Identification and classification of power quality disturbances using wavelet-based neural network

Abstract

The paper presents a method of identification and classification of power system disturbances by neural network based on the multi-resolution analysis wavelet technique. Various transient phenomena of voltage sag, voltage swell, harmonics, flicker, interruption, capacitor bank switching transients were simulated by transient caculation software ATP-EMTP, then exported to Matlab for further analysis by wavelet technique. Application of Parseval‘s theorem of energy calculation at various decomposition level will extract interesting features from signals in analysis. Probabilistic neural network technique was then used to recognize and classify automatically transient signals obtained from numerical experiment. Results obtained from numerical experiments shows the merit, as well as the feasibility of the proposed approach.