IDENTIFICATION OF FAULTS IN DC MICROGRID USING ARTIFICIAL NEURAL NETWORKS

This research presents a comprehensive and intelligent framework for the automated identification and classification of faults within a DC microgrid environment using Artificial Neural Networks (ANN). As the transition toward sustainable energy accelerates, DC microgrids have emerged as a superior alternative for integrating photovoltaic arrays, fuel cells, and battery energy storage systems due to their higher efficiency and reduced conversion stages. However, the low impedance of DC lines leads to extremely rapid fault current discharge, posing a significant challenge for conventional protection devices. This study addresses these challenges by developing an ANN-based diagnostic model capable of high-speed fault detection. The methodology employs a multi-layered perceptron (MLP) architecture trained using the backpropagation algorithm. Input features are derived from the transient current and voltage profiles extracted during various operating states. Extensive simulations were conducted on a test DC microgrid layout to capture data for pole-to-pole (PTP) and pole-to-ground (PTG) faults, considering variations in fault resistance, location, and power source intermittency. The results demonstrate that the proposed ANN model achieves a classification accuracy exceeding 99%, effectively distinguishing between internal faults and external disturbances such as sudden load switching or capacitor bank charging. Furthermore, the system exhibits remarkable robustness against measurement noise, ensuring that protection coordination remains reliable even under non-ideal sensing conditions. The significance of this work lies in its potential to replace time-delayed traditional relaying with a proactive, data-driven protection scheme, thereby enhancing the overall stability, safety, and resilience of next-generation DC distribution infrastructures in smart city applications.