In order to realize online condition monitoring of transformer by vibration response method

this paperproposes one kind of transformer fault diagnosis method based on the improved agglomerative hierarchical cluster-ing algorithm. First of all

using ensemble empirical mode decomposition（EEMD）method to get vibration signalcharacteristic values of transformer. Then

agglomerative hierarchical clustering algorithm is used to classify vibra-tion signal characteristic values of test transformers and get 4 kinds of characteristic values

including the normalstate of transformer

winding axial deformation

winding radial deformation and iron core fault state. Meanwhile

toovercome the shortcoming of hierarchical clustering algorithm

which has a large amount of computation

improve-ments are made to speed on transformer classification. Finally

by testing practical examples

the experimental re-sults prove that

this method can recognize transformer’s state immediately and effectively

which can realize on-line monitoring and fault diagnosis for transformer winding and core.