Comparative Analysis on Solar Panel Defect Detection Using Deep Learning Approaches

Solar energy is one of the widely used renewable energy for commercial as well as house hold purposes. Even though solar energy is widely used, still one time installation change for the panel is high. The production of panel is not yet cost effective. In India, several panel manufacturers are trying to bring defect free product into the market. There are several methods adopted to make the panel defect free during manufacture as well as in working condition when installed in solar farms. Artificial Intelligence can be used to tackle the complex problems of designing solar plants and can save diverse power system issues like planning, control, scheduling, forecasting and so on. It can handle the demanding requirements that applications in current big power systems confront so as to meet increasing load demand. Solar farm owners usually used a team of employees to physically check solar panels for faults. This process is time-consuming, not cost effective and also less accurate. Maintenance visit to solar farms is very expensive and are simply not feasible to perform on daily basis for an entire solar development. The inspection process can be speedup by turning into an artificial intelligence powered inspection. It entails using algorithms to detect solar panel defects from photographs, which is quicker and more precise than human inspection. Eventhough artificial intelligence technology helps a lot to find the solar panel defect perfectly, it has some challenges that needs to be overcome. These challenges are addressed by using the deep learning methods like MobileNetV2, ResNetV1 and inceptionv3. The present paper analyses MobileNetV2, ResNetV1 and inceptionv3 for detecting solar panel defect. The simulation result shows MobileNetV2 provides better accuracy compared to others.