Abstract: During the long-term operation of the hydraulic station of coal mine hoist，oil leakage of⁃ ten occurs due to aging of seals and pipeline damage，which not only wastes resources but also has poten⁃ tial safety hazards. The traditional manual inspection method has low efficiency and high missed detection rate. Based on the YOLO target detection model，this paper designs an oil leakage monitoring system based on video AI. The system first collects the image data of the hydraulic station under normal and oil leakage conditions. Then，the image data set is constructed by data preprocessing method，and the anomaly detec⁃ tion and recognition model is constructed by extracting the preprocessed image features and machine learn⁃ ing model. Finally，according to the actual environment of the hydraulic station，the risk warning and re⁃ sponse of the abnormal state are realized. In practical application，the accuracy of oil leakage detection is 92.1 % ，and the average response time is 0.2 s. It shows good robustness and real-time performance， which is significantly better than the traditional detection method. The system can significantly improve the detection efficiency and accuracy in the complex practical application environment of coal mines，effective⁃ ly reduce the potential safety hazards caused by missed detection，ensure the normal operation of hydraulic stations，and reduce the dependence on manual inspection，save human resources and reduce economic losses.

Key words: oil leakage detection, YOLO target detection mode, video AI, intelligent monitoring