Neuro-fuzzy access for detection of faults in an underground cable distribution system

Abstract

Nowadays, the interest of power system engineers In Indian Power System has increased towards the use of underground cables with the advent of cross-linked polyethylene (XLPE) insulated cables having high capacity for transmission of power. Underground cables are preferred in the densely populated regions where there is environmental constraint and right of way poses a big problem. The key limitation of underground cable is to locate and detect different types of faults in view of the fact that the cables are lying down under the surface. It is necessary that the fault must be cleared in minimum time on account of protection issues. As conventional methods for detection and classification of faults are time consuming, so, this work uses intelligent techniques for fast and more accurate detection of location and classification of faults in underground cables. Overall work has been performed in three steps, the first step is to develop a MATLAB/Simulink Model of a distribution system using underground cable with a provision to develop fault. In second step, an Artificial Neural Network (ANN) using DWT is used for fault detection & classification. In third step, ANN is hybridized with fuzzy system and discrete wavelet transform (DWT) methods to improve its performance. The training sets of adaptive neuro fuzzy inference system (ANFIS) are energy components of three phases of cable under fault (used as inputs) and fault type or different distances of faults in the cable(used as outputs). All the simulations have been carried out in MATLAB/ SIMULINK environment.