Study On Intelligent Fault Diagnosis Of Traction Transformer Based On Dissolved Gases

Traction transformer is an important part of the train vehicle,which is related to the safety of the train operation.Due to the problems of inaccurate fault diagnosis and overmaintenance of vehicles existing in the current maintenance system,China’s maintenance system is moving towards the goal of more efficient state repair.On this road,intelligent algorithm is playing a positive role in promoting the combination of traditional diagnosis methods and intelligent algorithm,can make the train traction transformer fault diagnosis accuracy significantly improved.The author combined the traditional dissolved gas analysis method with many intelligent algorithms to carry out the work.Firstly,the gases commonly used in the traditional fault diagnosis of dissolved gases in oil are discussed,and five dissolved gases with better fault performance are selected as the characteristic input gases.Then the intelligent algorithm fault diagnosis model is established,and the data are divided into training set and test set,and the model is trained and tested.Finally,the advantages and disadvantages of the analysis method of dissolved gas in oil,intelligent algorithm,fault diagnosis results after optimization of genetic algorithm,shallow machine learning algorithm and deep learning are compared.In this paper,the data are normalized to make the results more accurate.Then,two machine learning algorithms,support vector machine model and extreme learning machine,are used to establish a model based on the five characteristic gases of the ratio of three as the input quantity to make fault diagnosis for the train traction transformer.In order to improve the effect of fault diagnosis,the key parameters of support vector machine(SVM)optimized by genetic algorithm are used to build a model for fault diagnosis.Finally,in order to further explore,the deep belief network in the deep learning algorithm is used to construct its structure and parameters to establish a fault diagnosis model based on the deep belief network,and the established model is used to make fault diagnosis of the train traction transformer.Through comprehensive analysis of the above algorithms,it is found that the deep belief network has the best comprehensive effect on transformer fault diagnosis and the strongest data mining ability.The application of deep belief network to the fault diagnosis of train traction transformer can effectively improve the accuracy of fault diagnosis and make an attempt for the goal of train condition repair.