Unified Machine Learning Based Fault Detection Strategy Through Voltage-Sensing for Both AC and DC Side Faults in Photovoltaic Farms

Photovoltaic (PV) farms, consisting of a vast number of solar panels, are widely recognized as a viable source of renewable power generation. However, fault detection and protection remain critical challenges in PV farms. Conventional methods such as over-current relays prove to be inadequate, leading to system disruptions. The intricate nature of PV arrays necessitates fast fault detection. This paper proposes a unified machine learning based approach which can detect various types of faults on both DC (PV) and AC (grid) sides, using just one voltage sensor per each side. Moreover, our proposed approach offers a viable solution for fault detection, classification, and protection solution in a unified framework. The proposed method has been evaluated with a simulation case study of a 500kW grid-connected PV farm model and achieves 94.33% accuracy in fault detection on AC side while could perfectly detect faults on the DC side.