A DWT Approach for Detection and Classification of Transmission Line Faults

IJIRST International Journal for Innovative Research in Science & Technology Volume 3 Issue 02 July 2016 ISSN (online): 2349-6010 A DWT Approach for Detection and Classification of Transmission Line Faults Prasad P. Kawale PG student Department of Electrical Engineering SND College of Engg. & RC, Yeola, (M.S.) India Prof. C. Veeresh Assistant Professor Department of Electrical Engineering SND College of Engg. & RC, Yeola, (M.S.) India Abstract The rapid growth of electric power systems has resulted in a large increase of the number of lines in operation and their total length. These lines are exposed to faults because of many reasons such as a result of lightning, short circuits, faulty equipments, miss-operation, human errors, overload, and aging etc Due to these faults.long term power outages for customers and may lead to significant losses. Therefore fast detection and classification of transmission line faults is important in maintaining a reliable power system operation and to ensure quality performance of the power system. This paper aims at detecting and classifying the transmission line faults by using Discrete Wavelet Transform (DWT) and Artificial neural network (ANN). Various types of fault conditions such as Single line-to-ground faults (L-G), Line-to-Line faults (L-L) and Double Line-to-ground faults (L-L-G) are simulated in Power System Computer Added Design (PSCAD) software. An extremely large data set of current and voltage signals is generated by simulating various types of fault conditions by varying the system parameters. Then an advanced signal processing tools such as discrete wavelet transform (DWT) is used for calculating detail coefficients energy of the fault signals. Depending upon the detail coefficients energy the fault will be detected. A properly configured Artificial Neural Network (ANN) can be utilized for classification of the faults based on the DWT signal. Keywords: Power System Computer Added Design, Discrete Wavelet Transform, Artificial Neural Network, Transmission line fault detection, fault type classification I. INTRODUCTION Now-a-days for the fault detection, a high frequency components technique is used. This new technique is also known as transient based techniques. In this technique, it is essential that the fault signal has to be analyzed accurately. Wavelet transform has been used extensively for signal processing in recent years. It has been found that the wavelet transform is capable of investigating the transient signals generated in a power system. Wavelet theory is the mathematics, which deals with building a model for non-stationary signals, using a set of components that look like small waves, called wavelets. The wavelet transformation is a tool which helps the signal to analyze in time as well as frequency domain effectively. It uses short windows at high frequencies, long windows at low frequencies. Wavelet transform has the advantage of fast response and increased accuracy as compared to conventional techniques. Using multiresolution 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 currents and voltages. On the other hand, properly configured Artificial Neural Network (ANN) can be utilized for classification of the faults based on the DWT signal. The neural networks have the ability to learn, generalize and parallel processing, have made their applications for many systems ideal. The use of neural network as pattern classifiers is among their most common and powerful applications. II. DISCRETE WAVELET TRANSFORM In DWT a time- scale representation of a discrete signal is obtained using digital filtering technique. The desired signal which to be analyzed is passed through different filters having different cut off frequencies at different scales. In discrete wavelet transform the scale is changed by up sampling and down sampling. Normally half band high pass and low pass filters are used. The DWT is computed by successive lowpass and high pass filtering of the discrete time-domain signal as shown in figure1. This is called the Mallat algorithm or Mallat-tree decomposition. Its significance is in the manner it connects the continuous-time multiresolution to discrete-time filters. In the figure-1, the signal is denoted by the sequence x[n], where n is an integer. The low pass filter is denoted by G0 while the high pass filter is denoted by H0. At each level, the high pass filter produces detail information[n], while the low pass filter associated with scaling function produces approximations, a[n]. All rights reserved by www.ijirst.org 102

Fig. 1: Wavelet Decomposition up to 7 th level containing different frequency bands. For the system taken here, we used 10kHz as sampling frequencies. The typical wavelets used to detect power disturbance are mainly Haar, Daubechies, Coiflets, Bi-orthogonal, Morlet and Symlets. In wavelet analysis, a MOTHER WAVELET is chosen as the prototype for generating other basis window functions. i.e., all the window functions are obtained by translating and scaling the mother wavelet. III. METHODOLOGY In this work of detection and classification of EHV transmission line faults is carried out by using following steps. A 765kV EHV transmission system between UNNAO - ANPARA shown in Fig.3 is simulated using PSCAD software. Different types of faults are created at different locations with different inception angle. Voltage and current signals are captured with 10 KHz sampling frequency. Construction of modal signal. Modal signal is decomposed upto 7 th level using DWT. The energy is computed from the detail coefficients of the modal signal which is used for the detection of the faults. Preparation of data sheet of 7th level energies and importing it to ANN. Training of ANN and validation of the trained ANN using test patterns to check its correctness and generalization. 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. Classification of faults is done by using ANN. All rights reserved by www.ijirst.org 103

Fig. 2: Algorithm for classification of faults. IV. SIMULATED SYSTEM Fig-3 shows the 765kv single line diagram of 765kV transmission system between Unnao and Anpara (U.P.). The length of transmission line is extended upto 430km. Fig. 3: Single line diagram of 765kV Transmission system between Unnao and Anpara Fig.-4 shows the same transmission system is simulated in PSCAD software with detailed specification. All rights reserved by www.ijirst.org 104

Fig. 4: PSCAD simulated network of 765kV transmission line. V. FAULT ANALYSIS In this Paper 765KV network is simulated in PSCAD software with a sampling frequency of 10 KHZ. Different types of fault LG, LL and LLG are created at different locations on the 430 km long transmission line at an interval of 50 km including the different inception angle 0, 90, and 180 degrees. So for simplicity 5 different locations are considered here for the analysis of fault i.e. 50km, 100km, 150km, 200km and 215 km from the sending end. Hence, 5 fault locations, three inception fault angles and three different types of faults (5*3*3=45) constitute 45 cases. For the analysis of faults the three-phase line currents, the three-phase sending end and receiving end voltages are recorded from the simulation. The data generated from the simulation is very large. Therefore to avoid the complexity of handling of such huge data, a modal signal of current and sending end and receiving end voltages is used. The three phase voltages and three phase current signals for different fault conditions for different faults are taken into account. In this paper for the fault classification different conditions are used. Fig. 5 shows the simulation waveforms of line current (I), sending end voltage (V4) and receiving end voltage (V5) of 765kv transmission line for LG fault at a distance of 215km from the source and at 90 0 instant of blue phase and Fig. 6 shows the modal signal of LG fault. Fig. 7 to 9 shows the wavelet decomposition for modal current signal, sending end & receiving end modal voltage signal upto 7th level using db4 as mother wavelet. Similarly fig. 10 to 19 shows for the LL and LLG fault resp. Fig. 5: Simulation waveforms of V & I of LG Fault at 215km with 90 0 inception angle. All rights reserved by www.ijirst.org 105

Fig. 8: Wavelet decomposition for modal voltage signal of sending end (VS) of LG fault. Fig. 9: Wavelet decomposition for modal voltage signal of receiving end (VR) of LG fault. All rights reserved by www.ijirst.org 107

Fig. 18: Wavelet decomposition for modal voltage signal of sending end (VS) of LLG fault. Fig. 19: Wavelet decomposition for modal voltage signal of receiving end (VR) of LLG fault. All rights reserved by www.ijirst.org 112

VI. ANN USED AS FAULT CLASSIFIER Artificial neural networks are composed of simple elements which operate in parallel with interconnection between them. The weights of connection determine the network function. It is considered as the simplest kind offered forward network. An Artificial neural network when created, has to be configured which is done using training function. The elements of the network are adjusted automatically to get a particular target output for specific input. A network can have several layers. Each layer has a weight matrix, a bias vector and an output vector. Each neuron in one layer has direct connections to the neurons of the subsequent layer. The second class of feed forward neural network distinguishes itself by the presence of one or more hidden layers, whose computation nodes are called hidden neurons or hidden units. By increasing the number of layers and neurons the network is enabled to extract higher order. In this work, Principal component analysis (PCA) network of ANN is used as fault classifier. VII. RESULTS AND DISCUSSION After training, the artificial neural network based fault classifier is extensively tested using independent data set consisting of different types of faults, different fault locations with different inception angles. The result of classification for a given system are as shown in table-1. To inquire into the accuracy of the proposed method in these cases, 100% accurate results are found for LG, LL and LLG type of faults. Hence principal component analysis, modal based fault classifier classifies the types of faults with an accuracy of 100% in a very fast and effective manner. The upper part of Table-1 shows the number of readings taken by ANN for classification of respective fault for 4th processing element. From the table it is observed that the sum of the readings taken for three faults is 11 which is 25% of total readings input to ANN i.e. 45 readings and the lower part of table shows the various errors and percentage of fault classification accuracy for 4th processing element. Table 1 Output / Desired LG LL LLG LG 6 0 0 LL 0 2 0 LLG 0 0 3 Performance LG LL LLG MSE 0.038605979 0.006681175 0.152392245 NMSE 0.155710783 0.044912343 0.768310901 MAE 0.169711048 0.055192952 0.322101867 Min Abs Error 0.023260345 0.00295256 0.010735036 Max Abs Error 0.343454236 0.241157773 0.621906662 R 0.980436039 0.97734324 0.663213496 Percent Correct 100 100 100 VIII. CONCLUSION The work presented in this paper provides a new technique for EHV transmission line of 765kV and fault classification is done by using db4 as mother wavelet. Decomposition of modaling end voltage at level 7th differentiate clearly the types of fault by observing the decomposition waveform diagram. The time required for classification of the faults after constructing modal signal is less so number of inputs is reduced and it required less memory space because they user of modal signal complexity for taking adecision for ANN is reduced. The proposed strategy with the use of DWT-ANN based algorithm is promising and suggests that this approach could lead to useful application in an actual power system. REFERENCES [1] D.Thukaram, Senoir Member, IEEE and Dharmesh Yelamanchi Department of Electrical Engineering Indian Institute of Science Bangalore-560012, INDIA Performance Analysis of 765 kv system under Steady State and Transient Conditions with varying Reactive Compensation Fifteenth National Power Systems Conference (NPSC), IIT Bombay, December 2008. [2] K.M.Silva B.A.Souza andn.s.d.brito Fault Detection and Classification in Transmission Lines Based on Wavelet Transform & ANN IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 21, NO. 4, OCTOBER 2006 pp.2058-2063. [3] P. S. Bhowmik, P. Purkait Member and K. Bhattacharya A Novel Wavelet Assisted Neural Network for Transmission Line Fault Analysis Electrical Power and Energy Systems 31 (2009) pp.213 219 [4] M. Jayabharata Reddy and D.K. Mohanta A DSP based frequency domain approach for classification of transmission line faults Science Direct Digital Signal Processing 18 (2008) pp. 751 761 [5] Prince Jose, Bindu V.R, Wavelet-Based Transmission Line Fault Analysis, International Journal of Engineering and Innovative Technology (IJEIT) Volume 3, Issue 8, February 2014. [6] Dash PK and Samantaray SR A novel distance protection scheme using time frequency analysis and pattern recognition approach. Int J Elect Power Energy system 2006 All rights reserved by www.ijirst.org 113

[7] Dalstein and B. Kulicke, Neural network approach to fault classification for high speed protective relaying, IEEE Trans. Power Del., vol.10, pp. 1002-1009, 1995. [8] G. Sudha, T. Basavaraju B.I.T, V.T.U, Bangalore, India A comparison between different approaches for fault classification in transmission line IET- UK International Conference on Information and Communication Technology in Electrical Sciences (ICTES 2007) Dr. M.G.R. University, Chennai, Tamil Nadu, India. Dec. 20-22- 2007. pp.398-403. [9] M. A. Beg, M. K. Khedkar, S. R. Paraskar, G. M. Dhole Classification of fault originated transients in high voltage network using DWT PCA approach International Journal of Engineering, Science and Technology Vol. 3, No. 11, 2011, pp. 1-14 [10] P.P.Kawale and R.K.Jha, A DWT approach for detection of transmission line faults, ARDIJEET, VOLUME 04 ISSUE 02, 01 April 2016. All rights reserved by www.ijirst.org 114