Research On Fault Diagnosis And Prediction Method Of Traction Transformer Based On Bayesian Network

As the core equipment of traction power supply system,traction transformer is essential to ensure its safe and reliable operation.In the field analysis of the traction transformer status,there are problems such as fuzzy index relationships,differences in index testing,and missing status data,which greatly reduces the accuracy of the traction transformer status analysis results.In order to solve the above problems,this paper studies the state evaluation method based on variable weight coefficient and Bayesian network,the fault diagnosis method based on association rules and Bayesian network,and the fault prediction method based on improved maximum Lyapunov exponent and Bayesian network.In order to solve the problem of fuzzy relationship between state indexes in traction transformer condition evaluation based on expert experience,the method of combining variable weight coefficient with Bayesian network is studied.The variable weight coefficient analysis method is used to determine the weight coefficient of each index of traction transformer,so that the index weight coefficient can be changed according to the change of index score,and the state level can be clearly divided according to the score.At the same time,the Bayesian network method is used to comprehensively consider the historical state and the current state,and the evaluation results of the current state are corrected through the historical state,so that the evaluation results are closer to the actual operation state of the traction transformer,and the reliability of the evaluation results is improved.Aiming at the problem of the difference of state index caused by the different test items of traction transformer in different regions,the combination of association rules and Bayesian network method is used for fault diagnosis of traction transformer.The association rules method is used to calculate the association relationship among index set,fault set and condition set.The Bayesian diagnosis network is constructed by the relationship of each state quantity.In the case of incomplete symptom types,Bayesian network is used for fault diagnosis to effectively solve the problem of different input data types caused by different field tests.In view of the adaptability of the current traction transformer fault prediction methods to the missing values of oil color time series data,this paper uses Newton interpolation method to process the missing values and fill in the missing date data to obtain the complete time series data.The improved maximum Lyapunov exponent prediction method is used to get the prediction value of the next time,and the prediction value is corrected by error.Finally,Bayesian network is used to predict the fault of the prediction data,and the predicted fault type is obtained.In this paper,the problems of fuzzy relationship between indexes,difference of state indexes and incompleteness of state data in the process of condition evaluation,fault diagnosis and fault prediction of traction transformer are considered.This paper studies the corresponding state analysis method,comprehensively considers the field situation,and improves the accuracy of the analysis results,which can guide the maintenance work of the field operation and maintenance personnel,so as to improve the maintenance efficiency and save the maintenance cost.