Power Quality Management and Classification for Smart Grid Application using Machine Learning

Abstract

The Efficient Wavelet-based Convolutional Transformer network (EWT-ConvT) is proposed to detect power quality disturbances in time-frequency domain using attention mechanism. The support of machine learning further improves the network accuracy with synthetic signal generation and less system complexity under practical environment. The proposed EWT-ConvT can achieve 94.42% accuracy which is superior than other deep learning models. The detection of disturbances using EWT-ConvT can also be implemented into smart grid applications for real-time embedded system development.