AI-Powered VGG16-based Method for Detecting, Classifying, and Visualizing Solar Panel Defects

In this study, a deep learning-based approach is proposed for detecting and localizing defects in photovoltaic (PV) solar panels using thermal imaging. Given the increasing complexity of solar energy systems, automated fault detection is crucial for maintaining efficiency and reliability. The proposed method employs an enhanced VGG16-based convolutional neural network (CNN) to classify PV panel defects and a heatmap-based localization technique to identify fault regions. The classification model, fine-tuned using transfer learning, achieves an F1-score of 84.12%, while the heatmap-based localization achieves a Mean Minimum Precision (MMP) score of 71%. To further strengthen the model, comparative evaluations with advanced architectures, including YOLOv8 and Transformer-based methods such as Vision Transformers (ViTs) and Swin Transformer, are conducted. The study also incorporates an extended dataset with additional thermal images and advanced data augmentation techniques to enhance model generalization. This approach reduces the need for manual inspections, providing a more reliable and scalable solution for solar panel defect detection.