Power quality disturbances classification using convolutional neural networks

Abstract

This thesis focuses on the application of 1D Convolutional Neural Networks (CNN) for the classification of power quality disturbances. As the demand for electricity continues to rise, ensuring the reliability and efficiency of power systems has become paramount. Power quality disturbances pose a significant challenge in maintaining system stability and integrity. The aim of this research is to explore the potential of 1D CNN in accurately classifying various types of power quality disturbances, thereby contributing to the enhancement of power system reliability. This thesis provides an overview of power quality disturbances, including an exploration of the state-of-the-art research. It delves into the field of pattern recognition, specifically focusing on the detailed architecture of 1D CNN. The proposed 1D CNN model for power quality disturbance classification is presented in detail. Three different datasets were used in this work which are noiseless dataset, dataset with 30dB noise and dataset with random noise. The accuracy results were 100%, 97.18% and 93% respectively. The 1D CNN model proposed showed effective classification ability even in the case of noise, and also a good generalization for it to be used as prediction model.