A NOVEL LEARNING METHOD FOR THE CLASSIFICATION OF POWER QUALITY DISTURBANCES USING THE DEEP CONVOLUTIONAL NEURAL NETWORK

Abstract

Since a multitude of power systems are integrated into a single interconnected system, there is a growing risk of deteriorating electricity quality at all stages of electricity generation (production, transportation, distribution and in the final stage of its use. Automatic classification of energy quality distortions is the starting point for solving the problem of power quality. From the outset, the process of identifying power quality disturbances should be stratified into three independent levels, namely: analyzing the disturbances, selecting them, and classifying the signal characteristics. However, some defects that may occur are inherent in the signal analysis and thus requires a procedure of manual selection of characteristics which is demanding and inaccurate, which leads to low accuracy in the classification of multiple disturbances and poor immunity to electromagnetic noises. from the network.