The problem with regard to the uncertainty and the relationship between the components exists in the model-based diagnosis method. Aiming at this issue, based on the idea of complementary integration, this paper proposes a method of mode-based diagnosis combining with fuzzy Petri Nets to diagnose traction transformer. For the external faults of the traction transformer, the model-based method is directly implemented to build a two-layer model based on the structure and function using voltage and current as variables. The minimal conflict set candidates and the minimal conflict sets are introduced by offline search and on-line recognition to realize the diagnosis. For the internal faults of the traction transformer, the model-based method is implemented to obtain the fault component localization and the fuzzy Petri Nets is used to realize the regional modeling and diagnostic reasoning. By diagnosing the three-phase V/x connected traction transformer faults of Wuhan-Guangzhou High-Speed Railway, the feasibility and effectiveness of the proposed method are verified. This work is supported by National Natural Science Foundation of China (No. U1134205, No. U1434203, and No. 51377136).