Energy consumption forecast is an essential component in making energy plan. In the light of the complexity and nonlinearity of energy consumption system, the gray forecast model and neural network model are respectively established by using the energy consumption historical data of certain province. Then their advantages and disadvantages are analyzed. Lastly, the method of optimal combination is applied in this paper in order to obtain accurate forecast model and forecast value. The forecast results of example show that the model can be regarded as effective tool of energy consumption forecast.