DISCRETE WAVELET TRANSFORM AND MULTIRESOLUTION ANALYSIS ALGORITHM WITH APPROPRIATE FEEDFORWARD NEURAL NETWORK CLASSIFIER FOR POWER SYSTEM TRANSIENT DISTURBANCES

Power system transient (PST) disturbances occur for few cycles, which are very difficult to be identified and classified by digital measuring and recording instrumentation. They cause serious disturbances in the reliability, safety and economy of power system. The transient signals possess the non- stationary characteristics in which the frequency as well as varying time information is compulsory for the analysis. Hence, it is essential, first to detect and classify the type of transient fault and then to mitigate them in an efficient way. This article recommends the methodology of discrete wavelet transform (DWT) a time-frequency technique with multiresolution analysis (MRA) algorithm as detection and feature extractions and three types of feedforward neural network (FFNN) namely multilayer perceptron (MLP), radial basis function (RBF) and probabilistic neural networks (PNN) as classifier for power system transient problems. Different models of almost all major categories of transient disturbances are developed in Matlab, de- noised, and decomposed with the help of DWT-MRA algorithm and then selecting distinctive features to get optimal vector as input for training of three types of FFNN (MLP-RBF-PNN) as classifiers. The simulation results with proposed methodology using Matlab/Simulink/Wavelet Toolbox/Neural network prove their simplicity, accuracy and efficiency for the automatic classification of power system transient signal disturbances and propose PNN as the most suitable classifier.