AI Deep VGG19-based Heatmap Analysis and Thermal Imaging for Photovoltaic Panel Fault Detection

This study presents a novel deep learning-based approach for fault uncovering in photovoltaic (PV) panels by means of thermal imaging. With the rapid expansion of photovoltaic energy then the growing difficulty of solar power plants, there is an increasing demand for advanced condition monitoring systems that ensure operational reliability and efficiency. The proposed method leverages thermal images acquired via thermographic cameras and integrates two convolutional neural network (CNN) models. The first model, based on the VGG19 architecture and enhanced through transfer learning, classifies fault types in PV panels and achieves an F1 score of $\mathbf{8 3 . 2 5 \%}$. The second model, utilizing the Faster R-CNN framework, performs object detection to localize faulty regions within the panels, attaining a mean Average Precision (mAP) of $67 \%$. This dual prototypical system significantly reduces the need for manual inspection, enables early detection of faults, and contributes to improved maintenance strategies and operational efficiency in solar power installations. The results highlight the potential of combining thermal imaging and deep learning for scalable, automated PV panel fault diagnosis.