Defect detection of solar panel using CNN

The efficiency of solar panel power production is greatly influenced by the state of the panel. Therefore, it can be said that it is an important factor to check whether a solar cell is damaged in managing solar power. Existing solar cell damage detection technology detects whether the solar cell is damaged by an electroluminescence phenomenon or by checking the degree of scattering by shining a laser. In this study, we attempt to detect the damage of the solar cell by using the deep learning method and taking the surface image of the solar cell without additional measuring device. When using the existing electroluminescence phenomenon, a separate device must be configured to check for damage, and there is a disadvantage in that the use of the connected cell must be stopped to confirm a part. The method of checking the degree of scattering by illuminating the laser has a disadvantage that many areas cannot be seen at once. To improve this, this paper presents a detection method based on deep learning. We present an algorithm that is based on CNN (Convolutional Neural Network). Experimental results according to model size and data augmentation are given to show the feasibility of proposed method.