FaultyVoltaMan: Ensemble Learning Model for Accurate Fault Detection and Classification of PV-Integrated Systems

Using intelligent models is critical in rapidly identifying and categorizing faults occurring in renewable energy sources. Their sophisticated algorithms optimize system efficiency and preventatively detect deviations, thereby reducing periods of inactivity and facilitating the generation of sustainable, pure energy. The proposed study introduces FaultyVoltaMan, an ensemble learning model engineered explicitly for accurate defect identification and categorization in photovoltaic (PV) systems. FaultyVoltaMan generates a heterogeneous ensemble of decision trees by employing the Random Forest algorithm. Each tree is trained on an arbitrary subset of features, enhancing robustness and reducing the risk of overfitting. The model undergoes evaluation utilizing an extensive dataset obtained from a residential rooftop PV facility. This dataset comprises temperature data, electrical metrics, and solar irradiance. By conducting comprehensive simulations, FaultyVoltaMan showcases its efficacy in precisely discerning and classifying a wide range of fault categories, including intermittent shading and uniform shading. The dependability of predictions is ensured by the majority voting mechanism of the ensemble, whereas feature importance analysis offers valuable insights into the critical contributors to defect detection. Due to its capacity to manage a wide range of fault scenarios and its scalability and efficiency, FaultyVoltaMan is regarded as a highly prospective solution for improving the performance and dependability of PV systems in practical contexts.