Abstract: In order to achieve the accurate forecasting of dynamic gas emission in coal mining and heading face, firstly, we utilize the gray model GM (1,1) and the Autoregressive Integrated Moving Average Model (ARIMA) to forecast the concentration of gas emission respectively based on the time series, and then a new combination forecasting model, ARIMA-GM, built by the method of variance reciprocal weighting is represented to forecast and analyze the concentration of gas emission, finally, the gas pre-warning is obtained according to forecasting results. An application example of forecasting in 15101 heading face of Xinshun Coal Mine is presented and its results show that the combination forecasting model of ARIMA-GM provides a better fitting effect and a higher forecasting accuracy compared with single model.