To scientifically predict energy demand, the economic development level, population scale, urbanization rate, industrial structure and level of technical progress were selected as the influencing factors of energy demand, the weight of each model based on Deviation Maximization Method was calculated, and the combined model of ARIMA, Multiple Regression, Grey GM(1,1) and Support Vector Regression were established to calculate energy demand from 2005 to 2011 respectively. The result shows that the combined model based on deviation maximization method has higher prediction accuracy.Finally predict China energy demand from 2012 to 2020 based on the combined model, the prediction results show that China energy demand increases at an average growth rate of 3.42% from 2012 to 2020, China energy demand in 2020 will be about 30% more than that in 2012. 