The global installed capacity of Solar Photovoltaic (PV) systems surpassed 1.6 TW in 2024, with an expected annual growth rate of 12% over the next five years. However, environmental factors such as bird droppings, dust accumulation, snow coverage, and structural damages significantly reduce energy yield and lifespan of PV modules. Existing manual inspection methods are time-consuming, error-prone, and unable to provide realtime analysis for large-scale solar farms. To address these challenges, this study proposes an Intelligent IIoT-Driven Thermal Heat Analysis framework for automated fault detection in Solar PV arrays. The system integrates IIoT-based real-time data acquisition from both visual and thermal imaging sensors, followed by advanced image preprocessing techniques including noise reduction, contrast enhancement, and adaptive resizing to standardize diverse input sources. Thermal heat maps are analyzed to identify abnormal temperature variations indicative of faults such as hot spots and localized heating. While existing K-Nearest Neighbour (KNN) and Random Forest Classifier (RFC) models are used as benchmarks, the core novelty lies in the proposed hybrid VGG19 with Deep Neural Network (DNN) architecture, enabling deep feature extraction from both visible and thermal channels. The extracted features are fused and classified into six distinct fault categories: Bird-Drop, Clean, Dusty, Electrical Damage, Physical Damage, and Snow-Covered.
Saikumar et al. (Thu,) studied this question.