A photovoltaic panel defect detection framework enhanced by deep learning

With the rapid development of the solar photovoltaic industry, the efficient and stable operation of PV modules is crucial for the reliability of energy systems. However, PV panels are prone to various defects such as cracks, micro-cracks, and hot spots during manufacturing, installation, and operation, which can significantly reduce power generation efficiency and shorten equipment lifespan. Therefore, fast and accurate defect detection has become a vital technical demand in the industry. This paper proposes a lightweight PV defect detection algorithm based on an improved YOLOv11n architecture. Building upon the original YOLOv11n framework, two modules are introduced to enhance model performance: (1) the CFA module (Channel-wise Feature Aggregation), which improves feature representation of subtle defects through the integration of context and local pathways, channel attention guidance, and residual correction mechanisms; and (2) the C2CGA module (Cross-Channel and Cross-Group Attention), which strengthens the model’s perception and robustness in complex backgrounds by enabling intra-group channel interaction and cross-group context fusion. Experimental results on a self-constructed dataset comprising various types of PV defects demonstrate that the proposed model outperforms YOLOv11n and other mainstream lightweight detection algorithms in terms of mAP, precision, and recall, while maintaining high inference speed and deployment efficiency. This study verifies that the proposed method effectively balances detection accuracy and computational cost, offering a practical and efficient solution for intelligent quality inspection systems in industrial applications.