Edge-integrated IoT framework for real-time fault diagnosis and performance degradation analysis in photovoltaic modules

Conventional diagnostic techniques for photovoltaic (PV) module faults, such as I–V curve tracing and infrared thermography, are often reactive, lacking the resolution and real-time responsiveness needed for proactive fault mitigation. This study introduces an innovative, low-power IoT-integrated diagnostic system for assessing damage-induced performance degradation in PV modules, emphasizing thermal, electrical, and exergy-based metrics. The experimental setup features side-by-side field analysis of physically fractured and intact monocrystalline PV modules. Results reveal that the damaged module exhibited a 29.49 % current loss and 30.77 % power output reduction, with surface temperatures peaking at 60.3 °C. Electrical and exergy efficiencies declined by 33 % and 27.5 %, respectively. An energy-autonomous hotspot mitigation circuit, consuming just 10 mW, was deployed to suppress localized overheating without external power. For intelligent fault classification, signal features extracted via Stockwell Transform and Stationary Wavelet Transform were reduced using PCA and classified using SVM, ANN, and KNN. KNN yielded the highest accuracy (94.73 %) and F1-score (94.92 %) in simulated conditions, whereas ANN proved more robust in real-world testing (accuracy: 90.52 %). This study uniquely bridges thermal imaging, circuit-level protection, and edge-implemented machine learning without cloud dependence. It addresses a critical research gap by enabling embedded, real-time diagnostics tailored for cost-constrained and remote PV deployments. The proposed framework demonstrates scalable potential for enhancing PV system reliability, reducing maintenance overhead, and prolonging operational lifespan, offering a novel contribution to the scientific niche of intelligent, multi-modal PV health monitoring systems.