Fault diagnosis method of transformer based on cloud theory and entropy weight

Abstract

In this paper, we propose a method to identify potential faults in power transformers. Firstly, the cloud distribution model of gases under different fault types are established respectively, which are the basis for building the cloud knowledge base. Secondly, the membership grades between test sample and different fault types can be calculated by weighting the gases using entropy weight method. Finally, the effectiveness and the superior data learning ability of this method can be verified by comparing the diagnostic accuracy with three-ratio method introduced by IEEE and the existing cloud method under different amount of samples. The result of this method can provide effective reference for the maintenance planning of transformer.