It is easy for transformer winding to generate deformation after short-circuit fault, and traditional frequency response analysis or short circuit impedance analysis is one-sided during the process of the winding deformation detection. This paper proposes a diagnosis method for transformer winding deformation fault based on fault information fusion and Cuckoo Search Optimization based Support Vector Machine (CS-SVM). It integrates the inspection data related to winding deformation into SVM input sample and puts it in the CS-SVM trained according to artificial experience for diagnosis. According to Matlab simulation result, this method has preferable anti-interference property, which can diagnose the state of the transformer winding better. Finally, corresponding verification is conducted by combining specific example of certain transformer.