DEVELOPMENT OF A FAULT DETECTION AND CLASSIFICATION SYSTEM IN MICROGRID USING ARTIFICIAL NEURAL NETWORKS

Abstract

As microgrids grow in significance, enabling consumers to manage their energy demands themselves, they present complexities such as fluctuating loads, two-way (bidirectional) power flow, and low fault currents that make traditional protection approaches such as Overcurrent protection and Impedance-based techniques less effective. This study explores the development of a fault detection and classification system based on Artificial Neural Networks (ANN) specifically designed for microgrids to address these complications. This research simulates a range of fault scenarios applicable in a microgrid setting using MATLAB/Simulink, extracts relevant voltage and current data, develops and trains an ANN model, and assesses its ability to accurately identify and categorize faults. The performance analysis shows the detection and classification models demonstrate remarkable accuracy, achieving low Mean Squared Error (MSE) values of 3.72 × 10⁻¹⁰ and 0.014393 and regression correlations as high as 1 and 0.998 across training, validation, and testing datasets. This research has improved microgrid systems' reliability through artificial neural networks by simultaneously reducing downtime. The findings establish a foundation for the future development of fault detection systems and incorporation into smart grid technologies.