Research On Condition Monitoring Of JTC Compensation Capacitor Based On Wavelet Packet And Cluster Analysis

In recent years,with the development of railway technology,jointless track circuit is widely used in existing railway lines and high-speed railways because of its good transmission performance and high safety performance.Among them,the compensation capacitor,as an important part of the track circuit,weakens the influence caused by rail inductive reactance.The fault of compensation capacitor may cause "red light band" in the track circuit or "code dropping" in the locomotive,thus affecting the track circuit.For the compensation capacitor fault,at present,the detection method after compensation capacitors fauilure is widely used in China,which can not meet the transformation need of railway site from "fault repair" to "state repair".Therefore,based on the system structure and working principle of track circuit,through modeling and simulation analysis,this thesis proposes a method to realize the condition monitoring of compensation capacitor in jointless track circuit based on wavelet packet and cluster analysis.The main research contents are as follows:(1)Through analyzing the signal transmission process of ZPW-2000 A track circuit,the adjustment state and shunt state models of track circuit are established based on transmission line theory and circuit principle,so as to obtain the calculation expressions of adjustment state receiving voltage and shunt state shunt current of track circuit.(2)Based on the adjustment state receiving voltage and the minimum shunt current of shunt state of all disconnection fault types of compensation capacitors,the reliability truth table of track circuit is established.Then through calculating the importance coefficients of compensation capacitor at different positions in the reliability truth table,the positions of compensation capacitors that have a great impact on the track circuit are determined.The simulation results show that the second and third compensation capacitors near the receiving end have a great influence on the track circuit.This method can assist the field maintenance personnel to determine the maintenance priority of each compensation capacitor,and provide a basis for the key monitoring of compensation capacitor.(3)Taking the normalized shunt current curves when compensation capacitors are all normal and occur different faults as the research object,the fault features analysis method of compensation capacitor based on wavelet packet is proposed.Through the analysis of wavelet packet denoising effect,the optimal wavelet basis and decomposition layers are selected to decompose and reconstruct the normalized shunt current curves when the compensation capacitors are all normal,occur the line break fault and the capacitor value falling fault,so as to obtain the corresponding wavelet packet coefficient curves.Then,through comparing and analyzing these wavelet packet coefficient curves,the frequency bands with obvious fault features are select.Finally,the instantaneous energy of each compensation capacitor corresponding to the selected frequency bands is calculated.By comparing the instantaneous energy of each compensation capacitor when the compensation capacitors are all normal with the instantaneous energy of each compensation capacitor when the compensation capacitor occurs the line break fault and the capacitor value falling fault,the fault features of the compensation capacitor are extracted.The simulation results show that the wavelet packet can be used to analyze the fault features of the compensation capacitor,and the decrease percentage of the capacitor value of the compensation capacitor can roughly estimate,thus providing data support for the condition monitoring of the compensation capacitor.(4)Taking the wavelet packet energy spectrum of each frequency band corresponding to different capacitor value drop percentage of compensation capacitor as the research object,the optimal clustering number of the compensation capacitor state is determined based on the improved elbow method.At the same time,based on K-means clustering,hierarchical clustering and fuzzy C-means clustering,the capacitor value drop state of compensation capacitor is divided.By analyzing the clustering results,the appropriate division result is selected.The simulation results show that the compensation capacitor state can be divided into five capacitor value drop states and set five early warning signal colors accordingly,so as to provide a basis for the subsequent condition monitoring and early warning of the compensation capacitor.(5)Based on the fault feature analysis results of wavelet packet,the division results of final clustering and the early warning signal color corresponding to each state of compensation capacitor in track circuit,the design of compensation capacitor state monitoring and early warning process are completed.At the same time,the accuracy and feasibility of the algorithm are verified based on the field measured data and known monitoring results.The results show that the state monitoring and early warning of compensation capacitor can be realized by wavelet packet and clustering algorithm.To sum up,through combining the fault characteristics of compensation capacitor extracted by wavelet packet and the division results finally determined by clustering,the condition monitoring and early warning of compensation capacitor can be realized.At the same time,based on the field data,the accuracy and feasibility of algorithm are verified.