Abstract: In view of the problems of insufficient monitoring accuracy, limited early warning capability and others of the ventilation system in Yicheng Shanghe Coal Industry of Jinneng Holding Coal Industry Group, an intelligent monitoring system based on LSTM and random forest hybrid algorithm is proposed. A cloud edge end collaborative architecture is adopted by this system, lightweight models are deployed at the edge layer to conduct real-time anomaly preliminary screening, and complete models are deployed at the cloud end to achieve precise prediction. By establishing a three-level early warning mechanism and a multi-channel redundant transmission mechanism, the system achieves a data acquisition success rate of 94.8% and a prediction accuracy rate of 93.2%. 6 months after being put into operation, 47 times of ventilation abnormal events are successfully alerted, and the average detection time of abnormal status is shortened from 15 minutes to 90 seconds, 3 times of safety hidden dangerous are effectively avoided. During the system operation period, the ventilation parameter configuration is optimized through data analysis, 12% of reduction in the operation energy consumption of the ventilation system is achieved, and the annual electricity cost is saved by 850 000 yuan, meanwhile, the manual inspection workload is reduced by 80%, and the annual labor cost is saved by 1.2 million yuan approximately.