The three-ratio method has been used widely in fault diagnosis of transformer

however its codes are discontinuous. Noting different sample has different effect in the fuzzy C-means（FCM）clustering algorithm

each sample is endowed with a suitable weight to reflect its impact on clustering results. We apply the weighted FCM clustering algorithm to the three-ratio method to diagnose transformer fault. Moreover

at the beginning of clustering

we set a more appropriate initial membership matrix to accord with the actual situation of transformer fault diagnosis. Experimental results show that the diagnosis method of transformer fault based on the weighted FCM clustering algorithm is effective and feasible

and it achieves faster convergence speed and the fault cluster center closer to the actual one

compared with the diagnosis method based on FCM clustering algorithm.