研究目的

to eliminate the human visual identification in the analysis of solar panel characteristic by PV cell defect detection and recognition process; to detect solar panel defective portion (solar cells) from EL image; to utilize digital image processing for feature extraction of a detected defective portion (cells) of the solar panel; and to recognize the detected portion (solar cells) of PV panel or module through support vector machine (SVM).

研究成果

The proposed method simplifies the representation of thermal images into an easier-to-analyze format for photovoltaic applications, showing a clear difference between non-defective and defective cells. The SVM classifier achieved an accuracy of 97.5%, outperforming other classifiers tested.

研究不足

The study focuses on EL imaging and SVM classification for solar cell defect detection, which may not cover all types of defects or conditions. The accuracy of 95% indicates room for improvement, and the method's effectiveness on larger datasets or different types of solar panels was not explored.