Improving Fault Classification in Photovoltaic Arrays through Sequential Temporal Assessment of Current and Voltage Characteristics

This study scrutinizes the utility of involving more than one current-voltage ($\mathrm{I}-\mathrm{V}$) sample for photovoltaic array faults detection and classification. Our research focuses on examining how the temporal change of I-V curve can comprehend faults that may present with similar characteristics using two most known artificial intelligence algorithms categories, namely, Machine Learning (ML) and Ensemble Learning (EL) algorithms. The present analysis underscores the critical importance of actively monitoring I-V curves to detect subtle deviations and enhance the process of anomaly detection in solar panels. The Random Forest (RF) algorithm, one of the ML algorithms used in this study, demonstrated a significant increase in accuracy, improving from 89.89% to $\mathbf{9 5. 2 4 \%}$. Similarly, all the EL algorithms showed marked improvements, clearly highlighting the effectiveness of the proposed approach.