Deep Learning for Solar Panel Fault Detection: Integrating GAN and ResNet Models
V. Manimegalai, B. Oviya, S. Umesh Kargvel, P. L. Vilashini, V. Mohanapriya, A. Elakya · 2025
The rapid expansion of solar energy implementation has highlighted critical challenges related to temperature variations and environmental factors impacting photovoltaic (PV) systems. This paper addresses the need for continuous monitoring to ensure the efficiency and reliability of solar panels. PV panels frequently encounter operational issues including surface contamination, irregular shading patterns, and thermal irregularities, which can result in hotspot formation, leading to substantial power losses and accelerated panel deterioration. The Machine learning algorithms play a crucial role in detecting defects in solar panels, such as cracks, shading, and hotspots, by analyzing both thermal and visible light images. The implementation of deep learning architectures, specifically ResNet for feature extraction and Generative Adversarial Networks (GANs) for pattern recognition, enables precise real-time fault detection. The demonstrated success of this integrated deep learning approach establishes a new benchmark for PV system monitoring, enhancing both the reliability and longevity of solar installations while minimizing operational downtime.