Abstract: By the analysis of running status of current mine safety monitoring system, underground gas gathering data was deep-level studied, secondary developed and utilized to realize real time monitoring of gas concentration abnormal increase process and early warning. This paper presented gas abnormal warning method based on single measurement point, which combined inspection concentration value and the inspection time. Based on mathematical statistics, the paper put forward the method of using mean confidence interval estimation method to dynamically set alarm limit value of gas abnormal warning model, which was verified by combing the actual situation of mine.