U2013 Proceedings of Intemational Conference on Modelling, Identification & Control (ICMIC)

Cairo, Egypt, 31stAug.- 2ndSept. 2013

Distance Relay Protection for Short and Long Transmission line

Mohamed M Ismail and M A. Moustafa Hassan

Abstract- The advanced application of Artificial Neural in operation and their total length. These lines are exposed to
Network was introduced recently in Protection of Transmission faults [2] due to lightning, short circuits, faulty equipments,
lines in Electric Power Systems. In this proposed research, the miss operation, human errors, overload, and aging. Many
application of ANFIS and ANN for Distance Relay Protection electrical faults manifest in mechanical damages, which must
for short and long Transmission line, under different loading be repaired before returning the line to service. It is
conditions, in Electrical Power systems will be introduced and
practically impossible to avoid consequences of natural
discussed. Considering different loading conditions, the
events, physical accidents, and equipment failure or miss
suggested and proposed technique deals with fault detection,
operation which results in the loss of power; voltage dips on
classification, and location in short and long Transmission lines.
All these tasks will be addressed in details in this article. It the power system. The protective relay is preventive device
considers; firstly, detecting the fault occurrence in very short which operates after a fault has occurred and it helps in
time and isolate the faulty section of the short and long minimizing the duration of trouble and limiting the outage
transmission lines. Secondly to classify the fault type and deduce time and related damage. The protective relays can be
which of the three phases are exposed to the fault. Finally, electromechanical, solid-state or digital [3]. Natural events
locating the fault will be achieved easily. The input training data can cause short circuits i.e. faults which can either be single
of the detection units are firstly derived from the fundamental
phase to ground or phase to phase or phase to phase to ground
values of the voltage and current measurements (using digital
or a three phase fault [4] . The transmission line is modeled by
signal processing via Discrete Fourier Transform (DFT». These
various techniques according to, mainly, its length.
measurements were simulated considering different loading
conditions.
Therefore, there are three main types to represent the
transmission line [5], Short T.L (less than 80 km or 50 miles)
index Terms- ANFIS, ANN, Artificial Intelligent
, medium T.L (from 80 to 240 km or from 50 to 150 miles)
Techniques, Distance Protection, Fault Detection, Fault
and long T.L (longer than 240 km or 150 miles). When faults
Classification, Fault Location, short and Long
occur in the power system, they usually provide significant
Transmission Lines
changes in the system quantities like over-current, over or
I. INTRODUCTION
under-power, power factor, impedance, frequency and power
An electric power system comprises of a number of
or current direction. These abnormalities, faults and
power-generating plants, transmission lines, and substations.
intolerable conditions must provide a distinguishable
Electric power is normally generated at II-25kV in a power
difference from the normal operating or tolerable conditions.
station. To transmit over long distances, it is then stepped-up
In this paper, the captured faulted voltage and current signals
to 220kV / 500kV levels as necessary. Power is carried
from instrument transformers have to be filtered and analyzed
through a transmission network of high voltage lines.
using digital signal processing tools. Then, the filtered signals
Usually, these lines run into hundreds of kilometers and
are used for detection, classification, and location of the fault.
deliver the power into a common power pool called the grid.
Fault detection, classification and location have been a goal
The grid is connected to load centers (cities / towns) through a
of power system engineers since the creation of distribution
sub-transmission network of normally 33kV / 66kV lines.
and transmission systems. Quick fault detection can help
These lines terminate into a substation where the voltage is
protect equipment by allowing the disconnection of faulted
stepped-down to 22kV / IIkV / 6.6kV etc., for power
lines before any significant damage is done. Accurate fault
distribution to load points through a distribution network of
classification and location can help utility to remove
lines. This can be shown in Figure (1) [1]:
Color I<ey:
Black: Generation
persistent faults and locate areas where faults regularly occur,
,I�r-rF-'-----' sUbg��;;,���'UlUS reducing the frequency and length of power outages. For
Blue: Transmission
Green: Distribution

Generating Station the purpose of fault detection, classification and location in

P <
t-FF"'------1 �
"ma'YcuS °t;�msmission lines, distance relays can be c nsidered as the
Generating
Step up
I�n�� seconda'YcusNfflt attempt to realize this aim [6, 7]. As they can provide
Transformer

Fig I : Typical Power System adequate protection especially for the transmission lines and
they can be employed for initiating the protection reaction
The rapid growth of electric power systems over the past few
after the fault inception as soon as possible. On the other
decades has resulted in a large increase of the number of lines
hand, this protection scheme requires more accurate and
sophisticated computation routines due to the increasing
complexities of modem power system networks [8, 9] . Thus,
Manuscript received March 1, 2013.
Mohamed .M. Ismail . Faculty of Engineering , Helwan University , 1 while fault detection, classification and location schemes
shref street - Helwan Cairo , Egypt , department of electrical power and have been developed in the past, a variety of algorithms
machines (e-mail: m_m_ismail@yahoo.com ) continue to be developed to perform this task more accurately
M.A.Moustafa . Faculty of Engineering , Cairo University ,Giza , Egypt
and more effectively. Regarding the fault detection and
department of electrical power and machines (e-mail:
mmustafa 98@hotmail.com ) classification, the problem was solved some time ago by

204
Authorized licensed use limited to: Universiti Malaysia Perlis. Downloaded on November 14,2024 at 15:37:41 UTC from IEEE Xplore. Restrictions apply.
 U2013 Proceedings of International Conference on Modelling, Identification & Control ( lCMIC)
Cairo, Egypt, 31 S1Aug.- 2ndSept. 2013

introducing the traditional relaying principles such as over detecting the fault and the noise that can be obtained from the
current, distance, overvoltage, under voltage, differential, etc. measured signals lead to decreasing the accuracy . Although
All of these principles are based on a comparison between the there are previous researches, as in [12-16], that utilize the
measurements and predetermined setting calculated taking artificial intelligence techniques in fault detection in double
into account only predetermined operating conditions and circuit T.Ls, there isn't any research that deals with the fault
fault events. Consequently, if the actual power system classification and location in these lines using AI tools.
conditions deviate from the anticipated ones, the Moreover, the zero sequence mutual coupling between the
measurements and settings determined by the classical relay parallel lines introduces many complexities in establishing
design have inherent limitations in classifying certain fault the ANFIS fault units. However, this is solved in our paper by
conditions, and the performance of classical protective relays using the zero sequence current to be an input from the inputs
may significantly deteriorate . While for fault location, of the ANFIS and ANN classifier and it is also helpful in
traveling wave phenomena was early employed for its differentiating between the ground and non ground faults.
purposes. However, these schemes suffered from different The previous work obtained in [17-21] is also improved in
shortcomings associated with the propagation and this paper by using ANN technique as well as ANFIS method
economical problems. Then, the impedance-based fault and the simulations in our article was applied on short and
locators were developed depending on the available long of transmission lines
measured quantities at the locator position. Later, the II. TRANSMISSION LINE MODELING
revolution of digital technologies and microprocessor Short Transmission Line
applications [lO] were employed to develop these schemes A single line diagram for the protected T.L is shown in
into their digital forms. However, these digital versions are Figure (2). It is consists of two circuits of 80 km length,
usually structured following almost the same basic functions 66 kV voltage level and 2 GVA short circuit level.
that are used by the conventional schemes. Hence, these new
versions in most cases suffered from the associated
shortcomings with the function essence. Different factors are
�'I ! ::1i--
:---
xom
responsible for these drawbacks as these approaches are
based on deterministic computations on a well defmed model
Fig 2 : Single Line Diagram For The T.L
of the system to be protected. In general, the steps in the
protective relaying algorithm are detecting the fault , Classify Since the T.L is considered a link between the short and
the fault and fmally Locate the fault medium T.Ls, it represented in ATP by single 7[ model so that
In this paper , the proposed distance protective relay was more accurate data can be acquired. The single phase 7[ model
applied on two different transmission , first system is double in ATP is shown in Figure (3) [28] where R, L, and Care
circuit transmission line with 66 kV, 2 GVA, and 80 km and scalar values. The capacitance value Con the branch card is
the second system is Single circuit transmission line with 500 split up into two equal parts and connected to NODE k and
kV, 20 GVA, and 160 km. The fault detectors, classifiers, and NODEm.
locators of both lines are firstly instituted by the ANN and
ANFIS training, and then they are tested in variety of system R L

conditions to ensure the robustness and the comprehensive of

'k. ---T
...--.r----c"'--
-' ....
.rr "'il.
r " ,.
,... .,...
' ... --- .,. ----,r----- IU!t

the proposed protection scheme. Protection acts to open and -L _

-- 2
�, c
I
I
close circuit breakers, thus changing the structure of the
power system, whereas the control functions act continuously
to adjust system variables without changing its structure of
..... wL: .... ::lit .......... tMACKSi, � 2-

the power system. Distance relays are normally used to Fig 3 : Single Pi Model in ATP
protect transmission lines,. They respond to the impedance Long Transmission Line
between the relay location and the fault location. Distance After the implementation of the proposed protection scheme
relays are preferred than over current relays in the line on short transmission line with 66 kV considering constant
protection because they are not nearly so much affected by loading conditions, the proposed algorithm will now utilized
changes in short-circuit-current magnitude as over current for protecting actual overhead transmission line in Egypt. The
relays are, and, hence, are much less affected by changes in T.L is single circuit 500 kV and 160 km in length. It connects
generating capacity and in system configuration. This is between the two substations; Abu Zabal and the Suez. Since
because distance relays achieve selectivity on the basis of all the 500 kV T.Ls in Egypt are single circuits up till now,
impedance rather than current. The main distance relay likewise the selected one, the difficulty of the mutual
schemes could be classified in [9-11]. Our work is an coupling between parallel lines is vanished. Therefore, the
improvement of [12], [13], [14] , [15] and [16] such that we proposed protection relaying is applied on this line to
are using short and long transmission line at different loading accomplish the three fault tasks at a verity of loading
conditions with two techniques ANFIS and ANN, also in this conditions even though this line is twice the length of the first
article we can locate the fault with very good accuracy and we line and it is at much higher voltage level. The MATLAB
are using digital signal processing for filtering the noise , program represents the transmission lines by either
while the previous work deal only with classification and distributed parameter line method or a series combination of

205
Authorized licensed use limited to: Universiti Malaysia Perlis. Downloaded on November 14,2024 at 15:37:41 UTC from IEEE Xplore. Restrictions apply.
U2013 Proceedings of Intemational Conference on Modelling, Identification & Control (ICMIC)
Cairo, Egypt, 31 S1Aug.- 2ndSept. 2013

pi models. In this paper the latter method is used because the Switching in the power systems , Faults in the power systems
selected line is considered a medium T.L. Therefore, even a with their different types , non linear elements in the power
single pi model could be sufficient. However, in order to get systems; e.g. static shunt compensators, and magnetic
more accurate results from the T.L model, a series saturations in the transformers , Shunt capacitance of long
combination of pi models is adopted. In spite of this, it is transmission lines , dc-components in the power systems due
found that the adequate number of pi models that required to to the energy stored in the inductance that decays
emulate the T.L. is depending on the maximum frequency of exponentially in the system resistances. In most studies in the
the harmonics that are imposed on the fundamental power system analysis, the necessitated voltage and the
components and the length of the transmission line. There is current measurements are the fundamental components of the
a fonnula that relates this maximum frequency and the voltage and current signals. Therefore, several steps, as
number of pi models [14], which is shown in Figure (6), such that Analog to digital conversion,
Nv Digital low pass filter , Fourier analysis and Scaling.
fmax =--

81 (1) Digital Signal Pl"Ocessing

Where N is Number of PI sections , \) is Propagation speed in

=
km/s L in H/km, C in F/km and I is Line length in km . In
order to fmd this t;nax in the suggested power system model,
Fig 6 : Digital Signal Processing
the T.L is simulated with different number of pi models
starting from one pi model and increasing till the Total IV. THE DISTANCE PROTECTION SCHEME
Harmonic Distortion (THD) of the simulated model becomes On the occurrence of a fault, the fault detection unit activates
constant. The harmonic distortion when the T.L is the fault classification unit and the fault location unit. The
represented by single pi model where the THD in this case classification unit is used to classify the fault; if the fault is
equals 0.04%. While, the harmonic distortion of using five pi single phase to ground, phase to phase, double phase to
models to represent the T.L equals 0.05%. These steps were ground or three-phase fault. The function of fault location
repeated till the THD becomes constant by modeling the T.L unit in the protective relaying system is to predict the distance
with twenty pi units and it equals 0.06%. . Therefore each unit from the relay to the fault point. Therefore, the control block
will be exactly 8 km, 5% of the line length. In Figure (4) the derives the decision of trip or no-trip from the fault detector
transmission line model with 20 pi models is presented. And and in the fault condition it receives the output signals of
the whole protected system is shown in Figure (5) where classifier and locator units to determine the fault type and the
there are the two substations and the transmission line location at which the fault occurs, as illustrated in Figure (7).
between them, in addition to their electrical parameters. In Figure (8), a detailed protection algorithm is represented
where the Digital Signal Processing (DSP is shown for the
three main fault tasks. It also illustrates the input
measurements for the faults units. The training process for
the Fault Detection (FD) Unit continuously applies the DSP
every 20 msec which is equivalent to full wave power cycle.
Besides, the fault, regardless its type, is initiated at fault time
(Tc) equals 5 msec, taking the phase voltage Va as reference.
=====.:'
:r= �I--
== �':=="'
., -- : ==1:'1:1
--, -- ==
Regarding the Fault Classification (FC) and Location (FL)
Units, their training procedure carries out DSP only for one
Fig 4: The Transmission Line Model
power cycle during the fault existence.

�.r-C::::::: :::::::: :::::::: :::: �:: ���: :::::::::: :::::::: ::::::::::: D�su, ��
500kV·160kmTransmissionline

Short CirCUit level r: 002046 anmlkm Short CirCUit level

vonage & Current Voltage & Current
X= 0314 onmlkm 40 kA for 1sec
40 kA for 1sec Measurements Measurements
XJR,18 b � 3 683e-6 mhomn XJR'18

Fig 5 : The Whole Protected System

III. DIGITAL SIGNAL PROCESSING
Since the protective relay's development and operation
depend mainly on the voltage and current measurements,
these data should be processed in order to present the required
accurate information for the relay. This is carried out through
Fig 7 : Proposed Protection Scheme
disposing of any distortion or noise that imposed on the signal
measurements. Besides these distortion and noise, there are
hannonic frequencies of multiple integer of the fundamental
component, i.e. 50 Hz, and DC values which disturb the
fundamental signal because of the disturbances due to

206
Authorized licensed use limited to: Universiti Malaysia Perlis. Downloaded on November 14,2024 at 15:37:41 UTC from IEEE Xplore. Restrictions apply.
U2013 Proceedings of International Conference on Modelling, Identification & Control (ICMIC)
Cairo, Egypt, 31stAug.- 2ndSept. 2013

Fault detection unit:

COlltrol The training data used to train the ANFIS of the fault
detection unit are taken at the no fault conditions and fault
conditions. The fault conditions are carried out at all different
Ullit fault types at fault distance (Dr) 5%, 40% and 80% of the line
with inception fault time (Tc) 5 msec. and fault resistances
Activatioll (Rr) 0 and 100 ohms. The output data from ANN and ANFIS
are 0 for no fault conditions and 1 for fault conditions. Table
1 is the testing data for fault detection , Tr is the fault
inspection time, Rr is the fault resistance , Dr is the fault
distance in Km. . The impedance of the three phase are Za ,
Zb and Zc And O/P1 is the ANFIS output and O/P2 is the
ANN output .
Table 1: Testing Data of the Fault Detection Unit for the Fault
Conditions

Fig 8 : Detailed Protection Scheme �

.,.
� �
.,. .,.
V. NEURO FUZZY CONTROLLER i: i: i: I'
N
"0
During the past three decades, fuzzy logic has been an area of
heated debate and much controversy. The first
o
implementation of Zadeh's idea was accomplished in 1975 by ;:, U>
o
o
....
Mamdani, which demonstrated the viability of fuzzy logic
control (FLC) for a small model steam engine. After this
pioneer work, many consumer products as well as other high
technical applications have been developed using fuzzy
.... - o
technology. U>
o .... '" U.
o ....
A list of industrial applications and home appliances based on
FLC can be found in several recent references [22 to 28].

VI. ARTIFICIAL NEURAL NETWORK

o o - o , -
;:, U>
00 ;::! ;:, ....
U>
o
....
o
U> ... v:> .... �
A neural network (NN) is a machine like human brain with
properties of learning capability and generalization. It
requires a lot of training to understand the model of plant. The
basic property of this network is capability to learn the -
o .... .... o
U>
characteristic of nonlinear dynamic system mappings. The o ... .... .... U.
.... o i.H
neural network consists as shown in figure (15) of three
layers, an input layer, one or more hidden layers and an
output layer. Neurons of hidden and output layers have an
activation functions. The knowledge of NN can be achieved o , o
;:, U> p ....
'"'
through a learning algorithm process [29]. o
....
o .... ....
U.
....

VII. FAULT DETECTION, CLASSIFICATION AND

LOCATION USING ARTIFICIAL INTELLIGENT
TECHNIQUE
o
o
a) Short transmission line o 00

Since that the ATP is used to model the T.L and also to create
both the training and testing data, therefore, the training
procedure of fault classification (FC) and fault location (FL) ,
o - ....
units are established separately from that of fault detection ;:,
.... - o
00
o :... o '" .... :...
.... 00 :... ....
u.
....
(FD) unit. This is because of the difficulties which will be
inserted to the model in ATP to implement different DSP
processes in the same model at the same running cycle.
However, these separated training procedures allow
instituting the distance protection scheme at much variety of
the system conditions.

207
Authorized licensed use limited to: Universiti Malaysia Perlis. Downloaded on November 14,2024 at 15:37:41 UTC from IEEE Xplore. Restrictions apply.
U2013 Proceedings of Intemational Conference on Modelling, Identification & Control (ICMIC)
Cairo, Egypt, 31stAug.- 2ndSept. 2013

:.-
== Table 3: Testing Data of the Fault Classification Unit for the SLO Faults
,
0
;:, �
t-J 0 ...
-
0 N
'" 0 U; -
:; ;;> ;:, ;:, ... '" N
0
N
0 Q<, '" '" Q<,
N N t
� "'l N
.. "0 0 0
-

0 � -
N N N I' �
... �
-I
...., ::::
....,
-c
C:
q
-c
'"
.£
-c
C:
.E.
'?
�
-c
.n.
-c �
*
= I' C: I' C:
Fault classification unit: N N
"0 0;;
Fault Classification Unit is the subsequent unit to be
activated from the Fault Detection Unit in case of fault.
0
b -
;;> .... 0 .... 0 '" 0 N
The output of ANN and ANFIS fault classification unit is ;:,
0
0 e
0J0 �
N
Q<,
;:,
....
0J0
'"
'"
0J0 !" '"
... 0
'" :.. N -
....

• 1 for Single Line to Ground (SLG) fault

• 2 for Double Lines (DL) fault ;0
• 3 for Double Lines to Ground (DLG) fault Cl
;;> - ....
� ... � 0J0 0 .... � � N - - -
• 4 for Triple Lines (TL) I Triple Lines to Ground S 0J0 N
0J0 � .... '" ....
U.
0J0
'"
-.J
'"
'"
0J0
0
N '" N
(TLG)fault
In this stage a Control Circuit, Figure (9), is employed to
approximate the output (O/Pl) of the ANFIS and the output
(O/P2) of ANN classifiers, so that the final output (O/P*) Table 4: Testing Data of the Fault Classification Unit for the TL / TLO Faults

will referred to the accurate number of the fault type (Nc)

.--.---,----,,--,--.---,
"'l N

0
..
� N N N
"'0
I' �
N
�
...., q i: .E. � .n.
::l :::: -c -c
'"
-c -c -c -c
C: C: C: I' C: I' C:
�
N N
..0 "'"0

Tt. Output from

.. , :.-
tt ANFIS D t.cto,
.... ..
0" TTh�· Cc:,���;;lt gi�:::" ==
DIP'
0 0 h - ....
;:, N
0J0 N
0J0
!"
�
...
:::;
'"
e
'"
!I'
'"
e
'"
QO
'"
.... N
....
....

:;-
;0
0 0 0 0 .... .... 0
Fig 9 : Control Circuit (C.C) ;:, N
0J0 N
0J0 Cl ;; ;:,
Q<,
;-I
....
E
'"
�
-
;:,
'"
Q<,
'"
f- f-
�
....

Each subsystem in Figure (9) represents a certain fault

type where it contains the minimum and maximum
boundaries for each Nc as shown in Figure (10). The
boundaries for the fault types will be as follows: 1) Fault location unit:
• If 0.5::; Nf < 1.5, therefore, Nf"" 1 -7 SLG fault.
Fault Location Unit process is working in parallel with the
• If 1.5::; Nf <2.5, therefore, Nf"" 2 -7 DL fault. Fault Classification Unit after the activation from the
• If 2.5::; Nf <3.5, therefore, Nf"" 3 -7 DLG fault. Detection Unit. Its function is to estimate the normalized fault
• If 3.5::; Nf < 4.5, therefore, Nf"" 4 -7 TLlTLG distance Dc within the suggested transmission line. The
fault. results are indicated in tables 5 and 6.

Table 5 : Testing Data of the Fault Location Unit for Nr= 3

N
n 0 0
0 � �
0 N N N 1' IV
..., .E. <:J' .n.
::;; � .", .", .", .",
i::
The Outpulfrom i:: I' i:: I' i::
IheANFIS Detector N N
DIP "0 0;;

0
0 0 -.J 0 0 0
-.J � Vo
Vo
00
'='
0 00 :..., '=' '=' 00 :...,
Fig JO : Contents of Each Subsystem in C.C IV '" '" :,. IV :..., :;;: '0 -.J

The testing data are selected at different fault distances,

resistances, inception times and types and the results of the
simulation is indicated in tables 3 and 4

208
Authorized licensed use limited to: Universiti Malaysia Perlis. Downloaded on November 14,2024 at 15:37:41 UTC from IEEE Xplore. Restrictions apply.
U2013 Proceedings of Intemational Conference on Modelling, Identification & Control (ICMIC)
Cairo, Egypt, 31stAug.- 2ndSept. 2013

F�ntJt Detection Uuit

Vab
o o o
Iabc
'"
'" '"
'" :::;

0
0 0 0 0 0
0
'" :x; '" '" 00
'" N
00 ...
:;;: V. -..l
:.. 00 ;:;; ;:;; 101'
ActivatiOl\ or
0
Deactivation

Fault Chllssificntiou Unit Fnult LocntiOIl Unit

Table 6: Testing Data of the Fault Location Unit for Nr= 4

"N0 0 0
0
...., N N N I' � �
...
.E. � .n.
....,
-I
� 'C
'? 'C '?
i:: " I' i:: I' "
N N
"0 "'"
0

0 0 0 0 ... 0 ... 0 0 Unit

0
0 ....
Ut
""
u. :...
-..l
....
...
�
-.I
Ut
'" � '" U.
Ut
U.
Control

-..l

Figure 12 : The Protection Scheme

Fault detection units

...
...
o
:...
...
o
The input data to the ANFIS and ANN detection units are the
impedances of the three phases (magnitude and phase i.e. 6
inputs) after the digital signal processing is achieved as
b) Long transmission line shown previously. While the output data from ANFIS and
ANN units are 0 for no fault conditions and 1 for fault
For the long transmission line that indicated in figure 8. The conditions.
complete line system with the protection algorithm is Fault classification units:
indicated in figure 11. The Fault Classification and Fault The input data to the first ANFIS and ANN classification
Location units apply DSP for only power cycle during the units are the impedances of the three phases (magnitude and
fault existence. This process is shown in Figure (12) where it phase) and the zero sequence components of the currents, i.e.
is regulated by three control blocks which are The "Fault 7 inputs, after the digital signal processing is achieved.
Instance" block which is responsible for the occurrence of the While the output data from it are 1 for single phase to ground
fault at 5 msec by taking in consideration the phase voltage of faults., 2 for double phases to ground faults and 3 for three
phase A is the reference , The "FD Control" block which phases to ground faults.
allows the continuous DSP of the training data for the ANFIS On the other hand, the input data to the second ANFIS and
detectors and The "FC and FL Control" block implements ANN classification units are the impedances of the three
the DSP of the training data for the ANFIS classifiers and phases (magnitude and phase i.e. 6 inputs) after the digital
locators for only one power cycle during fault condition, i.e. signal processing is achieved. While the output data from it
20 msec. are 1 for double phases faults and 2 for three phases faults.
__ -,f�
l:U.tt loclll o n Fault location units:

·�t
Training data for fault location units:
I �

.�� The training data used to train the ANFIS and ANN of the
Fault Location Units are as same as those of Fault
�
Abu l� Sub:stwUlm :iOO kV· 160 t:m TrlnlmluJon line �
-, Classification Units and they also classified into different
� RlidIlII lllUt d,=====,.:a
: : �:. :�:bJt� section depending on either if the zero sequence current is
' --------. , ======� fl ..... -- ., .-- .. oe=----...roc
zero or not.
• "'t.J
...
.:1 · t
The input data to the ANFIS and ANN Location units are the
tl� "--=-1
impedances of the three phases (magnitude and phase i.e. 6
inputs) after the digital signal processing is achieved. While
" --. the output data from ANFIS locators are the nonnalized fault
distance value.

Fig I I : Complete Transmission line system with the Protection algorithm

209
Authorized licensed use limited to: Universiti Malaysia Perlis. Downloaded on November 14,2024 at 15:37:41 UTC from IEEE Xplore. Restrictions apply.
 U2013 Proceedings of International Conference on Modelling, Identification & Control (ICMIC)
Cairo, Egypt, 31stAug.- 2ndSept. 2013

B. Testing Procedure: performance in this application than ANN detector. The

The control unit will receive the output from the different performance of the artificial intelligent detectors in this paper
ANFIS detectors, classifiers, and locators and [mally, it will is improved by using the digital signal processing.
carry out the proper control action.In the testing procedure of
the ANFIS Detectors, the faulty condition is considered if the I. CONCLUSIONS

output of the ANFIS detector is greater than or equals 0.5.

Otherwise, it will be treated as un faulted situation. The proposed intelligent distance relay demonstrate
On the other hand, In the testing process of ANFIS and ANN successful performance for the three main protection tasks;
classifiers a Control Circuit, Figure (16), is employed to fault detection, fault classification and fault location. The
approximate the output (O/P) of the ANFIS and ANN results obtained in this scheme are very supportive. All the
classifiers, so that the final output (O/P*) will referred to the testing data for the ANFIS and ANN detectors in the fault and
accurate number of the fault type (Nc) as mentioned before in no fault conditions give the correct output within the
the training step. Each subsystem in Figure (16) represents a mentioned periods. For the fault classification task, all the
certain Nr where it contains its minimum and maximum testing data give the correct output number which represents
boundaries as shown in Figure (17). The boundaries for the the fault type by approximating it to the nearest integer
Nr in this case will be as follows: number. For the fault location task, the maximwn error is
about 8% for ANFIS detector and 10% for ANN detector for
• If 0.5:S Nf < 1.5, therefore, Nf"" 1 -7 SLG fault or
different fault resistances, different fault inception times and
DL fault according to 10.
different fault types. Moreover, the testing procedure takes
• If 1.5:SNr <2.5, therefore, Nr"" 2 -7 DLG fault or into account the randomness of the faults on T.Ls with respect
TL fault according to 10. to the time of occurrence, fault location, type and resistance
• If 2.5:SNr <3.5, therefore, Nr"" 3 -7 TLG fault. and even for the loading conditions.
In the following in Tables (7) and (8)., there are the testing
data that are used to test the proposed protection scheme.
REFERENCES
These data are chosen randomly at different fault inception
time, Tr, fault type, Nf, power angle, 8, fault resistance, Rr,
fault location, Df [I] Sunil S. Rao, "Switchgear and Protection", 10th Edition,
Table 7 : Testin Data at Sil nal Line to Ground Fault KHANNA Publishers, Delhi, 1994.
I:' "'"l !"l "'"l "'"l t""
"'"l Z [2] AREVA T&D Automation & Information Systems, Network
I:'
I:' I:'
I:'
I:'
r;; !"l
I:'
r;; Protection and Automation Guide, www.areva-td.com
-I � ;0:1 »-
...., ...., � �. z »- � »- �.
:; Z Z :; [3] A. M. Abd EI Aleem, "Design and Implementation of A
"
"'"l
r;; z �. z "
Q. Q. Q. Programmable Simulation for the Estimation of the Over current
Protection Relays Immunity Against Decaying DC offset", M.
Sc. Dissertation, Office of Graduate Studies of Cairo University,
2007.
... .... [4] K. Kasinathan, "Power System Fault Detection and
U> U>
Classification using Wavelet Transforms and Adaptive
Resonance Theory Neural Networks.", M. Sc. Dissertation,
Office of Graduate Studies of Kentucky University, 2006.
[5] w. D. Stevenson, Elements of Power System Analysis, Fourth
Edition, McGraw-Hili, Inc., Singapore, 1982.
N
o .... o 0 [6] Stanley H. Horowitz and Arun G. Phadke, Power System
= U> o 00 N N N
00 o N Relaying, Third Edition, Research Studies Press Limited,
England, 2008.
[7] A. Russell Mason, The Art and Science of Protective Relaying,
John Wiley & Sons, Inc., New York, 1956.
[8] L. Hewitson, M. Brown, B. Ramesh, Practical Power Systems
00 N
Protection, First Edition, Integra Software Services Pvt. Ltd,
:;: o U>
Pondicherry, India, 2004.
[9] Sunil S. Rao, Switchgear and Protection, Tenth Edition,
KHANNA Publishers, Delhi, 1994.
[10] Power Systems Relaying Committee (PSRC) of the IEEE
Power Engineering Society, "IEEE Guide for Protective Relay
.... Applications to Transmission Lines", IEEE Std C37.113, 1999.
.... ....
o [II] Fernando Calero, "Mutual Impedance in Parallel Lines:
U> ...
Protective Relaying and Fault Location Considerations",
Schweitzer Engineering Laboratories, Inc., 2008.
[12] Huisheng Wang and W. W. Keerthipala, "Fuzzy-Neuro
Approach to Fault Classification for Transmission Line
Protection" IEEE Transactions on Power Delivery, Vol. 13, No.
The simulation results for the short and long transmission 4, pp.1093-1104, 1998.
[13] B. V. Coury and D. C. Jorge, "Artificial Neural Network
line, we can find that the fault detection, classification and
Approach to Distance Protection", IEEE Transactions on Power
location can be demonstrated successfully using ANFIS and Delivery, Vol. 13, No. 1, pp. 102-108, 1998.
ANN controllers, while the ANFIS detector gives better [14] Bhavesh R. Bhalja and Rudra Prakash Maheshwari,
"High-Resistance Faults on Two Terminal Parallel Transmission

210
Authorized licensed use limited to: Universiti Malaysia Perlis. Downloaded on November 14,2024 at 15:37:41 UTC from IEEE Xplore. Restrictions apply.
U2013 Proceedings of International Conference on Modelling, Identification & Control (ICMIC)
Cairo, Egypt, 31 S1Aug.- 2ndSept. 2013

Line: Analysis, Simulation Studies, and an Adaptive Distance

Relaying Scheme", IEEE Transactions on power delivery, vol. Table 8: Testing Data at Triples Line to Ground Fault
22, no. 2, April 2007.
[15] K. Dash, A. K. Pradhan, and G. Panda, "A Novel Fuzzy Neural ""l
o ""l
Network Based Distance Relaying Scheme", IEEE Transactions o
»­
on Power Delivery, Vol. 15, No. 3, pp.902-907, 2000. z »­
""l z
[16] M. Jayabharata Reddy and D.K. Mohanata, "Performance iii z

Evaluation of Adaptive Network Based Fuzzy Inference System

Approach for Location of Faults on Transmission Lines Using
Monte Carlo Simulation", IEEE Transactions on Fuzzy
Systems, Volume 16, Issue 4, pp. 909 - 919, Aug. 2008.
[17] Tamer S. Kamel, M. A. Moustafa Hassan, A. EI-Morshedy;
N
N
N
'" ;;,
o
"Application of Artificial Intelligent Approach in Distance
Relay for Transmission line Protection", Proceeding of the
International Scientific & Technical Conference ' Actual Trends
in Development of Power System Protection and Automation'
being held on, P. 28-36, Moscow, Russia, 7-10 September
2009.
[18] Tamer S. Kamel, M. A. Moustafa Hassan, A. EI-Morshedy;
"Using a Combined Artificial Intelligent Approach In Distance
Relay For Transmission line Protection In EPS", Proceedings of
Fifth International Conference on Soft Computing, Computing
with Words and Perceptions in System Analysis, Decision and
o ... ... o o
Control (ICSCCW 2009), Famagusta - North Cyprus, P. ....
o :...
o
00
o
00
o
:...
o
:...
...
222-227, Sept 2-4, 2009.
[19] Tamer S. Kamel, and M. A. Moustafa Hassan, "On-Line
Trained Adaptive Neuro-Fuzzy Inference System for Distance
Relay of Transmission line Protection", Proceedings of "The
2009 World Congress on Electronics and Electrical
Engineering, (WCEEENG'09), Cairo, Egypt April 6-9, 2009. N
....
...
'"
[20] Tamer S. Kamel, and M. A. Moustafa Hassan, "Adaptive
Neuro Fuzzy Inference System For Fault Location in
Transmission Lines", Presented in "The 2009 World Congress
on Electronics and Electrical Engineering, (WCEEENG'09),
Proceedings of "The 2009 World Congress on Electronics and
Electrical Engineering, (WCEEENG'09), Cairo, Egypt April -
'"
...
'"
o
00
N
00
N
00
o o o
6-9, 2009.
[21] Tamer S. Kamel, M. A. Moustafa Hassan, A. EI-Morshedy;
"Advanced Distance Protection Scheme for Long Transmission
Lines In Electric Power Systems Using Multiple Classified
ANFIS Networks", Presented at Fifth International Conference
on Soft Computing, Computing with Words and Perceptions in ...
00
System Analysis, Decision and Control (ICSCCW 2009), P.
228-232, Famagusta - North Cyprus, Sept 2-4, 2009.
[22] S. M, "Industrial applications of fuzzy control,"
North-Holland, Amsterdam, 1985.
[23] M. II.R, "Fuzzy logic technology and applications," IEEE
o o ...
Press, Piscataway" 1994. ...
:.. :.. 00
o
o ... o
[24] M. Gupta and S. N., "Intelligent control systems," Theory and
Applications, IEEE Press, Piscataway, NJ" 1996.
[25] K. B., "Fuzzy engineering," Prentice Hall, Upper Saddle
River, NJ" 1997.
[26] R. A, "Applied research in fuzzy technology," Kluwer
Academic Publishers, Boston, 1994.
[27] K. A. and Gupta, "Introduction to fuzzy arithmetic theory and
applications," Van Nostrand Reinhold, 1985.
[28] T. R. Nguyen H, Sugeno M and Y. R, "Theoretical aspects of
fuzzy control," John Wiley and Sons, 1995.
[29] MATLAB R2009a help

211
Authorized licensed use limited to: Universiti Malaysia Perlis. Downloaded on November 14,2024 at 15:37:41 UTC from IEEE Xplore. Restrictions apply.