Abstract: To provide the basis for the optimization of China's energy structure, coal consumption forecasting model of coal consumption is studied. Based on ARIMA,GM and ANN, 3 single forecasting models and 8 compound forecasting models are built to forecast the coal consumption of China. To forecast and analyze the development trend of China's coal consumption in the next 10 years The optimal model would be selected through the parameter evaluation of R, MAE, MAPE and RMSE. The results show that the parameters MAE, MAPE and RMSE of optimal combination weighting model are smaller, and the prediction effect is obviously better than single prediction and simple combination model. The weight of optimum weighted composition model ARIMA-GM-ANN for China's coal consumption is (0.73,0.09,0.18). The growth trend of coal consumption is divided into three stages: "slow rise", "rapid growth period" and "stable period". After 2013, coal consumption growth tends to be stable, and the growth rate of the chain is more moderate, about 1.0% ~ 2.0 %.