Abstract: A light-weight convolutional neural network model with batch standardized T-VGG（Tiny Visual Geometry Group）was proposed to integrate attention mechanism and Ghost block into the EL image of solar cells. Using of Ghost convolutional layer to replace the conventional convolutional layer,followed by the introduction of attention and batch standardization,so as to achieve high precision and high-speed detection of battery defects. The experimental results show that the accuracy of the convolutional neural network model for defect detection is 99.15%,The detection accuracy of defect type is 96.28%,and the time is 0.032 s/piece,which not only ensures high precision and high efficiency,but also has universality.