Loss of main detection in distribution generation system based on hybrid signal processing and machine learning technique

Abstract

This manuscript presents a novel approach of islanding detection based on hybrid signal processing techniques comprised with mathematical morphology and wavelet transform. The combination of mathematical morphological filter and translational invariant wavelet is considered as a preprocessing filter unit to decrease the random noises in the original power signal. This technique uses the 2 basic mathematical morphology operator erosion dilation difference filter and opening-closing difference operator to operate on negative sequence voltage and current signal received at target distribution generation bus for the extraction of different useful features. An extreme learning machine-based machine learning technique is further used as a classifier to discriminate islanding events from nonislanding events. Apart from that, another islanding detection method based on the combination of proposed preprocessing filter unit and wavelet transform is presented. The comparative result demonstrates improved accuracy for the proposed approach of islanding detection and classification, even under certain extremely noisy conditions.