Addressing Power Quality Issues through Deep Learning

Power quality (PQ) issues refer to the different types of distortions in the sinusoidal voltage waveforms in electric power systems, also known as “power quality disturbances”. These disturbances can cause many severe problems, including losses in power supply and damage to the electrical equipment; hence, their timely detection and prevention are crucial. The first step in addressing these issues would be to identify the disturbance correctly; a total of 29 types are considered for this task. In this paper, we take some popular pre-trained Convolutional Neural Networks (CNNs) based on deep algorithms in MATLAB environment and train them to perform a new task of classifying these disturbances. An integral mathematical model is used to create a synthetic dataset for this purpose, the software version of which is publicly available. Based on the evaluation of the classification performance of these networks, we propose an automated solution to address these quality issues through deep learning.