A new machine learning method-support vector machine(SVM)was introduced and some of its characteristics were discussed.After linearly scaling and principal components analyzing(PCA), the data that affect the mid-long-term electric power load were trained by the tools of Libsvm.Com- pared to GM(1,1),multi-variables regression,BP neural network and fuzzy clustering-ISODATA,the result of SVM is excellent.Based on the model presented the mid-long-term load forecasting can be im- proved.