Detection of PQ Short Duration Variations using Wavelet Time Scattering with LSTM

Abstract

In the electrical power system, the detection of power quality disturbances (PQDs) is a critical mission. In this paper, two-step methodology is used to solve PQDs detection; features extraction and classification. The features extraction step uses wavelet time scattering and the classification step uses the long short-term memory (LSTM) techniques. To assess the efficacy of the proposed approach, various simple PQ disturbances such as sag, swell, harmonics, and interruption, as well as complicated power quality events such as sag with harmonics and swell with harmonics, are produced using the MATLAB programming framework. A comparison using several methodologies is provided. The results demonstrate that wavelet scattering with LSTM can decrease classification computation complexity. Furthermore, it may significantly shorten classification time while assuring classification accuracy better than different approaches.