Rajveer Singh Int. Journal of Engineering Research and Applications www.ijera.com ISSN : 2248-9622, Vol. 3, Issue 6, Nov-Dec 2013, pp.1993-1999 www.ijera.com 1993 | Page Artificial Neural Network & Wavelet Transform for Identification and Classification of Faults in Electrical Power System Rajveer Singh 1 1 Asstt. Prof. Deptt. of Electrical Engineering, Jamia Millia Islamia, New Delhi, India. Abstract In a distributed Electrical Power System Faults are the major problem for regular supply to the consumers. A low impedance fault in electrical power distribution system is distinguished by a non-linear and unstable varying fault current due to type of fault. In this combined approach of Wavelet and Artificial Neural Network is used for identification and classification of all types of faults in power distribution system. Wavelet transform identify the types of fault in the form of change in energy in the current waveform and ANN used for classification of faults. IEEE 13-Bus system and 17 bus actual radial distribution system is used to test and verifying the results. The proposed method is implemented and tested in Matlab ® / Simulink environment. Keywords—Fault Identification, Wavelet transform, ANN, Electrical distribution system, fault classification. I. INTRODUCTION In modern electrical power system either transmission system or distribution power system carries a large amount of power and is very complicated system. Fault identification and type of fault is necessary for fast clearance and restoration of supply to improve the power quality. Fuzzy logic based technique for fault type and identification in electrical distribution system need both voltage and current signals to analyse the system and fault inception angle also [1]. Fault classification technique for power system in case of shunt faults are used for multiple transmission lines [2]. In case of high impedance fault where the fault current is very low and can-not detected easily by the protection equipments, this situation also increased the complexity of the protection system [3-4]. Identification of fault which based on a decision tree method, identifies the fault types (i.e., L-G, L-L, L-L- G and L-L-L faults), the drawback in this method is that it cannot identify the phase or phases involved in this fault [5]. In electrical power system load is changing in nature and more sensitive to power quality disturbances. The supply of electricity for consumers with the superior quality is a major concern today. A novel approach to detect and locate power quality disturbance in distributed power system combining wavelet transform and neural network is proposed in [6]. Faults in electrical distribution system are classified as temporary or permanent. Temporary faults in overhead lines are usually caused by lightning where automatically restored of service within millisecond. Faults by objects falling on the overhead line are permanent [7]. Till today three approaches are used in the industry for fault analysis these are Classical Symmetrical Components, Phase Variable Approach and Complete Time Domain Simulations [8], but symmetrical components based techniques do not provide accurate result for power distribution systems. This proposed method is very accurate and can easily detect the faults in the distribution system and can classify all 10 types of faults i.e., AG, BG, CG, AB, BC, CA, ABG, BCG, CAG and ABC (3-Φ) symmetrical fault. The advantage of this proposed method is that it utilises only three line currents. The proposed technique is tested in IEEE-13 bus and 17 bus local feeders with the help of MATLAB ® / Simulink. II. FAULT IDENTIFICATION BY WAVELET In this proposed method the discrete wavelet transform is used for fault identification in electrical power distribution system, and obtained result from wavelet transform used as input to ANN model for fault classification. A. WAVELET TRANSFORM The wavelet transform analysis has emerged recently as a powerful tool for signal processing in different applications, particular for electrical power system. The discrete characteristics of wavelets can be employed to exact the information and effective analysis of the signals with complex frequency-time plane. Moreover, the wavelet analysis can accommodate uniform and non uniform bandwidths both. The bandwidth which is higher, make it possible to implement the wavelet analysis through different levels of decimation in a filter bank [9]. The wavelet transform of a signal does not change the information content present in the signal. The Wavelet Transform gives a time-frequency representation of the signal. Wavelet Transform uses RESEARCH ARTICLE OPEN ACCESS