Abstract:The stable operation of distribution lines can effectively improve the order of the power system. Vulnerable line defects are the main cause of cascading failures and blackouts in distribution networks. The artificial recognition method has obvious defects. With the help of UAV, an automatic detection method of fragile line defect image is designed. Based on the comprehensive vulnerability index of transmission lines, the vulnerable lines of distribution network are identified by constructing the vulnerable line data set. Establish the classification standard of vulnerable line defect characteristics in distribution network, and use image enhancement technology to improve the imaging effect of vulnerable line defect images. The contrast limited adaptive histogram equalization method is used to balance the color and contrast of the fragile line defect image, and the wavelet transform is used to denoise the balanced fragile line defect image. The convolution neural network is used to input the de-noised fragile line defect image into the convolution layer to complete the automatic detection of fragile line defects. Through experimental tests, it is found that the highest recall rate of the proposed method is 89.32%, the highest accuracy rate is 98.20%, and the lowest error detection rate is 0.98%. It can identify the vulnerable line defects in the minimum range, which fully proves that the proposed algorithm has high detection efficiency.