You are on page 1of 210

Fault detection, classification and

location in Underground Cables

By

Shimaa Hassan Sayed Barakat

A Thesis Submitted to Faculty of Engineering - Fayoum University


In Partial Fulfillment of the Requirements for Degree of

Master of Science
In
ELECTRICAL POWER ENGINEERING

Faculty of Engineering - Fayoum University


Fayoum - Egypt
2014

Fault detection, classification and location in


Underground Cables
By

Shimaa Hassan Sayed Barakat

A Thesis Submitted to Faculty of Engineering - Fayoum University in


Partial Fulfillment of the Requirements for Degree of

Master of Science
In
ELECTRICAL POWER ENGINEERING

Under the Supervision of


Main Supervisor
Prof. Dr. Mahmoud Magdy Bahgat Eteiba
Professor in Electrical Engineering Department
Faculty of Engineering - Fayoum University

Co-Supervisor

Dr. Wael Ismael Wahba


Lecturer in Electrical Engineering Department
Faculty of Engineering - Fayoum University

Faculty of Engineering - Fayoum University


2014

Fault detection, classification and location in Underground


Cables
By

Shimaa Hassan Sayed Barakat

A Thesis Submitted to Faculty of Engineering - Fayoum University in


Partial Fulfillment of the Requirements for Degree of

Master of Science
In
ELECTRICAL POWER ENGINEERING

Approved by the Examining Committee:


Prof. Dr Mahmoud Magdy Bahgat Eteiba

"Professor in Electrical Engineering Department - Faculty of Engineering Fayoum University" (Main Advisor)

Prof. Dr Essam El-Din Mohamed Aboul Zahab


"Professor in Electrical Power Engineering Department - Faculty of
Engineering - Cairo University"

Ass.Prof. Dr. Amr Mohamed Refaat Gody

"Associate Professor in Electrical Engineering Department - Faculty of


Engineering - Fayoum University"
Faculty of Engineering - Fayoum University
2014

ACKNOWLEDGEMENT

I would like to express my sincere gratitude to my


supervisors Prof. Dr. Mahmoud Magdy Bahgat Eteiba
and Dr. Wael Ismael Wahba for their advices, guidance,
continuous encouragement and their generous dedication
of precious time throughout the work of this thesis.
Furthermore, I would like to thank all of my
friends for their help and support.
Finally I dedicate the thesis to all the members of
my family for their moral support and patience during
this research work.

ABSTRACT
Electric power systems have rapidly grown for the
past fifty years. As a result, the number of cables in
operation and their total length had largely increased.
Underground cables have been widely implemented due
to reliability and environmental concerns. To improve
the

reliability

of

distribution

system,

accurate

identification of a faulted segment is required in order to


reduce the interruption time during fault. Speedy and
precise

fault

accelerating

location

plays

an

system restoration,

important

reducing

role

outage

in

time,

reducing great financial loss and significantly improving


system reliability.
Therefore,

this

study

presents

fault

identification, classification and fault location estimation


method

based

on

Discrete

Wavelet

Transform

and

Adaptive Network Fuzzy Inference System (ANFIS) for


medium voltage cable in distribution system.
Different faults and locations are simulated by
ATP/EMTP, and then certain selected features of the
wavelet transformed signals are used as an input for a
ii

training process on the ANFIS. Then an accurate fault


classifier and locator algorithm was designed, trained
and

tested

using

current

samples

only.

The

results

obtained from ANFIS output were compared with the


real output. From the results, it was found that the
percentage error between ANFIS output and real output
is less than three percent. Hence, it can be concluded that
the proposed technique is able to offer high accuracy in
both of the fault classification and fault location.

iii

TABLE OF CONTENTS

Title

Page

Acknowledgment

Abstract

ii

Table of Contents

iv

List of Tables

List of Figures

xii

List of Symbols and Abbreviations

xvii

1.0

2.0

Introduction
1.1

Problem of Statement

1.2

Objectives of The Thesis

1.3

Scope of the Thesis

1.4

Outline of thesis

1.5

Summary of Work

Literature Survey
2.1
1.1

Introduction

2.2

Methods of Fault Classification

2.2.1 Wavelet based method

2.2.2 ANFIS based method

iv

2.3

Methods of Fault Location

11

2.3.1 Circuit Theory Based Method

11

2.3.2 Traveling Wave Theory Based Method

13

2.3.3 Artificial Intelligence (AI) Based


Methods

3.0

4.0

17

General Fault Characteristics


3.1

Introduction

22

3.2

Definition of Fault

22

3.3

Distribution faults

23

3.4

Symmetrical and Asymmetrical Fault

25

3.5

Fault types

25

Adaptive Neuro Fuzzy Inference System (ANFIS)


4.1

Introduction

28

4.2

Fuzzy Logic

28

4.2.1

Fuzzy Sets

29

4.2.2

Membership function formulation


and parameterization

4.3

4.2.3

Combining fuzzy sets

33

4.2.4

Fuzzy If-Then Rules

37

The concept of adaptive Neuro fuzzy


inference system

4.4

31

ANFISs learning processes

38
39

4.5

5.0

4.4.1

Fuzzification

42

4.4.2

Stage of IF-THEN rule

44

4.4.3

Normalization

44

4.4.4

Defuzzification

45

4.4.5

Neuron addition

45

Good features of the ANFIS

46

Modeling and Simulation of the underground cable


5.1

Cable types and their characteristics

47

5.2

Types of power cable in medium voltage

48

5.2.1

Paper-insulated lead covered (PILC)


cables

5.2.2

Cross-linked polyethylene (XLPE)


cables

48

49

5.3

Software Implementation

50

5.4

The Alternative Transients Program (ATP)

50

5.4.1 Operating principles and capabilities of


ATP

5.5

5.4.2 Typical EMTP applications

53

Wavelet transform analysis

54

5.5.1

Theory of discrete wavelet


transform (DWT)

5.6

52

Modeling and Simulation of the underground


cable

vi

54

57

5.6.1 Cable modeling

57

5.6.2 ATP Implementation

62

5.6.2.1

No Fault (normal conditions)

62

5.6.2.2

Three Phase to Ground Fault

63

5.6.2.3

Single Line to Ground Fault

65

5.6.2.4

Double Line to Ground


Fault

5.6.2.5
5.7

Line to Line Fault

Discrete wavelet transform for fault


parameterization

5.8

Fault detection, Classification and Location


scheme

67
69

74

5.8.1 Fault detection scheme

74

5.8.2 Fault classification scheme

75

5.8.3 Fault Location scheme

81

5.8.3.1

Three Phase Fault

81

5.8.3.2

Single Line to Ground Fault

85

5.8.3.3

Line to Line Fault

88

5.8.3.4

Double Line to Ground


Fault

5.9

66

Fault Location System Tool


5.9.1 Components of the fault location
system tool

vii

91
94
95

6.0

Simulation Results
6.1

Introduction

101

6.2

ANFIS Fault Type classification Results

101

6.3

ANFIS Fault Location Estimation Training


Results
Three phase to ground ANFIS fault

6.3.1

location model
Double Line to Ground ANFIS fault

6.3.2

location model

6.3.3

Line to Line ANFIS fault location


model

6.3.4

Single Line to Ground ANFIS fault


location model

6.4

ANFIS Fault Location Estimation Testing


Results
6.4.1

ANFIS Fault Location Estimation


Testing Results

104
105
109
113

117

121

122

6.4.11

Three phase to ground fault

122

6.4.1.2

Double Line to Ground fault

122

6.4.1.3

Line to Line fault

122

6.4.1.4

Single Line to Ground fault

122

viii

7.0

Conclusion and Recommendation


7.1

Conclusions

133

7.2

Recommendation

134

135

REFERENCES

Appendix A
A.1

Development of the fault location system tool


(FL)

A.2

Creating and Training the ANFIS MODEL

146
158

Appendix B
B.1

ANFIS Fault Type classification Training


Results

ix

159

LIST OF TABLES

TABLE

TITLE

PAGE

3-1

Number of phases involved in each


fault measured in the EPRI fault study

24

5-1

System Data

58

5-2

Training scenarios for the simulation

61

5-3

Number of simulations

61

5-4

ATP simulation setting

62

6-1

Training target for ANFIS fault type


classification

103

6-2

The percentage average accuracy for


fault types (testing and training)

104

6-3

Three phase ANFIS fault location


training output results

107

6-4

Double Line to Ground (ABG) ANFIS


fault location training output results

111

6-5

Line to Line (AB) ANFIS fault


location training output results

115

6-6

Single Line to Ground (AG) ANFIS


fault location training output results

119

6-7

Three phase ANFIS fault location


testing output results

123

6-8

Double Line to Ground ANFIS fault


location testing output results

125

6-9

Line to Line ANFIS fault location


testing output results

127

6-10

Single Line to Ground ANFIS fault


location testing output results

129

xi

LIST OF FIGURES
FIGURE

TITLE

PAGE

1-1

Summary of work

4-1

Triangular function with a = 3, b = 6


and c = 8.

32

4-2

Generalized bell function with a = 2,


b = 4 and c = 6.

32

4-3

Symmetric Gaussian function with


= 2 and c = 5.

33

4-4

The concept of A B

34

4-5

(a) two fuzzy sets A and B; (b) A ; (c)


A B ; (d) A B.

36

4-6

A basic ANFIS model with two inputs


data and two MFs

43

5-1

Wavelet decomposition tree

56

5-2

Single line diagram for underground


cable model

57

5-3

Cable Configuration

58

5-4

Cable data (international cables Co.)

60

5-5

single line diagram of at no fault


(normal conditions)

63

5-6

current waveform at no fault (normal


conditions)

63

5-7

Single line diagram of three phase to


ground fault.

64

xii

5-8

Current waveforms at three phase to


ground fault (with fault
resistance=10 , Inception Angle
=135 and fault location=1.5 Km).

64

5-9

Single line diagram of single line to


ground fault

65

5-10

Current waveform at single line to


ground fault. (with fault
resistance=50 , Inception Angle
=45 and fault location=1.5 Km).

66

5-11

Single line diagram of double line to


ground fault

67

5-12

Current waveform at double line to


ground fault. (with fault
resistance=50 , Inception Angle
=90 and fault location=5 Km).

67

5-13

Single line diagram of line to line


fault

68

5-14

Current waveforms at line to line


fault. (with fault resistance=50 ,
Inception Angle =135 and fault
location=4 Km).

68

5-15

The DWT detailed coefficients at level 1


to level 5 for a three phase to ground
fault (Decomposition of phase A, with
fault resistance=100 , Inception Angle
=90 and fault location=1.375Km).

70

5-16

level 5 detailed coefficients of Three


Phase to Ground Fault case

71

xiii

5-17

level 5 detailed coefficients of Single


Line to Ground Fault case

72

5-18

level 5 detailed coefficients of


Double Line to Ground Fault case

72

5-19

level 5 detailed coefficients of Line to


Line Fault case

73

5-20

Input 1 Membership Functions of


ANFIS classification model

78

5-21

Input 2 Membership Functions of


ANFIS classification model

78

5-22

Input 3 Membership Functions of


ANFIS classification model

79

5-23

Input 4 Membership Functions of


ANFIS classification model

79

5-24

Input 1 Membership Functions of


(LLLG - ANFIS)

83

5-25

Input 2 Membership Functions of


(LLLG - ANFIS)

83

5-26

Input 3 Membership Functions of


(LLLG - ANFIS)

84

5-27

Input 4 Membership Functions of


(LLLG - ANFIS)

84

5-28

Input 1 Membership Functions of


(AG - ANFIS)

86

5-29

Input 2 Membership Functions of


(AG - ANFIS)

86

5-30

Input 3 Membership Functions of


(AG - ANFIS)

87

xiv

5-31

Input 4 Membership Functions of


(AG - ANFIS)

87

5-32

Input 1 Membership Functions of


(AB - ANFIS)

89

5-33

Input 2 Membership Functions of


(AB - ANFIS)

89

5-34

Input 3 Membership Functions of


(AB - ANFIS)

90

5-35

Input 4 Membership Functions of


(AB - ANFIS)

90

5-36

Input 1 Membership Functions of


(ABG - ANFIS)

92

5-37

Input 2 Membership Functions of


(ABG - ANFIS)

92

5-38

Input 3 Membership Functions of


(ABG - ANFIS)

93

5-39

Input 4 Membership Functions of


(ABG - ANFIS)

93

5-40

No Fault (normal conditions)

96

5-41

Three phase to ground fault

97

5-42

Single Line to ground fault

98

5-43

Double Line to ground fault

99

5-44

Line to Line fault

100

6-1

The performance graph for the


ANFIS Fault Type classification
model

103

xv

6-2

The performance graph for the three


phase ANFIS fault location model

106

6-3

Percentage error of three phase to


ground fault

109

6-4

The performance graph for the


ANFIS Double Line to Ground fault
location model

110

6-5

Percentage error of Double Line to


ground fault

113

6-6

The performance graph for the


ANFIS Line to Line fault location
model

114

6-7

Percentage error of Line to Line fault

117

6-8

The training performance graph for


the ANFIS Single Line to Ground
(AG) fault location model

118

6-9

Percentage error of Single Line to


ground fault

121

6-10

Percentage error of Three phase to


ground fault (testing).

125

6-11

Percentage error of Double Line to


Ground fault (testing).

127

6-12

Percentage error of Line to Line fault


(testing).

129

6-13

Percentage error of Single Line to


Ground fault (testing).

132

xvi

LIST OF ABBREVIATIONS

Wavelet approximation

ANFIS

ANN

Adaptive Neuro Fuzzy Inference


System
Artificial Neural Network

AI

Artificial Intelligence

ATP/EMTP

Alternative Transients Program /


Electromagnetic Transients Program

BPNN

Back Propagation Neural Network

Capacitance

Wavelet detail

db4

Daubechies 4

DWT

Discrete Wavelet Transform

EPRI

Electric Power Research Institute

ES

Expert System

Frequency

FACTS

FIS

Flexible Alternating Current


Transmission Systems
Fuzzy Inference System

FL

Fault location

GUI

Graphical User Interface

HIF

High Impedance Faults

HP

High Pass Filter

HVDC

High Voltage Direct Current


xvii

Inductance

L-G

Single Line to Ground

LIF

Low Impedance Faults

L-L

Line to Line

L-L-G

Double Line to Ground

L-L-L-G

Three Phase to Ground

xviii

CHAPTER 1
Introduction
Electrical

power

distribution

systems

are

responsible for supplying power to dispersed residential,


commercial and small industrial customers in a safe,
reliable and economical fashion. This is achieved by
maintaining a reliable voltage level, correcting the power
factor through use of reactive compensation and offering
as close to continuous service as possible in order to
meet demand.

1.1 Problem of Statement


Electric utilities often face the problem of finding
the exact location of a failure in a distribution cable.
These failures often occur at the worst possible time and
cause the maximum amount of inconvenience to utility
customers. The utility must quickly find and isolate the
failure to restore electric service.
Therefore, a fastest cable fault location technique
has to be studied to detect and locate cable fault as soon
as

possible

in

the

system.

Through

fault

location

1 - Introduction
prediction of cable, troubled cable can be detected and
located earlier and accurately.

1.2 Objectives of the thesis


The objectives of the thesis are:
1. To

review

classification

the
and

approach
location

of

for

fault

underground

cable.
2. To simulate the network using ATP/EMTP
and MATLAB software.
3. To classify the type of fault in underground
cable.
4. To estimate fault location of underground
cable.

1.3 Scope of the Thesis


The scope of this thesis is to explore an approach
to fault location estimation and classification for medium
voltage underground cable by using Adaptive Network
Fuzzy Inference System (ANFIS). With the ANFIS Fault
Location

Estimator

capabilities,

fault

area

can

be

determined and with the ANFIS Fault Type Classifier,


fault type can be determined. The network was simulated
using ATP/EMTP and MATLAB software.
2

1 - Introduction
1.4 Outline of thesis
This thesis consists of seven chapters. In first
chapter, it discusses about the background of the thesis,
statement of problem, objective of the thesis, scope of
the thesis, outline of thesis as long as summary of work.
While

Chapter

2,

discusses

more

on

theory

and

literature reviews that have been done.


Chapter 3 well discusses about definition of fault,
the types of fault, and the types of power cable in
medium voltage cable.
Chapter

well

discusses

general

view

of

Adaptive Network Fuzzy Inference System (ANFIS).


Chapter

discusses

on

the

methodology

and

software implementation of this thesis. In this thesis,


ATP and MATLAB were used.
Chapter 6 discusses on the result and discussion for
both software. On the other hand, Chapter 7 discusses on the
conclusion and recommendation for the thesis.

1.5 Summary of Work


Figure 1.1 represents the summary of work for this
thesis. During first stage, works are more focusing on
3

1 - Introduction
theoretical studies to enhance knowledge on the thesis done.
Besides that, softwares that are used for the thesis also are
study.
In the second stage the works are more focusing on
simulation and modification. Results obtained from the
simulation of both software (ATP and MATLAB) were discuss
and then tested.

1 - Introduction
Construct underground cable model in ATP/EMTP

Generate current samples using ATP/EMTP


simulation of different fault conditions

Preprocessing of the data

Perform Wavelet Decomposition of data and obtain


detailed coefficients at level 5

Select training and test sets

Select number of inputs and grouping data into patterns

Design the fault classification and fault location


ANFIS models

Training and testing

Compare the ANFIS output with the target

FIGURE 1-1: Summary of work


5

CHAPTER 2
Literature Survey
2.1 Introduction
Various fault classification and location methods
have been proposed and they can be categorized as
analytical

methods,

artificial

intelligence

(AI)

based

methods, travelling wave methods and software based


methods.

2.2 Methods of Fault Classification


Fault classification identifies the type of fault on
the fault lines in the system. There are different methods
proposed for the fault classification. This section reviews
wavelet based method; ANFIS based method, and hybrid
method which are useful in detecting the type of fault.
2.2.1 Wavelet based method
The wavelet based approach for fault classification
is implemented using multi resolution analysis (MRA) of
voltage and current signals [2-9]. MRA is the process of
decomposing a signal into different levels of resolutions
[1]. Initially, the signal is passed through two discrete
wavelet transform (DWT) filters high pass (HP) filter
6

2 - Literature Survey
and low pass (LP) filter. The samples of output signal
from the HP filter are known as detail coefficients, and
those from the LP filter are known as approximated
coefficients. This is considered the first level. Next, the
obtained samples from LP filter (first level approximated
coefficients) are further processed through HP and LP
filters forming the second level. In this way, the original
signal can be processed through many levels depending
on the need of resolution of signal. Reference [1]
proposes

wavelet

based

technique

for

fault

classification in transmission lines.


In [5], in order to detect, classify the fault and
discriminate the transients and the reflected signal from
noise,

new

algorithm

based

on

multiple

scale

correlation of wavelet transform was presented using


current signal from one end. By this algorithm, faulty
phase

can

be

detected

and

classified

by

the

approximation components of three phases on 4th level.


Then the details at first level are rescaled until the clear
transients are identified. It proved that the noises can be
significantly removed by the proposed algorithm.

2 - Literature Survey
The applications of the discrete wavelet transform
(DWT) for identifying the phase with fault appearance
along the underground cable distribution system have
been investigated in [6]. Daubechies4 (db4) is employed
as mother wavelet in order to decompose high frequency
components from fault signals. Coefficients of positive
sequence current signals are calculated and employed in
fault detection decision algorithm. The results show that
the proposed algorithm can indicate fault types with the
accuracy higher than 90%.
In

[7],

technique

using

discrete

wavelet

transform in combination with back-propagation neural


network (BPNN) in order to classify of fault type in
underground

distribution

system

has

been

proposed.

Daubechies4 (db4) is employed as mother wavelet in


order to decompose high frequency components from
fault signals. The maximum values from the first scale at
cycle of phase A, B, C and zero sequence of post-fault
current

signals

obtained

by

the

discrete

wavelet

transforms have been used as an input for the training


process of the BPNN in a decision algorithm.

2 - Literature Survey
[8]

shows

that

studying

on

respective

characteristics of wavelet analysis and curve fitting, a


flat coefficient is defined to distinguish the mutation of
the signal, then combining wavelet analysis with curve
fitting method to detect the pulse starting point of
traveling wave, and then calculating the fault distance.
[9] Proposes a fault detection and classification
scheme based on discrete wavelet analysis for power
cables. The property of multi resolution in time and
frequency provided by wavelets allows an accurate time
location of fault transients. Numerical test results of the
scheme are very encouraging.
2.2.2 ANFIS based method
ANFIS is based on fuzzy logic modeling and uses
artificial neural network as the learning algorithm. The
fuzzy logic can map inputs to outputs based on the
predefined if then rules. In the power system, the fault
data can be uncertain to make a decision regarding fault
location. This type of data is known as fuzzy data. Fuzzy
logic methods can model the uncertainty in the available
data and can help in estimating the fault classification
and location [4,10-12,16 ].
9

2 - Literature Survey
[10] Introduces the application of Adaptive NeuroFuzzy Inference System (ANFIS) for fault classification
in transmission lines. The proposed method uses two
ANFIS units. The fault detection unit is built by using
various training data at fault and no fault conditions. The
fault classification unit is built at different situations of
all fault types (i.e. single line to ground, double lines,
double lines to ground and three lines fault). Then the
unit is tested using testing data different from those of
the training stage.
[11] Proposed an ANFIS based fault classification
scheme in neutral non-effectively grounded distribution
system. The transient currents are obtained by wavelet
transform after faults occur. According to the statistic
characteristic
types,

the

of
fault

transient

currents

identifiers

are

in

different

defined.

The

fault
fault

identifiers can characterize the traits of fault type and


show different disciplinarian in different fault types.
They are inputted into three ANFISs to obtain the fault
type. The proposed approach only needs the voltages and
currents measured at substation, and can identify ten
types of short-circuit fault accurately.

10

2 - Literature Survey
In

[12],

an

algorithm for

fault

detection

and

classification of low impedance faults (LIF) and high


impedance

faults

(HIF)

has

been

developed

using

ANFIS. The proposed algorithm comprises of three basic


steps,

acquire

three-phase

currents,

calculate

zero

sequence current and deduce fault type using ANFIS.

2.3 Methods of Fault Location


After

fault

classification,

the

next

step

is

to

determine the fault location. The fault location identifies


the physical position of fault in the power system. This
information is very useful for isolating the fault and
restoring the power immediately. So far, there are three
major approaches for determining the location of fault.
They are:
Circuit theory based method
Traveling wave theory based method
Artificial Intelligence (AI) based method
2.3.1 Circuit Theory Based Method
The circuit theory based method locates the fault
using voltage and current values and impedance changes
[13-19]. After finding the fault type, fault voltage and
11

2 - Literature Survey
current values are used to calculate apparent impedance.
Based on the equation of apparent impedance, fault
location is estimated [13].
Reference [14] shows the implementation of the circuit theory based
method for rural distribution feeders. Depending on the fault type,
specific voltage and current values are selected to find out the
apparent impedance. The equation of apparent impedance consists
of unknown terms like fault resistance and distance of fault. By
separating the apparent impedance equation into real and imaginary
parts, the fault location is calculated. In this method, the equations
are shown only for a line-to-ground fault but a similar analysis can
be performed for other types of faults also. Instead of eliminating
the fault resistance, the equation of distance of fault is expressed as
a function of fault resistance in [15]. As the equation is in the form
of a square root of quadratic equation, the minimum value of fault
resistance is calculated and substituted in the equation of distance of
the fault to find the location of the fault. In [16], the fault location is
estimated by comparing the reactance of different sections of line
with apparent reactance. If the modified reactance of a section is not
greater than the apparent reactance, then the fault location is
considered to be present after that section. Initially, fault location is
assumed between two points based on the information of fault type
and sequence voltage and currents. Then, the modified reactance of
section is compared with the apparent reactance to find the accurate

12

2 - Literature Survey
location of fault between those two assumed points. If there are any
subsections on the lines such as laterals, then, an equivalent radial
system is considered and the relationship between voltage and
current is analyzed to find the location of fault.
As this involves more calculations, another approach for fault
location is presented in [17]. For balanced systems, the sequence
components of phase voltages, phase currents and impedances are
used to form a quadratic equation in terms of fault distance and fault
resistance. Unknown fault resistance is calculated by the imaginary
part of this equation and the fault resistance is substituted in the real
part of the equation to obtain the fault location. For unbalanced
systems, direct circuit analysis is used to form a quadratic equation
in terms of fault distance and fault resistance. In between, matrix
inverse lemma technique is used for calculating inverse operations
of matrices. Then, the quadratic equation is solved to find the fault
location.

2.3.2 Traveling Wave Theory Based Method


The traveling wave theory based method uses the
information of voltage and current traveling waves for
locating the fault. At the point of fault location, voltage
and current waveforms get suddenly disturbed. Such
disturbance causes them to propagate through the power
system. Based on the total length of the faulted line,
13

2 - Literature Survey
propagation velocity and time of waveforms, the location
of fault is known. This technique is mostly used for extra
high voltage transmission lines than distribution lines
because distribution system contains many subsections
such as laterals and feeders. These subsections may
create disturbances for the traveling waves on the lines.
Reference [20] explains fault location technique based
on the traveling wave theory. It includes two methods
the single ended method and the double ended method.
In the single ended method, only one fault locator is
installed at one substation.
Depending

on

the

arrival

time

of

waves

at

that

substation, the location of the fault is estimated. In


double

ended

method,

there

are

two

fault

locators

installed at two substations and depending on the arrival


time of waves at both stations, the fault location is
calculated. This process involves using costly equipment
such as GPS (Global Positioning System) to record the
exact arrival times. This paper [20] also states that the
double ended method worked better than the single
ended method when applied to a distribution system.

14

2 - Literature Survey
In

[21],

wavelet

transform is

used

to

analyze

the

recorded waveforms at one end or both ends of the line.


For double ended method, wavelet transform is used to
examine the waveforms with a better resolution that are
recorded at two ends. Then, based on the velocity and
arrival times of waves, fault location is found. A similar
approach is used for the single ended method, but due to
the

problems

in

reflected

waveforms,

separate

approaches are derived for ungrounded and grounded


faults. Another traveling wave theory approach in [22]
locates the fault by using the high frequency components
in the recorded waveforms of voltages and currents at
the substation. The recorded waveforms are decomposed
into a particular level of wavelet coefficients. The level
is decided based upon the selected high frequency. By
examining the wavelet coefficients, the faulted section in
the distribution system is identified. Then, based upon
the equivalent impedances of the remaining sections, an
equivalent

system

is

modeled,

and

the

apparent

impedance based technique explained in [14] is used for


locating the fault.
Also, the traveling wave theory can be applied for
transmission

lines

and

underground
15

cables.

In

[23],

2 - Literature Survey
wavelet analysis based single ended method is used to
locate the fault on the transmission line. In [24], double
ended based method is used to find the fault location in
underground cables.
In [25] a new single ended fault location method is
proposed for underground cable combined with overhead
lines. In this algorithm fault clearing high frequency
transients are used instead of fault-generated transients
and the line parameters are not needed. In the proposed
algorithm, samples just from voltage transients generated
by fault clearing action of circuit breaker are taken from
the sending end of the cable line. Applying wavelet
transform, the first three inceptions of traveling waves to
the fault locator are detected. Using these, the proposed
algorithm at first identifies fault section, overhead or
cable, and then wave speed is calculated and at last
location of fault is determined accurately. Because of
using only voltage samples taken from one terminal, it is
simple and economic and does not need to GPS and data
communication and synchronization.

16

2 - Literature Survey
2.3.3 Artificial Intelligence (AI) Based Methods
The following are the three major approaches in
AI based methods.
Artificial Neural Networks (ANN) based method.
Fuzzy Logic based method.
Expert System based method.
ANN based method detects the location of fault in
power system by recognizing the patterns of voltages
and currents for different fault locations. References [26,
27] suggest separate fault locators for each type of fault
whereas references [28, 29] propose only one single fault
locator

for

all

types

of

faults.

Depending

on

the

complexity in the problem, the design of the fault locator


is decided.
Reference

[27]

shows

that

the

Marquardt-

Levenberg technique gives better results when compared


with the backpropagation technique. But in [27], faulted
phase voltages

and

currents

along

with

the

neutral

current are used as inputs for training the three layer


feed-forward ANN to estimate the fault location in the
cable transmission.

17

2 - Literature Survey
Also, in [26], voltages and currents of faulted
phases along with the neutral current are used as inputs
for a three layer feed-forward ANN to detect the fault
location

in

medium

voltage

cables.

The

Marquardt-

Levenberg technique is used for training all the ANN


fault locators.
Reference [30] explains the expert system method
for distribution system faults, and the whole program is
developed by using the artificial language PROLOG.
This method only detects the faulted section. Also, the
expert system in [31] finds the only faulted section in the
transmission network. But in [32], the exact location of
the fault is determined by an expert system for high
voltage underground distribution cables.
[33] Presents an approach to fault location for a
radial cable power distribution network. A pipe type
single core cable including core and ungrounded sheath
is modeled in LCC support routine of the ATP software,
and voltage and current signals are obtained at the
sending end. The transient signals are then decomposed
into their modal components. At the second stage, useful
information is extracted by signal processing using the
18

2 - Literature Survey
wavelet

transform

and

finally

the

fault

position

is

localized through a fuzzy logic system.


[34] Presented an algorithm for locating faults in
a

combined

overhead

transmission

line

with

underground power cable using ANFIS. The proposed


method uses 10 ANFIS networks and consists of 3
stages, including fault type classification, faulty section
detection and exact fault location. In the first part, an
ANFIS is used to determine the fault type, applying four
inputs,

i.e.,

currents

and

fundamental
zero

component

sequence

current.

of

three

Another

phase
ANFIS

network is used to detect the faulty section, whether the


fault is on the overhead line or on the underground cable.
Other eight ANFIS networks are utilized to pinpoint the
faults (two for each fault type). Four inputs, i.e., the dc
component of the current, fundamental frequency of the
voltage and current and the angle between them, are used
to train the neuro-fuzzy inference systems in order to
accurately locate the faults on each part of the combined
line.
[35] Proposed a fault location approach based on
the current waveforms measured at the power substation,
19

2 - Literature Survey
the knowledge of protective device settings and ANFIS
nets. Wavelet transform is used to obtain the current
patterns

in

this

paper.

The

proposed

methodology

consists on training the ANFIS system with a fault


database registers obtained from the power distribution
system. The input data includes the current patterns
obtained from the measured signals registered at the
power

distribution

substation.

The

output

data

corresponds to the zone where the fault is located in the


power system. The performance of the ANFIS nets was
good and the 99.14% of the current patterns were
correctly classified.
In [36], a methodology based on the wavelet
transform

(WT)

and

ANFIS

techniques

for

fault

diagnosis in series compensated transmission lines has


been proposed. One feature is typically extracted for
each MRA level in literature studies including the WT
based fault diagnosis. In proposed technique, the number
of features is less than selected MRA levels. Therefore,
the proposed approach needs less memory space and less
computing time at both the training and testing processes
for ANFIS algorithms. The feasibility of the proposed
approach has been tested on an extensive data set of
20

2 - Literature Survey
23,436 fault cases covering a wide range of operating
conditions. Results of this paper show that the algorithm
has an effective structure in fault diagnosis for series
compensated transmission lines.
In [37], The Proposed ANFIS uses only post-fault
three-phase RMS currents as inputs, and outputs were XY points and integers between 1 and 10. The results
show that the approach can accurately locate the faults
and identify the faults types in the DISCO-Net 47-nodess
test practical distribution system. Its accuracy has been
validated by comparing the actual fault locations and
types

of

fault

obtained

from

simulations and with ANN approach.

21

the

PSS-ADEPT

CHAPTER 3
General Fault Characteristics
3.1 Introduction
Underground cables are used in 33kV, 11kV and
415V distribution networks. Faults are major disturbance
to the power system. Hence the need to locate the faulty
point in an underground cable in order to facilitate
quicker

repair,

improve

the

system

reliability

and

reduced outage period.


Power cable fault location techniques are used in
power

system for

accurate

pinpointing

of

the

fault

positions. The benefits of accurate location of fault are:


1. Fast repair to restore back the power system.
2. Improve

the

system

availability

and

performance.
3. Reduce operating cost and save the time
required

by

the

crew

searching

in

bad

weather, noisy area and tough terrains.

3.2 Definition of Fault


One of the definitions of a fault is (ANSI/IEEE
Std. 100-1992):
22

3 - General Fault Characteristics


Fault: A physical condition that causes a device, a
component, or an element to fail to perform in a required
manner; for example, a short circuit or a broken wire. A
fault almost always involves a short circuit between
energized phase conductors or between a phase and
ground. A fault may be a bolted connection or may have
some

impedance

in

the

fault

connection.

The

term

fault is often used synonymously with the term short


circuit defined as (ANSI/IEEE Std. 100-1992):
Short circuit: An abnormal connection (including an arc)
of relatively low impedance, whether made accidentally
or

intentionally,

between

two

points

of

different

potential. (Note: The term fault or short-circuit fault is


used to describe a short circuit.)
3.3 Distribution faults
Distribution faults occur on one phase, on two
phases, or on all three phases. Single-phase faults are the
most common. Almost 80% of the faults measured
involved only one phase either in contact with the
neutral or with ground (see Table 3.1).

23

3 - General Fault Characteristics


Table 3-1 Number of phases involved in each fault measured in
the EPRI fault study

As another data point, measurements on 34.5-kV


feeders found that 75% of faults involved ground (also
54% were phase to ground, and 15% were phase to
phase) [39]. Most faults are single phase because most of
the overall length of distribution lines is single phase, so
any fault on single-phase sections would only involve
one phase. Also, on three-phase sections, many types of
faults tend to occur from phase to ground. Equipment
faults and animal faults tend to cause line-to-ground
24

3 - General Fault Characteristics


faults. Trees can also cause line-to-ground faults on
three-phase structures, but line-to-line faults are more
common. Lightning faults tend to be two or three phases
to ground on three-phase structures.

3.4 Symmetrical and Asymmetrical Fault


Symmetrical or balanced fault is defined as the
simultaneous short circuit across all the three phases. It
happens rarely and the analysis for this fault is easy.
In contrast, asymmetrical fault is when the fault is no
longer

balanced.

In

practice,

most

faults

in

power

systems are unbalanced. However, asymmetric faults are


difficult to analyze. The common types of asymmetric
faults are single phase to ground fault, line to line fault,
double line to ground fault [40].

3.5 Fault types


Faults

are

either

temporary

or

permanent.

permanent fault is one where permanent damage is done


to the system. This includes insulator failures, broken
wires,

or

failed

equipment.

Virtually

all

faults

on

underground equipment are permanent. Most equipment


fails to a short circuit. A temporary fault does not
permanently

damage

any
25

system

equipment.

If

the

3 - General Fault Characteristics


circuit is interrupted and then reclosed after a delay, the
system

operates

normally.

Temporary

faults

cause

sustained interruptions if the fault is downstream of a


fuse, and fuse saving is not used or is not successful. For
temporary faults on the feeder backbone, all customers
on the circuit are momentarily interrupted. Faults that are
normally temporary can turn into permanent faults. If the
fault is allowed to remain too long, the fault arc can do
permanent damage to conductors, insulators, or other
hardware.
The most common types of cable fault are either
contact faults or breaks. The contact fault is described as
a short or shunt fault and is a connection or part
connection between core and core or between cores to
sheath. The value of fault resistance varies from zero
ohms to many mega ohms. On the other hand, break or
also called an open circuit or series fault can be a clean
break in a conductor, with infinite or very high resistance
reading across the break and to adjacent metal. There can
also be a dirty break where there is a measurable
resistance across the gap and/or to adjacent metal.
A partial break can occur when some of the
strands of a conductor are broken or burnt through.
26

3 - General Fault Characteristics


Another type of cable fault is flashing fault. This type of
fault does not manifest itself at lower voltages but
flashes over at a certain higher voltage threshold. This
may be several hundreds of volts or several kilovolts up
to a maximum which is the accepted dc test voltage for
the cable. Such a fault is acting like a spark gap. Lastly is
the ingress of moisture. Moisture usually produces a
contact fault involving all cores. Water enters a cable at
some point of damage and maybe present in one limited
stretch or it may spread many meters along the cable,
typically as far as next joint. Plastic cable can often
perform when wet but not for paper cables. Change of
characteristic impedance occurs at the wet section.

27

CHAPTER 4
Adaptive Neuro Fuzzy Inference System (ANFIS)
4.1 Introduction
Adaptive neuro fuzzy inference system (ANFIS) is
based on fuzzy logic modeling and uses artificial neural
network as the learning algorithm. The system can teach,
change the data environment or respond to the remote
stimulus for adapting to the change of data environment
[41]. ANFIS produces constant and linear target by using
respective zero and first-order polynomial equations and
is also known as a Sugeno-type of fuzzy inference
system (FIS).

4.2 Fuzzy Logic


L. Zadeh who published his first work on fuzzy
sets in 1965 [19] introduced the theory of fuzzy logic.
The basic idea of fuzzy logic is to allow not only the
values 1 and 0, corresponding to true and false but the
whole interval [0, 1] as degrees of truth. This leads to a
radical extension of classical logic. Zadeh was not the
first one to introduce a multi-valued logic calculus. In
the twenties, J. Lukasiewicz [20] had already developed
multi-valued

logic

calculus
28

but

its

application

was

4 - Adaptive Network Fuzzy Inference System (ANFIS)

limited until the introduction of computer technology in


the late fifties.
Although

critics

continually

state

that

all-

important problems can be dealt with by classical means,


use of fuzzy logic has become more wide spread. In
particular, this is indicated by the success of fuzzy logic
in the discipline of control techniques. In the beginning
of the seventies Zadeh [44] introduced the concept of
fuzzy logic control. In recent years it has been shown
that,

with

special

continuous

function

kind

of

fuzzy

controllers,

on

compact

set

can

each
be

approximated to any degree of accuracy [45].

4.2.1 Fuzzy Sets


We

obtain

fuzzy

sets

by

extending

the

membership predicate "" to the interval [0, 1] instead


of only using the classical truth-values 0 and 1. This
means that a set can contain points with a certain degree.
This

degree

of

membership

can

be

considered

in

different ways. On the one hand the membership grade


can be interpreted as a grade of probability; on the other
hand it can be regarded as a grade of possibility. In any
case fuzzy set theory allows us to deal with data
29

4 - Adaptive Network Fuzzy Inference System (ANFIS)

associated with uncertainty. This is an advantage we can


use within neural networks.
Definition 4.1 Fuzzy sets and membership functions
If X is a collection of objects denoted generically
by x, then a fuzzy set A in X is defined as a set of
ordered pairs:
=

( ) |

(4.1)

( ) is called the membership function

where

(MF) for the fuzzy set A. The MF maps each element of


X

to

membership

grade

(or

membership

value)

between 0 and 1.
The definition of a fuzzy set is a simple extension
of

the

definition

of

classical

set

in

which

the

characteristic function is permitted to have any values


between 0 and 1. If the value of the membership function
( ) is restricted to either 0 or 1, then A is reduced to a

classical set and

( ) is the characteristic function of

A. For clarity, we shall also refer to classical sets as


ordinary sets, crisp sets, non-fuzzy sets, or just sets.

30

4 - Adaptive Network Fuzzy Inference System (ANFIS)

Usually X

is

referred

to

as

the

universe

of

discourse, or simply the universe and it may consist of


discrete objects or continuous space.

4.2.2

Membership

function

formulation

and

parameterization
Since most fuzzy sets in use have a universe of
discourse X consisting of the real line R, it would be
impractical to list all the pairs defining a membership
function. A more convenient way to define a MF is to
express it as a mathematical formula. Here are some of
the MFs used in the fuzzy sets literature
Definition 4.2 Triangular function [46]
A triangular MF is specified by three parameters
{a,b,c} as follows:

( , , , )=

,
,

,
,

(4.2)

The parameter {a, b, c} (with < < ) determine the


x coordinates of the three corners of the underlying
triangular MF.

31

4 - Adaptive Network Fuzzy Inference System (ANFIS)

FIGURE 4.1 Triangular function with a = 3, b = 6 and c = 8.

Definition 4.3 The generalized bell function [46]


The generalized bell function depends on three
parameters a, b, and c as given by
( , , , )=

1+

(4.3)

where the parameter b is usually positive. The


parameter c locates the center of the curve.

FIGURE 4.2 Generalized bell function with a = 2, b = 4 and c = 6.

32

4 - Adaptive Network Fuzzy Inference System (ANFIS)

Definition 4.4 The symmetric Gaussian function [46]


The symmetric Gaussian function depends on two
parameters and c as given by
( ,

, )=

(4.4)

FIGURE 4.3 Symmetric Gaussian function with = 2 and c = 5.

4.2.3 Combining fuzzy sets


Fuzzy sets are combined in the application of
fuzzy reasoning. The combination of fuzzy sets can be
obtained

using

complement

intersection

(NOT)

(AND),

operations.

union
The

(OR)

and

mathematical

description of these operations for a set x and a subset A


are given in definitions 4.5 to 4.8.

33

4 - Adaptive Network Fuzzy Inference System (ANFIS)

Definition 4.5 Containment or subset


Fuzzy set A is contained in fuzzy set B (or,
equivalently, A is a subset of B) if and only if A (x)
B (x) for all x in X. Figure 4.4 illustrates the concept of
A B. In symbols,

A B A (x) B (x) x X

(4.5)

Figure 4.4 The concept of A B

Definition 4.6 Union (disjunction)


The union of two fuzzy sets A and B is a fuzzy set
C, written as C = A B or C = A OR B, whose MF is
related to those of A and B by

C (x) = max ( A (x) , B (x) ) = A (x) B (x) x X

34

(4.6)

4 - Adaptive Network Fuzzy Inference System (ANFIS)

A more intuitive but equivalent definition of union is the


"smallest" fuzzy set containing both A and B. Alternatively, if
D is any fuzzy set that contains both A and B, then it also
contains A B.

Definition 4.7 Intersection (conjunction)


The intersection of two fuzzy sets A and B is a
fuzzy set C, written as C = A B or C = A AND B,
whose MF is related to those of A and B by
C (x) = min ( A (x) , B (x) ) = A (x) B (x) x X

As in the case of the union, it is obvious that the

intersection of A and B is the "largest" fuzzy set which is


contained in both A and B. This reduces to the ordinary
intersection operation if both A and B are non-fuzzy.
Definition 4.8 Complement (negation)
The complement of fuzzy set A, denoted by A ( ,

NOT A), is defined as

(x) = 1 - A (x) x X

(4.8)

Figure 4.5(a) illustrates two fuzzy sets A and B;


Figure 4.5(b) is the compliment of A; Figure 4.5 (c) is

35

(4.7)

4 - Adaptive Network Fuzzy Inference System (ANFIS)

the union of A and B; and Figure 4.5 (d) is the


intersection of A and B.

Figure 4.5 (a) two fuzzy sets A and B; (b) A ; (c) AB ; (d) A B .

These fuzzy set operations perform exactly as the


corresponding operations for ordinary sets if the values
of the membership functions are restricted to either 0 or
1. However it is understood that these functions are not
the

only

possible

generalizations

of

the

crisp

set

operations. For each of the operations above, several


different classes of functions with desirable properties
36

4 - Adaptive Network Fuzzy Inference System (ANFIS)

have been proposed subsequently in the literature.

4.2.4 Fuzzy If-Then Rules


A fuzzy if-then rule (also known as fuzzy rule,
fuzzy

implication,

or

fuzzy

conditional

statement)

assumes the form

Where A and B are linguistic values defined by


fuzzy

sets

on

universes

of

discourse

and

Y,

respectively. Often x is A is called the antecedent or


premise, while y is B is called the consequence or
conclusion.

Examples

of

fuzzy

if-then

rules

are

widespread in our daily linguistic expressions, such as


the following:
If pressure is high, then volume is small.
If

the

road

is

slippery,

then

driving

is

dangerous.
If a tomato is red, then it is ripe.
If the speed is high, then apply the brake a
little.

37

4 - Adaptive Network Fuzzy Inference System (ANFIS)

4.3 The concept of adaptive neuro fuzzy inference


system [47]
ANFIS approach targets only one output from
several given inputs. The target is manipulated through
the

performance

according
parameters

to
are

of

the

membership

particular
identified

data
based

function

input.
on

the

The

curve
curve

respective

weighted values via the product in between the created


learning rules. A ratio between the individual and overall
weighted values is calculated. The ratio is gained by
using the parameters of output membership function
then, finally ANFIS predicts the target by producing an
overall gained value as an output. Membership function
parameters in input and output sides are adjusted through
a learning process to get the targeted values. ANFIS uses
hybrid algorithm that consists of a combination between
back-propagation and least-square estimation techniques
[48]. The techniques are implemented in artificial neural
network as a learning algorithm that gives very fast
convergence and more accurate in ANFIS target.

38

4 - Adaptive Network Fuzzy Inference System (ANFIS)

4.4 ANFISs learning processes


The

ANFIS

model

exhibits

predicted

target

whenever it is trained by using at least two columns of


data. The last column is the target data and also as an
output of the trained ANFIS, while the rest of the
columns are the input data. Thus, an ANFIS structure
has a single output with at least one column of input
data. For the best prediction and high reliability of its
performance, the model needs more elements in the
column of the input data. However, this situation will
also cause the processing time for learning to be slow.
For that reason, the ANFIS has to be configured in
a high speed processor. Every element in each row of the
input data is called data variable in which the linguistic
values of the relationship between them is by the rule of
IF-THEN. A total of the rule is proportional to the
membership function value and the number of column
data is linked by the following equation:

where,

(4.9)

D: Total number of column for the input data


39

4 - Adaptive Network Fuzzy Inference System (ANFIS)

F: Number of membership functions


P: Number of rules
The data is classified as training data and testing
data in ANFISs learning process. Testing data should be
in the range of training data for the purpose of testing
procedures. The number of training epoch also gives a
good result in predicting the target. Accurate targets
consider a minimum prediction error from the result of
ANFIS training. The error can be reduced by adjusting
the variable membership function and epoch parameters.
With increasing in number of MF and epoch, the error
will reduce accordingly. Sometimes, no reduction in
error

can

be

noticed

even

though

the

epoch

was

increased up to 5000 and above. This is due to the way


the data is assembled. Therefore effective input data
assembly will result good prediction.
During the training process, MF parameters are
varied so as to yield the ANFISs output as target values.
The minimum error percentage is a small difference
between target and prediction values and it is used to
measure the success level of a training process. ANFIS
performs a hybrid learning algorithm in the training
40

4 - Adaptive Network Fuzzy Inference System (ANFIS)

process

which

is

combination

of

two

algorithms

namely back-propagation and least square estimate [48].


The

hybrid

method

improves

individual algorithm and both

the

bad

features

are popular in

of

ANN

implementation.
In hybrid learning algorithm, MF parameters are
adjusted to identify the best prediction value. There are
MF curve in input and output parts of ANFIS model.
Back-propagation algorithm takes responsibility to vary
MF parameter in input side of the model, whereas least
square estimate (LSE) takes into consideration on the
output side as a linear line. In MF parameters, the input
side varies, whereas for output, they are static and vice
versa.
The prediction values are performed after the MF
parameters in both sides of the ANFIS model converge
the values according to the given training epoch [49].
ANFISs

learning

process

can

be

implemented

easily by using the provided source code in Matlab such


as newfis, evalfis and editfis. A trained ANFIS
model is formatted with a file extension of .fis to
represent an ANFIS module. On the other hand, the file
41

4 - Adaptive Network Fuzzy Inference System (ANFIS)

is represented as a module for a particular task where all


modules are configured accordingly based on a hierarchy
layout

to

form a

fault

diagnosis

system in

power

distribution network. The .fis file is also a flexible


module that can reform when the data changes or new
data is added without restructuring the model physically.
ANFIS has a capability of producing very fast result in
prediction even when handles a large size of input data.
Therefore, the system is compliable to most application
especially in adaptive control as well as ANFIS in
implementing fault diagnosis. Each ANFIS module for a
particular task is programmed by using source codes in
Matlab. The programming is developed for every task in
fault diagnosis and then the tasks are integrated in
another programming to perform a simulation tool for
fault diagnosis in power distribution network.
4.4.1 Fuzzification
Referring to FIGURE 4.6, the fuzzification stage
is located at the first stage of receiving of the input
signal. Its function is to convert the input signal to fuzzy
signal in which the signal is yielded via the input side of
the MF curve. The curve is performed by using the
42

4 - Adaptive Network Fuzzy Inference System (ANFIS)

following equations:
( )=
( )=

1+
1+

(4.10)

(4.11)

where, Xi(x) and Yi(y) are fuzzied values for each input
data, whereas ai, bi and ci are MF parameters for
respective representative of middle, width and slope of
the curve as shown in Figure 4.6. These parameters are
varied accordingly to get a suitable curve in order to get
fuzzy signal.

FIGURE 4-6 A basic ANFIS model with two inputs data and two
MFs.

43

4 - Adaptive Network Fuzzy Inference System (ANFIS)

4.4.2 Stage of IF-THEN rule


An

output

signal

from

the

fuzzification

stage

becomes an input to the stage of the IF-THEN rule. In


this stage, the fuzzy signal is gained by using equation
(4.12) up to (4.15).
=

( )

( )

(4.12)

( )

( )

(4.14)

=
=

( )

( )

( )

(4.13)

( )

(4.15)

R1, R2, R3 and R4 are real values for every IF-Then


rule.
4.4.3 Normalization
Next, the output signal from the stage of IFTHEN rule will be an input signal to the normalization
stage. In this stage, every gained signal is divided to the
total of gained signal by the following equation,
=

= 1,2,3,4

where, RT = R1 + R2 + R3 + R4
44

(4.16)

4 - Adaptive Network Fuzzy Inference System (ANFIS)

4.4.4 Defuzzification
The next process is signal defuzzification in which
the output signal from the normalization stage becomes
an input signal to this defuzzification stage. In this stage,
a normalized signal is gained again through a linear
equation that is formed from the MF of the output signal
as shown in the following equation,

= 1,2,3,4

(4.17)

with pi, qi and ri being the MF parameters for the linear


signal.
4.4.5 Neuron addition
The last process in the ANFIS operation is called
neuron addition in which all defuzzification signals, Gi
are added together as shown below:
=

= 1,2,3,4

OT is a predicted value.

45

(4.18)

4 - Adaptive Network Fuzzy Inference System (ANFIS)

4.5 Good features of the ANFIS


The advantages of ANFIS are compared to other
artificial

intelligent

techniques

such

as

an

artificial

neural network (ANN) and an expert system (ES). The


advantages are as follows;
i.

ANFIS

gives

classification

and

high
prediction

precision
models.

in
This

precision when compared to the index error


that is presented between ANFIS and ANN
show the error of 0.036 and 0.32 respectively
[48].
ii.

ANFIS has adaptive features to solve wrong


data

problem

that

involves

new

power

network configuration. The scenario is rather


difficult to solve using expert system due to
fixed rules.
iii.

ANFIS has an effective learning process on


the

training

data

while

considering

optimization in its implementation [48].

46

CHAPTER 5
Modeling and Simulation of the Underground
Cable
5.1 Cable types and their characteristics [50]
Cable

types

are

basically

defined

as

low-,

medium- and high voltage cables.


1. Low voltage: limited to 2 kV. Primarily used as
service cables.
2. Medium voltage:

211

kV.

Primarily used

to

supply distribution transformers.


3. High voltage: above 11 kV. Primarily used to
supply substation transformers.
According

to

the

cable

type,

different

requirements to cable testing, cable fault location as well


as

maintenance

strategy

are

defined.

Three-conductor

cables have been in use in the lower voltage ranges. The


tendency of the last years show the shifting to singlecore systems as they are lower in price, lower in weight
and cheaper in regards to repair costs. Furthermore oil
impregnated or oil filled cables are used less and less, as
the environmental sustainability cannot be guaranteed.
47

5 Modeling and Simulation

Especially in industrialized countries, these cable


types have been replaced and are no more installed. On
the other hand a high demand for maintenance of those
cables is given as the installed oil-insulated networks do
show up a lifetime of 50 years and more.
Today mainly XLPE insulated cables are used.
The

improvement

of

the

XLPE

insulation

material

combined with the modern design of the cable enable to


manufacture cables even for the extra high voltage level.

5.2 Types of power cable in medium voltage


The

types

of

power

cable

mainly

used

for

distribution purposes in medium voltage range are paperinsulated lead covered (PILC) cables and Cross-linked
Polyethylene (XLPE) cables.
5.2.1 Paper-insulated lead covered (PILC) cables
PILC cables have provided reliable underground
power

delivery

for

decades.

The

PILC

cables

are

manufactured using layers of paper impregnated with


compound mineral oil as insulating medium, both as
individual core and overall insulation. A lead sheath is
constructed as an outer core layer to mainly provide a
seal for the compound in the paper layers, and also for
48

5 Modeling and Simulation

excellent corrosion protective properties as well as to


provide additional mechanical protection.
A steel tape layer or steel wires are used for the
main mechanical protection and it may also be used as a
return path for earth currents. The outer sheath may be a
PVC layer or other tape of insulating and waterproof
material. PILC cables are generally used for 6.6kV up to
33kV

applications.

The

designation

6.35/11

kV

for

instance, means that the cable has an insulation level of


6.35kV between each core and earth, 11kV between
phases [51].
5.2.2 Cross-linked polyethylene (XLPE) cables
XLPE

provides

partial

insulation

as

well

as

electrical stress relieving. This insulation possesses very


good electrical, mechanical and thermal characteristics
in medium voltage networks. This type of insulation has
excellent chemical resistance and is also resistant to cold
temperatures.

Besides

that,

the

XLPE

cables

are

maintenance free, environmental friendly, invisible, not


affected by weather conditions and are widely use as a
solution for densely populated area. In comparison to
paper-insulated cables, the advantage of XLPE-insulated
49

5 Modeling and Simulation

medium voltage power cables is that they possess a low


dielectric factor and overcome the susceptibility of PILC
cables to ingress of moisture. Due to various advantages
stated, the XLPE-insulated type has vastly displaced the
traditional classic paper-insulated types in many sectors.
XLPE cables are used from low voltage to 132kV
applications [51].

5.3 Software Implementation


In this thesis we use ATP/EMTP, the Wavelet
Transform and Adaptive Neuro Fuzzy Inference System
(ANFIS) approach for fault detection, classification and
location in underground cables.
For software implementation, Alternate Transients
Program (ATP) is used to simulate underground cable
model to obtain currents data at different fault location
with

different

MATLAB

is

types

of faults. On

used

for

wavelet

the other

decomposition

hand,
and

ANFIS training and testing process.

5.4 The Alternative Transients Program (ATP)


The ATP is the PC version of the Electromagnetic
Transients Program (EMTP). The EMTP is primarily a
50

5 Modeling and Simulation

simulation program of the electric power industry. It can


predict

variables

of

interest

within

electric

power

networks as functions of time, typically following some


disturbance such as the switching of a circuit breaker, or
a fault [52].
It

was

developed

at

the

Bonneville

Power

Administration in the late 1960s as a replacement for the


Transient Network Analyzer (TNA), which was a large
analogue simulator used for transient analysis. What
began

as

approximately

5,000

lines

of

code

used

primarily for switching studies grew into a 70,000 line


multipurpose program by the early 1980s [53].
ATPDraw for Windows is a graphical, mousedriven

preprocessor

Electromagnetic

to

the

Transients

ATP

version

Program

of

the

(EMTP).

In

ATPDraw the user can construct the digital model of the


circuit to be simulated using the mouse and selecting
predefined

components

from

an

extensive

palette,

interactively. Then ATPDraw generates the input file for


the ATP simulation in the appropriate format based on
"what you see is what you get". Circuit node naming is

51

5 Modeling and Simulation

administrated by ATPDraw, thus the user needs to give a


name only to nodes having special interest.
5.4.1 Operating principles and capabilities of ATP [54]
The ATP program predicts variables of interest
within electric power networks as functions of time,
typically

initiated

by

some

disturbances.

Basically,

trapezoidal rule of integration is used to solve the


differential equations of system components in the time
domain. Non-zero initial conditions can be determined
either automatically by a steady-state phasor solution or
they can be entered by the user for simpler components.
Symmetrical

or

unsymmetrical

disturbances

are

allowed, such as faults, lightning surges and several


kinds of switching operations including commutation of
valves.
The model-library of ATP at present consists of
the following components:
Uncoupled and coupled linear, lumped R, L, C
elements.
Transmission lines and cables with distributed and
frequency-dependent parameters.
52

5 Modeling and Simulation


Nonlinear

resistances

and

inductances,

hysteretic

inductor, time-varying resistance,


TACS/MODELS controlled resistance.
Components

with

nonlinearities:

transformers

including saturation and hysteresis, surge


Arresters (gapless and with gap), arcs.
Ordinary

switches,

time-dependent

and

voltage-

dependent switches, statistical switching


Valves (diodes, thyristors, triacs), TACS/MODELS
controlled switches.
Analytical

sources:

step,

ramp,

sinusoidal,

exponential surge functions, TACS/MODELS


Defined sources.
Rotating machines: 3-phase synchronous machine,
universal machine model.
5.4. 2 Typical EMTP applications
ATP-EMTP is used world-wide for switching and
lightning

surge

analysis,

protective

relay

modeling,

harmonic and power quality studies, HVDC and FACTS


modeling.

53

5 Modeling and Simulation

5.5 Wavelet transform analysis


This thesis aims to use a combination of discrete
wavelet transform (DWT) and ANFIS to classify and
locate faults on underground cables.
Fault current signals in each case are extracted to
several

scales

certain

selected

on

the

wavelet

features

of

the

transforms,
wavelet

and

then

transformed

signals are used as an input for a training process on the


ANFIS.
5.5.1 Theory of discrete wavelet transform (DWT)
A wavelet is a waveform of effectively limited
duration that has an average value of zero.
Compare wavelets with sine waves, which are the
basis of Fourier analysis. Sinusoids do not have limited
duration they extend from minus to plus infinity. And
where sinusoids are smooth and predictable, wavelets
tend to be irregular and asymmetric.

54

5 Modeling and Simulation

Fourier analysis consists of breaking up a signal


into

sine

waves

of

various

frequencies.

Similarly,

wavelet analysis is the breaking up of a signal into


shifted and scaled versions of the original (or mother)
wavelet.
Just looking at pictures of wavelets and sine
waves, you can see intuitively that signals with sharp
changes might be better analyzed with an irregular
wavelet than with a smooth sinusoid.
Wavelet analysis is capable of revealing aspects of
data that other signal analysis techniques miss; aspects
like trends, breakdown points, discontinuities in higher
derivatives, and self-similarity. Furthermore, because it
affords a different view of data than those presented by
traditional
compress

techniques,
or

de-noise

wavelet
a

signal

analysis
without

can

often

appreciable

degradation.
In DWT, a time-scale representation of a digital
signal is obtained using digital filtering techniques. In
this case, filters of different cut-off frequencies are used
to analyze the signal at different scales [56]. The signal
is passed through a series of high pass filters to analyze
55

5 Modeling and Simulation

the high frequencies, and it is passed through a series of


low pass filters to analyze the low frequencies. Hence
the

signal

(S)

components
approximation

is

decomposed

approximation
is

(A)

the

high

into
and

two
detail

scale,

types
(D).

of
The

low-frequency

component of the signal. The detail is the low-scale,


high-frequency components. The decomposition process
can be iterated, with successive approximations being
decomposed in turn, so that one signal is broken down
into many lower resolution components. This is called
the

wavelet

decomposition

tree

which

is

shown

in

FIGURE 5.1. As decompositions are done on higher


levels,

lower

frequency

components

are

filtered

progressively.

FIGURE 5-1: Wavelet decomposition tree


56

out

5 Modeling and Simulation

For example, level 3 approximate signal (cA3) contains


lower frequency components as compared to level I
approximate signal (cAl).

5.6 Modeling and Simulation of the underground cable


5.6.1 Cable modeling:
ATP

is

used

to

simulate

medium

voltage

underground cable model [55] with various types of


faults and distances to obtain current data. The single
line diagram of the cable is shown in FIGURE 5.2 while
the cable configuration is shown in FIGURE 5.3.
The

components

used

are

three

phase

voltage

source, cable and fixed load.

FIGURE 5-2 Single line diagram for underground cable


model.

57

5 Modeling and Simulation

FIGURE 5-3 Cable Configuration.


Table 5-1 System Data
Specification of MV underground cable material
( XLPE Stranded Copper Conductor - 6 Km - Bergeron model )
Radius (mm)

r1= 6.75, r2 = 10.15, r3 = 12.05


r4 = 12.2, r5 = 13.8

Core conductor

c = 1.7 E-8 .m , c = 1.0

Insulation

i = 1.0 , i = 2.7

Sheath

sh = 2.5 E-8 .m , sh = 1.0

58

5 Modeling and Simulation

Where:
c : Resistivity of the conductor material.
sh : Resistivity of the sheath material.
c: Relative permeability of the conductor material.
sh: Relative permeability of the sheath material.
i : Relative permeability of the insulating material
outside the conductor.
i : Relative permittivity of the insulating material
outside the conductor.
Three phase voltage source:
V= 11KV with f = 50 Hz
Load:
Three-Phase Grounded-Wye load with parallel R, L
elements:

2 MVA

power factor = 0.85

R = 71.157

L = 365.475mH

Simulation of MV underground cable faults depends on


four main fault parameters: (fault type, fault distance, fault
resistance and inception angle).

59

5 Modeling and Simulation

FIGURE 5-4: Cable data (international cables Co.)


60

5 Modeling and Simulation


Table 5-2 Training scenarios for the simulation
Single line to ground (A-G)
Double line to ground (AB-G)

Fault type

Line to line (A-B)


Three phase to ground
Fault resistance

0 ,10 ,30 ,50, 100 , 200

Inception angle

0 , 45 , 90 , 135 ,180

Training
(21)
Fault distance
Testing
(20)

[0.5 , 0.75 , 1 , 1.25 , 1.5 , 1.75 , 2 ,


2.25 , 2.5 , 2.75 , 3 , 3.25 , 3.5 ,
3.75 , 4 , 4.25 , 4.5 , 4.75 , 5 , 5.25
, 5.5] Km from the sending end.
[0.625 , 0.875 , 1.125 , 1.375 ,
1.625 , 1.875 , 2.125 , 2.375 ,
2.625 , 2.875 , 3.125 , 3.375 ,
3.625 , 3.875 , 4.125 , 4.375 ,
4.625 , 4.875 , 5.125 , 5.375] Km
from the sending end.

Table 5-3 Number of simulations

Three phase to ground


Single line to ground (A-G)
Line to line (A-B)
Double line to ground (A-B-G)
Total
61

Training

Testing

630
630
630
630
2520

123
180
121
120
544

5 Modeling and Simulation

Table 5-4 ATP simulation setting


sampling frequency

200 KHz

Time step delta T

5e-6 sec

End time of simulation

0.1 sec

power frequency

50 Hz

5.6.2 ATP Implementation


Single line diagram and results of waveform for
various types of fault obtained from ATP simulation are
shown in FIGURES 5.5 to 5.9. In these figures, the three
phase currents are produced.
5.6.2.1 No Fault (normal conditions)

FIGURE 5.5 shows the single line diagram, and


FIGURE 5.6 shows the waveform of current at the
sending end at no fault (normal conditions).
During no fault (normal conditions), the waveform
of the current is constant and stable.

62

5 Modeling and Simulation

F IGURE 5-5 single line diagram of at no fault (normal


conditions)

FIGURE 5-6 Current waveforms at no fault (normal


conditions)
5.6.2.2 Three Phase to Ground Fault

FIGURE 5.7 shows the single line diagram and


FIGURE 5.8 shows the waveform of current at the
sending end at three phase to ground fault. Based on
FIGURE 5.8, when three phase to ground fault occur, all
the three phases current for the faulted phases are higher
than during normal condition.
63

5 Modeling and Simulation

FIGURE 5-7 Single line diagram of three phase to ground


fault.

FIGURE 5-8 Current waveforms at three phase to ground fault

(with fault resistance=10 , Inception Angle =135 and fault


location=1.5 Km).

64

5 Modeling and Simulation


5.6.2.3 Single Line to Ground Fault

FIGURE 5-9 shows the single line diagram and


FIGURE 5-10 shows the waveform of current at the
sending end at single line to ground fault at phase A.
Based on FIGURE 5-10, when single line to ground fault
occur

the

current

for

the

faulted

phase

is

higher

compared with the healthy phase. This can be seen in


FIGURE 5-10 where the phase A current is much higher
than phase B and phase C current.

FIGURE 5-9 Single line diagram of single line to ground


fault

65

5 Modeling and Simulation

FIGURE 5-10 Current waveform at single line to ground


fault.
(with fault resistance=50 , Inception Angle =45 and fault
location=1.5 Km).

5.6.2.4 Double Line to Ground Fault

FIGURE

5.11

shows

the

single

line

diagram

FIGURE 5.12 shows the waveform of current at the


sending end at double line to ground fault between phase
A and phase B. Based on FIGURE 5.12 when double
line to ground fault occur the current for the faulted
phases is higher compared with the healthy phases. This
can be seen in FIGURE 5.12 where the phase A and
phase B current is much higher than phase C current.

66

5 Modeling and Simulation

FIGURE 5-11 Single line diagram of double line to ground


fault.

FIGURE 5-12 Current waveform at double line to ground fault.


(with fault resistance=50 , Inception Angle =90 and fault
location=5 Km).
5.6.2.5 Line to Line Fault

FIGURE 5.13 shows the single line diagram and


FIGURE 5.14 shows the waveform of current at the
sending end at line to line fault between phase A and
phase B. Based on FIGURE 5.14, when line to line fault
67

5 Modeling and Simulation

occur the current

for the faulted

phases

is

higher

compared with the healthy phases. This can be seen in


FIGURE 5.14, where the phase A and phase B current is
much higher than phase C current.

FIGURE 5-13 Single line diagram of line to line fault.

FIGURE 5-14 Current waveforms at line to line fault.

(with fault resistance=50 , Inception Angle =135 and fault


location=4 Km).

68

5 Modeling and Simulation


5.7

Discrete

wavelet

transform

for

fault

parameterization
In this study, the line current signals are used as the
input signals of the wavelet analysis. The DWT, using a
Daubechies-4 wavelet (db4), performed better at identifying
the start and end of a disturbance [57- 60]. The Daubechies
wavelet is very well suited for identifying short-time, high
frequency transients, as well as low frequency behavior over
longer periods of time. In both cases the signals are nonperiodic or non-stationary [61].
The fault transients of the study cases are analyzed
through discrete wavelet transform at levels one to five. Both
approximation and details information related fault current are
extracted from the original signal with the multi-resolution
analysis.
When a fault occurs in the cable, it can be seen that
variations within the decomposition coefficient of the current
signals contain useful fault signatures. FIGURE 5-15 shows the
DWT detailed coefficients at level 1 to level 5 for a particular
type of fault studied in the work. Nature of the plot of detailed
coefficients at level 1 shows a sharp spike corresponding to the
fault initiation. According to DWT theory, this spike represents
69

5 Modeling and Simulation

the highest frequency available in the fault signal. It is


however, not wise to try to identify fault based on this spike
only since such spikes will occur every time there is a sudden
change in the cable current signal [56]. This will thus not be
able to clearly differentiate between faults of different types
and at different locations.

FIGURE 5-15 The DWT detailed coefficients at level 1 to level 5 for a


three phase to ground fault (Decomposition of phase A,
with fault resistance=100 , Inception Angle =90 and fault
location=1.375Km).

70

5 Modeling and Simulation

To probe further, the nature of level 5 detailed


coefficients (figures 5-16 to 19) shows that along with the high
spike, there is a certain side band containing some smaller
spikes. The nature of this side band along with the dominant
spike has been observed to change appreciably with variations
in fault types and locations.

FIGURE 5-16: level 5 detailed coefficients of Three Phase to Ground


Fault case

71

5 Modeling and Simulation

FIGURE 5-17: level 5 detailed coefficients of Single Line to Ground


Fault case

FIGURE 5-18: level 5 detailed coefficients of Double Line to Ground


Fault case

72

5 Modeling and Simulation

FIGURE 5-19: level 5 detailed coefficients of Line to Line Fault case

Detailed coefficients at still higher levels, however, have


been found to contain much wider side bands that complicated
correlation with possible fault types and locations.
It was thus decided that to start with, it will be better to
extract some meaningful features from levels 1 to 5 detailed
coefficients that can be correlated to possible fault types and
locations. With this in view, in the present work, the
summation of the square DWT detailed coefficients from level
1 to level 5 of three phase and zero sequence currents have
been used as features of the fault detection scheme, while the
maximum detailed energy of three phase and zero sequence
73

5 Modeling and Simulation

currents have been used as features of the fault classification


and location scheme.

5.8 Fault detection, Classification and Location scheme


5.8.1 Fault detection scheme
The proposed scheme is based on measuring the three
phase currents
as well as zero sequence current I0.
The wavelet transform is then calculated as well as the
summation of the squares of the 1-D wavelet signals of the
three phase currents and the zero sequence as follows:
1. Calculate zero sequence current I0 :

2. Perform a multilevel wavelet decomposition of


at level 5.
MATLAB code:
[C0,L0] = wavedec(I0,5,'db4');
[Ca,La] = wavedec(A,5,'db4');
[Cb,Lb] = wavedec(B,5,'db4');
[Cc,Lc] = wavedec(C,5,'db4');

74

5 Modeling and Simulation

3. Calculate summation of the squares of the 1-D


wavelet:

4. For the three phase currents, if there are three


consecutive absolute values of D5p are greater
than

1,

claim

fault.

Where

is

predetermined threshold, and 1=15.


MATLAB code:
% Fault Detection
if (da5<15) && (db5<15) && (dc5<15)
disp('no fault')
Fault=0;
else
Fault=1;
disp('fault')
end
5.8.2 Fault classification scheme
1. Calculate energy for 1-D wavelet decomposition
and calculate the maximum detail energy for

75

5 Modeling and Simulation

MATLAB code:
[Ea0,Ed0] = wenergy (C0,L0);
Ed0_max = max (Ed0);
[Eaa,Eda] = wenergy(Ca,La);
Eda_max = max (Eda);
[Eab,Edb] = wenergy(Cb,Lb);
Edb_max = max (Edb);
[Eac,Edc] = wenergy(Cc,Lc);
Edc_max = max (Edc);
Where:
Ea: the percentage of energy corresponding to the
approximation.
Ed: the vector containing
energy corresponding to the details.
2. Construct

the

fault

the

percentages

classification

of

system with

ANFIS:
Data preparation
The training and the testing data sets consisted of
four
for

inputs

(the

maximum

detail

energy

) and the desired output (fault type).


Number of training samples = 2520 sample
Number of testing samples = 544 sample
76

5 Modeling and Simulation

Output is set as:

Three Phase Fault = 1


Double line to ground Fault =2
line to line Fault = 3
Single line to ground Fault = 4

The ANFIS model details:

Name = class
Type = sugeno
Number of Inputs=4
Number of Outputs=1
Number of membership functions = 4
Membership function type = generalized bell
Number of Rules=256

The membership functions of the four inputs of the of


ANFIS classification model are shown figures 5-20 to 23

77

5 Modeling and Simulation

FIGURE 5-20 Input 1 Membership Functions of ANFIS


classification model

FIGURE 5-21 Input 2 Membership Functions of ANFIS


classification model

78

5 Modeling and Simulation

FIGURE 5-22 Input 3 Membership Functions of ANFIS


classification model

FIGURE 5-23 Input 4 Membership Functions of ANFIS


classification model

79

5 Modeling and Simulation

MATLAB code:
if Fault==1
FUZ = readfis('class');
clas = evalfis([Ed0_max Eda_max Edb_max Edc_max ], FUZ)
if (clas>0.5)&& (clas<1.5)
disp('Three Phase Fault ')
th_ph_f= 1;
elseif (clas>1.5)&& (clas<2.5)
disp(' Double line to ground Fault between phase A and B')
dlg_ab=1;
elseif (clas>2.5)&& (clas<3.5)
disp(' line to line Fault between phase A and B')
l2l_ab=1;
elseif (clas>3.5)&& (clas<4.5)
disp('Single line to ground Fault at phase A')
slf_A=1;
end

80

5 Modeling and Simulation

Data is then stored in a database file where :


if the fault is L-L-L-G then th_ph_f = 1
if the fault is L-L-G then dlg_ab = 1
if the fault is L-L then l2l_ab = 1
if the fault is L-G then slf_A = 1
5.8.3 Fault Location scheme
5.8.3.1 Three Phase to Ground Fault

Data preparation
The training and the testing data sets consisted of
four

inputs

for

(the
)

maximum

and

the

desired

detail
output

location in Km).

Number of training samples = 630 sample

Number of testing samples = 120 sample


The ANFIS model details:
name = anfis_3ph_new2.fis
Type=sugeno
Number of Inputs = 4
Number of Outputs =1
81

energy
(fault

5 Modeling and Simulation


Number of membership functions = 6
Number of Rules=1296
Membership function type = generalized bell
The membership functions of the four inputs of the of
ANFIS model are shown figures 5-24 to 27.

82

5 Modeling and Simulation

FIGURE 5-24 Input 1 Membership Functions of (LLLG - ANFIS)

FIGURE 5-25 Input 2 Membership Functions of (LLLG ANFIS)

83

5 Modeling and Simulation

FIGURE 5-26 Input 3 Membership Functions of (LLLG ANFIS)

FIGURE 5-27 Input 4 Membership Functions of (LLLG ANFIS)

84

5 Modeling and Simulation

MATLAB code:
if th_ph_f==1
FUZ = readfis('anfis_3ph');
D = evalfis([Ed0_max Eda_max Edb_max Edc_max ], FUZ)
end
5.8.3.2 Single Line to Ground Fault

Data preparation
The training and the testing data sets consisted of four
inputs (the maximum detail energy for
the desired output (fault location in Km).

Number of training samples = 630 sample

Number of testing samples = 180 sample


The ANFIS model details:

Anfis file = anfis_slg_new6.fis


Type=sugeno
Number of Inputs = 4
Number of Outputs =1
Number of membership functions = 6
Number of Rules=1296
Membership function type = generalized bell

85

) and

5 Modeling and Simulation

The membership functions of the four inputs of the of


ANFIS model are shown figures 5-28 to 31.

FIGURE 5-28 Input 1 Membership Functions of (AG - ANFIS)

FIGURE 5-29 Input 2 Membership Functions of (AG - ANFIS)

86

5 Modeling and Simulation

FIGURE 5-30 Input 3 Membership Functions of (AG - ANFIS)

FIGURE 5-31 Input 4 Membership Functions of (AG - ANFIS)

87

5 Modeling and Simulation

MATLAB code:
if slf_A == 1
FUZ = readfis('anfis_slg');
D = evalfis([Ed0_max Eda_max Edb_max Edc_max ],
FUZ)
end
5.8.3.3 Line to Line Fault

Data preparation
The training and the testing data sets consisted of four
inputs (the maximum detail energy for
the desired output (fault location in Km).

Number of training samples = 630 sample

Number of testing samples = 120 sample


The ANFIS model details:

Anfis file = l2l_new6.fis


Type=sugeno
Number of Inputs = 4
Number of Outputs =1

Number of membership functions = 6


Number of Rules=1296
88

) and

5 Modeling and Simulation


Membership function type = generalized bell
The membership functions of the four inputs of the of
ANFIS model are shown figures 5-32 to 35.

FIGURE 5-32 Input 1 Membership Functions of (AB - ANFIS)

FIGURE 5-33 Input 2 Membership Functions of (AB - ANFIS)

89

5 Modeling and Simulation

Figure 5-34 Input 3 Membership Functions of (AB - ANFIS)

Figure 5-35 Input 4 Membership Functions of (AB - ANFIS)

90

5 Modeling and Simulation

MATLAB code:
if l2l_ab == 1
FUZ = readfis('anfis_l2l');
D = evalfis([Ed0_max Eda_max Edb_max Edc_max ],
FUZ)
end
5.8.3.4 Double Line to Ground Fault

Data preparation
The training and the testing data sets consisted of four
inputs (the maximum detail energy for
) and
the desired output (fault location in Km).

Number of training samples = 630 sample

Number of testing samples = 120 sample


The ANFIS model details:
Anfis file = anfis_dlg_4.fis
Type=sugeno
Number of Inputs = 4
Number of Outputs =1
Number of membership functions = 6
Number of Rules=1296
91

5 Modeling and Simulation


Membership function type = generalized bell
The membership functions of the four inputs of the of
ANFIS model are shown figures 5-36 to 39.

FIGURE 5-36 Input 1 Membership Functions of (ABG - ANFIS)

FIGURE 5-37 Input 2 Membership Functions of (ABG - ANFIS)

92

5 Modeling and Simulation

FIGURE 5-38 Input 3 Membership Functions of (ABG ANFIS)

FIGURE 5-39 Input 4 Membership Functions of (ABG ANFIS)

93

5 Modeling and Simulation


MATLAB code:
if dlg_ab ==1
FUZ = readfis('anfis_dlg');
D = evalfis([Ed0_max Eda_max Edb_max Edc_max ],
FUZ)
end

5.9 Fault Location System Tool


Development of the fault location system tool (FL) is
based on Matlab GUI as shown in Figure 5.40. The purposes of
building the GUI tool are to:
Test the data using ANFIS models,

Display the original signal, the wavelet approximation


and detail coefficients at level 5,

Produces fault types,


Calculate fault location and
Calculate the percentage error.
The FL needs only a standard data format of three-phase
current as an input to pin point the fault.
94

5 Modeling and Simulation

5.9.1 Components of the fault location system tool


The fault location system tool (FL) is consists of three
sections:
Section 1:
It shows the original three phase signals (A , B
and C) and the corresponding wavelet
approximation signals at level 5 (Aa5 , Ab5 and
Ac5).
Section 2:
It shows the wavelet detail signals at level 5 (Da5,
Db5 and Dc5).
Section 3:
It displays the fault types ,calculate the fault
location and finally calculate the percentage error.
Figures 5.40 to 44 show some applications of the fault
location system tool on the proposed work.

95

5 Modeling and Simulation

96
FIGURE 5.40: No Fault (normal conditions)

5 Modeling and Simulation

97
FIGURE 5.41: Three phase to ground fault

5 Modeling and Simulation

FIGURE 5.42: Single Line to ground fault


98

5 Modeling and Simulation

FIGURE 5.43: Double Line to ground fault


99

5 Modeling and Simulation

FIGURE 5.44: Line to Line fault


100

CHAPTER 6
Simulation Results
6.1 Introduction
For data analysis; location of fault and type of fault is
considered. The input data, sending end current obtained from
the ATP are chosen randomly and are arranged in Microsoft
Office Excel before entered into program design in MATLAB.
Results are analyzed by comparing the real fault location and
the estimation fault location from the network based on the
right type fault type classification.
By using MATLAB ANFIS models are trained for fault
classification and fault location. Once trained, the network
performance was tested using testing data that were different
from the training data.

6.2 ANFIS Fault Type classification Results


An M-file is developed for the training stage which
produces the ANFIS model and plots the training performance
graph.
The performance graph for the ANFIS Fault Type
classification model is shown in FIGURE 6.1 with Root Mean
Squared Error equal to 0.029 at epoch 300.
101

6 Results

On the other hand, Table 6.1 indicates training target for


ANFIS fault type classification. From FIGURE 6.1, the
training output results obtained are similar with the training
target values. This shows that ANFIS is able to recognize and
classify fault correctly. Refer Appendix B for full ANFIS Fault
Type classification Training Results and Appendix A for the
command.

(a) Training

102

6 Results

(b) Testing
FIGURE 6.1: The performance graph for the ANFIS Fault
Type classification model.

Table 6.1: Training

target for ANFIS fault type

classification
Fault Type

Target

L-L-L-G

L-L-G

L-L

L-G

4
103

6 Results

Table 6.2: The percentage average accuracy for fault types


(testing and training)
Number of case studied
Fault type
L-G (AG)
L-L-G (ABG)
L-L (AB)
L-L-LG(ABC)

Average accuracy

training

testing

630
630
630

180
120
121

100 %
100 %
100 %

630

123

100 %

6.3 ANFIS Fault Location Estimation Training


Results
In ANFIS Fault Location Estimation, all the four types
of fault (ABCG, ABG, AB and AG) were trained separately.
As a result, a total of four networks involved to estimate the
fault distance. The results for each network are presented
below.
The location error is defined as[62]

=6
104

6 Results

Training target for the ANFIS fault location model


{0.5 ; 0.75 ; 1 ; 1.25 ; 1.5 ; 1.75 ; 2 ; 2.25 ; 2.5 ; 2.75 ; 3 ; 3.25 ;
3.5 ; 3.75 ; 4 ; 4.25 ; 4.5 ; 4.75 ; 5 ; 5.25 ; 5.5}
6.3.1 Three phase to ground ANFIS fault location model
The performance graph for the three phase ANFIS fault
location model is shown in FIGURE 6.2 with Root Mean
Squared Error equal to 0.0134 at epoch 250.
On the other hand, Table 6.3 indicates some of the fault
location training output results for three phase fault. Refer to
Appendix A for the command.

105

6 Results

(a) Training

(b) Testing
FIGURE 6.2: The performance graph for the three phase
ANFIS fault location model
106

6 Results
Table 6.3: Three phase ANFIS fault location training output results
inception angle = 0
Fault
resistance

*X
(KM)

*D
(KM)

*E%

inception angle = 45

inception angle = 90

inception angle = 135

inception angle = 180

*X
(KM)

*X
(KM)

*X
(KM)

*X
(KM)

*D
(KM)

*E%

*D
(KM)

*E%

*D
(KM)

*E%

*D
(KM)

*E%

0.5

0.502 0.033

1.5

1.504 0.069

0.5

0.500 0.002

1.5

1.501 0.010

0.5

0.499 0.024

0.75

0.748 0.028

1.995 0.088

0.75

0.750 0.007

1.999 0.016

2.75

2.750 0.000

3.002 0.026

3.5

3.497 0.054

3.000 0.001

3.5

3.499 0.013

4.75

4.748 0.025

3.25

3.242 0.129

3.75

3.751 0.011

3.25

3.250 0.001

3.75

3.750 0.004

5.25

5.251 0.025

1.75

1.750 0.001

0.5

0.500 0.001

1.75

1.752 0.038

0.5

0.500 0.007

0.75

0.751 0.011

1.999 0.012

0.75

0.750 0.001

2.002 0.028

0.75

0.748 0.027

1.000 0.005

4.25

4.250 0.002

3.000 0.004

4.25

4.254 0.066

2.998 0.025

3.5

3.500 0.002

4.5

4.500 0.003

3.25

3.250 0.005

4.5

4.504 0.070

3.25

3.250 0.005

5.002 0.028

0.999 0.019

1.5

1.500 0.005

1.003 0.045

1.5

1.495 0.081

1.5

1.497 0.057

1.25

1.251 0.016

2.25

2.252 0.030

1.25

1.248 0.038

2.25

2.249 0.023

2.25

2.250 0.001

5.250 0.007

3.999 0.016

4.997 0.057

3.997 0.051

4.001 0.022

5.25

5.500 0.006

5.5

5.494 0.105

5.25

5.247 0.057

5.5

5.497 0.054

5.25

5.253 0.047

107
10

30

6 Results

inception angle = 0
Fault
resistance

*X
(KM)

*D
(KM)

*E%

inception angle = 45
*X
(KM)

inception angle = 90

Fault

*D
*X
resistan
(KM)
(KM)

*D
(KM)

ce

*E%

inception angle = 135


*X
(KM)

inception angle = 180

Fault

*D
*X
resistan
(KM)
(KM)

*D
(KM)

ce

*E%

1.5

1.498 0.027

2.5

2.502 0.028

1.5

1.477 0.391

2.5

2.500 0.002

1.75

1.748 0.033

2.25

2.250 0.008

2.75

2.718 0.534

2.25

2.269 0.320

2.75

2.746 0.063

1.993 0.116

3.998 0.032

4.75

4.773 0.377

4.026 0.429

4.75

4.751 0.013

4.25

4.256 0.102

5.5

5.510 0.169

5.25

5.212 0.627

5.5

5.468 0.534

5.25

5.246 0.060

5.5

5.487 0.218

2.5

2.494 0.104

1.032 0.536

2.5

2.468 0.536

1.006 0.104

2.5

2.497 0.053

2.75

2.741 0.158

1.25

1.239 0.190

2.75

2.767 0.286

1.25

1.246 0.072

2.994 0.098

4.75

4.755 0.075

5.000 0.001

4.75

4.783 0.554

5.011 0.182

3.25

3.253 0.049

5.25

5.239 0.186

5.25

5.207 0.722

5.25

5.263 0.224

5.25

5.239 0.189

4.5

4.499 0.023

1.5

1.504 0.064

1.75

1.748 0.035

1.5

1.486 0.240

1.75

1.746 0.060

1.5

1.506 0.101

1.997 0.049

1.976 0.399

2.004 0.060

1.998 0.030

2.000 0.001

3.5

3.499 0.020

4.25

4.240 0.159

3.5

3.498 0.031

4.25

4.249 0.019

3.5

3.499 0.022

3.75

3.748 0.032

4.5

4.494 0.098

3.75

3.768 0.299

4.5

4.501 0.011

3.75

3.745 0.084

50

108
100

200

* X(KM): Actual fault location .


* D(KM) : Calculated fault location.
* E% : Percentage error.

6 Results

1.8
1.6
1.4

Percentage error

1.2
1
0.8
0.6
0.4
0.2
1
28
55
82
109
136
163
190
217
244
271
298
325
352
379
406
433
460
487
514
541
568
595
622

0
Number of Smulation

FIGURE 6.3: Percentage error of three phase to ground


fault.
From FIGURE 6.3 it can be notice that the Percentage error
of three phase to ground fault is less than 1.65%.
6.3.2 Double Line to Ground ANFIS fault location model
The performance graph for the double Line to Ground
ANFIS fault location model is shown in FIGURE 6.4 with Root
Mean Squared Error equal to 0.0105 at epoch 200.
On the other hand, Table 6.4 indicates some of the fault
location training output results for Double Line to Ground fault.
Refer to Appendix A for the command.

109

6 Results

(a) Training

(b) Testing
FIGURE 6.4: The training performance graph for the ANFIS
Double Line to Ground Fault Location model.

110

6 Results
Table 6.4: Double Line to Ground (ABG)ANFIS fault location training output results
inception angle = 0
Fault
resistance

inception angle = 45

inception angle = 90

inception angle = 135

inception angle = 180

*X
(KM)

*D
(KM)

*E%

*X
(KM)

*D
(KM)

*E%

*X
(KM)

*D
(KM)

*E%

*X
(KM)

*D
(KM)

*E%

*X
(KM)

*D
(KM)

*E%

0.5

0.500

0.000

1.5

1.497

0.042

0.5

0.500

0.000

1.5

1.499

0.009

0.5

0.500

0.003

0.75

0.735

0.251

2.000

0.000

0.75

0.750

0.003

2.000

0.000

2.75

2.749

0.016

2.965

0.583

3.5

3.502

0.033

3.000

0.002

3.5

3.499

0.012

4.75

4.749

0.012

3.25

3.200

0.837

3.75

3.750

0.008

3.25

3.249

0.015

3.75

3.750

0.004

5.25

5.248

0.036

1.75

1.751

0.012

0.5

0.500

0.000

1.75

1.750

0.001

0.5

0.500

0.000

0.75

0.750

0.003

2.000

0.000

0.75

0.750

0.002

2.000

0.000

0.75

0.750

0.000

1.000

0.000

4.25

4.249

0.010

3.000

0.004

4.25

4.250

0.001

3.249

0.018

3.5

3.499

0.011

4.5

4.500

0.006

3.25

3.250

0.000

4.5

4.511

0.175

3.25

3.498

0.041

5.000

0.000

1.000

0.005

1.5

1.500

0.005

1.000

0.003

1.5

1.504

0.065

1.5

1.500

0.000

1.25

1.253

0.053

2.25

2.249

0.021

1.25

1.242

0.131

2.25

2.250

0.007

2.25

2.250

0.004

5.000

0.005

4.000

0.001

5.000

0.001

4.000

0.000

4.000

0.000

5.25

5.249

0.022

5.5

5.500

0.000

5.25

5.242

0.131

5.5

5.500

0.000

5.25

5.251

0.010

111
10

30

6 Results

inception angle = 0
Fault
resistance

inception angle = 45

inception angle = 90

inception angle = 135

inception angle = 180

*X
(KM)

*D
(KM)

*E%

*X
(KM)

*D
(KM)

*E%

*X
(KM)

*D
(KM)

*E%

*X
(KM)

*D
(KM)

*E%

*X
(KM)

*D
(KM)

*E%

1.5

1.506

0.101

2.5

2.498

0.025

1.5

1.507

0.124

2.5

2.496

0.059

1.75

1.755

0.090

2.25

2.254

0.066

2.75

2.750

0.001

2.25

2.257

0.120

2.75

2.746

0.066

2.000

0.001

4.001

0.012

4.75

4.751

0.020

4.006

0.107

4.75

4.752

0.038

4.25

4.256

0.096

5.5

5.500

0.000

5.25

5.247

0.047

5.5

5.500

0.000

5.25

5.231

0.309

5.5

5.500

0.000

2.5

2.493

0.109

1.000

0.001

2.5

2.506

0.103

1.000

0.000

2.5

2.498

0.036

2.75

2.742

0.131

1.25

1.252

0.041

2.75

2.753

0.052

1.25

1.250

0.001

3.000

0.001

4.75

4.760

0.169

5.000

0.002

4.75

4.745

0.077

5.000

0.006

3.25

3.250

0.003

5.25

5.267

0.284

5.25

5.255

0.085

5.25

5.265

0.251

5.25

5.234

0.268

4.5

4.502

0.028

1.5

1.501

0.015

1.75

1.749

0.021

1.5

1.505

0.082

1.75

1.750

0.008

1.5

1.498

0.027

2.000

0.001

2.000

0.004

2.001

0.020

2.000

0.004

2.000

0.001

3.5

3.495

0.091

4.25

4.249

0.019

3.5

3.488

0.206

4.25

4.251

0.012

3.5

3.505

0.086

3.75

3.745

0.082

4.5

4.501

0.023

3.75

3.741

0.148

4.5

4.502

0.035

3.75

3.753

0.056

50

112

100

200

* X(KM): Actual fault location .


* D(KM) : Calculated fault location.
* E% : Percentage error.

6 Results

1
0.9
0.8
Percentage error

0.7
0.6
0.5
0.4
0.3
0.2
0

1
31
61
91
121
151
181
211
241
271
301
331
361
391
421
451
481
511
541
571
601

0.1

Number of Smulation

FIGURE 6.5: Percentage error of Double Line to


ground fault.
From FIGURE 6.5 it can be notice that the Percentage error
of Double Line to ground fault is less than 1%.
6.3.3 Line to Line ANFIS fault location model
The performance graph for the Line to Line ANFIS fault
location model is shown in FIGURE 6.6 with Root Mean Squared
Error equal to 0.0251 at epoch 200.
On the other hand, Table 6.5 indicates some of the fault
location training output results for Line to Line fault. Refer to
Appendix A for the command.

113

6 Results

(a) Training

(b)Testing
FIGURE 6.6: The performance graph for the ANFIS Line to Line Fault
Location model.

114

6 Results
Table 6.5: Line to Line (AB) ANFIS fault location training output results
inception angle = 0
Fault
*X
*D
*E%
resistance (KM) (KM)

inception angle = 45 inception angle = 90 inception angle = 135 inception angle = 180
*X
*D
*E%
(KM) (KM)

*X
*D
*E%
(KM) (KM)

*X
*D
*E%
(KM) (KM)

*X
*D
*E%
(KM) (KM)

0.5

0.502

0.028

1.5

1.498

0.026

0.5

0.498

0.034

1.5

1.500

0.001

0.5

0.499

0.018

0.75

0.747

0.047

2.000

0.005

0.75

0.750

0.006

2.000

0.002

2.75

2.750

0.004

3.003

0.046

3.5

3.500

0.005

2.998

0.030

3.5

3.500

0.005

4.75

4.750

0.005

3.25

3.246

0.065

3.75

3.750

0.000

3.25

3.248

0.040

3.75

3.750

0.000

5.25

5.250

0.002

1.75

1.749

0.015

0.5

0.503

0.047

1.75

1.749

0.018

0.5

0.511

0.185

0.75

0.750

0.003

2.000

0.000

0.75

0.745

0.088

2.000

0.000

0.75

0.734

0.263

1.000

0.000

4.25

4.250

0.002

2.999

0.009

4.25

4.249

0.014

2.992

0.130

3.5

3.509

0.144

4.5

4.501

0.009

3.25

3.236

0.228

4.5

4.506

0.102

3.25

3.254

0.072

5.005

0.082

1.000

0.000

1.5

1.508

0.128

1.000

0.000

1.5

1.500

0.002

1.5

1.501

0.020

1.25

1.228

0.360

2.25

2.250

0.001

1.25

1.236

0.239

2.25

2.250

0.001

1.75

1.728

0.361

4.987

0.223

3.975

0.409

5.012

0.204

4.007

0.124

2.984

0.259

5.25

5.249

0.020

5.5

5.500

0.006

5.25

5.248

0.032

5.5

5.500

0.000

5.25

5.244

0.102

115
10

30

6 Results

inception angle = 0
Fault
*X
*D
*E%
resistance (KM) (KM)

50

116
100

200

inception angle = 45 inception angle = 90 inception angle = 135 inception angle = 180
Fault
*X
*D
*X
*D
resista
*E%
(KM) (KM)
(KM) (KM)
nce
2.500 0.000
1.492 0.129
2.5
1.5

Fault
*X
*D
*X
*D
resista
*E%
(KM) (KM)
(KM) (KM)
nce
2.500 0.000 1.75 1.765 0.245
2.5

1.5

1.480

0.334

2.25

2.250

0.001

2.75

2.743

0.117

2.25

2.252

0.026

2.75

2.743

0.111

2.000

0.000

3.997

0.046

4.75

4.729

0.357

3.995

0.076

4.75

4.755

0.089

4.25

4.251

0.019

5.5

5.500

0.000

5.25

5.252

0.033

5.5

5.500

0.000

5.25

5.251

0.017

5.5

5.500

0.000

2.5

1.000

0.000

1.000

0.000

2.5

2.497

0.044

1.000

0.000

2.5

2.500

0.005

2.75

1.248

0.038

1.25

1.254

0.072

2.75

2.711

0.648

1.25

1.244

0.099

3.021

0.347

4.75

4.727

0.385

4.982

0.308

4.75

4.814

1.063

5.022

0.362

3.25

3.258

0.138

5.25

5.262

0.201

5.25

5.263

0.222

5.25

5.262

0.208

5.25

5.273

0.387

3.993

0.117

1.5

1.477

0.377

1.75

1.759

0.143

1.5

1.498

0.028

1.75

1.747

0.055

1.5

1.522

0.372

1.999

0.024

2.000

0.004

2.000

0.003

1.999

0.013

2.005

0.088

3.5

3.525

0.410

4.25

4.248

0.037

3.5

3.492

0.136

4.25

4.246

0.066

3.5

3.493

0.113

3.75

3.749

0.014

4.5

4.506

0.101

3.75

3.750

0.007

4.5

4.498

0.035

3.75

3.731

0.323

* X(KM): Actual fault location .


* D(KM) : Calculated fault location.
* E% : Percentage error.

6 Results

2.0
1.8
Percentage error

1.6
1.4
1.2
1.0
0.8
0.6
0.4
0.2
1
35
69
103
137
171
205
239
273
307
341
375
409
443
477
511
545
579
613

0.0
Number of Smulation

FIGURE 6.7: Percentage error of Line to Line fault.


From FIGURE 6.7 it can be notice that the Percentage error
of Line to Line fault is less than 2%.

6.3.4 Single Line to Ground ANFIS fault location model


The performance graph for the

Single Line to Ground

ANFIS fault location model is shown in FIGURE 6.8 with Root


Mean Squared Error equal to 0.03059 at epoch 250.
On the other hand, Table 6.6 indicates some of the fault
location training output results for Single Line to Ground (AG)
fault. Refer to Appendix A for the command.

117

6 Results

(a)Training

(b) Testing
FIGURE 6.8: The performance graph for the ANFIS Single Line to
Ground (AG) Fault Location model.

118

6 Results
Table 6.6: Single Line to Ground (AG) ANFIS fault location training output results
inception angle = 0
Fault
resistance

119
10

30

*X
(KM)

*D
(KM)

*E%

0.5

0.500

0.002

0.75

0.750

inception angle = 45

inception angle = 90

inception angle = 135

inception angle = 180

*X
*D
(KM) (KM)

*E%

*X
(KM)

*D
(KM)

*E%

*X
(KM)

*D
(KM)

*E%

*X
(KM)

*D
(KM)

*E%

1.5

1.494

0.107

0.5

0.520

0.340

1.5

1.498

0.038

0.5

0.497

0.050

0.001

1.994

0.102

0.75

0.835

1.424

2.005

0.087

3.75

3.773

0.384

3.000

0.000

3.5

3.500

0.002

2.946

0.892

3.5

3.493

0.113

4.75

4.712

0.626

3.25

3.265

0.242

3.75

3.807

0.952

3.25

3.241

0.142

3.75

3.817

1.117

5.25

5.265

0.253

1.75

1.751

0.011

0.5

0.500

0.003

1.75

1.750

0.002

0.5

0.500

0.008

0.75

0.750

0.001

2.000

0.006

0.75

0.750

0.003

1.996

0.066

0.75

0.747

0.043

0.999

0.012

4.25

4.250

0.005

3.000

0.003

4.25

4.250

0.002

2.991

0.147

3.5

3.500

0.001

4.5

4.501

0.011

3.25

3.250

0.008

4.5

4.500

0.001

3.25

3.248

0.036

5.000

0.000

0.984

0.262

1.5

1.547

0.787

0.999

0.013

1.5

1.524

0.393

1.5

1.565

1.081

1.25

1.271

0.353

2.25

2.233

0.277

1.25

1.253

0.044

2.75

2.744

0.105

2.25

2.255

0.084

5.000

0.001

4.000

0.002

5.002

0.037

3.992

0.126

3.990

0.174

5.25

5.258

0.141

5.5

5.500

0.001

5.25

5.251

0.012

5.5

5.491

0.155

5.25

5.248

0.026

6 Results

inception angle = 0
Fault
resistance

50

120
100

200

inception angle = 45

inception angle = 90

Fault
*X
*D
*X
resistanc
(KM) (KM)
(KM)
e
2.499 0.021
2.5
1.5

inception angle = 135

inception angle = 180

*D
(KM)

*E%

*X
(KM)

1.506

0.096

2.5

Fault
*D
*X
resistan
(KM)
(KM)
ce
2.522 0.374
1.75

2.25

2.245

0.086

2.75

2.756

0.098

0.052

4.009

0.148

4.75

4.737

5.263

0.217

5.5

5.508

0.129

5.25

1.010

0.164

2.5

2.495

0.085

0.267

1.25

1.240

0.173

2.75

2.745

4.763

0.213

5.025

0.417

4.75

5.25

5.298

0.802

5.25

5.277

0.443

1.5

1.464

0.604

1.75

1.762

2.033

0.550

3.5

3.441

0.984

3.75

3.814

1.061

*X
(KM)

*D
(KM)

*E%

*D
(KM)

*E%

1.5

1.482

0.296

1.734

0.264

2.25

2.271

0.348

2.75

2.754

0.066

1.998

0.027

4.017

0.287

4.75

4.747

0.217

4.25

4.251

0.021

5.5

5.485

0.255

5.25

5.278

0.470

5.5

5.488

0.204

2.5

2.530

0.501

1.007

0.109

2.5

2.484

0.260

2.75

2.734

0.077

1.25

1.270

0.326

3.033

0.554

4.75

4.745

0.078

5.033

0.548

3.25

3.267

0.286

5.25

5.253

0.057

5.25

5.206

0.737

4.5

4.471

0.478

0.194

1.5

1.470

0.507

1.75

1.798

0.796

1.5

1.593

1.549

1.995

0.087

2.004

0.066

1.989

0.192

1.952

0.803

4.25

4.274

0.394

3.5

3.486

0.238

4.25

4.254

0.064

3.5

3.594

1.558

4.5

4.497

0.051

3.75

3.760

0.164

4.5

4.382

1.968

3.75

3.769

0.308

* X(KM): Actual fault location .


* D(KM) : Calculated fault location.
* E% : Percentage error.

6 Results

3.0

Percentage error

2.5
2.0
1.5
1.0
0.5

1
33
65
97
129
161
193
225
257
289
321
353
385
417
449
481
513
545
577
609

0.0
Number of Smulation

FIGURE 6.9: Percentage error of Single Line to ground fault.

From FIGURE 6.9 it can be notice that the Percentage error


of Single Line to ground fault is less than 3% .

6.4 ANFIS Fault Location Estimation Testing Results


In order to test the ability of the network that combining the
ANFIS type classifier and ANFIS fault location estimation in
classifying and estimate fault location, 540 numbers of
simulations were done. The data must be different with training
data.

121

6 Results

6.4.1 ANFIS Fault Location Estimation Testing Results


6.4.1.1 Three phase to ground fault

120 current samples were used to test the three phase


ANFIS fault location model, Table 6.7

represent the testing

results. From the table, the error produce in estimating the fault
distance is less than 0.7 %, which is a quit good accuracy for the
fault location estimation.
6.4.1.2 Double Line to Ground fault

120 current samples were used to test the double line to ground
ANFIS fault location model, Table 6.8

represent the testing

results. From the table, the error produce in estimating the fault
distance is less than 1.7%, which is a quit good accuracy for the
fault location estimation.
6.4.1.3 Line to Line fault

120 current samples were used to test the line to line ANFIS
fault location model, Table 6.9 represent the testing results. From
the table, the error produce in estimating the fault distance is less
than 1.8%, which is a quit good accuracy for the fault location
estimation.
6.4.1.4 Single Line to Ground fault

180 current samples were used to test the single line to


ground ANFIS fault location model, Table 6.10 represent the
122

6 Results

testing results. From the table, the error produce in estimating the
fault distance is less than 3%, which is a quit good accuracy for
the fault location estimation.
Table 6.7: Three phase ANFIS fault location testing output
results
inception angle = 0

inception angle = 45 inception angle = 90

Fault
*X
resistance (KM)

*D
(KM)

*E%

*D
(KM)

*E%

*D
(KM)

*E%

0.625

0.625

0.001

0.627

0.041

0.601

0.394

0.875

0.875

0.004

0.875

0.007

0.873

0.026

1.125

1.125

0.005

1.126

0.009

1.126

0.008

1.375

1.375

0.008

1.375

0.002

1.376

0.017

1.625

1.625

0.007

1.625

0.000

1.624

0.019

1.875

1.875

0.008

1.875

0.000

1.877

0.026

2.125

2.125

0.006

2.124

0.009

2.132

0.122

2.375

2.375

0.002

2.375

0.004

2.372

0.043

2.625

2.626

0.010

2.625

0.001

2.626

0.009

2.875

2.875

0.005

2.875

0.002

2.878

0.048

3.125

3.125

0.001

3.125

0.000

3.129

0.064

3.375

3.375

0.001

3.375

0.004

3.374

0.016

3.625

3.624

0.016

3.625

0.004

3.622

0.044

3.875

3.875

0.002

3.875

0.002

3.878

0.047

4.125

4.125

0.000

4.125

0.006

4.127

0.030

4.375

4.375

0.006

4.375

0.002

4.372

0.045

4.625

4.625

0.004

4.625

0.001

4.629

0.063

4.875

4.875

0.002

4.875

0.000

4.882

0.110

10

123

6 Results

inception angle = 0

inception angle = 45 inception angle = 90

Fault
*X
resistance (KM)

*D
(KM)

*E%

*D
(KM)

*E%

*D
(KM)

*E%

5.125

5.125

0.001

5.125

0.006

5.122

0.051

5.375

5.375

0.004

5.375

0.002

5.380

0.089

0.625

0.631

0.102

0.641

0.261

0.634

0.146

0.875

0.873

0.041

0.851

0.392

0.854

0.350

1.125

1.125

0.008

1.141

0.274

1.140

0.253

1.375

1.376

0.017

1.364

0.182

1.365

0.167

1.625

1.617

0.131

1.619

0.105

1.623

0.031

1.875

1.889

0.237

1.876

0.010

1.875

0.004

2.125

2.123

0.025

2.094

0.518

2.103

0.366

2.375

2.380

0.092

2.388

0.218

2.382

0.122

2.625

2.620

0.080

2.614

0.189

2.616

0.148

2.875

2.871

0.074

2.905

0.503

2.899

0.401

3.125

3.119

0.097

3.130

0.077

3.129

0.066

3.375

3.372

0.057

3.386

0.175

3.404

0.491

3.625

3.608

0.276

3.647

0.359

3.639

0.232

3.875

3.884

0.145

3.874

0.012

3.870

0.076

4.125

4.113

0.196

4.165

0.669

4.138

0.215

4.375

4.364

0.186

4.380

0.087

4.383

0.128

4.625

4.619

0.093

4.626

0.021

4.624

0.012

4.875

4.872

0.057

4.875

0.003

4.873

0.037

5.125

5.127

0.040

5.112

0.219

5.115

0.159

5.375

5.382

0.125

5.357

0.301

5.349

0.427

10

50

* X(KM): Actual fault location .


* D(KM) : Calculated fault location.
* E% : Percentage error
124

0.8
0.7
0.6
0.5
0.4
0.3
10
0.2
0.1
0.0
1
6
11
16
21
26
31
36
41
46
51
56
61
66
71
76
81
86
91
96
101
106
111
116

Percentage error

6 Results

Number of Smulation

FIGURE 6.10: Percentage error of Three phase to ground fault


(testing).

Table 6.8 : Double Line to Ground ANFIS fault location


testing output results
inception angle = 0

Fault
resistance

10

*X
(KM)
0.625
0.875
1.125
1.375
1.625
1.875
2.125
2.375
2.625
2.875
3.125
3.375
3.625
3.875

*D
(KM)
0.630
0.878
1.125
1.377
1.623
1.874
2.124
2.375
2.626
2.873
3.125
3.376
3.623
3.876

*E%
0.086
0.042
0.002
0.035
0.037
0.018
0.012
0.001
0.010
0.040
0.008
0.018
0.029
0.017

125

inception angle = 45 inception angle = 90

*D
(KM)
0.626
0.875
1.125
1.376
1.624
1.875
2.125
2.374
2.625
2.875
3.124
3.375
3.627
3.876

*E%
0.016
0.001
0.004
0.014
0.022
0.000
0.007
0.013
0.000
0.003
0.014
0.007
0.025
0.017

*D
(KM)
0.637
0.862
1.124
1.373
1.627
1.875
2.122
2.368
2.618
2.883
3.133
3.373
3.628
3.851

*E%
0.200
0.224
0.013
0.031
0.030
0.005
0.056
0.109
0.110
0.135
0.136
0.027
0.044
0.392

6 Results

inception angle = 0

Fault
resistance

10

200

*X
(KM)
4.125
4.375
4.625
4.875
5.125
5.375
0.625
0.875
1.125
1.375
1.625
1.875
2.125
2.375
2.625
2.875
3.125
3.375
3.625
3.875
4.125
4.375
4.625
4.875
5.125
5.375

*D
(KM)
4.125
4.375
4.623
4.876
5.122
5.363
0.643
0.888
1.132
1.378
1.625
1.872
2.119
2.368
2.617
2.866
3.118
3.368
3.618
3.872
4.123
4.377
4.630
4.882
5.138
5.391

*E%
0.005
0.007
0.033
0.015
0.055
0.201
0.306
0.215
0.113
0.046
0.004
0.056
0.094
0.113
0.130
0.143
0.121
0.117
0.111
0.042
0.025
0.033
0.091
0.119
0.212
0.268

inception angle = 45 inception angle = 90

*D
(KM)
4.125
4.377
4.624
4.874
5.126
5.377
0.638
0.885
1.131
1.377
1.626
1.874
2.121
2.371
2.620
2.868
3.119
3.370
3.619
3.872
4.123
4.375
4.627
4.880
5.133
5.386

* X(KM): Actual fault location .


* D(KM) : Calculated fault location.
* E% : Percentage error

126

*E%
0.005
0.028
0.011
0.024
0.012
0.037
0.219
0.163
0.101
0.034
0.024
0.017
0.064
0.067
0.078
0.109
0.092
0.086
0.107
0.057
0.028
0.008
0.037
0.091
0.139
0.188

*D
(KM)
4.122
4.366
4.627
4.898
5.107
5.384
0.672
0.909
1.146
1.384
1.627
1.870
2.114
2.361
2.610
2.858
3.109
3.361
3.613
3.868
4.123
4.379
4.634
4.890
5.148
5.404

*E%
0.048
0.142
0.040
0.375
0.299
0.156
0.791
0.560
0.346
0.152
0.026
0.089
0.183
0.230
0.254
0.279
0.260
0.225
0.196
0.112
0.034
0.063
0.147
0.255
0.388
0.486

6 Results

1
6
11
16
21
26
31
36
41
46
51
56
61
66
71
76
81
86
91
96
101
106
111
116

Percentage error

1.8
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
Number of Smulation

FIGURE 6.11: Percentage error of Double Line to Ground fault


(testing).

Table 6.9: Line to Line ANFIS fault location testing output


results
inception angle = 0

Fault
resistance

10

*X
(KM)
0.625

*D
(KM)
0.609

0.875

inception angle =
45

0.263

*D
(KM)
0.723

0.866

0.156

1.125

1.226

1.375

inception angle =90

1.638

*D
(KM)
0.590

0.591

0.936

1.021

0.861

0.241

1.678

1.157

0.530

1.114

0.191

1.341

0.567

1.371

0.070

1.373

0.026

1.625

1.571

0.893

1.611

0.233

1.631

0.092

1.875

1.890

0.250

1.846

0.483

1.902

0.458

2.125

2.152

0.444

2.088

0.612

2.128

0.046

2.375

2.331

0.742

2.340

0.578

2.375

0.006

2.625

2.660

0.577

2.647

0.372

2.622

0.047

2.875

2.892

0.285

2.853

0.361

2.874

0.024

3.125

3.136

0.180

3.115

0.166

3.126

0.011

3.375

3.394

0.310

3.403

0.470

3.367

0.138

*E%

127

*E%

*E%

6 Results

inception angle = 0

Fault
resistance

10

200

*X
(KM)
3.625
3.875
4.125
4.375
4.625
4.875
5.125
5.375
0.625
0.875
1.125
1.375
1.625
1.875
2.125
2.375
2.625
2.875
3.125
3.375
3.625
3.875
4.125
4.375
4.625
4.875
5.125
5.375

*D
(KM)
3.565
3.885
4.128
4.357
4.627
4.863
5.120
5.310
0.577
0.858
1.127
1.347
1.599
1.864
2.117
2.376
2.638
2.886
3.144
3.398
3.641
3.903
4.151
4.397
4.653
4.893
5.118
5.350

*E%
1.008
0.165
0.055
0.293
0.036
0.200
0.076
1.080
0.800
0.284
0.030
0.474
0.439
0.182
0.131
0.021
0.215
0.180
0.313
0.381
0.260
0.459
0.432
0.371
0.470
0.295
0.118
0.416

inception angle =
45

*D
(KM)
3.583
3.912
4.160
4.340
4.642
4.862
5.101
5.372
0.659
0.643
0.886
1.127
1.381
1.627
1.888
2.135
2.380
2.633
2.867
3.125
3.383
3.612
3.875
4.123
4.354
4.617
4.879
5.097

* X(KM): Actual fault location .


* D(KM) : Calculated fault location.
* E% : Percentage error

128

*E%
0.701
0.613
0.576
0.583
0.288
0.222
0.404
0.049
0.571
0.294
0.191
0.030
0.093
0.040
0.221
0.159
0.078
0.138
0.134
0.005
0.138
0.214
0.004
0.038
0.357
0.127
0.059
0.466

inception angle =90

*D
(KM)
3.617
3.867
4.120
4.381
4.624
4.889
5.117
5.308
0.642
0.876
1.115
1.376
1.622
1.872
2.115
2.361
2.615
2.856
3.111
3.366
3.607
3.873
4.127
4.375
4.636
4.888
5.141
5.400

*E%
0.139
0.127
0.076
0.102
0.019
0.226
0.128
1.119
0.279
0.025
0.173
0.010
0.044
0.048
0.169
0.240
0.170
0.321
0.233
0.143
0.299
0.032
0.040
0.006
0.188
0.209
0.266
0.409

6 Results

1.8
1.6
Percentage error

1.4
1.2
1.0
0.8
0.6
0.4
0.2
1
7
13
19
25
31
37
43
49
55
61
67
73
79
85
91
97
103
109
115

0.0
Number of Smulation

FIGURE 6.12: Percentage error of Line to Line fault (testing).

Table 6.10: Single Line to Ground ANFIS fault location


testing output results
inception angle = 0

inception angle = inception angle =


45
90

Fault
resistance

*X
(KM)

*D
(KM)

*E%

*D
(KM)

*E%

*D
(KM)

*E%

10

0.625
0.875
1.125
1.375
1.625
1.875
2.125
2.375
2.625
2.875
3.125

0.623
0.874
1.126
1.375
1.627
1.874
2.123
2.375
2.625
2.875
3.125

0.025
0.018
0.024
0.002
0.027
0.012
0.040
0.002
0.003
0.005
0.006

0.625
0.875
1.124
1.375
1.624
1.875
2.125
2.375
2.625
2.875
3.125

0.006
0.006
0.010
0.003
0.011
0.006
0.005
0.003
0.004
0.002
0.004

0.625
0.872
1.133
1.373
1.624
1.879
2.116
2.371
2.631
2.875
3.128

0.005
0.044
0.133
0.036
0.021
0.062
0.153
0.059
0.107
0.001
0.049

129

6 Results

inception angle = 0

Fault
resistance

10

50

inception angle =
45

inception angle =
90

*X
(KM)

*D
(KM)

*E%

*D
(KM)

*E%

*D
(KM)

*E%

3.375
3.625
3.875
4.125
4.375
4.625
4.875
5.125
5.375
0.625
0.875
1.125
1.375
1.625
1.875
2.125
2.375
2.625
2.875
3.125
3.375
3.625
3.875
4.125
4.375
4.625
4.875
5.125

3.376
3.624
3.875
4.128
4.372
4.628
4.871
5.123
5.376
0.559
0.826
1.091
1.368
1.620
1.885
2.143
2.390
2.655
2.902
3.161
3.412
3.646
3.914
4.144
4.377
4.632
4.848
5.087

0.020
0.023
0.002
0.050
0.042
0.053
0.072
0.031
0.019
1.105
0.815
0.570
0.118
0.086
0.170
0.307
0.249
0.500
0.457
0.604
0.625
0.355
0.651
0.322
0.037
0.122
0.442
0.641

3.375
3.625
3.874
4.127
4.375
4.625
4.876
5.122
5.376
0.629
0.875
1.122
1.379
1.630
1.880
2.130
2.381
2.631
2.868
3.161
3.379
3.611
3.889
4.132
4.367
4.634
4.871
5.125

0.005
0.006
0.011
0.032
0.001
0.002
0.015
0.047
0.024
0.075
0.006
0.058
0.070
0.086
0.089
0.080
0.097
0.101
0.121
0.604
0.067
0.226
0.232
0.123
0.137
0.142
0.073
0.002

3.379
3.621
3.890
4.104
4.358
4.624
4.798
5.099
5.374
0.620
0.892
1.123
1.389
1.621
1.868
2.128
2.371
2.622
2.871
3.138
3.382
3.629
3.909
4.127
4.376
4.637
4.833
5.091

0.060
0.059
0.245
0.345
0.285
0.010
1.276
0.433
0.019
0.081
0.288
0.040
0.227
0.064
0.124
0.050
0.065
0.049
0.066
0.215
0.109
0.063
0.566
0.036
0.015
0.192
0.702
0.564

130

6 Results

inception angle = 0

Fault
resistance

200

inception angle =
45

inception angle =
90

*X
(KM)

*D
(KM)

*E%

*D
(KM)

*E%

*D
(KM)

*E%

5.375
0.625
0.875
1.125
1.375
1.625
1.875
2.125
2.375
2.625
2.875
3.125
3.375
3.625
3.875
4.125
4.375
4.625
4.875
5.125
5.375

5.320
0.538
0.739
1.033
1.356
1.599
1.885
2.150
2.371
2.681
2.918
3.174
3.448
3.650
3.933
4.167
4.375
4.639
4.833
5.068
5.290

0.911
1.445
2.264
1.531
0.318
0.439
0.163
0.418
0.068
0.928
0.713
0.813
1.222
0.424
0.973
0.694
0.007
0.226
0.701
0.944
1.411

5.356
0.670
0.882
1.115
1.386
1.627
1.868
2.129
2.365
2.620
2.872
3.130
3.379
3.624
3.900
4.130
4.378
4.637
4.843
5.101
5.336

0.320
0.753
0.116
0.174
0.188
0.041
0.122
0.061
0.166
0.091
0.050
0.078
0.065
0.016
0.418
0.083
0.047
0.195
0.541
0.402
0.648

5.350
0.463
0.895
1.128
1.401
1.628
1.871
2.135
2.369
2.620
2.880
3.138
3.381
3.639
3.910
4.129
4.379
4.634
4.826
5.072
5.275

0.422
2.704
0.341
0.056
0.434
0.057
0.070
0.162
0.094
0.082
0.088
0.220
0.096
0.233
0.576
0.072
0.065
0.146
0.823
0.875
1.668

* X(KM): Actual fault location .


* D(KM) : Calculated fault location.
* E% : Percentage error.

131

6 Results

Percentage error

2.5
2
1.5
1

1
9
17
25
33
41
49
57
65
73
81
89
97
105
113
121
129
137
145
153
161
169
177

0.5

Number of Smulation

FIGURE 6.13: Percentage error of Single Line to Ground


fault (testing).

132

CHAPTER 7
Conclusion And Recommendation

7.1 Conclusion
Nowadays,

electrical

energy

is

being

increasingly

demanded and in order to maintain reliability and security


to an acceptable level, new technologies for protection and
control of power schemes are needed. For distribution
system, underground system is very important these days.
In this thesis, a method that employs Wavelet Transform
and Adaptive Network Fuzzy Inference System (ANFIS)
for fault classification and location in medium voltage
underground system is designed.
To

begin,

this

approach

is

based

on

computer

simulation and designing. For this reason, ATP/EMTP and


MATLAB software are used for simulation. Different types
of faults with different fault locations are simulated by
ATP/EMTP software to prepared input data. On the other
hand, in MATLAB software, five ANFIS based models
which is fault classifier and fault location estimation were
designed.

Models

were

trained

different set of data.


133

and

then

tested

with

7 Conclusion And Recommendation


From the results, different types of faults could be
classified and located on the cable with high accuracy. The
percentage error between ANFIS output and real output is
less than three percent. This shows that the proposed
technique is able to offer acceptable accuracy in both of the
fault classification and fault location estimation. Moreover,
ANFIS could be used as a part of a new generation of high
speed advanced fault locators.

7.2 Recommendation
Since this study had only focused on distribution
level of voltage, it is recommended that further studies be
carried out on variety level of voltage in power system. The
network should be able to perform in any case of power
system. More studies can be done on the effect of cable
aging on fault location and for locating faults in a combined
overhead transmission line with underground power cable.
Moreover,

to

increase

accuracy

in

results,

it

is

recommended that more data for training and testing were


used.

134

REFERENCES

REFERENCES
[1]

Sawatpipat,P., Tayjasanant,T., Fault classification


for Thailands transmission lines based on discrete
wavelet transform, International Conference on
Electrical
Engineering/Electronics
Computer
Telecommunications and Information Technology
(ECTI-CON), Page(s): 636 640, 2010.

[2] M. Jaya Bharata Reddy, D. Venkata Rajesh, D.K.


Mohanta,
"Robust
transmission
line
fault
classification
using
wavelet
multi-resolution
analysis.", omputers & Electrical Engineering,
Volume 39, Issue 4, Pages 1219-1247. , May 2013
[3] Lei Ye, Dahai You, Xianggen Yin, Ke Wang,
Junchun Wu," An improved fault-location method
for distribution system using wavelets and support
vector regression ", International Journal of
Electrical Power & Energy Systems, Volume 55,
Pages 467-472, February 2014.
[4] Ali Rafinia, Jamal Moshtagh, " A new approach to
fault
location
in
three-phase
underground
distribution system using combination of wavelet
analysis with ANN and FLS ", International Journal
of Electrical Power & Energy Systems, Volume 55,
Pages 261-274 , February 2014.

135

REFERENCES
[5]

C.K. Jung, J.B. Lee, X.H. Wang, Y.H. Song


"Wavelet based noise cancellation technique for
fault location on underground power cables" EPSR,
77, pp. 13491362 (2007).

[6] C. Apisit and A. Ngaopitakkul, "Identification of


Fault Types for Underground Cable using Discrete
Wavelet
transform"
Proceedings
of
The
International MultiConference of Engineers and
Computer Scientists 2010, pp.1262-1266, March
2010.
[7] S. Kaitwanidvilai, C. Pothisarn, C. Jettanasen, P.
Chiradeja and A. Ngaopitakkul, "Discrete Wavelet
Transform and Back-propagation Neural Networks
Algorithm for Fault Classification in Underground
Cable" , International MultiConference of Engineers
and Computer Scientists 2011, pp996-1000 , March
2011.
[8] Qi-neng Jiang, Shu-tao Zhao, Jun Zhao and Chenguang Zhao, "A power cable fault location method
combining with wavelet analysis and curve fitting",
China
International
Conference
on
Electricity
Distribution (2010).
[9] W. Zhao, Y.H. Song, Y. Min, "Wavelet analysis
based scheme for fault detection and classification
in underground power cable systems", Electric
Power Systems Research 53, pp. 2330 (2000).

136

REFERENCES
[10] Tamer S. Kamel M. A. Moustafa Hassan, "Adaptive
Neuro Fuzzy Inference System (ANFIS) For Fault
Classification in the Transmission Lines", the
Online Journal on Electronics and Electrical
Engineering (OJEEE), Vol. (2), No. (1), January,
2010.
[11] J. Zhang, Z.Y. He, S. Lin, Y.B. Zhang, Q.Q. Qian,
"An ANFIS-based fault classification approach in
power distribution system ", Electrical Power and
Energy Systems, pp. 243252, vol. 49 (2013).
[12] S.M. Yeo, C.H. Kim, K.S. Hong, Y.B. Lim, R.K.
Aggarwal, A.T. Johns, M.S. Choi, " A novel
algorithm for fault classification in transmission
lines using a combined adaptive network and fuzzy
inference system ", Electrical Power and Energy
Systems, pp. 747758, vol. 25 (2003).
[13]M.Mirzaei, M.Z. A Ab Kadir, E.Moazami, H.Hizam,
Review of fault location methods for distribution
power system, Australian Journal of Basic and
Applied Sciences, 3(3), page(s): 2670-2676, 2009.
[14]Adly
A.Girgis,
Christopher
M.Fallon,
David
L.Lubkeman, A fault location technique for rural
distribution feeders, IEEE Transactions on Industry
Applicatuions, Vol.29, No.6, Page(s):1170-1175,
1993
[15]Santoso, S., Dugan, R.C., Lamoree, J., Sundaram,
A., Distance estimation technique for single line137

REFERENCES
to-ground faults in a radial distribution system,
IEEE Power Engineering Society Winter Meeting,
Vol.4, Page(s): 2551-2555, 2000
[16] Das, R., Sachdev, M.S., Sidhu, T.S., A fault
locator for radial subtransmission and distribution
lines, IEEE Power Engineering Society Summer
Meeting, Vol.1, Page(s): 443-448, 2000
[17] Myeon-Song Choi, Seung-Jae Lee, Duck-Su Lee,
Bo-Gun Jin, A new fault location algorithm using
direct circuit analysis for distribution systems,
IEEE Transactions on Power Delivery, Vol.19,
No.1, Page(s):35-41, 2004.
[18] A.H.A. Bakar, M.S. Ali, ChiaKwang Tan, H.
Mokhlis, H. Arof, H.A. Illias, " High impedance
fault location in 11 kV underground distribution
systems using wavelet transforms", International
Journal of Electrical Power & Energy Systems,
Volume 55, Pages 723-730, February 2014.
[42] Wanjing Xiu, Yuan Liao, "Novel fault location
methods for ungrounded radial distribution systems
using measurements at substation ", Electric Power
Systems Research, Volume 106, Pages 95-100,
January 2014.
[43]

Thomas, D.W.P., Christopoulos, C., Carvalho,


R.J.O., Pereira, E.T., Single and double ended
travelling-wave fault location on a MV system,
Eighth
IEE
International
Conference
on
138

REFERENCES
Developments in Power System Protection, Vol.1,
Page(s): 200-203, 2004
[21] Magnago, F.H., Abur, A., Fault Location using
wavelets, IEEE Transactions on Power Deleivery,
Vol.13, No.4, Page(s): 1475-1480, 1998
[22] Magnago, F.H., Abur, A., A new fault location
technique for radial distribution systems based on
high frequency signals, IEEE Power Engineering
Society Summer Meeting, Vol.1, Page(s): 426-431,
1999
[23] Thomas, D.W.P., Christopoulos, C., Tang, Y., Gale,
P., Stokoe, J., Single ended travelling wave fault
location scheme based on wavelet analysis, Eighth
IEE International Conference on Developments in
Power System Protection, Vol.1, Page(s): 196-199,
2004
[24] Chul-Hwan Kim, Woo-Gon Chung, Myong-Chul
Shin, Sung-Wan Park, Myong-Hee Lee, Yong-Pil
Seo, Hae-Sul Choi, A study on the expert system
for fault location estimation in underground cable,
International Conference on Energy Management
and Power Delivery, Proceedings of EMPD, Vol.1,
Page(s): 67-72, 1995
[25] Ismail Niazy, Javad Sadeh, " A new single ended
fault location algorithm for combined transmission
line considering fault clearing transients without
using line parameters ", International Journal of
139

REFERENCES
Electrical Power & Energy Systems, Volume 44,
Issue 1, Pages 816-823, January 2013.
[26] H.Khorashadi-Zadeh, M.R.Aghaebrahimi, A novel
approach to fault classification and fault location for
medium voltage cables based on artificial neural
network, International Journal of Computational
Intelligence, Vol.2, No.2, pp. 90-93, 2005
[27]

Zadeh-Khorashadi, H. Naveh-Hosseini, S., An


Accurate Fault Locator for Cable Transmission
Using
ANN
Electrotechnical
Conference,
MELECON 2004. Proceedings of the 12th IEEE
Mediterranean, Vol.3, Page(s):901-904, 2004.

[28]

Pranav.D.Raval, ANN based classification and


location of faults in EHV transmission line,
Proceedings of the International MultiConference of
Engineers and Computer Scientists, Vol.1, 19-21
March, 2008

[29] Jain, A. Thoke, A.S. Patel, R.N., Double Circuit


Transmission Line Fault Distance Location Using
Artificial Neural Network, World Congress on
Nature & Biologically Inspired Computing, Page(s):
13-18, 2009
[30] Yuan-Yih Hsu Lu, F.-C. Chien, Y. Liu, J.P. Lin,
J.T. Yu, P.H.S. Kuo, R.R.T., An Expert System for
Locating
Distribution
System
Faults,
IEEE
Transactions
on
Power
Delivery,
Vol.6,
Page(s):366-372, 1991
140

REFERENCES
[31] H.H. Sherwali, P.A.Crossley, Expert system for
fault location and protection analysis, Fourth
International Conference on Power System Control
and Management, Conference Publication No.421,
Page(s): 181-186, 1996
[32] K.K. Kuan, K.Warwick, Real-time expert system
for fault location on high voltage underground
distribution cables, IEE Proceedings C, Generation,
Transmission and Distribution, Vol.139, No.3,
Page(s): 235- 240, 1992
[33] J. Moshtagh, R. K. Aggarwal, A new approach to
fault location in a single core underground cable
system using combined fuzzy logic & wavelet
analysis, The Eight IEE International Conference
on Developments In Power System Protection, pp.
228-231, April 2004.
[34] Javad Sadeh, Hamid Afradi, "A new and accurate
fault location algorithm for combined transmission
lines
using
Adaptive
Network-Based
Fuzzy
Inference
System",
Electric
Power
Systems
Research, pp. 15381545, vol. 79 (2009).
[57] J. J. Mora, G. Carrillo, "
Distribution Systems using
Patterns ", 2006 IEEE
Distribution Conference
America, Venezuela, 2006.

Fault Location in Power


ANFIS Nets and Current
PES Transmission and
and Exposition Latin

141

REFERENCES
[36] Huseyin Eristi, "Fault diagnosis system for series
compensated transmission line based on wavelet
transform
and
adaptive
neuro-fuzzy
inference
system ", pp. 393401, Measurement vol.46 (2013).
[37] Rasli A Ghani, MohdFauzi Othman, A.M. Hashim, "
ANFIS approach for locating precise fault points in
distribution system ", International Journal of
Emerging Technology and Advanced Engineering ,
Volume 2, Issue 6, June 2012.
[38] T.A Short, Electric power distribution handbook,
CRC PRESS, 2004.
[39] Johnston, L., Tweed, N. B., Ward, D. J., and Burke,
J. J., An Analysis of Vepcos 34.5 kV Distribution
Feeder Faults as Related to Through Fault Failures
of Substation Transformers, IEEE Transactions on
Power Apparatus and Systems, vol. PAS-97, no. 5,
pp. 187684, 1978.
[40] Mohd Wazir Mustafa, Power System Analysis,
Second edition, 2009.
[41] Michael Negnevitsky. Artificil Intelligent: A Guide
to Intelligent System, Addison-Wesley, ISBN 0321-20466-2, Harlow, England, (2005).
[42] L. A. Zadeh. Fuzzy sets, Information and Control 8,
338-353, (1965).
[43] J. Lukasiewicz. O logice trjwartociowej, Ruch
Filozoficzny 5, 170f, (1920).
142

REFERENCES
[44] L. A. Zadeh. A rationale for Fuzzy Control, Journal
of Dynamic Systems, Measurement, Control 94 (6),
3-4, (1972).
[45]

J. J. Buckley. Universal
Automatica, 28:1245-1248, (1992).

fuzzy

[46]

The MathWorks, Inc. "Fuzzy Logic


Users Guide" , (1995-2012).

controllers.
Toolbox

[47] Rasli, Hussain and Fauzi. "Fault Diagnosis in Power


Distribution Network Using Adaptive Neuro-Fuzzy
Inference System (ANFIS), Fuzzy Inference System
- Theory and Applications", Dr. Mohammad Fazle
Azeem, (2012).
[48] Jang, J.-S.R.. ANFIS: Adaptive Network Based
Fuzzy Inference System. IEEE Transaction on
Systems, Man and Cybernetic, Vol.23, No.3, pp.
665-685, (1993).
[49]

Mitra, P., Maulik, S., Chowdhury, S.P. &


Chowdhury, S.. "ANFIS Based Automatic Voltage
Regulator
with
Hybrid
Learning
Algorithm."
International Journal of Innovations in Energy
Systems and Power, Vol. 3, No.2, pp.1-5 , (2008)

[50] Tobias Nei, "Cable Fault Location in LV, MV and


HV Underground Cable Networks", April, 2009.
[51] Jan de Kock, Cobus Strauss, Practical Power
Distribution for industry, Newness, pp 30-31,
2004.
143

REFERENCES
[52] Canadian/American EMTP User Group, Alternative
Transients Program (ATP) Rule Book, CoChairmen W. S. Meyer and T. Liu, Portland,
Oregon, 1995.
[53] Gunther, E., Grebe, T., Rambabu, A., and Mader, D.,
Running EMTP on PCs, IEEE Computer Applications in
Power, pp. 33-38, Jan. 1993.
[54] Lszl Prikler, Hans Kristian Hidalen, "ATPDRAW
Users' Manual", Preliminary Release No. 1.0 November
2009.
[55]

International Cables Co. SAE, Web Site:


"http://www.intlcables.com/aProducts/MVCables.aspx".

[56] P.S. Bhowmik, P. Purkait , K. Bhattacharya, A novel


wavelet transform aided neural network based
transmission line fault analysis method, Electrical Power
and Energy Systems 213219, (2009).
[57] Santoso S, Powers EJ, Grady EJ, Hofmann P. "Power
quality assessment via wavelet transform analysis.", IEEE
Trans Power Delivery;11 (2):92430, 1996.
[58] Y. H. Gu and M. H. J. Bollen, "Time-frequency and timescale domain analysis of voltage disturbances," Power
Delivery, IEEE Transactions on, vol. 15, pp. 1279- 1284,
2000.
[59] C. David, Yu and S. H. Khan, High-impedance fault
detection using multi-resolution signal decomposition and
adaptive neural fuzzy inference system , IET Generation,
144

REFERENCES
Transmission & Distribution, Vol. 2, No. 1, pp. 110 - 118,
Jan. 2008.
[60] Zhaoa W, Songa YH, Minb Y. Wavelet analysis based
scheme for fault detection and classification in
underground power cable systems. Electric Power System
Research;53(1):2330, 2000.
[61] K. Chul Hwan and A. Raj, "Wavelet transforms in power
systems. I. General introduction to the wavelet
transforms," Power Engineering Journal [see also Power
Engineer], vol. 14, pp. 81-87, 2000.
[62] "IEEE Guide for Determining Fault Location on AC
Transmission and Distribution Lines", Jun. 2005, IEEE
Standard C37.114, (2004).

145

Appendix A
A.1 Development of the fault location system tool
(FL)
function varargout = fl(varargin)
gui_Singleton = 1;
gui_State = struct('gui_Name', mfilename,
...
'gui_Singleton',
gui_Singleton, ...
'gui_OpeningFcn',
@fl_OpeningFcn, ...
'gui_OutputFcn',
@fl_OutputFcn, ...
'gui_LayoutFcn', [] , ...
'gui_Callback',
[]);
if nargin && ischar(varargin{1})
gui_State.gui_Callback =
str2func(varargin{1});
end
if nargout
[varargout{1:nargout}] =
gui_mainfcn(gui_State, varargin{:});
else
gui_mainfcn(gui_State, varargin{:});
end
% --- Executes on button press in delete.
function delete_Callback(hObject, eventdata,
handles)
% hObject
handle to delete (see GCBO)
% eventdata reserved - to be defined in a
future version of MATLAB

146

Appendix A
% handles
structure with handles and user
data (see GUIDATA)
axes(handles.axes1); cla reset
axes(handles.axes2); cla reset
axes(handles.axes3); cla reset
set(handles.t1,'String',' ')
axes(handles.axes7); cla reset
axes(handles.axes8); cla reset
axes(handles.axes9); cla reset
set(handles.t2,'String',' ')
set(handles.t4,'String',' ')
set(handles.et1,'String',' ')
set(handles.t_error,'String',' ')
% --- Executes just before fl is made
visible.
function fl_OpeningFcn(hObject, eventdata,
handles, varargin)
handles.output = hObject;
guidata(hObject, handles);
% --- Outputs from this function are returned
to the command line.
function varargout = fl_OutputFcn(hObject,
eventdata, handles)
% varargout cell array for returning output
args (see
% Get default command line output from
handles structure
varargout{1} = handles.output;
% --- Executes on button press in b.
function b_Callback(hObject, eventdata,
handles)

147

Appendix A
% hObject
handle to b (see GCBO)
global th_ph_f;
global dlg_ab;
global l2l_ab;
global slf_A;
global D;
th_ph_f=0; dlg_ab=0; l2l_ab=0; slf_A=0; D=0;
file = uigetfile('*.csv');
newData1 = importdata(file);
% Create new variables in the base workspace
from those fields.
vars = fieldnames(newData1);
for i = 1:length(vars)
assignin('base', vars{i},
newData1.(vars{i}));
end
data = newData1.(vars{1});
t=data(1:20000,1);
A =data(1:20000,2);
B =data(1:20000,3);
C =data(1:20000,4);
I0 = (1/3)* (A+B+C);
I1 = (1/3)*(2*A - B -C);
I2 = (1/3)*(sqrt(3)*B - sqrt(3)*C);
% Perform a multilevel wavelet decomposition
of I0 at level 5
[C0,L0] = wavedec(I0,5,'db4');
[Ea0,Ed0] = wenergy(C0,L0);

148

Appendix A
% Extract approximation and detail
coefficients.
c0A5
c0A4
c0A3
c0A2
c0A1

=
=
=
=
=

appcoef(C0,L0,'db4',5);
appcoef(C0,L0,'db4',4);
appcoef(C0,L0,'db4',3);
appcoef(C0,L0,'db4',2);
appcoef(C0,L0,'db4',1);

c0D5
c0D4
c0D3
c0D2
c0D1

=
=
=
=
=

detcoef(C0,L0,5);
detcoef(C0,L0,4);
detcoef(C0,L0,3);
detcoef(C0,L0,2);
detcoef(C0,L0,1);

%Energy for 1-D wavelet or wavelet packet


decomposition
[Ea0,Ed0] = wenergy(C0,L0);
Ed0_max = max(Ed0);
% Reconstruct the Level 5 approximation and
the Level 1 to 9 details.
A05 = wrcoef('a',C0,L0,'db4',5);
D01
D02
D03
D04
D05

=
=
=
=
=

wrcoef('d',C0,L0,'db4',1);
wrcoef('d',C0,L0,'db4',2);
wrcoef('d',C0,L0,'db4',3);
wrcoef('d',C0,L0,'db4',4);
wrcoef('d',C0,L0,'db4',5);

% Evaluation of the square of the wavelet


coefficients
s_D01=D01.*D01;
s_D02=D02.*D02;
s_D03=D03.*D03;
s_D04=D04.*D04;

sum_s_D01=sum(s_D01);
sum_s_D02=sum(s_D02);
sum_s_D03=sum(s_D03);
sum_s_D04=sum(s_D04);

149

Appendix A
s_D05=D05.*D05;

sum_s_D05=sum(s_D05);

% caculation of the disotrted signal


energy%%%%
d01=sum_s_D01;
d02=d01+sum_s_D02;
d03=d02+sum_s_D03;
d04=d03+sum_s_D04;
d05=d04+sum_s_D05;
% Perform a multilevel wavelet decomposition
of phase A at level 5
[Ca,La] = wavedec(A,5,'db4');
% Extract approximation and detail
coefficients.
caA5
caA4
caA3
caA2
caA1

=
=
=
=
=

appcoef(Ca,La,'db4',5);
appcoef(Ca,La,'db4',4);
appcoef(Ca,La,'db4',3);
appcoef(Ca,La,'db4',2);
appcoef(Ca,La,'db4',1);

caD5
caD4
caD3
caD2
caD1

=
=
=
=
=

detcoef(Ca,La,5);
detcoef(Ca,La,4);
detcoef(Ca,La,3);
detcoef(Ca,La,2);
detcoef(Ca,La,1);

% Reconstruct the Level 5 approximation and


the Level 1 to 5 details.
Aa5
Da1
Da2
Da3

=
=
=
=

wrcoef('a',Ca,La,'db4',11);
wrcoef('d',Ca,La,'db4',1);
wrcoef('d',Ca,La,'db4',2);
wrcoef('d',Ca,La,'db4',3);

150

Appendix A
Da4 = wrcoef('d',Ca,La,'db4',4);
Da5 = wrcoef('d',Ca,La,'db4',5);
%Energy for 1-D wavelet or wavelet packet
decomposition
[Eaa,Eda] = wenergy(Ca,La);
Eda_max = max(Eda);
% Evaluation of the square of the wavelet
coefficients
s_Da1=Da1.*Da1;
sum_s_Da1=sum(s_Da1);
s_Da2=Da2.*Da2;
sum_s_Da2=sum(s_Da2);
s_Da3=Da3.*Da3;
sum_s_Da3=sum(s_Da3);
s_Da4=Da4.*Da4;
sum_s_Da4=sum(s_Da4);
s_Da5=Da5.*Da5;
sum_s_Da5=sum(s_Da5);
% caculation of the disotrted signal
energy%%%%
da1=sum_s_Da1;
da2=da1+sum_s_Da2;
da3=da2+sum_s_Da3;
da4=da3+sum_s_Da4;
da5=da4+sum_s_Da5;

% Perform a multilevel wavelet decomposition


of phase B at level 5
[Cb,Lb] = wavedec(B,5,'db4');
% Extract approximation and detail
coefficients.
cbA5
cbA4
cbA3
cbA2

=
=
=
=

appcoef(Cb,Lb,'db4',5);
appcoef(Cb,Lb,'db4',4);
appcoef(Cb,Lb,'db4',3);
appcoef(Cb,Lb,'db4',2);

151

Appendix A
cbA1 = appcoef(Cb,Lb,'db4',1);
cbD5
cbD4
cbD3
cbD2
cbD1

=
=
=
=
=

detcoef(Cb,Lb,5);
detcoef(Cb,Lb,4);
detcoef(Cb,Lb,3);
detcoef(Cb,Lb,2);
detcoef(Cb,Lb,1);

% Reconstruct the Level 5 approximation and


the Level 1 to 5 details.
Ab5 = wrcoef('a',Cb,Lb,'db4',1);
Db1
Db2
Db3
Db4
Db5

=
=
=
=
=

wrcoef('d',Cb,Lb,'db4',1);
wrcoef('d',Cb,Lb,'db4',2);
wrcoef('d',Cb,Lb,'db4',3);
wrcoef('d',Cb,Lb,'db4',4);
wrcoef('d',Cb,Lb,'db4',5);

%Energy for 1-D wavelet or wavelet packet


decomposition
[Eab,Edb] = wenergy(Cb,Lb);
Edb_max = max(Edb);

% Evaluation of the square of the wavelet


coefficients
s_Db1=Db1.*Db1;
s_Db2=Db2.*Db2;
s_Db3=Db3.*Db3;
s_Db4=Db4.*Db4;
s_Db5=Db5.*Db5;

sum_s_Db1=sum(s_Db1);
sum_s_Db2=sum(s_Db2);
sum_s_Db3=sum(s_Db3);
sum_s_Db4=sum(s_Db4);
sum_s_Db5=sum(s_Db5);

% caculation of the disotrted signal


energy%%%%
db1=sum_s_Db1;

152

Appendix A
db2=db1+sum_s_Db2;
db3=db2+sum_s_Db3;
db4=db3+sum_s_Db4;
db5=db4+sum_s_Db5;
% Perform a multilevel wavelet decomposition
of phase C at level 5
[Cc,Lc] = wavedec(C,5,'db4');
% Extract approximation and detail
coefficients.
ccA5
ccA4
ccA3
ccA2
ccA1

=
=
=
=
=

appcoef(Cc,Lc,'db4',5);
appcoef(Cc,Lc,'db4',4);
appcoef(Cc,Lc,'db4',3);
appcoef(Cc,Lc,'db4',2);
appcoef(Cc,Lc,'db4',1);

ccD5
ccD4
ccD3
ccD2
ccD1

=
=
=
=
=

detcoef(Cc,Lc,5);
detcoef(Cc,Lc,4);
detcoef(Cc,Lc,3);
detcoef(Cc,Lc,2);
detcoef(Cc,Lc,1);

% Reconstruct the Level 5 approximation and


the Level 1 to 5 details.
Ac5 = wrcoef('a',Cc,Lc,'db4',11);
Dc1
Dc2
Dc3
Dc4
Dc5

=
=
=
=
=

wrcoef('d',Cc,Lc,'db4',1);
wrcoef('d',Cc,Lc,'db4',2);
wrcoef('d',Cc,Lc,'db4',3);
wrcoef('d',Cc,Lc,'db4',4);
wrcoef('d',Cc,Lc,'db4',5);

153

Appendix A
%Energy for 1-D wavelet or wavelet packet
decomposition
[Eac,Edc] = wenergy(Cc,Lc);
Edc_max = max(Edc);
% Evaluation of the square of the wavelet
coefficients
s_Dc1=Dc1.*Dc1;
s_Dc2=Dc2.*Dc2;
s_Dc3=Dc3.*Dc3;
s_Dc4=Dc4.*Dc4;
s_Dc5=Dc5.*Dc5;

sum_s_Dc1=sum(s_Dc1);
sum_s_Dc2=sum(s_Dc2);
sum_s_Dc3=sum(s_Dc3);
sum_s_Dc4=sum(s_Dc4);
sum_s_Dc5=sum(s_Dc5);

% caculation of the disotrted signal


energy%%%%
dc1=sum_s_Dc1;
dc2=dc1+sum_s_Dc2;
dc3=dc2+sum_s_Dc3;
dc4=dc3+sum_s_Dc4;
dc5=dc4+sum_s_Dc5;
axes(handles.axes1);
plot(t,A );hold on; plot(t, Aa5,'r');
axes(handles.axes2);
plot(t,B,'.' );hold on; plot(t, Ab5,'r')
axes(handles.axes3);
plot(t,C );hold on; plot(t, Ac5,'r')
axes(handles.axes4);
imshow('panel.png')

axes(handles.axes5);
imshow('panel_b.png')

154

Appendix A

axes(handles.axes6);
imshow('panel_c.png')
set(handles.t1,'String',file)
axes(handles.axes7); plot(t,Da5);
axes(handles.axes8); plot(t,Db5);
axes(handles.axes9); plot(t,Db5);

% no fault
if (da5<15) && (db5<15) && (dc5<15)
set(handles.t2,'String','No Fault')
Fault=0;
else
Fault=1;
end
%%%%%%

Fault Classification

%%%%%
if Fault==1
FUZ = readfis('class');
clas = evalfis([Ed0_max Eda_max Edb_max
Edc_max ], FUZ)

if (clas>0.5)&& (clas<1.5)
set(handles.t2,'String','Three Phase
Fault')
th_ph_f= 1;
elseif (clas>1.5)&& (clas<2.5)
set(handles.t2,'String','Double line to
ground Fault between phase A and B')

155

Appendix A
dlg_ab=1;
elseif (clas>2.5)&& (clas<3.5)
set(handles.t2,'String','line to line
Fault between phase A and B')
l2l_ab=1;
elseif (clas>3.5)&& (clas<4.5)
set(handles.t2,'String','Single line to
ground Fault at phase A')
slf_A=1;
else
set(handles.t2,'String','Wrong')
end
%%%%%%%%
%%%%%%%%%%%%%%%

Fault location

if th_ph_f==1
FUZ = readfis('anfis_3ph');
D = evalfis([Ed0_max Eda_max Edb_max Edc_max
], FUZ)
set(handles.t4,'String',D);
end
if slf_A == 1
FUZ = readfis('anfis_slg');
D = evalfis([Ed0_max Eda_max Edb_max Edc_max
], FUZ)
set(handles.t4,'String',D);
end
if l2l_ab == 1

156

Appendix A

FUZ = readfis('anfis_l2l');
D = evalfis([Ed0_max Eda_max Edb_max Edc_max
], FUZ)
set(handles.t4,'String',D);
end
if dlg_ab ==1
FUZ = readfis('anfis_dlg');
D = evalfis([Ed0_max Eda_max Edb_max Edc_max
], FUZ)
set(handles.t4,'String',D);
end
end
function et1_Callback(hObject, eventdata,
handles)
function et1_CreateFcn(hObject, eventdata,
handles)
if ispc &&
isequal(get(hObject,'BackgroundColor'),
get(0,'defaultUicontrolBackgroundColor'))
set(hObject,'BackgroundColor','white');
end
% --- Executes on button press in calc_error.
function calc_error_Callback(hObject,
eventdata, handles)
d = get(handles.t4,'String');
d_val = str2double(d);
AFL = get(handles.et1,'String');
AFL_val = str2double(AFL);
error = 100 * abs(AFL_val-d_val)/6 ;
set(handles.t_error,'String',error);

157

Appendix A
A.2 Creating and Training the ANFIS MODEL
files_names = dir(fullfile(matlabroot,
'anfis'));
no_files = length (files_names);
for nf=3:no_files
fp = files_names(nf).name;
savefile = 'trnError_3ph.csv';
file_Path = ['C:\Program
Files\MATLAB2\R2011a\anfis\' , fp];
importfile_anfis(file_Path)
trnData = train_3ph;
numMFs = 5;
mfType = 'gbellmf';
in_fis = genfis1(trnData,numMFs,mfType);
[out_fis,trnError_class] = anfis(trnData,
in_fis,300);
plot(trnError_class , '*');grid;
xlabel('Epochs');
ylabel('RMSE (Root Mean Squared Error)');
title('Error Curves');
figure(2)
out_fis_o = evalfis(trnData(:,1:4),out_fis);
plot([out_fis_o ,trnData(:,5)], 'x');
end

158

Appendix B
ANFIS Fault Type classification Training Results
B.1 Three phase ANFIS Fault Type classification Training
Results.
Fault
resistance
()

ANFIS
ANFIS ANFIS ANFIS
ANFIS
Output
*X
Output
Target Output Output Output
IA =
(KM)
IA = 0 IA = 45 IA = 90
IA = 180
135

0.5
0.75
1
1.25
1.5
1.75
2
2.25
2.5
2.75
3
3.25
3.5
3.75
4
4.25
4.5
4.75
5
5.25
5.5

1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

1.004
1.004
1.002
1.002
1.001
1.000
0.999
0.998
0.998
0.997
0.997
0.997
0.997
0.997
0.999
0.998
1.000
1.001
1.001
1.005
1.005

0.978
0.988
0.992
1.000
1.008
1.014
1.019
1.021
1.022
1.019
1.014
1.008
0.998
0.990
0.983
0.978
0.978
0.981
0.990
1.006
1.025

* X(KM): Actual fault location .

0.977
1.001
1.008
1.010
1.012
1.009
1.005
1.003
0.999
0.996
0.995
0.995
0.994
0.995
0.997
0.997
0.999
1.001
1.001
1.002
1.003

1.054
1.022
0.991
0.964
0.959
0.958
0.960
0.976
0.989
1.006
1.017
1.028
1.035
1.039
1.044
1.035
1.024
1.012
0.989
0.963
0.935

1.021
1.012
1.003
0.995
0.992
0.990
0.989
0.991
0.992
0.994
0.997
1.000
1.003
1.005
1.007
1.007
1.007
1.005
1.002
0.999
0.991

* IA : Inception Angle

159

Appendix B

Fault
resistance
()

10

ANFIS
ANFIS ANFIS ANFIS
ANFIS
Output
*X
Output
Target Output Output Output
IA =
(KM)
IA = 0 IA = 45 IA = 90
IA = 180
135

0.5
0.75
1
1.25
1.5
1.75
2
2.25
2.5
2.75
3
3.25
3.5
3.75
4
4.25
4.5
4.75
5
5.25
5.5

1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

0.970
0.994
1.005
1.012
1.013
1.012
1.010
1.003
1.000
0.998
0.996
0.996
0.994
0.995
0.997
0.998
1.000
1.001
1.002
1.003
1.003

0.994
1.004
1.003
1.002
1.000
0.998
1.000
0.999
1.000
1.001
1.001
1.001
0.999
0.997
0.996
0.995
0.997
0.998
1.001
1.004
1.009

* X(KM): Actual fault location .

1.009
1.005
0.998
0.992
0.990
0.989
0.991
0.995
0.999
1.003
1.008
1.011
1.012
1.012
1.013
1.010
1.006
1.001
0.993
0.986
0.977

0.967
0.975
0.987
0.999
1.009
1.016
1.019
1.018
1.015
1.009
1.002
0.995
0.989
0.985
0.984
0.985
0.989
0.995
1.003
1.012
1.020

1.012
1.009
1.000
0.998
0.994
0.995
1.000
0.991
0.992
0.999
0.998
1.005
0.999
1.002
1.003
1.000
1.004
1.002
0.998
1.002
1.001

* IA : Inception Angle

160

Appendix B

Fault
resistance
()

30

ANFIS
ANFIS ANFIS ANFIS
ANFIS
Output
*X
Output
Target Output Output Output
IA =
(KM)
IA = 0 IA = 45 IA = 90
IA = 180
135

0.5
0.75
1
1.25
1.5
1.75
2
2.25
2.5
2.75
3
3.25
3.5
3.75
4
4.25
4.5
4.75
5
5.25
5.5

1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

1.012
1.008
1.005
1.002
0.999
0.997
0.995
0.994
0.993
0.993
0.993
0.993
0.994
0.995
0.997
0.999
1.001
1.003
1.005
1.010
1.013

1.000
1.000
1.000
1.001
1.002
1.002
1.003
1.001
0.998
0.999
1.000
0.997
0.993
0.995
0.994
0.993
1.001
1.003
0.971
1.021
1.026

* X(KM): Actual fault location .

0.980
0.985
0.993
0.995
1.003
1.008
1.008
1.013
1.011
1.012
1.015
1.011
1.010
1.006
1.006
1.002
1.001
0.996
0.961
0.989
0.984

0.999
0.999
1.000
0.999
1.000
1.000
0.999
1.002
1.001
1.002
1.002
1.003
1.003
1.003
1.004
1.003
1.004
1.003
0.996
0.995
1.001

0.996
0.997
0.997
0.997
0.997
0.998
0.998
0.999
0.999
0.999
0.999
1.000
1.001
1.001
1.002
1.002
1.003
1.003
1.002
1.003
1.005

* IA : Inception Angle

161

Appendix B

Fault
resistance
()

50

ANFIS
ANFIS ANFIS ANFIS
ANFIS
Output
*X
Output
Target Output Output Output
IA =
(KM)
IA = 0 IA = 45 IA = 90
IA = 180
135

0.5
0.75
1
1.25
1.5
1.75
2
2.25
2.5
2.75
3
3.25
3.5
3.75
4
4.25
4.5
4.75
5
5.25
5.5

1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

0.893
0.966
0.983
0.994
1.007
1.019
1.022
1.030
1.036
1.032
1.036
1.034
1.030
1.025
1.022
1.009
1.002
0.991
0.938
0.962
0.950

0.994
0.996
0.996
0.997
1.002
0.999
1.002
0.999
1.002
1.005
1.002
1.004
1.002
1.004
1.005
1.004
1.006
1.003
0.967
1.005
1.004

* X(KM): Actual fault location .

1.001
1.003
1.005
1.006
1.008
1.008
1.009
1.008
1.008
1.008
1.006
1.006
1.003
1.002
0.999
0.997
0.995
0.992
0.976
0.986
0.983

1.010
1.006
1.004
1.001
0.998
0.999
0.995
0.994
0.996
0.993
0.995
0.996
0.995
0.998
1.000
1.000
1.004
1.007
0.985
0.988
1.016

0.990
0.993
0.996
0.999
1.001
1.003
1.006
1.007
1.008
1.009
1.008
1.008
1.007
1.006
1.004
1.002
0.999
0.996
0.996
0.992
0.983

* IA : Inception Angle

162

Appendix B

Fault
resistance
()

100

ANFIS
ANFIS ANFIS ANFIS
ANFIS
Output
*X
Output
Target Output Output Output
IA =
(KM)
IA = 0 IA = 45 IA = 90
IA = 180
135

0.5
0.75
1
1.25
1.5
1.75
2
2.25
2.5
2.75
3
3.25
3.5
3.75
4
4.25
4.5
4.75
5
5.25
5.5

1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

0.970
0.976
0.979
0.981
0.983
0.985
0.987
0.990
0.991
0.993
0.995
0.997
0.999
1.001
1.002
1.005
1.006
1.008
1.002
1.012
1.014

0.952
0.958
0.963
0.969
0.975
0.980
0.985
0.990
0.995
1.000
1.005
1.009
1.014
1.018
1.022
1.026
1.030
1.034
1.020
1.041
1.044

* X(KM): Actual fault location .

1.043
1.039
1.034
1.031
1.027
1.022
1.018
1.014
1.010
1.004
1.000
0.994
0.989
0.983
0.979
0.973
0.968
0.962
0.928
0.951
0.946

0.973
0.975
0.980
0.986
0.991
0.996
1.001
1.004
1.005
1.010
1.015
1.018
1.022
1.025
1.027
1.028
1.031
1.032
0.961
0.961
1.039

0.982
0.986
0.988
0.989
0.992
0.994
0.996
0.997
1.001
1.003
1.005
1.007
1.009
1.011
1.014
1.016
1.017
1.019
1.041
1.043
1.025

* IA : Inception Angle

163

Appendix B

Fault
resistance
()

200

ANFIS
ANFIS ANFIS ANFIS
ANFIS
Output
*X
Output
Target Output Output Output
IA =
(KM)
IA = 0 IA = 45 IA = 90
IA = 180
135

0.5
0.75
1
1.25
1.5
1.75
2
2.25
2.5
2.75
3
3.25
3.5
3.75
4
4.25
4.5
4.75
5
5.25
5.5

1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

0.955
0.970
0.977
0.984
0.990
0.997
1.004
1.011
1.018
1.024
1.031
1.038
1.044
1.051
1.058
1.064
1.070
1.077
1.081
1.090
1.096

1.003
1.000
0.998
0.995
0.992
0.989
0.969
0.972
0.975
0.979
0.982
0.986
0.989
0.993
0.994
0.998
1.002
1.006
1.023
1.014
1.018

* X(KM): Actual fault location .

0.980
0.983
0.984
0.986
0.987
0.988
0.989
0.991
0.995
0.996
0.998
1.000
1.002
1.004
1.006
1.008
1.012
1.014
1.160
1.021
1.024

0.956
0.961
0.969
0.976
0.983
0.990
0.999
1.003
1.010
1.017
1.024
1.031
1.037
1.041
1.048
1.054
1.060
1.066
0.852
0.857
1.083

0.953
0.956
0.959
0.961
0.963
0.965
0.968
0.971
0.974
0.977
0.979
0.982
0.985
0.989
0.991
0.994
0.997
1.000
1.058
1.061
1.010

* IA : Inception Angle

164

Appendix B
B.2 Double Line to Ground (ABG) ANFIS Fault Type
classification Training Results.
Fault
resistance
()

ANFIS ANFIS
ANFIS ANFIS ANFIS
Output Output
*X
Target Output Output Output
IA =
IA =
(KM)
IA = 0 IA = 45 IA = 90
135
180
0.5

2.000

1.973

2.048

2.107

2.031

0.75

2.005

2.011

2.009

2.022

1.995

1.975

2.019

1.994

1.992

1.980

1.25

2.005

2.022

1.982

1.974

2.000

1.5

2.004

1.987

2.005

1.978

2.002

1.75

2.004

2.017

1.967

1.951

2.003

2.031

2.023

1.964

1.950

1.979

2.25

2.003

2.012

1.965

1.951

2.005

2.5

2.000

2.010

2.003

2.010

2.005

2.75

2.000

1.999

1.973

1.964

2.005

2.001

1.989

1.982

1.974

2.006

3.25

2.000

1.992

1.991

1.985

2.005

3.5

1.997

1.998

1.968

1.956

2.005

3.75

1.997

1.992

2.017

2.016

2.004

1.965

1.991

2.032

2.029

2.009

4.25

1.993

1.986

2.047

2.049

2.002

4.5

1.992

2.214

2.000

2.003

2.001

4.75

1.994

2.006

1.993

2.013

1.999

2.001

1.985

2.082

2.088

1.991

5.25

1.992

1.998

1.989

1.999

1.996

5.5

2.009

1.913

2.011

1.978

2.028

* X(KM): Actual fault location .

* IA : Inception Angle

165

Appendix B
ANFIS
ANFIS ANFIS ANFIS
ANFIS
Fault
*X
Output
Target Output Output Output
Output
resistance (KM)
IA =
IA = 0 IA = 45 IA = 90
IA = 180
135

10

0.5

2.000

1.955

2.015

1.906

2.012

0.75

2.000

1.996

1.984

1.949

1.973

2.000

2.009

1.981

1.968

1.990

1.25

2.037

2.018

1.983

1.985

2.012

1.5

2.053

1.987

2.016

1.991

2.022

1.75

2.066

2.029

1.996

2.012

2.028

2.015

2.029

2.002

2.021

2.012

2.25

2.082

2.028

2.008

2.028

2.029

2.5

2.084

2.010

1.980

1.998

2.026

2.75

2.081

2.020

2.014

2.032

2.021

2.075

2.016

2.014

2.027

2.014

3.25

2.064

2.009

2.013

2.032

2.006

3.5

2.050

2.025

2.011

2.010

1.999

3.75

2.033

1.997

2.006

1.999

1.989

2.034

1.990

2.002

1.985

2.008

4.25

1.991

1.983

1.997

2.012

1.973

4.5

1.969

1.987

2.007

2.033

1.964

4.75

1.944

2.012

1.996

2.007

1.958

1.994

1.972

1.987

1.954

1.970

5.25

1.893

2.036

1.998

2.006

1.947

5.5

1.975

1.956

2.001

1.984

2.020

* X(KM): Actual fault location .

* IA : Inception Angle

166

Appendix B
ANFIS
ANFIS ANFIS ANFIS
ANFIS
Fault
*X
Output
Target Output Output Output
Output
resistance (KM)
IA =
IA = 0 IA = 45 IA = 90
IA = 180
135

30

0.5

2.012

1.980

2.033

1.963

2.023

0.75

1.995

2.000

2.016

1.978

1.961

2.014

2.008

2.010

1.984

1.968

1.25

2.000

2.015

2.004

1.990

1.974

1.5

2.002

2.038

1.968

2.005

1.980

1.75

2.004

2.022

1.996

1.998

1.985

2.036

2.023

1.993

2.002

1.962

2.25

2.007

2.024

1.991

2.004

1.996

2.5

2.009

2.026

2.050

1.993

2.001

2.75

2.010

2.018

1.991

2.008

2.006

2.011

2.015

1.991

2.008

2.010

3.25

2.011

2.008

1.993

2.009

2.015

3.5

2.012

2.022

1.991

2.006

2.019

3.75

2.012

1.993

1.999

2.009

2.023

1.993

1.985

2.003

2.009

2.007

4.25

2.011

1.974

2.007

2.009

2.030

4.5

2.011

1.992

2.029

1.986

2.034

4.75

2.010

2.010

2.005

2.007

2.037

1.959

1.950

2.019

2.007

2.010

5.25

2.007

1.973

2.014

2.003

2.043

5.5

2.000

1.973

1.963

2.002

2.017

* X(KM): Actual fault location .

* IA : Inception Angle.

167

Appendix B
ANFIS ANFIS
ANFIS ANFIS ANFIS
*X
Output Output
Target Output Output Output
(KM)
IA =
IA =
resistance
IA = 0 IA = 45 IA = 90
135
180
Fault

50

0.5

2.041

2.003

1.973

2.048

2.031

0.75

1.958

1.998

1.975

2.042

2.059

2.025

1.996

1.977

2.038

2.017

1.25

1.981

1.995

1.978

2.033

2.052

1.5

1.990

2.051

1.925

1.959

2.050

1.75

1.998

1.993

1.984

2.023

2.047

1.992

1.993

1.987

2.017

1.990

2.25

2.008

1.994

1.992

2.011

2.042

2.5

2.012

1.998

2.231

2.013

2.039

2.75

2.012

1.995

2.000

1.998

2.037

2.004

1.995

2.005

1.991

2.031

3.25

2.013

1.996

2.011

1.985

2.032

3.5

2.011

1.996

2.000

2.007

2.028

3.75

2.008

1.999

2.024

1.970

2.025

1.995

2.001

2.031

1.964

2.020

4.25

1.998

2.004

2.039

1.957

2.018

4.5

1.992

2.000

2.021

1.980

2.016

4.75

1.984

1.999

1.911

1.973

2.013

1.968

2.010

2.056

1.944

1.988

5.25

1.965

2.085

2.065

2.035

2.007

5.5

1.955

2.082

2.097

2.033

1.977

* X(KM): Actual fault location .

* IA : Inception Angle

168

Appendix B
ANFIS ANFIS
ANFIS ANFIS ANFIS
*X
Output Output
Target Output Output Output
(KM)
IA =
IA =
resistance
IA = 0 IA = 45 IA = 90
135
180
Fault

100

0.5

1.994

2.063

1.995

2.045

1.904

0.75

1.949

2.050

1.994

2.034

2.081

1.943

2.044

1.995

2.029

1.951

1.25

1.956

2.038

1.995

2.023

2.070

1.5

1.959

2.009

2.060

1.959

2.065

1.75

1.962

2.026

1.996

2.013

2.060

1.997

2.023

1.997

2.008

2.059

2.25

1.969

2.017

1.998

2.003

2.050

2.5

1.972

1.991

2.224

2.017

2.045

2.75

1.975

2.006

1.999

1.995

2.041

1.970

2.001

2.001

1.990

2.037

3.25

1.982

1.996

2.002

1.986

2.032

3.5

1.985

2.039

2.003

2.012

2.028

3.75

1.988

1.988

2.005

1.977

2.024

1.982

1.984

2.007

1.974

2.020

4.25

1.994

1.979

2.009

1.970

2.017

4.5

1.997

2.021

2.008

1.994

2.013

4.75

2.000

1.991

2.096

2.092

2.010

1.958

1.971

2.013

1.964

2.051

5.25

2.006

1.989

2.065

1.960

2.003

5.5

2.083

1.996

2.114

1.941

2.074

* X(KM): Actual fault location .

* IA : Inception Angle

169

Appendix B
ANFIS ANFIS
ANFIS ANFIS ANFIS
*X
Output Output
Target Output Output Output
(KM)
IA =
IA =
resistance
IA = 0 IA = 45 IA = 90
135
180
Fault

200

0.5

1.980

1.975

1.981

1.954

2.006

0.75

1.926

1.975

1.984

1.964

2.022

2.021

1.974

1.985

1.968

1.975

1.25

1.935

1.974

1.986

1.973

2.025

1.5

1.940

2.012

1.970

1.908

2.027

1.75

1.945

1.975

1.990

1.981

2.029

1.902

1.976

1.991

1.986

2.057

2.25

1.954

1.976

1.993

1.990

2.032

2.5

1.959

1.955

2.026

2.034

2.034

2.75

1.963

1.977

1.996

1.998

2.035

1.942

1.978

1.998

2.002

2.015

3.25

1.972

1.979

2.000

2.006

2.039

3.5

1.976

2.122

2.011

1.988

2.040

3.75

1.980

1.980

2.004

2.014

2.042

1.959

1.981

2.006

2.018

2.021

4.25

1.988

1.983

2.008

2.022

2.045

4.5

1.992

2.128

2.019

2.005

2.047

4.75

1.996

1.889

2.001

1.954

2.049

2.119

1.986

2.013

2.030

2.005

5.25

2.003

2.176

2.043

1.948

2.052

5.5

1.979

1.988

1.979

1.946

2.011

* X(KM): Actual fault location .

* IA : Inception Angle

170

Appendix B
B.3 Line to Line (AB) ANFIS Fault Type classification
Training Results.
Fault
*X
resistance (KM)

ANFIS ANFIS ANFIS ANFIS ANFIS


Target Output Output Output Output Output
IA = 0 IA = 45 IA = 90 IA = 135 IA = 180

0.5

3.015

3.013

3.000

3.033

2.989

0.75

3.007

3.008

3.004

3.016

2.974

2.996

3.000

3.016

3.001

3.001

1.25

3.007

3.000

3.006

2.992

3.006

1.5

3.004

2.997

3.005

2.984

2.988

1.75

3.002

2.991

3.004

2.978

3.001

3.008

2.996

2.947

2.989

2.988

2.25

3.003

2.993

2.993

3.003

2.997

2.5

3.004

2.993

3.046

3.008

3.020

2.75

2.998

2.995

2.994

2.976

3.000

3.001

2.995

2.991

2.979

2.988

3.25

3.002

2.987

2.988

2.982

3.037

3.5

3.005

2.993

2.988

2.987

3.002

3.75

3.000

3.000

3.000

3.002

3.000

2.998

2.973

2.993

2.998

2.988

4.25

3.000

3.018

3.007

3.004

3.007

4.5

3.000

3.011

3.003

3.010

3.004

4.75

2.992

2.991

2.996

3.017

2.993

2.991

3.009

3.002

3.023

2.999

5.25

3.003

3.010

3.008

3.030

3.019

5.5

3.000

3.000

3.000

2.997

3.000

* X(KM): Actual fault location .

* IA : Inception Angle

171

Appendix B

Fault
*X
resistance (KM)

10

ANFIS ANFIS ANFIS ANFIS ANFIS


Target Output Output Output Output Output
IA = 0 IA = 45 IA = 90 IA = 135 IA = 180

0.5

3.006

3.005

2.948

2.967

2.984

0.75

3.060

3.007

3.040

2.995

3.007

3.034

2.997

2.998

3.009

2.974

1.25

2.986

2.998

3.004

3.004

3.014

1.5

3.027

2.977

3.007

3.028

2.973

1.75

2.956

2.960

3.036

3.020

3.006

2.999

2.969

3.000

2.990

3.015

2.25

2.958

3.025

3.015

2.993

3.007

2.5

2.986

3.029

2.986

3.014

3.015

2.75

2.974

3.009

3.109

3.004

2.992

2.986

3.003

2.940

3.014

2.990

3.25

2.994

2.987

2.989

2.992

2.996

3.5

3.064

3.038

2.975

2.990

2.954

3.75

3.000

3.000

3.000

2.995

3.000

4
4.25
4.5
4.75
5
5.25
5.5

3
3
3
3
3
3
3

3.019
3.077
2.967
2.975
2.941
2.933
2.999

2.980
3.005
3.018
3.122
2.980
3.031
3.001

2.916
2.929
2.944
3.020
3.063
3.082
3.000

2.994
2.996
2.995
3.000
3.005
2.998
2.998

3.001
3.002
3.005
3.015
3.012
3.008
3.001

* X(KM): Actual fault location .

* IA : Inception Angle

172

Appendix B

Fault
*X
resistance (KM)

30

ANFIS ANFIS ANFIS ANFIS ANFIS


Target Output Output Output Output Output
IA = 0 IA = 45 IA = 90 IA = 135 IA = 180

0.5

3.000

2.972

3.016

3.077

3.026

0.75

2.952

2.983

2.979

2.975

3.022

3.020

2.994

3.013

2.992

2.969

1.25

3.016

2.992

3.064

3.041

3.021

1.5

3.039

3.001

3.054

2.982

3.019

1.75

3.031

2.989

3.047

2.987

2.990

2.942

3.003

2.978

2.964

3.053

2.25

3.038

3.007

2.974

3.008

3.000

2.5

2.975

3.000

3.028

3.055

2.989

2.75

2.996

2.995

3.017

3.025

2.935

2.992

3.028

2.983

2.910

3.012

3.25

3.036

3.044

2.933

3.043

2.899

3.5

2.907

3.012

2.881

3.052

2.888

3.75

3.001

3.000

3.006

2.992

2.999

3.054

3.001

3.001

3.065

2.975

4.25

3.042

2.996

3.006

3.054

2.977

4.5

2.989

3.029

2.966

2.906

3.013

4.75

3.051

2.993

3.020

3.008

2.974

2.991

2.990

3.040

3.075

3.000

5.25

2.966

2.983

2.947

2.830

3.019

5.5

3.000

3.000

2.997

3.002

3.003

* X(KM): Actual fault location .

* IA : Inception Angle

173

Appendix B

Fault
*X
resistance (KM)

50

ANFIS ANFIS ANFIS ANFIS ANFIS


Target Output Output Output Output Output
IA = 0 IA = 45 IA = 90 IA = 135 IA = 180

0.5

3.037

3.041

3.025

3.007

2.915

0.75

3.082

3.004

2.992

2.952

2.939

3.014

2.997

3.015

2.988

3.026

1.25

3.029

3.000

3.021

3.017

2.951

1.5

3.019

3.007

3.015

2.991

2.960

1.75

3.005

2.994

3.048

2.988

2.994

2.959

3.002

2.984

2.982

3.036

2.25

2.995

3.007

3.068

2.970

2.990

2.5

3.009

2.995

2.998

3.029

2.962

2.75

2.955

3.005

2.994

2.992

3.017

3.001

2.983

2.984

2.980

2.986

3.25

2.926

3.007

2.997

3.001

3.024

3.5

2.909

3.001

2.998

3.006

3.026

3.75

3.028

3.000

2.931

3.015

2.998

2.943

3.002

2.984

3.019

3.025

4.25

2.897

2.995

2.991

3.023

3.034

4.5

2.993

2.994

2.988

2.970

2.978

4.75

2.927

2.984

2.989

3.003

3.020

2.983

2.986

2.993

3.012

2.998

5.25

3.017

3.021

2.998

2.954

2.980

5.5

3.001

3.001

2.994

3.012

2.975

* X(KM): Actual fault location .

* IA : Inception Angle

174

Appendix B

Fault
*X
resistance (KM)

100

ANFIS ANFIS ANFIS ANFIS ANFIS


Target Output Output Output Output Output
IA = 0 IA = 45 IA = 90 IA = 135 IA = 180

0.5

2.774

2.963

2.862

3.016

2.938

0.75

2.937

3.111

2.936

3.012

2.967

3.005

3.000

3.087

2.951

3.002

1.25

2.809

3.044

2.921

3.025

2.977

1.5

2.834

2.932

2.922

3.024

2.981

1.75

2.944

3.053

3.061

3.030

2.960

3.037

2.998

2.826

2.873

3.002

2.25

3.003

2.988

2.964

2.921

2.928

2.5

2.989

3.012

2.954

3.040

3.041

2.75

2.965

3.015

3.039

3.018

3.015

2.917

3.082

3.038

3.014

3.023

3.25

2.949

3.040

3.042

3.010

3.032

3.5

2.954

3.038

3.039

3.006

3.039

3.75

2.989

3.001

3.103

3.076

3.021

2.982

2.946

3.034

2.995

3.062

4.25

3.100

2.918

3.045

2.988

3.075

4.5

2.958

2.977

2.992

2.976

3.061

4.75

3.070

3.004

3.110

2.972

3.110

2.972

2.946

3.138

2.961

3.093

5.25

3.045

2.897

2.974

2.941

3.072

5.5

2.995

2.996

2.963

2.972

2.995

* X(KM): Actual fault location .

* IA : Inception Angle

175

Appendix B

Fault
*X
resistance (KM)

200

ANFIS ANFIS ANFIS ANFIS ANFIS


Target Output Output Output Output Output
IA = 0 IA = 45 IA = 90 IA = 135 IA = 180

0.5

2.973

3.011

3.039

2.979

2.982

0.75

2.867

3.111

2.934

2.982

2.921

3.005

2.999

3.006

2.679

2.973

1.25

2.966

2.908

2.983

2.987

2.954

1.5

2.973

2.765

2.990

2.990

2.964

1.75

2.991

2.875

3.038

2.992

3.002

3.031

3.049

2.826

3.054

2.909

2.25

2.967

2.999

2.933

2.957

2.997

2.5

3.062

3.075

3.161

2.975

3.014

2.75

2.989

3.002

3.029

3.002

2.996

2.982

3.034

3.018

3.005

2.990

3.25

2.983

3.009

3.007

3.007

2.993

3.5

2.981

3.007

2.997

3.009

2.994

3.75

2.809

2.911

2.870

3.097

2.963

2.984

2.996

2.976

3.014

3.003

4.25

2.984

2.998

2.967

3.016

3.006

4.5

2.979

3.063

2.959

3.019

3.010

4.75

2.989

2.990

2.963

3.021

3.015

2.981

2.986

2.964

3.023

3.018

5.25

2.986

2.883

2.957

3.025

3.019

5.5

2.932

3.020

2.914

3.040

3.088

* X(KM): Actual fault location .

* IA : Inception Angle

176

Appendix B
B.4 Single Line to Ground (AG)
classification Training Results.

ANFIS Fault Type

ANFIS
ANFIS ANFIS ANFIS
ANFIS
Fault
*X
Output
Target Output Output Output
Output
resistance (KM)
IA =
IA = 0 IA = 45 IA = 90
IA = 180
135

0.5

4.024

4.001

4.013

3.998

4.008

0.75

4.013

4.000

4.013

3.992

3.983

4.006

3.997

3.997

3.997

4.001

1.25

4.002

3.994

3.996

3.998

4.000

1.5

4.000

3.991

3.996

4.000

4.000

1.75

4.001

3.989

3.998

4.002

4.000

4.002

3.986

4.007

4.003

4.000

2.25

4.012

3.995

4.009

3.997

3.990

2.5

4.006

3.983

4.007

4.005

4.000

2.75

4.009

3.995

4.009

3.998

3.992

3.937

3.993

4.007

3.999

4.005

3.25

4.005

3.991

4.005

3.999

3.994

3.5

4.035

4.015

3.946

3.950

3.994

3.75

4.002

3.988

4.003

4.000

3.995

3.966

3.987

3.972

3.971

4.001

4.25

4.018

3.998

3.984

3.981

4.004

4.5

4.013

3.994

3.995

3.991

4.006

4.75

4.008

3.992

4.004

3.999

4.008

3.996

3.987

4.001

4.002

4.026

5.25

4.002

3.999

4.012

4.011

4.006

5.5

3.999

4.009

4.009

4.013

4.002

* X(KM): Actual fault location .

* IA : Inception Angle.

177

Appendix B

ANFIS
ANFIS ANFIS ANFIS
ANFIS
Fault
*X
Output
Target Output Output Output
Output
resistance (KM)
IA =
IA = 0 IA = 45 IA = 90
IA = 180
135

10

0.5

4.017

4.000

3.997

4.005

3.996

0.75

4.000

3.999

3.998

4.008

3.995

4.006

3.999

3.991

4.009

4.000

1.25

4.002

4.000

3.996

4.012

4.003

1.5

4.000

3.998

3.992

4.014

4.001

1.75

4.000

4.000

4.009

4.015

4.010

4.003

3.997

4.001

4.014

4.013

2.25

3.982

3.991

4.002

4.003

4.002

2.5

4.013

3.985

4.002

4.005

4.014

2.75

3.989

4.000

4.005

3.998

4.000

3.993

4.004

4.005

3.996

4.001

3.25

3.997

4.007

4.006

3.993

3.996

3.5

4.002

3.999

3.976

3.987

3.985

3.75

4.004

4.011

4.007

3.988

3.989

4
4.25
4.5
4.75
5
5.25
5.5

4
4
4
4
4
4
4

4.007
3.991
3.988
3.987
4.017
3.991
3.998

4.013
4.010
4.003
4.004
3.974
4.001
4.008

4.007
4.010
4.003
4.007
3.973
4.001
4.004

3.985
3.985
3.990
3.986
4.013
3.994
4.000

3.991
3.994
3.996
3.996
4.023
3.996
3.994

* X(KM): Actual fault location .

* IA : Inception Angle.

178

Appendix B

ANFIS
ANFIS ANFIS ANFIS
ANFIS
Fault
*X
Output
Target Output Output Output
Output
resistance (KM)
IA =
IA = 0 IA = 45 IA = 90
IA = 180
135

30

0.5

4.002

4.010

4.012

4.010

4.009

0.75

4.001

3.982

3.984

3.961

3.998

4.002

3.994

4.009

4.008

4.001

1.25

4.002

4.003

4.006

4.008

4.000

1.5

4.003

4.009

4.004

4.008

4.001

1.75

4.005

4.013

4.004

4.010

4.002

4.005

4.004

4.006

4.012

4.004

2.25

3.994

4.004

3.996

3.988

4.001

2.5

4.006

4.003

4.022

4.017

4.010

2.75

3.994

4.003

4.000

3.997

4.002

4.003

4.008

4.001

4.011

4.010

3.25

3.995

4.003

4.003

4.005

4.003

3.5

3.996

3.996

3.994

3.994

3.996

3.75

3.995

4.002

4.006

4.013

4.005

4.001

3.961

3.986

3.995

3.993

4.25

3.994

3.997

3.985

3.995

3.992

4.5

3.993

3.997

3.986

3.996

3.991

4.75

3.992

3.998

3.989

3.998

3.992

4.007

4.019

4.015

4.030

4.008

5.25

3.991

4.002

3.997

4.001

3.997

5.5

3.990

4.004

3.998

4.003

4.002

* X(KM): Actual fault location .

* IA : Inception Angle.

179

Appendix B

ANFIS
ANFIS ANFIS ANFIS
ANFIS
Fault
*X
Output
Target Output Output Output
Output
resistance (KM)
IA =
IA = 0 IA = 45 IA = 90
IA = 180
135

50

0.5

3.994

3.975

4.024

3.998

4.005

0.75

3.998

4.004

4.008

3.983

3.990

4.000

3.999

4.004

3.999

4.007

1.25

4.003

3.994

3.994

3.999

4.008

1.5

4.007

3.991

3.986

3.998

4.008

1.75

4.010

3.990

3.984

3.997

4.009

4.013

3.991

3.990

3.996

4.008

2.25

4.000

3.993

4.001

3.994

3.996

2.5

4.016

4.007

4.029

3.996

4.003

2.75

4.001

3.998

3.998

3.997

3.998

4.002

4.000

3.996

3.999

3.998

3.25

4.003

4.002

3.995

4.000

3.999

3.5

4.012

4.002

4.032

3.993

4.005

3.75

4.005

4.006

3.992

4.002

4.001

4.006

4.008

3.990

4.003

4.001

4.25

4.006

3.992

3.987

3.995

4.010

4.5

4.002

3.994

3.987

3.995

4.009

4.75

3.999

4.014

3.993

3.995

4.007

4.009

4.016

3.983

4.006

4.003

5.25

4.010

4.007

4.008

3.993

4.004

5.5

3.994

4.009

4.013

3.990

4.002

* X(KM): Actual fault location .

* IA : Inception Angle.

180

Appendix B

ANFIS
ANFIS ANFIS ANFIS
ANFIS
Fault
*X
Output
Target Output Output Output
Output
resistance (KM)
IA =
IA = 0 IA = 45 IA = 90
IA = 180
135

100

0.5

3.996

4.007

4.015

3.996

3.996

0.75

3.999

3.986

4.015

3.997

3.995

4.000

3.986

4.010

4.003

3.999

1.25

4.001

3.986

4.001

4.003

4.000

1.5

4.002

3.986

3.993

4.002

4.001

1.75

4.001

3.989

3.988

4.000

4.002

4.000

4.011

3.989

3.998

4.002

2.25

4.005

4.011

4.007

3.997

3.995

2.5

3.997

4.030

4.019

3.994

4.002

2.75

4.006

4.010

4.004

3.997

3.995

4.006

4.010

4.003

3.996

3.995

3.25

4.006

4.009

4.002

3.996

3.995

3.5

3.995

4.030

4.021

3.997

3.998

3.75

4.007

4.008

3.999

3.996

3.995

4.008

4.008

3.997

3.995

3.995

4.25

4.000

3.995

3.990

4.003

3.999

4.5

4.001

3.993

3.993

4.004

3.998

4.75

4.000

3.994

3.999

4.004

3.998

4.009

4.006

3.992

3.994

3.996

5.25

3.997

3.994

4.007

4.000

3.996

5.5

4.010

3.990

4.002

3.995

3.994

* X(KM): Actual fault location .

* IA : Inception Angle.

181

Appendix B

ANFIS
ANFIS ANFIS ANFIS
ANFIS
Fault
*X
Output
Target Output Output Output
Output
resistance (KM)
IA =
IA = 0 IA = 45 IA = 90
IA = 180
135

200

0.5

3.992

4.007

4.006

3.998

3.994

0.75

3.998

3.999

4.010

4.003

4.000

4.002

4.000

4.010

4.006

4.000

1.25

4.004

3.999

4.003

4.007

4.004

1.5

4.004

3.996

3.995

4.006

4.003

1.75

4.003

3.996

3.990

4.004

4.002

4.000

3.988

3.990

4.002

3.999

2.25

4.005

3.988

4.006

4.001

4.000

2.5

3.993

4.025

4.015

3.997

3.992

2.75

4.005

3.987

4.004

4.000

4.000

4.005

3.987

4.004

3.999

4.000

3.25

4.005

3.987

4.003

3.999

4.000

3.5

3.994

3.987

4.012

3.999

3.993

3.75

4.005

3.987

4.002

3.998

4.000

4.005

3.986

4.001

3.998

3.999

4.25

4.005

3.986

3.987

3.997

4.003

4.5

4.005

4.002

3.991

4.007

4.004

4.75

4.005

3.986

3.997

4.007

4.004

4.005

3.986

3.998

3.996

3.999

5.25

4.005

4.004

4.000

4.003

3.998

5.5

3.992

3.996

3.990

3.998

3.993

* X(KM): Actual fault location .

* IA : Inception Angle.

182


:
.
) (Wavelet Transform

.
).(ANFIS

) (ATP/EMTP
MATLAB

183


ANFIS

.
,
, .
.

:
: .

:
.

:
.

:
).(ANFIS

184

.
: .

.
:
.

: .

185

/
) (

-

2014

/
-

. /

-

2014

:
. / ................................... .
" - - "" "

. / ..................................... .
" - - "

. / ...................................... .
" - - "

-

2014

You might also like