Abstract: As to the problem that the accuracy of medium and long-term electricity demand forecasting is affected by such factors as the singleness of traditional method and the weak inner link between different power prediction levels, a new medium and long-term electricity demand forecasting method is proposed based on industry classification. Firstly, electricity industry classification principles and methods are designed. Then, an appropriate optimized combination prediction model is built to forecast annual and quarterly electricity demand of the predicted district and its electricity consuming industries based on eight basic complementary predicting methods. Finally, a two-dimension and two-level coordinated model is designed based on multilevel forecasting and coordinated principle to correct the predicted value, to improve the predicting accuracy, and to obtain future annual and quarterly electricity demand of whole district and each industry. The actual data is used to test the effectiveness of the proposed method.