Neural Network Methods for Harmonic Analysis using Wavelet Transform Coefficients

J.S. Huang (Australia)


Neural networks, Wavelet transform, Power quality disturbances, Harmonic distortion


A neural network based method is proposed in the paper for harmonic analysis using wavelet transform (WT) coefficients. The research is motivated by the author's project on power quality monitoring. In the project, a classifier for recognizing power quality disturbances was developed. The classifier adopts wavelet transform for feature extraction of various distorted waveforms including those containing harmonics. Obviously, it is no longer necessary to use Fourier Transform for evaluating harmonics that are retrievable from the WT coefficients. Consequently, a number of algorithms have been developed to estimate the distortion contributions from different sub-band coefficients. By using neural networks having the architecture of Multi Layer Perceptron (MLP), furthermore, each harmonic component is directly retrieved. Both the distortion estimation and the evaluation of individual harmonic component have achieved high accuracies as shown by the simulation results.

Important Links:

Go Back