Abstract:To address the issues of low efficiency and poor accuracy in manual inspections in substations, this paper explores two types of substation hazard identification algorithms based on image analysis. Compared to the hazard identification algorithm based on scene change detection, the target detection-based algorithm is not limited by benchmark images and has better comprehensive recognition effects, enabling effective identification of various hazards in substations. After annotating a large number of hidden danger images of substations, the model was trained on the YOLOv8 deep learning framework. The trained model was tested on a test set consisting of some hidden danger images, and the results showed that the hidden danger detection rate was 82.10%, and the hidden danger false detection rate was 14.64%. The proposed target detection-based substation hazard identification algorithm has high recognition accuracy, fast recognition speed, and simple deployment, which can effectively replace manual inspection work. It significantly improves the efficiency and accuracy of hazard identification in substations, which is of great significance for ensuring the safe and stable operation of substations.