Fault diagnosis of transformer insulation based on compensated fuzzy neural network

Abstract

This paper introduces a kind of compensated fuzzy neural network based on fusion fuzzy theory and neural network technology. The compensated fuzzy neural network have fleet self-study algorithm and can perform compensated fuzzy reasoning. This method overcomes the critical value criterion defection problem that exists in traditional dissolved gas analysis. The method improves fault recognition capability by conversion fuzzy semantic to ration denotation applying features air diagnosis method. The method can resolve the transformer insulation's fuzzy phenomena. The method realizes fuzzy disposal of transformer fault diagnosis of feature gas by applying fuzzy neural network in the transformer insulation diagnosis knowledge base. The method increases the accuracy of the diagnosis and maneuverability by actual computation.