Detection and classification of power quality event using hybrid wavelet-Hilbert transform and extreme learning machine

Abstract

The objective of this paper is to integrate the wavelet transform (WT), Hilbert transform (HT) and extreme learning machine (ELM) for the purpose of classifying power quality (PQ) event signals. Non-stationary power quality event signals are appraised as the superimposition of various oscillating modes, and WT is used to distant out the decomposition and approximation coefficients. In this approach, the distinctive features of PQ event signals have been acquired by applying the HT on all the decomposed levels and in order to analysis the performance of the proposed method on noisy conditions, three types of PQ event data sets are constructed by accumulating noise of 25, 35 and 45 dB. ELM is an efficient learning algorithm for generalized single hidden layer feedforward networks (SLFNs), which is implemented to recognizing the various PQEs. Based on very high performance under ideal and noisy conditions, the proposed WHT-ELM method has robust recognition structure that can be used in real power systems.