Ijeet: International Journal of Electrical Engineering & Technology (Ijeet)
Ijeet: International Journal of Electrical Engineering & Technology (Ijeet)
Ijeet: International Journal of Electrical Engineering & Technology (Ijeet)
ISSN 0976 6553(Online) Volume 5, Issue 7, July (2014), pp. 32-44 IAEME
32
DETECTION CLASSIFICATION AND LOCATION OF FAULTS ON 220 KV
TRANSMISSION LINE USING WAVELET TRANSFORM AND NEURAL
NETWORK
R.P. Hasabe, A.P. Vaidya
Electrical Engineering Department,
Walchand College of Engineering, Sangli, Maharashtra. India
ABSTRACT
This paper presents a discrete wavelet transform and neural network approach to fault
detection and classification and location in transmission lines. The fault detection is carried out by
using energy of the detail coefficients of the phase signals and artificial neutral network algorithm
used for fault type classification and fault distance location for all the types of faults for 220 KV
transmission line. The energies of the all three phases A, B, C and ground phase are given in put to
the neural network for the fault classification. For each type of fault separate neural network is
prepared for finding out the fault location. An improved performance is obtained once the neutral
network is trained suitably, thus performance correctly when faced with different system parameters
and conditions.
Index Terms: Fault Detection, Fault Classification, Wavelet Transform.
I. INTRODUCTION
Transmission lines are a crucial part of an electrical power system as they allow bulk energy
to be transported from a group of generating units to an area where the energy is needed. Protecting
of transmission lines is one of the important tasks to safeguard electric power systems. For safe
operation of transmission line systems, the protection system need to be able to detect, classify,
locate accurately and clear the fault as fast as possible to maintain stability in the network. The
occurrence of any transmission line faults gives rise to transient condition. Fourier transform gives
information about all frequencies that are present in the signal but does not give any information
about the time at which these frequencies were present. Wavelet transform has the advantage of fast
response and increased accuracy as compared to conventional techniques. The wavelet
transformation is a tool which helps the signal to be analyzed in time as well as frequency domain
INTERNATIONAL JOURNAL OF ELECTRICAL ENGINEERING &
TECHNOLOGY (IJEET)
ISSN 0976 6545(Print)
ISSN 0976 6553(Online)
Volume 5, Issue 7, July (2014), pp. 32-44
IAEME: www.iaeme.com/IJEET.asp
Journal Impact Factor (2014): 6.8310 (Calculated by GISI)
www.jifactor.com
IJEET
I A E M E
International Journal of Electrical Engineering and Technology (IJEET), ISSN 0976 6545(Print),
ISSN 0976 6553(Online) Volume 5, Issue 7, July (2014), pp. 32-44 IAEME
33
effectively. It uses short windows at high frequencies, long windows at low frequencies. Using multi
resolution analysis a particular band of frequencies present in the signal can be analyzed. The
detection of fault is carried out by the analysis of the wavelets coefficients energy related to phase
currents. ANN based techniques have been used in power system protection and encouraging results
are obtained [1], [2], [3]. Neural networks are used for different applications as classification, pattern
recognition. In classification, the objective is to assign the input patterns to one of the different
classes [4], [5]. Fault location in a transmission line using synchronized phasor measurements has
been studied for a long time. Some selected papers are listed as [6][10]. Takagi et al. [6] use current
and voltage phasors from one terminal for their method based on reactive power. Girgis et al. [7],
Abe et al. [8], Jiang et al. [9] and Gopalakrishnan et al. [10] use voltage and current phasors from
both ends.
In this paper a scheme is propose for 220KV transmission line for fast and reliable fault
detection using energy of the detail coefficients of the phase signals, classification and location using
neural network. For fault classification current signals (Ia2, Ib2, Ic2, and IG) detail coefficients
energy values are given as input to the neural network. For each type of fault location separate neural
network with different combination of input signals are prepared. In each of these cases, the current,
voltage and ground phase current signals detail coefficients energies values of only phase involving
in the fault signals are given as input to the neural network. The MATLAB 7.10 version is used to
generate the fault signals and verify the correctness of the algorithm. The proposed scheme is
insensitive to variation of different parameters such as fault type, fault resistance etc.
II. DISCRETE WAVELET TRANSFORM
Discrete Wavelet Transform is found to be useful in analyzing transient phenomenon such as
that associated with faults on the transmission lines. The fault signals are generally non stationary
signals, any change may spread all over the frequency axis. The wavelet transform technique is well
suited to wide band signals that may not be periodic and may contain both sinusoidal and non
sinusoidal components. Multi-Resolution Analysis (MRA) is one of the tools of Discrete Wavelet
Transform (D.W.T), which decomposes original, typically non-stationary signal into low frequency
signals called approximations and high frequency signals called details, with different levels or
scales of resolution. The use of detail coefficients for fault detection is discussed in this paper.
Detail coefficients contain information about the fault, which is required for fault detection.
Fig.1: Wavelet filter Bank
In the first decomposition, signal is decomposed into D1 and A1, the frequency band of D1
and A1 is /4 /2, and 0 /4 respectively where the sampling frequency is
. The signal
of desired frequency component can be obtained from repetitive decompositions as shown by Fig.1.
The mother wavelet determines the filters used to analyze signals. In this paper Db4 (Daubechies 4)
wavelet was chosen because of its success in detecting faults [4], [5].
International Journal of Electrical Engineering and Technology (IJEET), ISSN 0976 6545(Print),
ISSN 0976 6553(Online) Volume 5, Issue 7, July (2014), pp. 32-44 IAEME
34
III. ARTIFICIAL NEURAL NETWORKS
Artificial Neural Networks simulate the natural systems behavior by means of the intercon-
nection of basic processing units called neurons. ANNs have a high degree of robustness and ability
to learn [8]. Once the network is trained, it is able to properly resolve the different situations that are
different from those presented in the learning process. The multilayered feed forward network has
the ability of handling complex and nonlinear input-output relationship with hidden layers. In this
method, errors are propagated backwards; the idea of back- propagation algorithm is to reduce errors
until the ANN learns the training data [13] [14]. The multilayered feed forward network has been
chosen to process the prepared input data obtained from the W.T.
IV. TRANSMISSION LINE MODEL
In Fig.2, model of 220kv, 90 km transmission line from P to Q is chosen. Generator of
500MW is connected at one end and 4 loads are connected at 13.8kv and 220kv.
Fig.2: Transmission Line Single Line Model
TABLE I: MODEL PARAMETERS
Various faults are simulated on that line by varying various parameters. Ratings of power
system model are shown in Table I. As shown in Fig.2 a transmission line model is prepared in
MATLAB7.10. The transmission line positive and zero sequence parameters are R1=0.10809/km,
R0=0.2188/km, L1=0.00092H/km, L0=0.0032H/km, C1=1.25*10
f/km, C0=7.85*10
f/km.
The distributed parameter model of transmission line is considered for analysis. The current signals
are sampled at sampling frequency of 20 kHz.
1. Generator 500MVA, 13.8kv, 50Hz, synchronous
generator pu model
2. Transformer1 13.8kv/220kv, 500MVA.
3. Transfomer2 220kv/13.8kv, 500MVA.
4. Load1 50MW, 220kv, 50MW, 1Mvar, RL load.
5. Load2 50MW, 220kv, 50MW, 1MVar, RL load
6. Load3 13.8kv, 40MW, RL load
7. Load4 13.8kv, 40MW, RL load
8 Transmission line Length=90 km.
Load
Load
Load3
Load4
G 90 km
P
Q
International Journal of Electrical Engineering and Technology (IJEET), ISSN 0976
ISSN 0976 6553(Online) Volume 5, Issue 7, July (2014), pp.
V. DESIGN OF FAULT DETECTION
The design process of proposed fault detection, classification and location approach is
shown in Fig.3 Combination of different fault conditions are to be considered and training patterns are
required to be generated by simulating different k
resistance, fault location, and fault type are changed to generate different training patterns.
Fig.3: Process of fault detection
VI. FAULT DETECTION
The signals taken from the scope are
DWT is applied up to level 5, and detail coefficients
detail coefficients energy is calculated.
amount of energy than the level 4 [11],
taken and decomposition is done and
data window. As the fault signals contain the high
signal increases at the occurrence of fault as shown in F
Here, for detecting the fault,
considered. The energy of detail coefficients for a
Where, k=window number, l=level of the DWT, N=length of Detail coefficients at level l.
accurately detecting the
Fig. 4: Energy of the detail level 5 vs. window number
Data acquisition of current signals
D.W.T multiresolution analysis,
calculation, fault detection based on energy
ANN based classification and Location of
International Journal of Electrical Engineering and Technology (IJEET), ISSN 0976
6553(Online) Volume 5, Issue 7, July (2014), pp. 32-44 IAEME
35
DESIGN OF FAULT DETECTION, CLASSIFICATION AND LOCATION
The design process of proposed fault detection, classification and location approach is
Combination of different fault conditions are to be considered and training patterns are
required to be generated by simulating different kinds of faults on the power system. The fault
resistance, fault location, and fault type are changed to generate different training patterns.
.
Process of fault detection, classification and Location
The signals taken from the scope are filtered, sampled at 20 kHz sampling frequency. Then
detail coefficients and approximate coefficients are calculated and
calculated. Then, we come to know that detail level 5 contains highest
[11], [12]. A moving data window of one cycle (
decomposition is done and energy of the details coefficients at level 5 is
the fault signals contain the high amount of harmonic components, the
ccurrence of fault as shown in Fig.4
, difference of energies between two adjacent windows
. The energy of detail coefficients for a window is given by equation (1),
(1)
Where, k=window number, l=level of the DWT, N=length of Detail coefficients at level l.
Energy of the detail level 5 vs. window number
Data acquisition of current signals
D.W.T multiresolution analysis, Energy
calculation, fault detection based on energy
Feature extraction
ANN based classification and Location of
faults
International Journal of Electrical Engineering and Technology (IJEET), ISSN 0976 6545(Print),
The design process of proposed fault detection, classification and location approach is as
Combination of different fault conditions are to be considered and training patterns are
inds of faults on the power system. The fault
resistance, fault location, and fault type are changed to generate different training patterns.
sampling frequency. Then
and approximate coefficients are calculated and
Then, we come to know that detail level 5 contains highest
cycle (400 samples) is
obtained for each
amount of harmonic components, the energy of the
difference of energies between two adjacent windows has been
(1),
(1)
Where, k=window number, l=level of the DWT, N=length of Detail coefficients at level l. For
International Journal of Electrical Engineering and Technology (IJEET), ISSN 0976
ISSN 0976 6553(Online) Volume 5, Issue 7, July (2014), pp.
Fig. 5: F.D index for single line to ground fault vs. window number
presence of faults, the difference between the two consecutive energies of the moving windows is
calculated by (2) and shown in Fig.5.
F.D (k) =
In this sampling frequency of 20 kHz
window slides taking only 1 new sample
cycle corresponds to nearly 400 data samples
The fault is present on R-phase and ground
phase, green colour shows the ground
shows the B phase.The Fault Detection value
data windows, and then decision is made whether f
Fault Detection values the faults can be
threshold values are set and the fault detection is achieved. The transient energy is present mainly
during fault inception and clearing. The high frequency content energy is smaller than the low
frequency content energy of the current signals.
VII. ANN BASED FAULT CLASSIFICATION
All different faults are simulated for different conditions and
from the energy values of the detail coefficients. The 4
selected. The two hidden layers are
network is selected. The average value
of fault are given as input to the neural network, along with the three lines energies, zero sequence
current energy is also given as fourth input to t
the three phases, if fault is present it is shown by the presence of 1
Similarly fourth output indicates the
by the presence of 1, otherwise it is presented b
different training patterns is done as shown in Table
Red
Green
Black
Blue
International Journal of Electrical Engineering and Technology (IJEET), ISSN 0976
6553(Online) Volume 5, Issue 7, July (2014), pp. 32-44 IAEME
36
F.D index for single line to ground fault vs. window number.
presence of faults, the difference between the two consecutive energies of the moving windows is
.
F.D (k) = F.D (k-1) + [E
d
(k) - E
d
(k - 400)] (2)
In this sampling frequency of 20 kHz gives 400 samples for each cycle of 20ms.
1 new sample at each move and keeping 399 previous
ponds to nearly 400 data samples.
phase and ground (G) for the present case. Red colour shows the R
reen colour shows the ground (G) phase, black colour represents the Y phase and blue colour
Detection value is compared with threshold value for consecutive
windows, and then decision is made whether fault is permanent or temporary
the faults can be accurately detected [7]. For different phases diffe
ld values are set and the fault detection is achieved. The transient energy is present mainly
during fault inception and clearing. The high frequency content energy is smaller than the low
frequency content energy of the current signals.
CLASSIFICATION
All different faults are simulated for different conditions and training patterns are generate
detail coefficients. The 4 input neurons and 4 output neurons are
s are selected. Feed forward multilayer back propagation neural
network is selected. The average values of energies of current signals, half cycle after the occurrence
of fault are given as input to the neural network, along with the three lines energies, zero sequence
current energy is also given as fourth input to the neural network. Three outputs show
it is shown by the presence of 1, otherwise with presence of 0
Similarly fourth output indicates the ground fault. If ground is involved in the fault will be indicated
by the presence of 1, otherwise it is presented by 0. This is shown in Table
ns is done as shown in Table III.
Red
Green
Black
Blue
International Journal of Electrical Engineering and Technology (IJEET), ISSN 0976 6545(Print),
.
presence of faults, the difference between the two consecutive energies of the moving windows is
400 samples for each cycle of 20ms. Here, moving
at each move and keeping 399 previous samples. So one
Red colour shows the R
black colour represents the Y phase and blue colour
is compared with threshold value for consecutive 10
ault is permanent or temporary. By using these
For different phases different
ld values are set and the fault detection is achieved. The transient energy is present mainly
during fault inception and clearing. The high frequency content energy is smaller than the low
training patterns are generated
output neurons are
forward multilayer back propagation neural
half cycle after the occurrence
of fault are given as input to the neural network, along with the three lines energies, zero sequence
show the statuses of
, otherwise with presence of 0.
round is involved in the fault will be indicated
is shown in Table II. Generation of
International Journal of Electrical Engineering and Technology (IJEET), ISSN 0976 6545(Print),
ISSN 0976 6553(Online) Volume 5, Issue 7, July (2014), pp. 32-44 IAEME
37
TABLE II
TARGET OUTPUTS
Fault Type A B C G
AG 1 0 0 1
BG 0 1 0 1
CG 0 0 1 1
AB 1 1 0 0
BC 0 1 1 0
CA 1 0 1 0
ABG 1 1 0 1
BCG 0 1 1 1
CAG 1 0 1 1
ABC 1 1 1 0
TABLE III
TRAINING PATTERNS
For training neural network different fault conditions are simulated, features are extracted and
network is trained. At 7 different locations on the transmission line fault is created, at 20, 30, 40, 50,
60, 70, 80% of the transmission line length from the sending end, 4 different values of fault resis-
tances can be used and total 10 different faults are created, and this gives 7*4*10=280 cases for
training neural network.
The different training algorithms are presented to train the neural network; they use the
gradient of the performance function to determine how to adjust the weights to minimize a
performance function. The gradient is determined using back propagation technique, which involves
performing computations backwards through the network. A variation of back propagation algorithm
called Levenberg-Marquardt (LM) algorithm was used for neural network training, since it is one of
the fastest methods for training moderate-sized feed forward neural networks.
LM algorithm to weight update is given by (3),
X
J I
e (3)
Where J is Jacobean matrix that contains first derivatives of the network error with respect to
the weights and biases, e is a vector of network errors. J
e is the gradient and is the scalar affecting performance function. LM algorithm based
method for training neural network is much faster than the other methods. Fig.6 shows the
Multilayered Feed forward Neural Network (M.F.N.N.)
Type of fault LG, LLG, LL, LLL.
Location of fault (%)
from busbar P.
20,30,40,50,60,70,80
Fault resistance 5,10,15,20 .
International Journal of Electrical Engineering and Technology (IJEET), ISSN 0976
ISSN 0976 6553(Online) Volume 5, Issue 7, July (2014), pp.
1
4
Fig.6: Multilayer feed forward network for fault classification
Fig.7: 4-22-22
Network with 2 hidden layers worked out to be better than the 1
4-22-22-4 configuration give better results than the 4
functions used for the hidden layers
respectively. The Fig.7 shows the neural network.
The data used for training data division is done randomly; training function used is LM
algorithm. Performance function used is Mean least square error
chosen is . Fig.8 shows the performance curve. F
we cannot distinguish between the faults with ground
VIII. TEST RESULTS
A validation data set consisti
line model shown in Fig.2. The validation test patterns were different than they were used for the
training of the neural network .For different faults on the model system
fault resistance values are changed to
proposed algorithm. Test results are as shown in
network for varying fault location values and
The output layer activation function used is Purelin, because of its success in the
classification of faults correctly. The tansig and logsig transfer functions did not show a good
classification capability. The output layer transfer function is fixed at
transfer function was changed.
If the transfer functions of the hidden layers I and II are chosen as 1) Tansig
Logsig. 3) Tansig-Logsig, the Table V test result shows that the accuracy obtained with the Ta
Ea
.
.
.
En
International Journal of Electrical Engineering and Technology (IJEET), ISSN 0976
6553(Online) Volume 5, Issue 7, July (2014), pp. 32-44 IAEME
38
1
2
3
22
1
22
1
4
Multilayer feed forward network for fault classification
22-4 Tansig, Logsig, Purelin configuration
Network with 2 hidden layers worked out to be better than the 1 hidden
4 configuration give better results than the 4-22-4, 4-10-4 configurations. Activation
layers I, II and output layer are tansig, logsig
the neural network.
Fig.8: Performance curve.
The data used for training data division is done randomly; training function used is LM
function used is Mean least square error method. The performance goal
shows the performance curve. For network configurations 4-
between the faults with ground without ground.
data set consisting of different fault types was generated using the
The validation test patterns were different than they were used for the
network .For different faults on the model system, fault type; fault location and
fault resistance values are changed to investigate the effects of these factors on the performance of the
results are as shown in Table IV. These results show the accuracy of neural
varying fault location values and varying fault resistance value.
The output layer activation function used is Purelin, because of its success in the
The tansig and logsig transfer functions did not show a good
classification capability. The output layer transfer function is fixed at Purelin and the hidden layer
If the transfer functions of the hidden layers I and II are chosen as 1) Tansig-Tansig. 2) Logsig
Logsig, the Table V test result shows that the accuracy obtained with the Ta
A
.
.
.
.
G
I II
International Journal of Electrical Engineering and Technology (IJEET), ISSN 0976 6545(Print),
hidden layer network.
4 configurations. Activation
logsig and purelin
The data used for training data division is done randomly; training function used is LM
performance goal
-22-4 and 4-10-4,
generated using the transmission
The validation test patterns were different than they were used for the
fault location and
investigate the effects of these factors on the performance of the
results show the accuracy of neural
The output layer activation function used is Purelin, because of its success in the
The tansig and logsig transfer functions did not show a good
Purelin and the hidden layer
Tansig. 2) Logsig-
Logsig, the Table V test result shows that the accuracy obtained with the Tansig-
International Journal of Electrical Engineering and Technology (IJEET), ISSN 0976 6545(Print),
ISSN 0976 6553(Online) Volume 5, Issue 7, July (2014), pp. 32-44 IAEME
39
Logsig type of neural network is more and it is having good generalization capability. The
classification results for almost all types of faults are satisfactory.
TABLE IV
TESTING RESULTS
TABLE V
COMPARISON OF TRANSFER FUNCTIONS
Transfer Functions for
hidden layers.
No. neurons in hidden
layers
Tansig-tansig.
22-22
Logsig-logsig.
22-22
Tansig-logsig.
22-22.
Performance error of test
results
2.9*10^(-7) 5.5*10^(-7) 5.39*10^ (-8).
IX. ANN BASED FAULT DISTANCE LOCATOR
In this paper single line to Ground fault locator explains in detail.
SINGLE LINE TO GROUND FAULTS LOCATOR
A. Selecting the right architecture
One factor in determining the right size and structure for the network is the number of inputs
and outputs that it must have. However, sufficient input data to characterize the problem must be
ensured. The network inputs chosen here are the magnitudes of the detail coefficients energies
fundamental components (50 Hz) of phase voltages and currents measured (AG-Ia2, Va2, IG,
BG-Ib2, Vb2, IG,) at the relay location. As the basic task of fault location is to determine the
distance to the fault, the distance to the fault, in km with regard to the total length of the line, should
be the only output provided by the fault location network. Thus the input and the output for the fault
location network are:
Input = different combinations of Va2, Vb2, Vc2, Ia2, Ib2, Ic2
and IG as per faults. (1)
Output Lf = Fault distance in KM. (2)
Fault
type
Fault
Location
from P(%)
Fault
Resistance
.
Output of neurons
A B C G
AG 30% 10 1.0001 2*10^-3 4*10^-3 1.00
BCG 50% 15 0 1.00 0.9989 1.00
CAG 50% 10 1 0 1.00 0.998
CG 50% 10 1*10^-3 0.000 1.00 1.00
ABC 30% 10 1.00 1.00 0.999 0.00
ACG 70% 5 0.9996 -3*10^-4 0.997 1.000
AB 70% 5 1.018 1.0847 0.1587 0.052
International Journal of Electrical Engineering and Technology (IJEET), ISSN 0976 6545(Print),
ISSN 0976 6553(Online) Volume 5, Issue 7, July (2014), pp. 32-44 IAEME
40
For each type of fault separate neural network is prepared for finding out the fault location.
The ANN architecture, including the number of inputs to the network and the number of neurons in
hidden layers, is determined empirically by experimenting with various network configurations.
Through a series of trial and error, and modifications of the ANN architecture, the best performance
is achieved by using a four layer neural network with 3 inputs and 1 output as shown in Fig. 9. The
number of neurons for the hidden layer is 10 and 5. The final determination of the neural network
requires the relevant transfer functions in the layers to be established. After analysing the various
possible combinations of transfer functions normally used, such as logsig, tansig and linear
functions, the tansig function was chosen as transfer function for the hidden layer, and pure linear
function purelin in the output layer.
1
3
1
2
3
10
1
5
1
Fig.9: Structure of the chosen ANN with configuration for LG fault
B. Learning rule selection
The back-propagation learning rule can be used to adjust the weights and biases of networks to
minimize the sum-squared error of the network. The simple back-propagation method is slow
because it requires small learning rates for stable learning, improvement techniques such as momen-
tum and adaptive learning rate or an alternative method to gradient descent, LevenbergMarquardt
optimisation, can be used. Various techniques were applied to the different network architectures,
and it was concluded that the most suitable training method for the architecture selected was based
on the LevenbergMarquardt (Trainlm) optimization technique.
C. Training process
To train the network, a suitable number of representative examples of the relevant phenome-
non must be selected so that the network can learn the fundamental characteristics of the problem
and, once training is completed, provide correct outputs in new situations not used during training.
To obtain enough examples to train the network, a software package MATLAB 7.10 is used. Using
SIMULINK & SIMPOWER SYSTEM toolbox of MATLAB all the ten types of fault at different
fault locations between 0-100% of line length and different fault resistance have been simulated as
shown below in Table VI. Feed forward back-propagation network have been surveyed for the
purpose of single line-ground fault location, mainly because of the availability of the sufficient
relevant data for training. In order to train the neural network, several single phase faults have been
simulated on transmission line model. For each of the three phases, faults have been simulated at
every 10 km on a 90 km transmission line. Total of 648 cases have been simulated with different
fault resistances 1, 2, 3 ohms respectively. In each of these cases, the current and voltage signals
detail coefficients energies of only phase involving in the fault and ground phase current signals
given as input to the neural network such as Ia2, Ib2, Ic2,Va2 ,Vb2, Vc2 and IG. The output of the
neural network is the distance to the fault from the sending end of the transmission line.
The ANN based fault distance locator was trained using LevenbergMarquardt training
algorithm using neural network toolbox of Matlab as shown in Fig. 10.
Ia2
Va2
IG
.
.
.
.
.
Lf
II
I
.
.
International Journal of Electrical Engineering and Technology (IJEET), ISSN 0976 6545(Print),
ISSN 0976 6553(Online) Volume 5, Issue 7, July (2014), pp. 32-44 IAEME
41
TABLE VI: TRAINING PATTERNS GENERATION
Sr.
No.
Parameter Set value
1 Fault type LG: AG-Ia2, Va2, IG BG- Ib2, Vb2, IG
CG -Ic2, Vc2, IG
LL: AB- Ia2, Va2, Ib2, Vb2,
LLG: ABG -Ia2, Va2, Ib2, Vb2, IG
LLL: ABC- Ia2, Va2, Ib2, Vb2, Ic2, Vc2
LLLG:ABCG- Ia2, Va2, Ib2, Vb2, Ic2, Vc2, IG
2 Fault location
(Lf in KM)
10, 20, 30, 80 and 90 km
3
Fault resistance
(Rf)
1, 2, 3 ohm
Once the network is trained sufficiently and suitably with large training data sets, the network
gives the correct output after one cycle from the inception of fault.
Fig. 10: Overview of the chosen ANN (3-10-5-1)
Fig. 11 plots the mean square error as a function of time during the learning process and it
can be seen that the achieved MSE is about 2.61.
Fig.11: MSE performance of the Network with configuration (3-10-5-1)
International Journal of Electrical Engineering and Technology (IJEET), ISSN 0976 6545(Print),
ISSN 0976 6553(Online) Volume 5, Issue 7, July (2014), pp. 32-44 IAEME
42
TEST RESULT OF ANN BASED FAULT DISTANCE LOCATOR
Once training was completed, ANN based Fault distance locator was then extensively tested
using independent data sets consisting of fault scenarios never used previously in training. For
different faults of the validation/test data set, fault type, fault location, and fault resistance were
changed to investigate the effects of these factors on the performance of the proposed algorithm. The
network was tested and performance was validated by presenting all the ten types of fault cases with
varying fault locations (Lf=0-90KM), fault resistances (Rf=1, 2, 3 etc).
TABLE VII Percentage errors as a function of fault distance and fault resistance for the ANN
chosen for single phase fault location.
TABLE VII
Fault Distance
(Km)
Measured
Fault Location (Km)
Percentage Error
(%)
RF=1 RF=4 RF=1 RF=4 RF=1 RF=4
9 9 8 7.3 1.11 1.8
18 18 15 15.5 3 2.7
54 54 52 51 2.22 3
63 63 60 57 3 6
72 72 70.5 70 1.6 2
Table VII shows some of the test results of ANN based fault locator under different fault
conditions. It can be seen that all results are correct with reasonable accuracy. At various locations
different types of LG faults were tested to find out the maximum deviation of the estimated distance
Lf measured from the relay location, from the actual fault location La. Then the resulted estimated
error e is expressed as a percentage of total line length L In all the fault cases, the results have
shown that the errors in locating the fault are less than 1.11% to +6%.
Table VII can show the percentage errors in fault location as a function of Fault Distance and
Fault resistance. Different cases are shown with different fault resistances. Thus, the neural network
performance is considered satisfactory and can be used for the purpose of single line- ground fault
location.
X. CONCLUSION
In this paper accurate fault detection, classification and location technique is discussed. This
technique depends upon the current and voltage signals. The features are extracted from the current
and voltage signals by using wavelet transform. The feature vector is then given as input to the neural
network. The capabilities of neural network in pattern classification were utilized. Simulation studies
were performed and the performance of the scheme with different system parameters and conditions
was investigated. The test result shows that the accuracy obtained for fault classification with the
tansig-logsig transfer function for hidden layers I and II is satisfactory. For fault location after
analysing the various possible combinations of transfer functions normally used, such as logsig, tansig
and linear functions, the tansig function was chosen as transfer function for the hidden layer I and II,
and pure linear function Purelin in the output layer gives satisfactory results.
International Journal of Electrical Engineering and Technology (IJEET), ISSN 0976 6545(Print),
ISSN 0976 6553(Online) Volume 5, Issue 7, July (2014), pp. 32-44 IAEME
43
REFERENCES
[1] K. Gayathri and N. Kumarapan, Comparative Study of Fault Identification and Classifica-
tion on EHV Lines Using Discrete Wavelet Transform and Fourier Transform Based ANN,
World Academy of Science, Engineering and Technology, pp.822-831.2008.
[2] D. V. Coury, D. C. Jorge, Artificial Neural Network Approach To Distance Protection of
Transmission Lines. IEEE Transactions on Power Delivery, Vol. 13, No. 1, January 1998.
[3] H. Khorashadi-Zadeh, M. R. Aghaebrahimi, A Novel Approach to Fault Classification and
Fault Location for Medium Voltage Cables Based on Artificial Neural Network, World
Academy of Science, Engineering and Technology, pp.1100-1103.2008.
[4] V. S. Kale, S. R. Bhide, P. P. Bedekar, Faulted Phase Selection on Double circuit
Transmission Line using Wavelet Transform and Neural Network, Third International
Conference On Power Systems, Kharagpur, INDIA, December 27-29.
[5] A. Abdollahi, S. Seyedtabaii, Comparison of Fourier & Wavelet Transform Methods for
Transmission Line Fault Classification, The 4th International Power Engineering and
Optimization Conf. (PEOC2010), Shah Alam, Selangor, MALAYSIA: 23-24 June 2010.
[6] K. Takagi, Y. Yomakoshi, M. Yamaura, R. Kondow, and T. Matsushima, Development of a
new type fault locator using the one terminal voltage and current data, IEEE Trans. Power
App. Syst., vol. PAS-101, pp. 28922898, Aug. 1982.
[7] A. Girgis, D. Hart, andW. Peterson, A new fault location technique for two- and three-
terminal lines, IEEE Trans. Power Delivery, vol. 7, pp. 98107, Jan. 1992.
[8] M. Abe, T. Emura, N. Otsuzuki, and M. Takeuchi, Development of a new fault location
system for multi-terminal single transmission lines, IEEE Trans. Power Delivery, vol. 10,
pp. 159168, Jan. 1995.
[9] J.-A. Jiang, J.-Z. Yang, Y.-H. Lin, C.-W. Liu, and J.-C. Ma, An adaptive PMU based fault
detection/location technique for transmission lines part I: Theory and algorithms, IEEE
Trans. Power Delivery, vol. 15, pp. 486493, Apr. 2000.
[10] A. Gopalakrishnan, D. Hamai, M. Kezunovic, and S. McKenna, Fault location using the
distributed parameter transmission line model, IEEE Trans. Power Delivery, vol. 15,
pp. 11691174, Oct. 2000.
[11] V. S. Kale, S. R. Bhide, P. P. Bedekar, G. V. K. Mohan, Detection and Classification of
Faults on Parallel Transmission Lines using Wavelet Transform and Neural Network,
International Journal of Electrical and Computer Engineering, pp.1063-1067, 16, 2008.
[12] N. Zamanan, M. Gilany, A Sensitive Wavelet-Based Algorithm for Fault Detection in Power
Distribution Networks ACEEE Int. J. on Communication, Vol. 02, No. 01, Mar 2011,
pp.46-50.
[13] S. Haykin, Neural Networks, IEEE Press, New York, 1994.
[14] Mamta Patel, R. N. Patel, Fault Detection and Classification on a Transmission Line using
Wavelet Multi Resolution Analysis and Neural.
[15] Suresh J. Thanekar, Waman Z. Gandhare and Anil P. Vaidya, Voltage Stability Assessment
of a Transmission System -A Review, International Journal of Electrical Engineering &
Technology (IJEET), Volume 3, Issue 2, 2012, pp. 182 - 191, ISSN Print: 0976-6545,
ISSN Online: 0976-6553.
[16] Soumyadip Jana, Sudipta Nath and Aritra Dasgupta, Transmission Line Fault Classification
Based on Wavelet Entropy and Neural Network, International Journal of Electrical
Engineering & Technology (IJEET), Volume 3, Issue 2, 2012, pp. 94 - 102, ISSN Print:
0976-6545, ISSN Online: 0976-6553.
International Journal of Electrical Engineering and Technology (IJEET), ISSN 0976 6545(Print),
ISSN 0976 6553(Online) Volume 5, Issue 7, July (2014), pp. 32-44 IAEME
44
AUTHORS DETAIL
R. P. Hasabe received the B.E. degree in electrical engineering and the M.E.
degree in electrical power systems from Shivaji University, Kolhapur, India in
2001 and 2006, respectively, and is currently pursuing the Ph.D. degree in
Electrical engineering at Shivaji University Kolhapur.
Currently, he is an Assistant Professor in the Department of Electrical
Engineering, Walchand College of Engineering, Sangli. His research interests
include power system protection, planning and design, system modeling, and
simulation.
A. P. Vaidya received the B.E. in electrical engineering and the M.E. in electrical
power systems from Shivaji University, Kolhapur, India, in 1983 and 1993
respectively, and the Ph.D. degree from the IISc, Bangalore in 2005.
Currently, he is Professor in the Department of Electrical Engineering,
Walchand College of Engineering, Sangli. He has published more than 10 papers
in journals and conferences at international and national levels. His research
interests include power system protection, automation, planning and design,
system modeling and simulation, and artificial intelligence.