Characterization of Power Quality Disturbances and Their Efficient Classification

Abstract

Characterization and classification of power quality (PQ) disturbances are an essential component in the field of power engineering to meet consumer demands. Accurate analysis of supply power, its processing, and distribution requires identification of the noise and disturbances associated during power generation, transmission, and distribution. In general, pure-tone power signals are non-stationary with both time- and frequency-varying statistical parameters. Thus, the use of either the time domain or frequency domain analysis cannot characterize or classify the PQ signal adequately. This motivates the authors to approach the problem domain employing time–frequency (TF) characterization using a spectrogram initially. TF distribution is one of the best application tools for PQ analysis and is emphasized using the short-time Fourier transform (STFT) and wavelet-based features in this paper. Finally, the TF-based features extracted from the normal and different PQ disturbances are applied to an efficient probabilistic neural network (PNN) model for classification. We have shown that PNN with TF-based wavelet features provides an efficient classification result as compared to other chosen techniques.