Detection of Soiling on PV Module using Deep Learning

Abstract

As a clean and sustainable energy source, solar energy is gaining popularity quickly and surpassing other generation methods. However, the accumulation of dirt on the surface of Photovoltaic panels (PV) is a significant barrier to harvesting solar energy. The panel's performance and energy output are negatively impacted by this soiling, which drastically reduces the panel's capacity to harvest sunlight effectively. Therefore, regular PV module cleaning is essential to minimise efficiency losses and maximize income during the system. In this work, an intelligent approach for monitoring soiling on PV panels utilising cutting-edge Artificial Intelligence (AI) methods in order to solve this problem. As AI continues to gain popularity and become an essential component of technological advancements, we employ a predictive maintenance strategy using deep learning for soiling. Our method utilizes real-time data collection and testing, unlike existing models requiring high computational power. We achieved similar results by comparing our approach to state-of-the-art computer vision architectures while significantly reducing computational costs. Experimental results demonstrate an impressive accuracy rate of 97% in classifying solar panels' soiling status. This indicates excellent performance in identifying when panels require cleaning. Therefore, our proposed method can help maintenance personnel determine optimal cleaning schedules for PV systems. By minimizing power loss and saving labour and time associated with long-term maintenance, our approach offers tangible benefits to the overall operation and efficiency of PV systems.