Study On Fault Location Of High-Speed Railway Traction Power Supply System

As a safe,fast and clean means of transportation,high-speed railway plays an important role in China’s transportation industry and plays a positive role in promoting the development of many fields and industries.Once high-speed rail breaks down,power cuts and shutdown will have a major impact on the economy and people’s livelihood.High-speed railway traction power supply system has frequent failures because of its complex structure,many parts and working environment.Therefore,the traction power supply system is in urgent need of accurate fault location system,in order to determine the fault location when the permanent fault occurs,quick repair;In case of non-permanent fault,feedback the fault information,arrange inspection in time,promote the safe and reliable operation of traction power supply system.AT present,the full-parallel AT power supply mode is widely used in China’s high-speed railway.With the efforts of many scholars,the research on fault location system of this power supply mode has made rich achievements,but there is still a lot of work to be done.In this paper,the fault location accuracy of full-parallel AT power supply mode is not high and the fault type cannot be accurately judged.The main work is as follows:1 Multi-scale decomposition of the fault electrical wave modulus signal is performed by using B-spline wavelet.The influence of different scale decomposition signals on the accuracy of fault location is analyzed,and the signal under scale 1decomposition is selected as the maximum of the mode to extract the signal.Then,the d-type traveling wave location method is adopted to realize the fault location of the traction network and improve the accuracy of fault location.2 In view of the low precision of the current single-ended fault location method,the single-ended fault current modulus signal is extracted,the data samples under different fault types and different fault distances are made,and the BP neural network is trained.During the training process,the influence of the algorithm,the proportion of data set and the number of nodes in the hidden layer on the network performance is analyzed in detail,and the optimal combination is obtained.The traction network fault location is realized through the network,and the fault location accuracy of the single-end method is improved.3 Support vector machine(SVM)neural network is used to realize the accurate discrimination of fault types.