Research On Typical Fault Recognition Algorithm For Traction Network Of High Speed Railway

Traction network is an important part of traction power supply system of high-speed railway,which covers a wide range of areas and operates in a complex environment.Faults can be easily caused by bad weather,external force damage and other causes.After the tripping of the traction network,it is of great significance to identify the fault types quickly according to the fault signal for fast maintenance in order to reduce the loss of the fault,improve the operation and maintenance level of the traction network and realize the intelligent management of high-speed railway.This paper focuses on the typical fault types of high-speed railway and carries out the following work: Combining with Yamenkou-Beijing field line structure and the traction network system simulation model are built where the traction substation,traction network line and electric locomotive module are established,and through simulating the two states of short circuit and open circuit of traction network,comparing with the recorded datas,the correctness of the model is verified;According to the comparative analysis of linear resistance and non-linear resistance simulation,a more practical method of non-linear resistance simulation analysis has been proposed;Aiming at the simulation analysis of typical faults of traction network in two states of no-load and locomotive load by combining time-domain analysis with frequency-domain analysis.According to the simulation comparison of the typical faults in different power supply sections,it is concluded that the main characteristics of faults are not affected by the line structure of traction network;Aiming at four kinds of non-lightning stroke faults with similar frequency characteristics,a feature extraction algorithm is designed.By analyzing the two states of no-load and locomotive load,four kinds of faults after power spectrum analysis are extracted by using wavelet packet energy spectrum;The simulation experiment platform and field experiment platform are built to collect the original data of kinds of non-lightning faults,and the feature vectors extracted by the algorithm are used as input to classify and recognize the fault types.The overall correct rate is up to 85%,which basically meets the classification and recognition of different forms of faults.