Classification of Defective Photovoltaic Modules in ImageNet-Trained Networks Using Transfer Learning

The quality inspection of photovoltaic modules is very important to power generation efficiency. Although electroluminescence can directly detect the defects in photovoltaic modules, it is difficult to get the desirable accuracy by traditional manual inspection. It thus necessitates an efficient defect identification method. This paper introduces ImageNet-trained networks for identifying the defective Photovoltaic modules. Six convolutional neural networks are realized by transfer learning. The experimental results present the Xception model has a higher accuracy for the classification of defective photovoltaic cells.