Research On Fault Prediction Of Traction Transformer Of EMU Based On Dissolved Gas In Oil

At the moment,High-speed rail of china has begun to transform from a large-scale construction period to a comprehensive operation and maintenance period.The traction transformer is one of the core equipment of the traction power supply system of the EMU,and its performance is directly related to the normal operation of the EMU.The failure prediction of the traction transformer of the EMU can not only realize the timely prediction and warning before the failure of the traction transformer and the determination of the fault type when the failure occurs,but also provide reasonable decision support for the formulation of the maintenance support plan for the traction transformer,and improve the safety of the EMU and reliability.Based on the analysis of the existing methods for predicting the concentration of dissolved gas in transformer oil,fault diagnosis and fault prediction,this paper uses dissolved gas analysis,big data analysis and intelligent algorithms to establish a fault prediction model for the traction transformer of an EMU.The specific research is as follows:(1)Prediction of dissolved gas concentration in transformer oilAiming at the problem of low prediction accuracy caused by long-term short-term memory network(LSTM)selecting parameters based on experience,this paper uses an improved artificial bee colony algorithm(ABC)to iteratively optimize the learning rate and the number of neurons in the hidden layer of LSTM,and establish an improvement ABC-LSTM model to predict the concentration of dissolved gas in transformer oil.First,the Levy mutation factor and adaptive search factor are added to the standard ABC algorithm to balance the global and local optimization capabilities and improve accuracy to obtain the improved ABC algorithm;then the improved ABC algorithm is used to improve the learning rate and hidden layer neural network in the LSTM model.The number of elements is iteratively optimized;Finally,an improved ABC-LSTM prediction model is constructed by using the number of hidden layer neurons and learning rate obtained by iterative optimization,so as to predict the dissolved gas concentration in transformer oil.The results of calculation examples show that the prediction method of dissolved gas concentration in transformer oil based on the improved ABC-LSTM model has higher prediction accuracy than other prediction methods.(2)Transformer fault diagnosisAnalysis of misjudgment cases of transformer fault diagnosis shows that there are many misjudgments of partial discharge,low-temperature overheating,low-energy arc discharge and overheating confusion in the traditional fault diagnosis methods based on the analysis of dissolved gas in oil.At the same time,the fault diagnosis is not considered.The difference in energy required for the malfunctioning gas.Improve the accuracy of transformer fault diagnosis,this paper takes into account the difference in energy required to produce different fault gases,and uses the respective weighting factors of the fault gases to appropriately weight them,and establishes an improved particle swarm(IPSO)with gas concentration as input optimize the transformer fault diagnosis model of BP neural network.By inputting the fault gas weighted as fault gas and appropriate energy,the influence of different gas energy levels on transformer fault diagnosis is studied.Based on the energy-weighted dissolved gas analysis,the IPSO-BP fault diagnosis model and the unweighted IPSO-BP fault diagnosis model are analyzed by examples and show that the former has better diagnostic capabilities.(3)Fault prediction of traction transformer of EMUCombine the established method for predicting dissolved gas concentration in transformer oil based on the improved ABC-LSTM model and the IPSO-BP transformer fault diagnosis model based on energy-weighted dissolved gas analysis to establish a traction transformer fault prediction model for EMU,and collect the collected The concentration of dissolved gas in the oil before the failure of the traction transformer is input into the established failure prediction model to verify the failure prediction performance of the model.It has been verified that the established EMU traction transformer failure prediction model can use the dissolved gas concentration in the traction transformer oil to predict the EMU traction transformer failure.