In order to solve the problem of correctly identifying fault classes in transformer fault diagnosis and improve the accuracy of the faule diagnosis, a novel fault diagnosis method of the transformer model based on fuzzy support vector machine(FSVM) in this paper. In this method, transformer faults are firstly divided into two types , and the gird search method based on cross-validation is chosen to determine the optimized parameters of the FSVM model .The membership value of the FSVM is obtained by fuzzy C-means clustering alaorithm. The experiment shows that the proposed method is more effective and accurate than the method of IEC and normal SVM, and this method is proper in fault diagnosis of transformer.