Partial discharge pattern recognition is an effective method to diagnose the insulation condition of high voltage electrical equipment. In order to improve the recognition accuracy of partial discharge, this paper presents a partial discharge recognition method which based on the statistical parameters and multi-classification SVM. In this paper, four typical kinds of transformer faults models are made in the laboratory, 27 statistical characteristic parameters of each partial discharge patterns are extracted. The binary classification of support vector machine is extended to multi classification, which the M-ary classification is applied to the support vector machine, thus, the computation of training and testing has greatly reduced. The test results show that the method is an effective and reliable method for partial discharge pattern recognition, which realize higher recognition rate and computing speed.