Global fault diagnosis method of traction transformer based on Improved Fuzzy Cellular Neural Network

Abstract

For compensating the deficiency of dissolved gases analysis (DGA) method of traction transformer fault diagnosis, a global fault diagnosis method of traction transformer based on improved fuzzy cellular neural network (IFCNN) is introduced in model building mode. Global fault diagnosis model is comprised of input space, fault diagnosis rule and output space. Input space is fault symptom set and output space is fault type set. As to input space, fault symptom is enriched by increasing water in oil, key device resistance and electric current besides using DGA analysis content. Fault diagnosis rule is depended on fuzzy integrated judging method and the combination between DGA and IFCNN fault diagnosis model designed in this paper. Output space is diagnosed fault types through defuzzification processing of diagnosis result. And this paper uses experiment to test fault diagnosis precision. The experiment result indicates that global fault diagnosis method has better practicable performance and high precision on analyzing causal relation of different fault, ascertains valid input and fault characteristic types, avoided localization of traction transformer fault diagnosis by DGA, and collectivity precision can reach 90.91%.