Application of Machine Learning Techniques for Fault Diagnosis in photovoltaic Arrays Using Temporal Changes in Current and Voltage Signals

This study explores the variability of two key signals, I (Photovoltaic Array Current) and V (Photovoltaic Array Voltage), within the context of diagnosing faults in photovoltaic systems. Our research focuses on how examining the temporal evolution of these two signals can differentiate faults that may initially present with similar characteristics. This analysis underscores the significance of continuous monitoring of these signals to identify any variation and enhance the fault diagnosis process in photovoltaic systems. In fact, K-Nearest Neighbor (K-NN) as one of the used models in this study. K-NN saw a significant increase in accuracy, passing from 80.81% to 87.80%, demonstrating the importance of using the proposed method.