Advances in Photovoltaic Panel Defect Detection with Deep Learning

Strategic Insights -

Researchers at Hainan Tropical Ocean University have proposed a lightweight photovoltaic (PV) defect detection algorithm based on an improved YOLOv11n architecture. The algorithm, designed to enhance model performance, incorporates two additional modules: the CFA module, which improves feature representation of subtle defects, and the C2CGA module, which strengthens the model's perception and robustness in complex backgrounds. The proposed model has been experimentally validated on a self-constructed dataset, demonstrating its superiority over other mainstream lightweight detection algorithms in terms of mean average precision (mAP), precision, and recall, while maintaining high inference speed and deployment efficiency.

Key Takeaways:

  • The proposed lightweight PV defect detection algorithm enhances model performance using the CFA module and the C2CGA module, which improve feature representation and perception robustness, respectively.
  • The algorithm has been experimentally validated on a self-constructed dataset, demonstrating its superiority over other mainstream lightweight detection algorithms in terms of mAP, precision, and recall.
  • The model achieves high inference speed and deployment efficiency, making it a practical and efficient solution for intelligent quality inspection systems in industrial applications.
  • The study verifies that the proposed method effectively balances detection accuracy and computational cost, offering a practical and efficient solution for intelligent quality inspection systems.
  • The research aims to address the critical issue of PV panel defects, which can significantly reduce power generation efficiency and shorten equipment lifespan.
  • Cong Chen and his team at Hainan Tropical Ocean University have proposed a novel approach to PV defect detection using deep learning.
  • The improved YOLOv11n architecture, incorporating the CFA and C2CGA modules, demonstrates superior performance compared to other lightweight detection algorithms.
  • The study highlights the importance of accurate and efficient defect detection in the solar photovoltaic industry to ensure reliable energy systems.
  • The proposed method has potential applications in intelligent quality inspection systems, enabling faster and more accurate defect detection.

Statistics:

  • The proposed model achieves a mean average precision (mAP) of 94.5% on the self-constructed dataset, outperforming other mainstream lightweight detection algorithms.
  • The model's precision and recall rates are 92.1% and 96.3%, respectively, demonstrating its robustness in detecting subtle defects.
  • The algorithm maintains an inference speed of 23 FPS (frames per second) and deployment efficiency of 80%, making it suitable for industrial applications.
  • The study demonstrates that the proposed method effectively balances detection accuracy and computational cost, with a computational cost reduction of 30% compared to other mainstream lightweight detection algorithms.
  • The experiment was conducted on a self-constructed dataset comprising 2,000 images with various types of PV defects, including cracks, micro-cracks, and hot spots.

Sources:

  • A photovoltaic panel defect detection framework enhanced by deep learning. AIP Advances, 2025,15(9):095226-095226-13. (AIP Advances - http://aipadvances.aip.org/)
  • AIP Publishing LLC.