Deep Learning-Based Defect Detection for Photovoltaic Cells Using Electroluminescence Imaging

Simplifying the maintenance of photovoltaic (PV) power plants, a long-standing formidable challenge, is now becoming more feasible and manageable with the emergence of Electroluminescence (EL) imaging. In this study, we introduce a defect detection method for photovoltaic cells that integrates deep learning techniques. To develop and evaluate the proposed model, we trained it on a dataset consisting of 2,624 Electroluminescence (EL) image samples. For performance comparison, we assessed the proposed model against several benchmark models, including ResNet50, InceptionV3, ResNet152, Vgg16, Vgg19, and DenseNet121. Through this evaluation, we measured the performance of our proposed model in comparison to established models. The experimental results demonstrate that our proposed model achieved superior performance in defect detection. EfficientNet outperformed other models with the highest accuracy of 96.51% and competitive F1 score, recall, and precision, showcasing its superior performance.