Photovoltaics Cell Anomaly Detection Using Deep Learning Techniques

Photovoltaic cells play a crucial role in converting sunlight into electrical energy. However, defects can occur during the manufacturing process, negatively impacting these cells’ efficiency and overall performance. Electroluminescence (EL) imaging has emerged as a viable method for defect detection in photovoltaic cells. Developing an accurate and automated detection model capable of identifying and classifying defects in EL images holds significant importance in photovoltaics. This paper introduces a state-of-the-art defect detection model based on the Yolo v.7 architecture designed explicitly for photovoltaic cell electroluminescence images. The model is trained to recognize and categorize five common defect classes, namely black core (Bc), crack (Ck), finger (Fr), star crack (Sc), and thick line (Tl). The proposed model exhibits remarkable performance through experimentation with an average precision of 80%, recall of 87%, and an [email protected] score of 86% across all defect classes. Furthermore, a comparative analysis is conducted to evaluate the model’s performance against two recently proposed models. The results affirm the excellent performance of the proposed model, highlighting its superiority in defect detection within the context of photovoltaic cell electroluminescence images.