Machine Learning-Driven Fault Detection System for Electrical Networks

This study investigates the application of Machine Learning (ML) algorithms for enhancing fault detection in complex electrical systems, influenced by the integration of distributed generations and microgrids. The research signifies a shift towards automated, self-healing electrical infrastructures, promising enhanced operational efficiency, reduced downtime, and economic benefits. The study aims to present a comprehensive analysis, providing a practical approach to modernizing power system protection strategies for the smart grid era. The effect of optimizing the hyperparameters of a neural network is presented in this study using Bayesian Optimization techniques. A total of three models were developed in which one and three hyperparameters were optimized. It was found that optimizing three hyperparameters with Bayesian optimization has yielded the highest accuracy of $\mathbf{9 4 . 2 \%}$.