Aiming at the problems of complex defect types, variable target scales and background interference of current substation equipment, a depth feature sensing method for substation equipment defect detection is proposed. Firstly, a C2f+module based on sliding window mechanism is designed to enable the backbone network to effectively process cross window feature information, expand the receptive field of the network, and strengthen the network's deep attention to local features and global feature processing ability. Then, integrate SimAM into the feature fusion stage of the network, and endow the network with the ability of autonomous learning and dynamic fine-tuning of attention weight to improve the perception performance of multi-scale defects. Secondly, in order to avoid the detection error introduced by equipment defects of different sizes and shapes, a loss function integrating NWD and CIOU is designed. By modeling the two-dimensional Gaussian distribution of the prediction box and the real box, the learning ability of the model to the defect location is improved. The experimental results on the substation equipment defect data set show that the average detection accuracy mAP of the proposed method for 11 defect categories reaches 98.27%, and the best accuracy is achieved in nine defect detection methods, which is superior to the other six mainstream target detection methods and power equipment defect detection methods. At the same time, FPS reaches 78.4, with high real-time performance.