WAVELET TRANSFORM AND ARTIFICIAL NEURAL NETWORK BASED NOVEL APPROACH FOR THE DETECTION AND CLASSIFICATION OF ELECTRICAL POWER QUALITY DISTURBANCES: A PROPOSED METHODOLOGY

It is the fact that electrical power quality disturbances (EPQDs) vary in a wide range of time and frequency with a broad range of disturbance categories from low-frequency dc offsets to high- frequency transients. This makes automatic detection and classification problems of power quality (PQ) often difficult and elusive to diagnose. Hence one of the most important issues in PQ problems nowadays is how to detect and classify PQDs waveforms automatically in an efficient manner. To improve electric PQ, sources and causes of disturbance must be specified before initiating any mitigating action. This requires monitoring, recognition and classification of disturbances. Detailed studies of literature survey indicate that lot of work; studies and research have been presented proposing the use of various signal processing and artificial intelligent techniques which have been implemented for the continuous monitoring, recognition and classification of PQ disturbances. But still the simplest software based novel approach methodology is required for detection and classification of all 15 types of EPQDs by combining discrete wavelet transform (DWT) based time- frequency analysis with multiresolution analysis (MRA) algorithm having different mother wavelet functions and signal to noise ratio for selecting suitable statistical parameters of feature extraction in order to produce optimal feature vector for for the input of feedforward neural network like radial basis function (RBF), multilayer perceptron (MLP) and probabilistic neural network (PNN) as classifier.