For diagnosing deformation fault and latent loosing status of winding, this study proposed a state diagnosis method of transformer winding deformation based on fusing vibration and reactance parameters. The wavelet energy-matrix norm transformation method and reactance identification method were proposed to extract vibration and reactance parameters. The multi-information collection system of transformer winding deformation state was designed. Aiming at the S-11-M-500/35 transformer, short circuit current impact experiment was made and three deformation fault windings were designed. Status database of multi-information for normal, loosing, deformation and specific fault types of winding were established. State diagnosis experiment on normal winding, loosing winding after short circuit current impact and deformation fault winding was made. Per unit value-space vector transformation method was proposed to fuse vibration and reactance parameters, fusing eigenvector was established, and the winding deformation state was diagnosed by extracting fusing feature. On the basis of the above research, regarding deformation fault state winding, fault types were diagnosed based on the relevance vector machine. Results indicated that latent loosing, deformation fault and three deformation fault types of winding can be diagnosed accurately. Fusing diagnosis method proposed in this study was superior to the single vibration or reactance parameter diagnosis method.