Identification of asymmetrical faults in electrical power systems based on signal processing and neural network

Abstract

In the present era, faults are the greatest interruption for the power system utility. Theses, faults on electrical power systems are unavoidable problems and will continue to happen. These faults are effects on the power system reliability and stability, hence, diagnosis and classification of such faults in rapid and accurate way is an important issue. In this paper, combination method of digital signal processing and multi-layer neural network has presented. The methodology has divided in two steps, firstly: wavelet transform has implemented in here for pre-processing the data, which is used to extract the useful information during the fault in both time and frequency domain, and calculate the features of coefficients which is used as input for neural network. Secondly: multi-layer neural network has adopted here to detect and classify the unsymmetrical faults in different conditions such as single line to ground fault, line to line to ground fault and double line fault. Power System Computer-Aided Design /Electromagnetic Transients with DC (PSCAD/EMTDC) used to simulate the three types of asymmetry faults. Simulation results reveal that the proposed method gives satisfactory results, and will be very useful in the development of a power system protection scheme