Abstract: This study analyzes the advantages of Support Vector Regression (SVR) in the prediction of energy demand, decides the set of input vectors and output vectors and then establishes the model of prediction of Energy demand by SVR based on Matlab technology Modeling and Simulation of Energy demand from 1985 to 2009. At last, we apply this method to predict the demand for energy of china in 2010 and 2020. The article drew following conclusions: On the one hand, with the development of economy of china, the demand of energy will gradually increase from 31.553 million tons in 2010 to 45.30 million tons in 2020, with an annual increase of about 2.39 %. On the other hand, the SVR better than bp neural network about forecast accuracy, the long existing problem with the small sample(4non-linear and pattern recognition of energy system will be soon solved.