Power Transformer Fault Integrated Diagnosis Based on Improved PSO-BP Neural Networks and D-S Evidential Reasoning

This paper demonstrates the shortcomings of existing transformer fault diagnosis methods and introduces the basic idea of information fusion into this field. In accordance with the characteristics and requirements of power transformer fault integrated diagnosis, a new type of fault decision model based on the information fusion technique is proposed in this paper using an advanced PSO-BP (particle swarm optimization-back propagation) algorithm to train the neural network and applying the D-S evidential reasoning. By combining the dissolved gas-in-oil analysis (DGA) with the results of conventional electrical tests and on-site experience in operation and maintenance, the above model is capable of drawing credible diagnosis conclusions. The method proposed is proved effective with examples.