Random Forest Empowered Shade and Fault Detection in Photovoltaic Panel

In recent years, the role of drones in disaster management has garnered substantial attention due to their ability to respond to a variety of hostile environments effectively. Unmanned aerial vehicles have demonstrated value in rapid damage assessment, search-and-rescue operations, and navigating challenging terrains. This paper explores integrating photovoltaic (PV) based power sources in drones for disaster response. While PV power sources offer sustainability and endurance advantages, they face challenges in low light conditions and susceptibility to physical damage or short circuits. To address these challenges, the incorporation of Complementary Metal-Oxide-Semiconductor (CMOS) embedded PV panels is proposed, enabling real-time reconfiguration of the electrical layout to optimize power generation. However, accurate detection of shade and faults becomes crucial for this strategy. This paper emphasizes applying machine learning (ML) classification techniques for reliable shade and fault detection. Notably, the Random Forest Technique is investigated for identifying faults in CMOS-embedded PV panels. Leveraging multiple decision trees, this ensemble learning approach enhances accuracy in fault classification. By training on diverse fault scenarios, the model can discern normal and faulty PV panel behavior, facilitating the detection of shaded, open circuit, and short circuit faults. This real-time response ensures optimal power generation and drone functionality. The presented ML classifier achieves 94.50% accuracy in detecting shade and faults, thus contributing to the enhanced reliability and effectiveness of PV-powered drones in disaster scenarios.