Research And Application Of Transformer Fault Diagnosis Method Based On PSO-BP Neural Network

Electricity is an unavoidable topic in people’s daily life.There are many factors affecting the daily power supply,but the transformer plays an important role in these factors.Because the design of the transformer itself is more complex,the manufacturing process is also relatively complicated,and most of the transformers are laid in outdoor conditions,so the problem of transformer fault has always been the focus of the power supply department.The traditional inspection and defect elimination methods are difficult to ensure that the transformer does not fail.Therefore,it is urgent to study the fault diagnosis and operation monitoring of the transformer,so as to find the faults and hidden dangers in the operation process of the transformer as soon as possible.Therefore,a transformer fault diagnosis method based on PSO-BP neural network is proposed.（1）The types of transformer faults and the changes of characteristic gas in oil are analyzed in detail.On this basis,combined with domestic and foreign references and common transformer fault diagnosis methods,the traditional transformer fault diagnosis methods and intelligent diagnosis methods are compared and analyzed in detail.（2）This paper introduces the basic mathematical basis and principle of BP neural network in detail,and integrates it into the detection of transformer fault,determines the input and output of BP neural network,selects the initial weight of network structure through the topology of transformer fault diagnosis network,and obtains the output result of neural network through simulation experiment.（3）In order to solve the problem that the extreme point of BP neural network is not unique due to the nonlinear model structure,and the traditional parameter optimization algorithm can not quickly find the optimal parameter value point,based on the introduction of the basic principle of particle swarm optimization algorithm,the gradient descent method in BP neural network back propagation is replaced by PSO algorithm,A transformer fault diagnosis method based on PSO-BP is proposed.（4）The accuracy of the proposed method is verified by training the transformer fault diagnosis model based on PSO-BP neural network by using the dissolved characteristic gas and fault data collected during the operation of the transformer.