Wavelet-based neural classification for power quality disturbances

The objective of this paper is to present a new method for automatically detecting, localizing and classifying various types of power quality disturbances. The new method is based on wavelet transform analysis, artificial neural networks, and the mathematical theory of evidence. The proposed detection and localization algorithm is carried out in the wavelet transform domain using multiresolution signal decomposition techniques and the proposed classification method is carried out in the sets of multiple neural networks using a learning vector quantization network. The outcomes of the networks are then integrated using a voting decision making scheme. The performance of the automatic detection and localization have 90.14% accuracy and the error is less than 5%.