Anomaly Detection and Classification in Solar Panels Using Infrared Image and Deep Learning

Maintaining the efficiency of solar panels is crucial for maximizing renewable energy generation. However, timely detection and addressing anomalies, such as hotspots or delamination, can be challenging. This research explores the potential of machine learning, specifically utilizing a ResNet-9 architecture with filter pruning, for anomaly detection in solar panels using infrared imagery. By analysing 20,000 labelled images from the Infrared Solar Modules dataset, the trained model achieved an accuracy of 80.2%. This research demonstrates the effectiveness of deep learning for solar panel anomaly detection, contributing to improved efficiency and sustainability of solar farms by enabling faster and more accurate anomaly identification.