Author(s)

Binhui Liu, Qiangrong Yang, Yurong Han

Corresponding Author

Binhui Liu

ABSTRACT

Electroluminescent (EL) plays an important role in the application of photovoltaic cell Defect detection. Traditional approaches for EL result analysis usually utilize visual inspection by technicians and have the drawbacks of low efficiency which can be improved by employing deep convolutional neural network (CNN) features that contain more semantic and structure information and thus possess more discriminative ability. Therefore, a defect detection method based on EL and GoogLeNet is proposed in this work. Firstly, a database of EL image samples for photovoltaic cell defects is built, then a deep convolutional neural network based on GoogLeNet is established. At last, the experiments and simulation tests prove that the presented defect detection approach is superior to the conventional methods. The detection precision is more than 85%, while the previous accuracy is under 67%. What’s more, the proposed method is more stable and efficient.

KEYWORDS

Photovoltaic cell defect detection, Convolutional Neural Network CNN, Electroluminescent (EL), GoogLeNet