A PSO Tuned CNN Approach for Accurate Fault Detection in PV Grid Systems

Photovoltaic (PV) systems are growing in popularity and are widely utilized in power systems to capture renewable energy. However, these systems remain susceptible to numerous defects caused by environmental variables, device lifespan, and failures in operation. The proposed system presents an effective methodology for fault diagnostics in PV-connected grid systems by combining Particle Swarm Optimization (PSO) with Convolutional Neural Networks (CNN). The system utilizes data to monitor crucial factors like voltage, current, and other indicators of performance. This data is processed through acquisition and preprocessing stages, including normalization and structured transmission, to ensure effective analysis for further use in decision-making processes. The hybrid model employs PSO to optimize CNN hyperparameters, enhancing the network's ability to accurately detect and classify faults. In this work, Python software implemented to diagnose faults in PV systems, with the effectiveness of the proposed approach evaluated using a "Solar Panel Fault Detection Dataset". The results are evaluated to ensure accurate and making an accuracy of 93.36%, precision of 94%, recall of 95%, and an F1 score of 95% for maintaining the stability and efficiency of PV-connected grid systems.