Performance Analysis of Deep Neural Network-based Fault Detection in Standalone Photovoltaic DC Ring Microgrids

This study explores the application of Deep Neural Networks (DNN) for fault detection in a standalone photovoltaic (PV)-based DC ring microgrid system. It follows a structured five-step methodology, beginning with the identification of various fault types, including short circuits, open circuits, hot spots, overheating, mismatch, and partial shading. Current and voltage signals undergo pre-processing steps such as data cleaning, normalization, and segmentation before being used to train the DNN model. The training and evaluation are conducted using simulation data from a PV-based DC ring standalone microgrid developed in Simulink. While the confusion matrix indicates challenges in accurately classifying faults like partial shading due to higher misclassification rates, the model achieves high diagnostic accuracy for hot spot faults with a test accuracy of 98%, along with strong precision and recall scores. The integration of DNN in the standalone PV-based DC ring micro grid, known for its looped topology and reliability, enables early fault detection and supports predictive maintenance, thereby enhancing system safety, reliability, and performance.