Wavelet based signal analysis in three phase circuits using neural network

Abstract

This article proposes the use of wavelet based neural network in recognizing the harmonics. The discrete wavelet transforms is used in calculating the effective values of three phase active and reactive power. The discrete wavelet transforms with probabilistic neural network (PNN) model is used in constructing the classifier of harmonics. There are four stages in identification of harmonics. First stage is simulation of a three phase inverter circuit. The second stage is the conversion of the three phase current and voltage by PQ theory into alpha–beta reference frame. The third stage is that these alpha–beta vectors are subjected to multi resolution analysis. The signals are then decomposed by discrete wavelet transform (DWT) at six levels. The final stage is that the coefficients generated are used in the calculation of active, reactive power and energy. The DWT coefficients are used in identifying the harmonics using PNN. The main advantage in this analysis is that the active and reactive power calculated is done at various frequency bands. This covers the harmonic power at various frequencies. Wavelet is used in this article since they can detect and locate the disturbance accurately. © 2009 Wiley Periodicals, Inc. Comput Appl Eng Educ 18: 537–546, 2010; View this article online at wileyonlinelibrary.com; DOI 10.1002/cae.20223