Download Fault Detection in double circuit transmission lines using ANN

Fault Detection in double circuit transmission
lines using ANN
Chhavi Gupta1,[email protected],+919536706788
Abstract- This paper discuss the fault detection in high voltage transmission line using artificial
neural network technique. The problem of faults in a transmission line has been going for a very
long time. It has been one of the major concerns of the power industry. Normally, protective
relays, recording devices and special control and protection software systems are responsible for
detecting the fault occurrences and isolating the faulted portion from the system. Thus it is
necessary for the faults to be detected quickly and precisely. It is also equally important to know
the details about the fault that has occurred so that it can be corrected soon. This paper focuses
on detecting faults on High Voltage electric power transmission lines. Fault detection have been
achieved by using artificial neural networks. Backpropagation algorithm has been employed
using feedforward networks for each of the three phases in the Fault detection process. Analysis
on neural networks with varying number of hidden layers and neurons per hidden layer has been
provided to validate the choice of the neural networks in each step. Taking this into account,
models for conventional Transmission line to generate voltage and current parameter is
developed in MATLAB/SIMULINK. A selected simplified power network is used to simulate all
possible fault scenarios and to generate test cases. From generated data test sets, ANN is trained
and tested for fault detection and fault classification. Simulation results have been provided to
demonstrate that artificial neural network based methods are efficient for detecting on High
Voltage transmission lines and achieve satisfactory performances.
Key words: Double circuit transmission line, ANN, Fault detection.
I.INTRODUCTION
The power transmission over a utility company network is a complicated process involving a
number of generating plants, transmission lines, distribution and substations. Electric power is
normally generated at 11-25kV in a power station. It is necessary to transmit over long distances, it is
then stepped-up to HV / EHV / UHV levels. Power is generally transmit through a transmission
network have high voltage lines. [5]
High Voltage transmission lines utilized in modern power systems to increase reliability
power transfer, and security for the transmission of electrical energy. The different configurations of
double circuit transmission lines with the effect of mutual coupling make their protection a problem.
The vast majority of faults (over 90%) [1] are of the single line to ground type.
The appropriate percentages of occurrences various faults are listed below: Single line to ground
fault – 70-80%, Line-Line to ground fault - 10-17%, Line-Line fault – 8-10%,three phase – 23%.The causes of temporary faults include lightning, conductors slapping together in the wind, tree
branches that fall across conductors and then fall or burn off and insulator flashovers caused by
pollution. Faults that are normally temporary can turn into permanent faults.
Validation data and exciting results were noticed by the proposed Neural Network.
The fault detection tested with problem being a major part of transmission and distribution
system automation is selected as the topic of in this paper. In this, a method for detection of
fault in high voltage transmission systems is devised. The various practical constraints,
which are the unique characteristics of the transmission system are considered in the fault
detection and classification problem. The work broadly covers:
“Development of a new approach for fault detection in transmission systems using
Feed Forward Neural Networks and Neural Network as a detector, using the
measurements available at the substation”.
The objectives of this paper is to develop the proposed 5-bus system model of
transmission system for generating data sets for ANN technique under steady state and different
faulty states like line to ground fault, line to line fault, double line to ground fault, and three
phase fault etc and to develop an Artificial Neural Network (ANN) model from data sets
obtained by the transmission model. The algorithm used in this paper employs the fundamental
components of voltage and current signals. A model of 250KV transmission line is developed
and simulated in MATLAB/SIMULINK for generating training data sets for ANN. The faults
are more easily detecting using ANN in a simple and efficient way.
II. SYSTEM CONFIGURATION
To gauge the performance of the proposed neural network-based fault detector, a 250 kV, 200
km transmission line extending between two sources has been considered in this study as shown
in figure 1.
Fig. 1 The system under consideration
The transmission line is represented by distributed parameters and the sampling frequency of the
line parameters is taken into account [3].The values of the three-phase voltages and currents are
measured at 5 bus system and the resulting data set is ultimately fed into the neural network as
inputs. A simulink model of the transmission line model shown in figure 2 which has been used
to generate the entire set of training and testing data for Neural Network. There have been 10
different fault cases simulated for the purpose of fault detection.
Fig. 2 System model simulated in MATLAB/Simulink
The first step in the process is fault detection.
(a). Network Simulation
The network is trained using “znntrain”. This function calls ANN functions included in the
MATLAB NN toolbox. The functions called are “newff” and “nftool”. The “newff” function
creates feedforward network and returns a network object. The “train” function trains the
network created.The syntax of “newff” function can be read from MATLAB help. The one used
in this work only uses some of its parameters namely “newff” (input, [N1, N2, … Nj], {F1, F2,
… Fj}, training algorithm). The above function creates a feedforward network with j layers
(input layer not involved). The Tolal numbers of nodes withevery layer are N1, N2, and Nj with
each layer having activation functions F1, F2, and Fj respectively.
Feedforward Networks
Where there is no feedback connection involved in the network is called feed forward
network and hence the information travel is unidirectional .A feedforward network with N0 input
and KR output signals is shown in Fig 3. The computation process in the ith layer can be
described by the following equation (3)
qi = fi( Wi gi-1)
Where qi = [qi1, qi2, qi3…….qni]T is the signal vector at the output of the ith layer.
Figure 3. Structure of a two-layered feedforward network
Wi =
Wi10
Wi11
…
Wi21
Wi22
…
Winio
Wini1
Wi1i-1
Wi2i-1
Wini2
is the weighing matrix between the ( i-1)th the ith layer.
A is the vector containing the input signals, f (i) is the activation function of the neurons
in the ith layer and R is the number of processing layers. All the neurons in a particular layer are
assumed to be similar in all aspects and the number of hidden layers can be more than one and is
usually determined by the purpose of the neural network. The output of the processed neural
network is represented by the output vector Y.[4]
Y= [ y1 y2……….. ynr
]
Proposed fault detector based on neural networks
Functional blocks of the proposed fault detector are shown in Figure .4 Voltage signals at
the transmission line end S (relay location) will be acquired by the relay at buses. After
preprocessing, It fed to the fault detector (FD) to detect a fault, and if the fault is detected by FD,
the fault classifier (FC) estimates the which type of the fault occur in the transmission line. The
proposed fault detector (FD) is designed to indicate the presence or absence of a fault. The
occurrence of the fault is classifying.
Figure 4: Major blocks constituting the fault detector [7]
Design Process of Ann Fault Detector
The design process of the ANN fault detector goes through the following steps:
1. Preparation of a suitable training data set that represents cases that ANN needs to learn to
process further.
2. Selection of a suitable ANN structure for a given application for better output.
3. Training the ANN with suitable data.
4. Evaluation of the trained ANN using test patterns until its performance is satisfactory or error
reduced.[5]
Fault Detector
In order to build up an ANN, the inputs and outputs of the neural network have to be
defined for recognition of pattern. The inputs of the network must provide a true representation
of the situation under consideration. The generation process of input patterns to the ANN fault
detector (FD) [6].
Inputs and Outputs
In order to build up an ANN, the inputs and outputs of the neural network have to be
defined for pattern recognition. The inputs to the network should provide a true representation of
the situation under consideration. The process of generating input patterns to the ANN fault
detector (FD).
The input (Voltage) patterns are generated as a string of samples corresponding to a
sampling frequency. These signals as to simulate a continue sampling process (62 samples per 50
Hz cycle). The ANN output is indexed with either a value of ≠ 0 in the case of fault present or 0
in the case of fault absence.
The input layer gives the three phase current and voltage. The outputs of the hidden layer with
the sigmoid activation function are calculated and transferred to the output layer, which is
composed of only one neuron. The output value of the neuron in the output layer with the
sigmoid activation function calculated to gives the state of the transmission line: ≠0 (the presence
of a fault) or 0 (the non-faulty situation) [8].
III. Fault Detection Results and discussion
Fault detection perform an important role in minimizing the damage to system due to
high short circuit fault or short circuit current and fast detection of fault conduces to quick
isolation of faulty transmission line from the system, hence protect it from the harmful effects of
voltage and current.For the purpose detection, various methods of Multi-Layer Perceptron have
been studied. The various factors that play a role in deciding the ideal methods are the network
size, the learning strategy employed and the training data set size. The back-propagation
algorithm has been decided as the ideal methods. The basic back-propagation algorithm is very
slow due to the small learning rates employed, few techniques can significantly enhance the
performance of the algorithm. The Levenberg-Marquardt optimization technique is a one such
strategy used. The selection of the network size is very vital because this not only reduces the
training time but also greatly enhance the ability of the neural network to represent the problem
in hand. Unfortunately there is no thumb rule that can dictate the number of hidden layers and
the number of neurons per hidden layer in a given problem.
Training the Fault Detection Neural Network
Best Validation Performance is 0.025747 at epoch 26
0
Mean Squared Error (mse)
10
Train
Validation
Test
Best
-1
10
-2
10
0
5
10
15
20
25
30
32 Epochs
Figure 5 Mean-square error performance of the network (6-10-1)
Best Validation Performance is 0.034774 at epoch 100
1
Mean Squared Error (mse)
10
Train
Validation
Test
Best
0
10
-1
10
-2
10
0
10
20
30
40
50
60
70
80
90
100
106 Epochs
Figure 6.Mean-square error performance of the network (6-25-1)
Best Validation Performance is 0.034774 at epoch 100
1
Mean Squared Error (mse)
10
Train
Validation
Test
Best
0
10
-1
10
-2
10
0
10
20
30
40
50
60
70
80
90
100
106 Epochs
Figure 7 Mean-square error performance of the network (6-35-1)
Best Validation Performance is 0.028682 at epoch 139
1
Mean Squared Error (mse)
10
Train
Validation
Test
Best
0
10
-1
10
-2
10
0
20
40
60
80
100
120
140
145 Epochs
Figure 8 Mean-square error performance of the network (6-40-1)
Best Validation Performance is 0.025449 at epoch 119
1
Mean Squared Error (mse)
10
Train
Validation
Test
Best
0
10
-1
10
-2
10
0
20
40
60
80
100
120
125 Epochs
Figure 9 Mean-square error performance of the network (6-50-1)
Discussion
For the fault detection, the network takes in six inputs at a time, which are the voltages
and currents for all the three phases for ten different faults and also no-fault case. The output of
the neural network is just a yes or a no (1 or 0) depending on whether or not a fault has been
detected.
After extensive simulations it has been decided that the desired network has one hidden
layer with 10 neurons in the hidden layer. For illustration purposes, several neural networks
(with varying number of hidden layers and neurons per hidden layer) that achieved satisfactory
performance are shown and the best neural network has been described further in detail.
Figures 5-6 show the error performance plots of neural networks with hidden layers 10
and 25 respectively. The various error performance plots have been shown in Figures 5 – 9.
Figure 5 shows the training performance plot of the neural network 6-10-1 (6 neurons in
the input layer, 1 hidden layer with 10 neurons in it and 1 neuron in the output layer). It can be
seen that the network did not achieve the desired Mean Square Error (MSE) goal by the end of
the training process.
Fig 6 shows the training performance plot of the neural network with 6-25-1
configuration (6 neurons in the input layer, 1 hidden layers with 25 neurons respectively and one
neuron in the output layer). It is to be noted that the neural network could not achieve the MSE
goal of 0.0001 by the end of the training process.
Figure 7 shows the training process of the neural network with 6-35-1 configuration (6
neurons in the input layer, 1 hidden layers with 35 neurons in them respectively and one neuron
in the output layer). From the above training performance plots, it is to be noted that very
satisfactory training performance has been not achieved by the neural network with the 6-35-1
configuration (6 neurons in the input layer, 1 hidden layers with 35 neurons in them respectively
and one neuron in the output layer).
Figure 8 training performance plots, it is to be noted that very good training performance
has been achieved comparison to last one. The overall MSE of the trained neural network is way
below the value of 0.0001. Hence this has been not chosen as the ideal ANN for the purpose of
fault detection.
By the neural network with the 6-35-1 configuration (6 neurons in the input layer, 1
hidden layers with 35 neurons in them respectively and one neuron in the output layer). It is to be
noted that very satisfactory training performance but did not achieved the satisfactory result by
this neural network.
Figure 9 training performance plots, it is to be noted that very satisfactory training
performance has been achieved comparison to result network (6-50-1).The overall MSE of the
trained neural network is near about the value of 0.0001. Hence this has been chosen as the ideal
ANN for the purpose of fault detection.
Testing the Fault Detection Neural Network
Validation: R=0.95651
Output ~= 0.92*Target + 0.041
Output ~= 0.92*Target + 0.041
Training: R=0.95871
Data
Fit
Y = T
1
0.5
0
-0.5
-1
-1
-0.5
0
0.5
Data
Fit
Y = T
1
0.5
0
-0.5
-1
1
-1
-0.5
Target
Data
Fit
Y = T
0.5
0
-0.5
-1
-1
-0.5
0
0.5
1
All: R=0.95768
Output ~= 0.92*Target + 0.041
Output ~= 0.92*Target + 0.044
Test: R=0.9541
1
0
Target
0.5
1
Data
Fit
Y = T
1
0.5
0
-0.5
-1
-1
-0.5
Target
0
0.5
1
Target
Figure10.Regression fit of the output s vs. targets for the network (6-10-1)
Discussion
Once the neural network has been trained, its performance has been tested by three
different factors. The first of these is by plotting the best linear regression that relates the targets
to the outputs as shown in Figure. The correlation coefficient (r) is a measure of how well the
neural network’s targets can track the variations in the outputs (0 being no correlation at all and 1
being complete correlation). The correlation coefficient in this case has been found to be 0.95768
in this case which indicates excellent correlation.
VII. CONCLUTIONS
The objective of this paper is to present a novel approach for detection faults in High
Voltage Transmission Line systems using ANN technique considering various practical
constraints. Power system Transmission Line model was developed in MATLAB/SIMULINK
platform. The proposed Neural Network can be used for a real time approach in the power
system analysis. From the results obtained, it can be concluded that the measurements made
during fault in a high voltage transmission system contain significant information about the
detection of faults. The measurements obtained at the substation are used to detect fault in
Transmission systems. The studies carried out in this work show that, for different network
configurations, the similar structure of the neural network can be used , and only the
weights of the neural networks are updated for different configurations.
References
[1] Anamika Yadav and A.S Thoke,”Classification of Single Line to Ground Faults on Double Circuit Transmission Line using
ANN” International Journal of Computer & Electrical Engineering 01/2009; DOI: 10.7763/IJCEE.2009.V1.30.
[2]Manohar singh, Dr. B.K Panigrahi Department of Electrical Engineering, Indian Institute of Technology,Delhi, India-110016
manoharsingh33@ gmail.com
[3]Aggarwal, R.K., Xuan Q.Y., Dunn, R.W., Johns, A.T. and Bennett, “A Novel Fault Classification Technique for Double-circuit lines Based on
a Combined Unsupervised/Supervised Neural Network”, IEEE Transactions on Power Delivery, VOL.14, October 1999
[4] Rajveer Singh," Fault Detection of Electric Power Transmission Line by Using Neural Network”,Asstt.Prof. Deptt. of
Electrical EngineeringDecember 2012, Jamia Millia Islamia, New Delhi, India
[5] Bouthiba T. and Maun J.C., “Relais à base de réseaux de neurones pour la protection des lignes de transport à THT”.Revue
Internationale de Génie Electrique, Vol. 6, No. 3–4, 2003 pp. 413–428.
[6] Feng Yan, Zhiye Chen, Zhirui Liang, Yinghui Kong, and Peng Li, “Fault Location Using Wavelet Packets”, Department Of Electric Power
Engineering, North China Electric Power University, Baoding Hebei, China, 2575-2579, 2002
[7]Anamika Jain, Thoke, A. S., and Patel, R. N.2009. Classification of Single Line to Ground Faults on Double Circuit
Transmission Line using ANN, International Journal Of Computer And Electrical Engineering, Vol.1 (No. 2), June 2009, 17938163.
[8] Murari Mohan Saha, Ratan Das,PekkaVerho, and Damir Novosel, “Review of Fault Location Techniques for Distribution Systems”, Power
Systems
And Communications Infrastructures For The Future, Beijing, September 2002.