minimize, F1(x) =system energy loss (3.30)

Minimize, F2(x) = total cost (3.31)

Multi-objective optimization function is,

f(x) =[minimize F1(x), minimize F2(x)] (3.32)

34
3.6.3.3 Selecting the best result

For the multi-objective optimization solution, a fuzzy-based technique is

employed in this report. From the obtained Pareto-optimal solution the best
compromise solution is selected by calculating the fuzzy membership function as
follows[ CITATION Moh131 \l 1033 ]:

f max
i −f i
μf = max min
(3.33)
i
f i −f i

Where f max
i ,f min
i are the maximum and minimum values of the ith objective
functions of all Pareto-optimal solutions. For each Pareto set k, the normalized
membership function is calculated as:
m

∑ μf
k

k
μ= i=1
D m
i

k
∑∑ μf i
k=1 i=1
(3.34)

Where D is the total number of Pareto solutions and m is the total number of
objective functions. The best solution is that having maximum value of μk .

3.6.3.4 Setting of parameters

The results obtained by genetic algorithm (GA) function are influenced by

several parameters like population size, selection function, crossover function,
mutation function, constraint parameters, stopping criteria, etc. GA accepts bounds,
linear constraints, or nonlinear constraints.
The lower and upper bounds variables are fixed by creating lower bound (lb)
and upper bound (ub) constraint vectors. The lower bound and upper bound element
are selected based on the allowable capacity of PV unit. The non-linear constraint
function is developed by considering the maximum allowable limit on PV penetration.
The options for size of population, maximum generation limit, function tolerance, plot
functions, etc. are selected to visualize the process.

35
 Chapter - 4

SOLAR ENERGY RESOURCE ASSESSMENT

4.1 General

This chapter explains about the solar energy potential of Kathmandu valley for
the study period of one year. The study period is further divided into four seasons:
spring, summer, autumn and winter. March, April and May are considered under
spring season. June, July and August under summer season. September, October and
November falls on autumn season and remaining three months December, January and
February are under winter season. The average hourly equivalent data of each season
are estimated to evaluate the irradiance. The day of the year is selected in accordance
with the seasonal days. Based on aforementioned data solar potential is estimated and
suitable size of PV is determined to connect at the load buses.

4.2 Estimation of solar radiation on tilted surface

Table 4.1: seasonal average global solar radiation (Wm-2) on horizontal surface

Time/
Spring Summer Autumn Winter
Season
6:00 0.0 10.0 0.0 0.0
7:00 40.0 70.0 15.0 0.0
8:00 130.0 160.0 85.0 25.0
9:00 270.0 290.0 200.0 100.0
10:00 415.0 380.0 320.0 200.0
11:00 560.0 440.0 425.0 300.0
12:00 620.0 480.0 490.0 370.0
13:00 650.0 500.0 450.0 395.0
14:00 550.0 460.0 375.0 340.0
15:00 450.0 370.0 290.0 275.0
16:00 300.0 255.0 180.0 180.0
17:00 180.0 150.0 70.0 65.0
18:00 70.0 50.0 0.0 0.0

36
 Solar radiation is the most important parameter for the design and development
of various solar energy systems. Generally, global solar radiation is measured on
horizontal surface and rarely on inclined surface. The solar radiation depends on the
latitude of the place, day of the year, hour of day and metrological conditions. The
initial solar radiation data are recorded on continuous basis and the data obtained from
the metrology department is daily averaged data.
The average seasonal global solar radiation on horizontal surface which were
obtained from the Department of Hydrology and Metrology (DHM), Nepal are given in
Table 4.1 and the pictorial representation is shown in Fig. 4.1.
700 Spring Summer Autumn
Winter
600
Solar radiation (Wm-2)

500

400

300

200

100

0
6 7 8 9 10 11 12 13 14 15 16 17 18
Local time (hours)

Fig. 4.11: Seasonal variation of solar radiation in horizontal surface

Maximum global solar radiation can be received by tilting the receiving surface
perpendicular to the incident radiation.
In that reason, receiving surface should be tilted to match with the latitude of
the earth. Following assumptions are made to translate the given solar radiation in
tilted surface.
Reflectivity of ground surface = g =0.2
Latitude () =27o
Tilt angle (s) = 30o
Number of days for:
Spring: 105; Summer: 196;
Autumn: 288; Winter: 15.

37
 From these values, tilt factor for beam radiation (R b), diffused radiation (Rd)
and reflected radiation (Rr)are determined[ CITATION Adh13 \l 1033 ].
(i) Declination angle of Sun:
The declination angle of the sun can be approximated using Cooper’s equation (4.1) as:

360 ( n+284 ) (4.1)

δ =23.45 sin [ 365 ] where, δ=declination angle of the sun in degree

n= day number of the year (from 1 to 365)

(ii) Time of day:
The solar time, t, is expressed as an hour of day,
ω=¿15o (t-12) (4.2)
(iii) Sunset hour angle is given by

−1
h s=cos (−tan ∅ tan δ ) (4.3)

(iv) Ho is the monthly average daily extraterrestrial radiation on the horizontal

surface given as follows:

Ho= ( 24π ) I E [ 180π h sin ∅ sin δ+cos ∅ cos δ sin h ]

sc o s s Where, I sc
(4.5)

=1353 Wh/m2 is the solar consta

Eo =1+0.033 cos ( 360 n

365 )
is the eccentricity constant.

(v) The clearness index, KT, which is the ratio of the horizontal insolation at the site
Hm to the extraterrestrial insolation on a horizontal surface above the site and
just outside the atmosphere Ho.

Hm
KT= (4.6)
Ho
(vi) Diffuse sunlight
The diffuse insolation is given by
H d= H [1-1.13KT]
(4.7)
H=total solar radiation on horizontal surface Wh/m2
(vii) Beam radiation is given by

38
 H b= H-Hd
(4.8)

(viii) Total solar radiation on tilted surface.

In order to estimate the solar radiation on tilted surface we need to determine
the tilt factor for all the components of the solar radiation. Tilt factor, R, is defined as
the ratio of solar radiation on tilted surface to the solar radiation of horizontal surface.
 Tilt factor for beam radiation

sin ( ∅−s ) sin δ+cos ( ∅−s ) cos δ cos ω (4.9)

Rb =
sin ∅ sin δ+cos ∅ cos δ cos ω
Where, s= angle of tilt
 Tilt factor for diffuse radiation

Rd = ( 1+cos
2 )
s (4.10)

 Tilt factor for reflected beam radiation

(4.11)
1−cos s
Rr =ρ g ( 2 )
Where, ρ g is the reflectivity of surrounding surface and is taken as 0.2.
(ix)Total global solar radiation on tilted surface is given as:
H t =H b R b + H d Rd + H r R r
(4.12)

39
 Spring Summer
Autumn Winter

700
600
Solar radiation(W/m2)

500
400
300
200
100
0
6 7 8 9 10 11 12 13 14 15 16 17 18
Time (hour)

Fig. 4.12: Seasonal global solar radiation in tilted surface.

Fig. 4.2 shows the average seasonal solar radiation in 30o tilted surface. These values
are used to determine the size of solar PV unit.

4.3 Selection of solar PV unit

The power rating of solar module is expressed in watt-peak(Wp), is the

maximum power produced by the module at standard test condition that is solar
radiation 1 kW/m2, cell temperature 25 oC and spectrum distribution 1.5 AM. The
manufacturer provides the characteristics of PV module under STC which are shown in
Table 4.2 below for selected modules.

Table 4.2: Specification of different modules

Parameters A B C D E F G
Peak Power Output
100 150 240 250 310 300 320
(Pmax) W
Module
14 14.5 14.4 15 16 15.5 16.5
Efficienty(η) %
Voltage at
maximum rating 17.6 18.1 30 30.7 36.3 35.8 37
(Vm) V
Current at
maximum rating 5.68 8.27 8.03 8.16 8.53 8.37 8.64
(Im) A

40
 Open circuit
21.8 22.5 37.3 38.1 45.6 45.2 46
voltage (Voc) V
Short circuit
6.15 8.68 8.52 8.58 8.99 8.86 9.18
current (Isc) A
NOCT (oC) 47 47 47 47 47 47 47
1667
Module 1202 x 1504 x 1667 x 1960 x 1960 1960 x
x
Dimensions ( L x B 661 x 661 x 1000 x 990 x x 990 990 x
1000
x T) in mm 38.5 38.5 33 40 x 40 40
x 33
Number of cell in
36 36 60 60 72 72 72
series (Ns)
Temperature
coefficient of Voc -2.931 -2.931 -3.305 -3.305 -3.2 -3.2 -3.2
(mV/oC)
Temperature
coefficient of Isc 0.442 0.442 0.638 0.638 0.5 0.5 0.5
(mA/oC)
0.758 0.748
Fill Factor (FF) 0.7456 0.7664 0.7663 0.7553 0.7570
0 2

The current-voltage characteristic of a PV module can be determined for a given

radiation level H and ambient temperature Ts using the following relations:

(4.13)
N oc−20 I=[Isc+αI(Tc-25)]Ht
T c =T s+ Ht (4.14)
800

V=Voc – αvTc (4.15)

Where, Tc represents cell temperature in oC, I represents for PV

module short-circuit current in A, αI represents short-circuit current temperature

coefficient in A/oC, V indicates open-circuit voltage in V and αv is open-circuit voltage

temperature coefficient in V/oC. The actual power output from a PV array can be

directly calculated as:

Pactual =FF .V . I (4.16)

41
 Where, Fill Factor, FF, depends on the material of PV module and is given by
the following relation:
(4.17)
V max I max
FF=
V oc I sc

The selected modules with their characteristics are given in the Table above.
The actual S. N. Module CF (%) output
capacity of 1 A 13.704 the module
2 B 13.688
is different from the
3 C 13.727
standard test 4 D 13.750 condition
which 5 E 13.711 depends
6 F 13.710
upon the available
7 G 13.081
solar radiation and
ambient temperature. The actual power output can be determined by the equation
(4.16).

Based on the actual power produced by the module in a year the capacity factor
(CF) of each module is calculated.

4.3.1 Capacity Factor (CF):

The capacity factor of a PV power plant is the ratio of the actual output over a
period of one year and its output if it has operated at nominal power in the entire year:

( Energy generated per annum(kWh) )

CF = (4.18) The determined CF for
( 8670 × Installed Capacity (k W p ) )
selected modules are given below in Table 4.3. The highest CF module is selected as
optimal size of the solar PV system. From the Table ‘D’ module has the highest CF
and hence it is selected for the integration in the grid.

Table 4.3: CF of the PV modules available

42
 The actual power output is calculated using solar radiation and temperature
during the study period. The seasonal output variation of module ‘D’ with 1.0 kWp
solar PV array is shown in Fig.4.3 below. It is found that the maximum power output is
0.65 kW at 13 PM in spring season and there is less power output at other seasons. In
winter season the output power is lowest among all season.

Spring Summer Autumn

700

600

500
Power outpur (p.u)

400

300

200

100

0
1 3 5 7 9 11 13 15 17 19 21 23
time(hour)

Fig. 4.13:Daily solar PV power output of a 1.0 kWp PV array

43
Chapter - 5 RES

ULT AND DISCUSSION

5.1 General:

This chapter presents the results and discussions related to the load flow analysis
comparing the energy losses obtained from the different size of PV units for 21 nodes
distribution feeder. The investment cost is calculated considering the different size of PV
units. Multi-objective optimization is formulated combining both the objectives. The
formulations are determined with the help of MATLAB Optimization Toolbox and the
results obtained will be discussed.

5.2 Case Study:

For locating and sizing of solar PV unit, a radial distribution network having 11 kV
voltage level is selected which is located in Kathmandu. The general layout of the feeder is
shown in the Fig. 5.1 below. The details of the line data and maximum connected load of
each bus which were obtained by Nepal Electricity Authority (NEA) are given in the Table
5.1.

1 2 3 4 5 6 7 8 9 10 11

Sub-station
12
13 14 15
16
20

17
21
18
19

Fig. 5.14: Distribution system with 21 bus

44
The main features of the distribution system are as follows:
 The average load of the system is determined as 1.4 MW whereas the peak system
demand is 2.12 MW. In winter and autumn, the system peak load occurs at 6 PM,
while in spring and summer the system peak is shifted around at 8 PM.
 It is short feeder highly unbalanced load.
 It has overhead lines that are three phase with different spacing between them.
 Autumn and winter peak is greater than the spring and summer peak load.
 During night time the system load is below average load and in day time the demand
is average load.

Table 5.4: Line Data and Load Data of the distribution system

Branch Sendin Receiving

R(Ω) X(Ω) PL (kW) QL (kVAR)
No g Bus Bus
1 1 2 0.272 0.149 90 60
2 2 3 0.407 0.223 90 40
3 3 4 0.244 0.134 100 80
4 4 5 0.272 0.149 60 30
5 5 6 0.472 0.258 60 20
6 6 7 0.984 0.288 40 30
7 7 8 1.114 0.326 10 5
8 8 9 0.820 0.240 75 30
9 9 10 0.738 0.216 90 70
10 10 11 0.191 0.056 100 40
11 11 12 0.492 0.144 190 45
12 2 13 0.191 0.056 200 50
13 3 14 0.104 0.030 100 45
14 6 15 0.301 0.088 190 50
15 12 16 0.258 0.141 100 80
16 16 17 0.020 0.011 100 50
17 17 18 0.566 0.310 200 50
18 18 19 0.245 0.134 100 70
19 16 20 0.191 0.056 120 60
20 17 21 0.984 0.288 100 75
45
 The seasonal load profile with hourly variation of demand for different nodes is given
in Fig. 5.2.

Fig. 5.15: Seasonal load profile for distribution network.

The system is assumed to be connected with the solar photovoltaic distributed

generation(DG) without storage system. Solar PV generated energy in day time only. During
the night time the demand is fulfilled by the grid itself. 1 kW capacity solar PV based DG
unit is formed by combining 4 number of modules having peak power 250 W P. The total
installed capacity of DG unit is restricted to 50% of the total peak demand of the feeder. The
investment cost per kW of DG is considered as Rs. 62,225 and unit running cost is assumed
Rs. 0.5 per unit. The project life cycle for PV is taken as 25 years for this study. The discount
rate is assumed 12% for this study.

5.3 Result and Discussion

This section presents the result obtained by considering the objective functions one by
one and then simultaneously forming the multi-objective function. After obtaining the results
a comparative study was carried out. Optimal location and size is obtained with the energy
loss value. From the multi-objective optimization function, different results are obtained and
the best results are selected from those values. Various parameters need to be specified for

46
GA before running the optimization program. Table 5.2 shows the GA parameters set in this
program.

Table 5.5: Parameters of Genetic Algorithm

Number of Population 100

Generation Number 50
Number of variables 20
Probability of Cross-over 0.5
Probability of Mutation 0.2
Selection Options Default (Roulette)

Plot Functions Plot Best Function, Plot Best Individual

5.3.1 Result of energy loss minimization

Proposed GA based program was run and the best result obtained to minimize the
energy loss is plotted in Fig. 5.3. When the maximum capacity of PV unit at selected bus is
set to 400 kW the best location/bus and the size for the system is given in Table 5.3.
Best: 397.926 Mean: 397.926
470

460
Annual Total Energy Loss (MWh)

450

440

430
Fig. 5.16: Fitness function evaluation for 21 Bus
420

410

400

390
0 5 10 15 20 25 30 35 40 45 50
Generation

47
 Fig. 5.17: Optimal size and allocation of PV units for 21 Bus
Current Best Individual
0.4
It is observed
0.35
from the Fig.
5.4 that 0.3 the GA
Size of PV per unit

based
0.25
approach
0.2 gives the
0.15 optimal
locations of PV as
0.1
10,11,12,
0.05 16,17,18,
19 & 21 with their
0
2 4 6 8 10 12 14 16 18 20 22
corresponding size
Number of buses
(rounded up to) as given in Table 5.3. The maximum size of PV is 296 kW at bus number 18
and the minimum size is 3.5 at bus number 19. The total system losses for time-varying load
conditions with PV and without PV connection to the buses is given in Table 5.4. The base
case load flow and after PV integration load flow MATLAB programs are given in Appendix
A, B & C. During base case load flow analysis, the total annual energy loss is obtained as
478.25 MWh. After integration of PV, the energy loss reduced to 394.64 MWh saving 18 %
of energy loss. The percentage of loss saving is less because at the peak load, there is no solar
power available. If the energy storage devices connected to the system the loss saving will be
increased.

Table 5.6: Result for PV size and location

Category Candidate Bus Number Total

Bus Number 8 9 10 11 12 16 17 18 19 20 21 11
Size of PV (kW) 10 27 29 23
0 0 67 19 60 4 0 1058
4 3 6 5

Table 5.7: Annual total energy loss reduction with PV

Base case energy loss(MWh) Energy loss with PV (MWh) % Reduction

48
 478.25 397.92 17

The impact of PV connection to the voltage profile is also studied in this report. The
comparison of voltage profile has been done for maximum load condition which is shown in
Fig. 5.5 below. The voltage profile of the 21-bus system is improved significantly after
connecting the PV units which is depicted in Fig. 5.5. The voltage regulation is improved and
better quality of voltage profile is obtained after integration of DG. The voltage regulation
before DG addition is 9.25 and this value is reduced below 2% after addition of the DG units.
The voltage regulation within 5% is acceptable limit.

Plot of Bus number & voltage profile

1.02 WithoutSPV
WithSPV

1
Bus Voltage Profile (in per unit)

0.98

0.96

0.94

0.92

0.9
2 4 6 8 10 12 14 16 18 20 22
Bus Number

Fig. 5.18: Comparison of voltage profile with PV during maximum load condition

5.3.2 Result of cost minimization

To minimize the cost the obtained optimal solution is when there is no DG connected
in the system. The result of cost minimization function is shown below:

49
 Table 5.8: PV location and size for cost minimization

Category Candidate Bus Number Total

Bus Number 8 9 10 11 12 16 17 18 19 20 21 11
Size of PV (kW) 0 0 0 0 0 0 0 0 0 0 0 0

5.3.3 Result of Multi-Objective Optimization

8
x 10 Pareto front
10

8
Annual PV cost (Rs.)

2
370 380 390 400 410 420 430 440 450
Annual Energy Loss (MWh)

Fig. 5.19: Multi-Objective Pareto front representation

Multi-objective optimization is employed to depict the annual energy loss and PV

cost of the system. Equal importance is given to both the objective functions. The result from
the multi-objective function is in the form of Pareto-optimal set which is shown in Fig. 5.6
above. The number of optimal solutions from the multi-objective functions is given in Table
5.6 below. The table shows the objective functions and candidate bus locations with the PV
capacity in kW.
Table 5.6: PV size and location for multi-objective optimization

50
 Objective Functions Optimal Bus Locations and Capacity (kW)
S.N Loss
. PV Cost
(MWh 9 10 11 12 13 17 18 19 20 21
(Rs. In lacs)
)

1 448.07 212.33 59 14 15 11 45 1 38 93 49 24

2 434.97 287.23 55 21 21 46 17 75 41 109 62 26

3 380.24 922.14 170 167 167 156 75 169 119 186 133 173

4 439.91 255.63 47 18 15 13 27 46 39 99 91 25

5 388.00 757.79 71 114 146 144 29 153 120 166 130 171

6 389.64 739.37 108 162 88 112 29 150 108 168 117 172

7 433.38 296.79 46 25 19 28 17 58 82 102 78 32

8 421.54 389.23 29 38 64 62 17 93 86 123 68 58

9 396.20 648.45 91 97 78 138 23 127 83 152 123 153

10 381.90 884.01 157 163 164 156 63 154 118 182 127 170

11 381.91 867.52 157 156 158 152 34 160 118 185 132 173

12 384.42 823.16 126 163 141 140 35 162 111 180 131 164

13 409.54 506.79 60 50 117 102 26 64 64 109 147 94

14 403.52 560.54 51 57 97 109 21 76 91 161 115 144

15 392.98 686.49 91 97 78 138 23 127 146 152 123 153

16 379.84 939.74 171 169 171 158 94 169 117 188 133 176

Objective Functions Optimal Bus Locations and Capacity (kW)

S.N PV Cost
. Loss
(Rs. In 9 10 11 12 13 17 18 19 20 21
(MWh)
lacs)
17 421.01 400.75 54 41 56 24 26 79 77 137 105 60

51
 18 445.33 229.87 54 43 16 13 41 18 39 75 54 25

19 432.27 310.96 46 23 36 26 26 79 60 113 71 30

20 427.50 340.78 29 32 27 60 18 98 84 85 79 48

21 443.61 244.68 68 15 18 13 47 32 48 84 51 25

22 410.15 493.41 59 52 62 75 18 123 95 115 119 92

23 379.84 939.74 171 169 171 158 94 169 117 188 133 176

24 378.43 968.27 171 185 171 158 94 169 117 188 164 176

25 386.71 776.81 71 114 146 175 29 153 120 166 130 171

26 379.72 958.76 171 169 171 158 125 169 117 188 133 176

27 396.96 646.00 126 77 57 140 32 139 103 160 115 112

28 401.72 579.42 46 124 47 98 18 122 103 118 127 149

29 395.65 663.56 77 106 162 32 37 146 102 174 124 131

30 415.40 443.82 48 36 62 63 19 92 88 158 83 79

31 387.90 762.04 112 162 101 118 25 147 127 169 126 164

32 385.52 803.34 154 109 157 155 28 162 117 176 108 154

33 427.45 356.46 76 36 39 47 33 92 56 100 77 31

34 385.87 789.30 94 132 145 150 23 164 119 168 129 173

35 447.43 214.76 56 24 15 13 41 2 38 88 51 25

The maximum and minimum ranges of objective function values along with the PV
capacity is summarized in Table 5.7 below:

Table 5.9: Summary of multi-objective optimization results

S.N. Energy Loss (MWh) PV Cost in Total PV capacity(kW)

52
 lacs(Rs.)
1 378.43 968.27 1591
2 448.07 212.33 349

The optimal range of PV capacity vary from 349 kW to 1591 kW. Similarly, the
corresponding optimal range of total annual energy loss varies from 448.07 MWh to 378.43
MWh. This shows that the higher the values of PV smaller is the energy loss and vice versa.
It is observed from the Fig. 5.7 below that the relationship between PV capacity and energy
loss is non-linear.

460

440
Energy Loss (MWh)

420

400

380

360

340
300 500 700 900 1100 1300 1500 1700
PV Installed Capacity (kW)

Fig. 5.20: Effect of PV capacity of the energy losses

53
 1000
900
800
PV cost in lacs(Rs.) 700
600
500
400
300
200
100
300 500 700 900 1100 1300 1500 1700
PV Installed capacity (kW)

Fig. 5.21: Effect of PV capacity on PV cost

The cost of PV in multi-objective optimization varies from Rs. 212.33 lacs to Rs.
968.27 lacs. It is observed from the Fig. 5.8 that the cost of PV is directly proportional to the
installed capacity of the PV units. At lowest cost, only 349 kW SPV can be installed and for
highest PV capacity as 1591 kW maximum cost is required.
The selection of the best solution depends upon the purpose of the study and
importance given to the objective function. From the fuzzy set theory, the best compromise
solution is obtained at the total penetration capacity of 952 kW which is the intermediate
range of PV capacity. The energy loss at the best capacity is 401.72 MWh and the
corresponding PV cost is Rs. 579.42 lacs. The best result is shown in Table 5.8 below:

Table 5.10: The best result based on fuzzy theory

Total
Objective Functions Optimal Bus Locations and Capacity (kW) Capacity
(kW)
PV Cost
Loss
(Rs. In 9 10 11 12 13 17 18 19 20 21
(MWh)
lacs)

401.72 579.42 46 124 47 98 18 122 103 118 127 149 952

54
 5.3.4 Result Analysis

The solar energy potential for 30o inclined solar panel has been determined in Fig. 4.2
and the corresponding per unit capacity for 1 kWp solar module is given in Fig. 4.3 for
different seasons. For each season a lot of solar power generation has determined to the
selected site. The production of solar power is higher in the seasons spring and autumn has
identified than the other seasons. The outputs from solar has obtained from 6 am to 6 pm in
all the seasons and there are no outputs from the solar during night times.
The integration of above mentioned PV units in the radial distribution system reduces
power loss in the day time only. Therefore, at the peak time period there is no share of the
solar power to the grid. Hence, relatively smaller percentage of savings in the system loss as
assumed to the whole day availability of the solar power/ storage system.
Integration of PV units in the buses reduce the current flowing through the branch of
the network and loss is minimized. Similarly, voltage drop in the line is reduced and hence
the profile of the voltage level increases. System loss is proportional to the size of PV unit.
But greater the size increases the cost of the installation and maintenance. The system
loadability has improved due to the integration of PV at the best location and of optimal size.
In addition, that, voltage stability is improved and hence the quality of the power increases
significantly.

55
 Chapter - 6

CONCLUSION AND FUTURE WORK

6.1 General

Solar photovoltaic (SPV) is a type of distributed generation (DG) mostly connected in

a distribution network. Proper sizing and location of DG are important aspects for
distribution network which are to be examined. The benefits that DG brings to the network
can be best utilized if these resources have been properly allocated in the system. DG has
significant impact on power loss minimization and voltage profile improvement in the
system. However, inappropriate location and capacity of DG may impact adversely.
Therefore, location and size of DG unit plays important role in a power system.

6.2 Conclusion

In this study, optimal location and sizing of PV units has been implemented. A multi-
objective function was formed for optimum PV placement, considering minimization of
annual energy loss and SPV system cost. The solar energy resource assessment has been
done considering hourly seasonal values of solar radiation and the load profile of the system
was also considered hourly basis for different seasons. The backward/forward sweep method
was implemented in MATLAB to perform the load flow analysis. The multi-objective GA
program has formed using Optimization toolbox in MATLAB to find the Pareto-optimal
solution. Finally, a fuzzy-based technique has been utilized to select the best solution. The
proposed method was tested on 21 bus radial distribution system which is situated in
Kathmandu.
From the single objective optimization, energy loss minimization is achieved up to
18% by installing 1058 kW of PV units with SPV system cost Rs. 643.91 lacs. While from
the multi-objective optimization energy loss minimization is around 16.3% with annual PV
cost Rs. 579.42 lacs. The annual loss saving is proportional to the size of the PV unit which is

56
proportional to the PV cost. By minimizing the loss of the system, voltage level of the system
is also significantly improved.
In multi-objective optimization case, the results obtained for each objective function
is higher than the corresponding single objective function. Because GA takes care about the
other function while computing one single objective function.

6.3 Future Work

This work deals with the finding of optimal location and size of PV units and it is
tested for 11 kV, 21 nodes radial distribution system which is located in Kathmandu. Future
work would be to find the location and size of the renewable resource as well as
dispatchability based DG for mesh type of network using real and larger realistic system data
and consider space constraint for particular site. Future work would also be to consider
different models of DG and compare the results. Optimization function can be extended to
include other objective functions such as switching/reconfiguration of the distribution
network and this can be tested with different optimization techniques like Particle Swarm
Optimization (PSO) and other intelligent techniques.

57
 REFERENCES

[1] S. Regmi and S. Adhikari, "Solar Energy Potential in Kathmandu Valley, Nepal,"
Journal of Hydrology and Metrology, vol. 8, no. 1, pp. 77-82, 2012.
[2] T. Ackermann, G. Anderson and L. Soder, "Distributed generation: a definition,"
Electric Power System Research, vol. 57, pp. 195-204, 2001.
[3] C. Jha, "Placement of distributed generation units in a distribution network," AHEC,
IIT Roorkee, Roorkee, 2014.
[4] R. Shivarudraswamy, D. N. Gaonkar and N. Jayalakshmi, "GA based Optimal
Location and Size of the Distributed Generators in Distribution System for Different
Load Conditions," 1st IEEE International Conference on Power Electronics,
Intelligent Control and Energy Systems (ICPEICES), pp. 1-4, 2016.
[5] C. Nave, "The P-N junction Physics world," 2012. [Online]. Available:
http://hyperphysics.phy-astr.gsu.edu/hbase/solids/pnjun.html#c3. [Accessed 28
April 2017].
[6] R. Boylestad and L. Nashelsky, Electronic devices and circuit theory, New Jersey:
Pearson Prentice Hall, 2009.
[7] D. Tan and A. K. Seng, Hand book for solar photovoltaic systems, Singapore: The
Energy Market Authority and the Building Construction Authority.
[8] M. Suri, T. Cebecauer, A. Skoczek, R. Marais, U. Mushwana, J. Reinecke and R.
Meyer, "Cloud cover impact on photovoltaic power production in South Africa,"
2014.
[9] J. H. Teng, "A direct approach for distribution system load flow solutions," IEEE
Transactions on power delivery, vol. 18, no. 3, July 2003.
[10] J. A. M. Rupa and S. Ganesh, "Power flow analysis for radial distribution system
using backward/forward sweep method," International Journal of Electrical,
Computer, Energetic, Electronic and Communication Engineering, vol. 8, no. 10,
2014.
[11] K. R. Adhikari, S. Gurung and B. K. Bhattarai, "Solar Energy in Nepal and Global
Context," Journal of Institute of Engineering, vol. 9, no. 1, pp. 95-106, 2013.
[12] R. Bhandari and I. Stadler, "Electrification using solar photovoltaic system in
Nepal," Applied Energy, pp. 458-465, 2011.
[13] S. Pokharel, "Promotional issues on alternative energy technologies in Nepal,"
Energy Policy 31, pp. 307-318, 2003.

58
[14] A. Gurung, A. K. Ghimerey and S. H. A. Hassan, "The prospects of renewable
energy technologies for rural electrification: a review from Nepal," Energy Policy
40, pp. 374-380, 2012.
[15] S. KC, S. K. Khanal, P. Shrestha and B. Lamsal, "Current status of renewable
energy in Nepal: opportunities and challenges," Renewable and sustainable Review
15, pp. 4107-4117, 2011.
[16] S. Dubey, J. N. Sarvaiya and B. Seshadri, "Temperature dependent photovoltaic
(PV) efficiency and its effect on PV production in the world," Energy Proceedings,
vol. 33, no. 1, pp. 331-321, 2012.
[17] E. Phil, "Concentrating solar power," Energy Comittee of Royal Swedish Academy
of Sciences, Stockholm, 2009.
[18] M. H. Haque, "Efficient load flow method for distribution systems with radial or
mesh configuration," 1996.
[19] G. Meerimatha, G. Kesavarao and N. Sreenivasulu, "A novel distribution system
power flow algorithm using forward backward matrix method," IOSR Journal of
Electrical and Electronics Engineering, vol. 10, no. 6, pp. 46-51, Nov-Dec 2015.
[20] D. Bhujel, b. Adhikary and A. K. Mishra, "A load flow algorithm for distribution
system with distributed generation," in IEEE ICSET, Nepal, 2012.
[21] M. Paulescu, E. Paulescu, P. Gravila and V. Badescu, "Weather modeling and
forecasting of PV systems operation," Springer-Verlag, London, 2013.
[22] J. Widen, E. Wackelgard, J. Paatero and P. Lund, "Impacts of distributed
photovoltaic on network voltages: stochastic simulations of three Swedish low
voltage distribution grids," Electric Power Systems Research, vol. 80, pp. 1562-
1571, 2010.
[23] J. Solanki, S. K. Solanki and V. Ramachandran, "Steady state analysis of High
penetration PV on utility distribution feeder," in IEEE Conference, 2012.
[24] R. Shaha, N. Mithulananthana, C. R. Bansalb and V. K. Ramachandaramurthy, "A
review of key power system stability challenges for large-scale PV integration,"
Renewable and Sustainable Energy Reviews, vol. 41, pp. 1423-1436, 2015.
[25] J. Schoene, V. Zheglov, D. Houseman, J. C. Smith and A. Ellis, "Photovoltaic in
distribution systems-integration issues and simulation challenges," in IEEE, 2013.
[26] H. L. Willis, Power distribution planning reference book, Raleigh, North Carolina:
CRC Press, 2004.
[27] Enernex, "Integration issues and simulation challenges of high penetration PV,"
Ernex, Knoxville, 2011.

59
[28] A. Helal, M. Amer and H. Eldosouki, "Optimal location and sizing of distributed
generation based on genetic algorithm," 2012.
[29] D. K. Khatod, V. Pant and J. Sharma, "Evolutionary programming based optimal
placement of renewable distribution generators," IEEE transactions on power
systems, vol. 28, no. 2, pp. 683-695, 2 May 2013.
[30] F. Banihashemi, H. Lesani and A. S. Makhzani, "Locating and capacity
determination of distributed generations using non-dominated sorting genetic
algorithm," IEEE, 2013.
[31] D. Sosic and I. Skokljev, "Optimal location and sizing of photovoltaic plant," in
International Conference on Renewable Energy Research and Applications,
Madrid, Spain, 2013.
[32] A. K. Saonerkar and B. Y. Bagde, "Optimized DG placement in radial distribution
system with reconfiguration and capacitor placement using genetic algorithm," in
IEEE International Conference on Advanced Communication Control and
Computing Technologies, 2014.
[33] T. Walla, J. Widen, J. Johansoon and C. Bergerland, "Determining and increasing
the PV hosting capacity for photovoltaic in the Swedish distribution grids," in
European Photovoltaic Solar Energy Conference and Exhibition, Frankfurt, 2012.
[34] D. Q. Hung, N. Mitulananthan and K. Y. Lee, "Optimal placement of dispatchable
and non-dispatchable renewable DG units in distribution networks for minimizing
energy loss," Electric Power and Energy Systems, vol. 55, pp. 179-186, 2014.
[35] N. Acharya, P. Mahat and N. Mithulananthan, "An analytical approach for DG
allocation in primary distribution network," International Journal of Electrical
Power and Energy Systems, vol. 28, no. 10, pp. 669-678, 2006.
[36] V. V. S. N. Murthy and A. Kumar, "Comparison of optimal DG allocation methods
in radial distribution systems based on sensitivity approaches," Electrical Power
and Energy Systems, vol. 53, pp. 450-467, 2013.
[37] D. Singh, R. K. Mishra and D. Singh, "Effect of load models in distributed
generation planning," IEEE Transactions on Power Systems, vol. 22, no. 4, pp.
2204-2213, November 2007.
[38] S. Abdi and K. Afshar, "Application of IPSO-Monte Carlo for optimal distributed
generation allocation and sizing," Electrical Power and Energy Systems , vol. 44,
pp. 786-797, 2013.
[39] R. A. Jabar and B. C. Pal, "Ordinal optimization approach for locating and sizing of

60
 distributed generation," IET Generation, Transmission & Distribution , vol. 3, no. 8,
pp. 713-723, 2009.
[40] A. A. Ameri, C. Nichita, T. Riouch and R. El-Bachitiri, "Genetic algorithm for
optimal sizing and Location of multiple distributed generations in electrical
network," Wroclaw, Poland, 2015.
[41] L. D. Arya, A. Koshti and S. c. choube, "Distributed generation planning using
differential evolution accounting voltage stability consideration," Electrical Power
and Energy Systems, vol. 42, pp. 196-207, 2012.
[42] I. Ziari, G. Ledwich, A. Gosh, D. Cornforth and M. Wishart, "Optimal allocation
and sizing of DGs in Distribution Networks," IEEE, 2010.
[43] M. A. Junjie, W. Yulong and L. Yang, "Size and location of distributed generation
in distribution system based on immune algorithm," Systems Engineering Procedia,
vol. 4, pp. 124-132, 2012.
[44] A. Kumar and W. Gao, "Optimal distributed generation location using mixed
integer non-linear programming in hybrid electricity markets," IET Generation,
Transmission & Distribution , vol. 4, no. 2, pp. 281-298, 2010.
[45] A. M. El-Zonkoly, "Optimal placement of multi distributed generation units
including different load models using particle swarm optimization," IET
Generation, Transmission & Distribution, vol. 5, no. 7, pp. 760-771, 2011.
[46] L. Wang and C. Singh, "Reliability constrained optimum placement of reclosers and
distributed generators in distribution networks using an Ant Colony system
algorithm," IEEE Transactions on Systems, Man and Cybernetics - Part C,
Applications and Reviews, vol. 38, no. 6, pp. 757-764, 2008.
[47] R. D. Mohammedi, A. Helal, S. Arif and M. Mosbah, "Optimal DG placement and
sizing in radial distribution in radial distribution system using NSGA-II for power
loss minimization and voltage stability enhancement," International Review of
Electrical Engineering, 2013.
[48] V. Vahidinasab, "Optimal distributed energy resources planning in a competitive
electricity market: multi-objective optimization and probabilistic design,"
Renewable Energy, vol. 66, pp. 354-363, 2014.
[49] D. Lamsal, A. K. Mishra and P. Gautam, "Optimal location and sizing of distributed
generation: B-coefficient matrix approach," 12th IEEE Conference on Control and
Automation (ICCA), pp. 810-815, 1-3 June 2016.
[50] J. Le and Y. Qiang, "Maximizing DG Penetration in Power Distribution Networks

61
Based on Cost/Benefit Analysis," Proceedings of the 35th Chinese Control
Conference, pp. 2872-2877, 2016.

62
 APPENDIX: A

BASE CASE LOAD FLOW PROGRAM IN MATLAB

% MATLAB source code for base case load flow using

%backward/forward sweep method forming BIBC matrix
clear all; format compact; format short; close all;
LD=xlsread ('distdata.xlsx', 1, 'A2:E21');
BD=xlsread ('distdata.xlsx', 1,'F2:G21');
br=length(LD);
no=length(BD);
xlp=xlsread ('CF calculation.xls',1,'A14:X17')/1000;
lp=xlsread ('loadprofile.xlsx', 3,'A3:E26'); %load profile
MVAb=1;
KVb=11;
Zb=(KVb^2)/MVAb;
% Per unit Values
for i=1:br
R(i)=(LD(i,4))/Zb;
X(i)=(LD(i,5))/Zb;
end
for i=1:no
BD(i,1) =((BD(i,1))/(1000*MVAb));
BD(i,2) =((BD(i,2))/(1000*MVAb));
end
tic;
%code for bus-injection to branch-current matrix
bibc=zeros(size(LD,1), size(LD,1));
for i=1: size(LD,1)
if LD(i,2) ==1
bibc(LD(i,3)-1, LD(i,3)-1) =1;
else

63
 bibc(: LD(i,3)-1) =bibc (: LD(i,2)-1);
bibc(LD(i,3)-1, LD(i,3)-1) =1;
end
end
Pload=BD (: ,1);
Qload=BD (: ,2);
Ns= [92 92 90 91];
vbr=ones (20,1);
for k=1:4
for j=1: size(lp,1)
Pld(:,j)=lp(j,k+1)*Pload;
Qld( :,j)=lp(j,k+1)*Qload;
s(:,j)=complex(Pld(:,j),Qld(:,j));
Vo=ones(20,1);% initial bus votage value
itr=0;
v(:,j)=Vo;
diff=1;
% loop starts
while (diff>1e-4)
Vt=v(:,j);
%backward sweep
I=conj(s(:,j)./v(:,j));% injected current
Z=complex(R,X);%branch impedance
ZD=diag(Z);%making it diagonal
IB=bibc*I; %branch current
itr=itr+1;
%forward sweep
TRX=bibc'*ZD*bibc;
v(:,j)=Vo-TRX*I;
dVB=(Vt-v(:,j));
diff=abs(max(dVB));
end

64
 if abs(min(v(:,j)))<=abs(min(vbr))
vbr=v(:,j);
end
Ibr(:,j)=abs(IB);
temp=Ibr(:,j).^2.*R';
tempq=Ibr(:,j).^2.*X';
Pl(j)=sum(temp);
Ql(j)=sum(tempq);
end
smp(k)=sum(Pl);
smq(k)=sum(Ql);
end
AEL=smp*Ns';
REL=smq*Ns';
VB=vbr;
VB=[1+0*i;VB];
Vbus=abs(VB);
theta=angle(VB)*180/pi;
toc;
fprintf('\nNumber of iteration=%d\n',itr);
% voltage profile plot
x=1:length(Vbus);
plot(x,Vbus,'mo--');
title('Minimum bus voltage at different nodes');
xlabel('Bus Number')
ylabel('Bus Voltage Profile (in per unit)')
axis([1 no 0.91 1.005]);
grid on
% daily average energy loss plot
fprintf('\nTotal Annual Active Energy Loss(MWh)=%.4f\n',AEL);
fprintf('\nTotal Annual Reactive Energy Loss(MVARh)=
%.4f\n',REL);

65
n=length(Pl);
y=1:n;
Fig.=Fig.ure;
plot(y, Pl'*1000, 'rx--', y, Ql'*1000,'c.-');
title('Plot of Daily Power Loss and time in hour');
xlabel('Time in hour');
ylabel('Daily Active Loss(MW)');
legend({'Active Loss(MW)','Reactive Loss(MVAR)'});
axis([1 24 0 140]);
grid on
save('test.mat');

66
 APPENDIX: B

SINGLE OBJECTIVE OPTIMIZATION USING MATLAB GA

TOOLBOX

% main function for single objective GA Toolbox

function [val]=obj_func(dg)
load('test.mat');
Ns=[92 92 90 91];
% annual energy loss calculation
for j=1:size(lp,1)
for k=1:4
Pld(:,j)=lp(j,k+1)*Pload;
Qld(:,j)=lp(j,k+1)*Qload;
xpv=xlp(k,j)*dg;
P_new(:,j)=Pld(:,j)-xpv';
s(:,j)=complex(P_new(:,j),Qld(:,j));
Vo=ones(20,1);% initial bus votage
v(:,j)=Vo;
diff=1;
while (diff>1e-4)
Vt=v(:,j);
%backward sweep
I=conj(s(:,j)./v(:,j));% injected current
Z=complex(R,X);%branch impedance
ZD=diag(Z);%making it diagonal
IB=bibc*I; %branch current
%forward sweep
TRX=bibc'*ZD*bibc;
v(:,j)=Vo-TRX*I;
dVB=(Vt-v(:,j));
diff=abs(max(dVB));

67
end
%power loss calculation
Ibr(:,j)=abs(IB);
temp=Ibr(:,j).^2.*R';
smp(k)=sum(temp)
end
Pl(j)=smp*Ns';
% handling of voltage constraint
kp=10000;
t=0;
for i=1:size(R)
if abs(vbr(i))<0.95
vlim(i)=0.95;
elseif abs(vbr(i))>1.0
vlim(i)=1.0;
else
vlim(i)=abs(vbr(i));
end
t=t+(abs(vbr(i))-vlim(i))^2;
end
val=sum(Pl)+t*kp;
end
%Non-linear constrain function
function [c, ceq]=myconst(x)
load('test.mat','Pload');
ceq=[];
c=sum(x)-0.5*sum(Pload);
%main program for PV location and size
load('test.mat');
%objective function
FitnessFunction=@obj_func;
rng default;

68
 Nvars=20;
Vbus(1)=[];
x0=zeros(1,20);
ubx=0.4*ones(1,20);
%find candidate bus location
Norm=abs(Vbus)./0.95
k=find(Norm<=1.00);
for i=1:20
i~=k;
ubx(k)=0;
end
lb=zeros(1,20);
UB=ubx;
%A*x <=b (linear inequalities)
A=[];
b=[];
%Aeq*x=b (linear equalities)
Aeq=[];
beq=[];
%Non-linear constrain function
nonlcon=@myconst;
%GA options
v_old=Vbus;
%setting options
options=gaoptimset('InitialPopulation',x0,'PopulationSize',...
100,'Generations',50,'FitnessLimit',0','TolFun',0.00000, ...
'PlotFcns',{@gaplotbestf, @gaplotbestindiv});
%calling genetic algorithm function for optimization
[xpv, ploss]=ga(FitnessFunction, Nvars,A,b,Aeq,beq,lb,UB,...
nonlcon, options);
P_new=Pload-xpv';
S=complex(P_new,Qload);% complex power

69
Vo=ones(Nvar,1); % initial bus votage
v=Vo;
diff=1;
%main loop
while (diff>1e-4)
Vt=v;
%backward sweep
I=conj(S./v);% injected current
Z=complex(R,X);%branch impedance
ZD=diag(Z);%making it diagonal
IB=bibc*I; %branch current
%forward sweep
TRX=bibc'*ZD*bibc;
v=Vo-TRX*I;
dVB=(Vt-v);
diff=abs(max(dVB));
end
Ibr=abs(IB);
for i=1:size(LD,1)
Pl(i)=power(Ibr(i),2)*R(i);
Ql(i)=power(Ibr(i),2)*X(i);
End
PL=sum(Pl);
QL=sum(Ql);
v_old=[1+0*i;v_old];
th_old=angle(v_old)*180/pi;
v=[1+0*i;v];
vnew=abs(v);
theta=angle(v)*180/pi;
x=1:length(v_old);
Fig.ure
hold on

70
plot(x,v_old,'mx-', x, vnew, 'b*-');
title('Plot of Bus number & voltage profile');
xlabel('Bus Number');
ylabel('Bus Voltage Profile (in per unit)');
axis([1 no 0.90 1.025]);
legend('WithoutDG','WithDG');
grid on
disp(xpv');
fprintf('The minimum loss of the energy is=%f
MWhr\n',ploss);

71
 APPENDIX: C

MULTI-OBJECTIVE OPTIMIZATION USING MATLAB GA

TOOLBOX

%multiobjective optimization function

function [val]=multiobj(dg)
%%
load('test.mat');
Ns=[92 92 90 91];
for j=1:size(lp,1)
for k=1:4
Pld(:,j)=lp(j,k+1)*Pload;
Qld(:,j)=lp(j,k+1)*Qload;
xpv=xlp(k,j)*dg;
P_new(:,j)=Pld(:,j)-xpv';
sr(:,j)=complex(P_new(:,j),Qld(:,j));
Vo=ones(20,1);% initial bus votage
v(:,j)=Vo;
diff=1;
%iteration
while (diff>1e-4)
Vt=v(:,j);
%backward sweep
I=conj(sr(:,j)./v(:,j));% injected current
Z=complex(R,X);%branch impedance
ZD=diag(Z);%making it diagonal
IB=bibc*I; %branch current
%forward sweep
TRX=bibc'*ZD*bibc;
v(:,j)=Vo-TRX*I;
dVB=(Vt-v(:,j));

72
 diff=abs(max(dVB));
end
Ibr(:,j)=abs(IB);
temp=Ibr(:,j).^2.*R';
smp(k)=sum(temp);
end
Pl(j)=smp*Ns';
% handling of voltage constraint
kp=10000;
t=0;
for i=1:size(R)
if abs(vbr(i))<0.95
vlim(i)=0.95;
elseif abs(vbr(i))>1.0
vlim(i)=1.0;
else
vlim(i)=abs(vbr(i));
end
t=t+(abs(vbr(i))-vlim(i))^2;
end
val(1)=sum(Pl)+t*kp;%eof first objective function
%% second objective function
r=0.12; %discount rate is 12%
n=25;
sr=(r*(1+r)^n/((1+r)^n-1));
pv=sum(dg)*1000;
val(2)=pv*(sr*12445*4+301130.4881*4*0.5);
end
%multi-objective genetic algorithm main program
load('test.mat');
%objective function
FitnessFunction=@multiobj;

73
 Nvars=20;
ubx=0.2*ones(1,20);
k=find(abs(Vbus)>=0.95);
for i=1:20
i~=k;
ubx(k)=0;
end
UB=ubx;
%A*x <=b (linear inequalities)
A=[];
b=[];
%Aeq*x=b (linear equalities)
Aeq=[];
beq=[];
%GA options
options =gaoptimset('PopulationSize',100,...
'Generations',50,'PlotFcns',@gaplotpareto);
[xpv,Fval] = gamultiobj(FitnessFunction,Nvars,...
A,b,Aeq,beq,lbz,UB,options);

74