Research On Parameter Identification For DC-link Capacitors In Traction Drive System

In order to realize the reliable operation of high-speed train,it is necessary to evaluate the health of DC-link capacitors in traction drive system and establish corresponding fault warning.This paper studies the parameter identification scheme of DC-link capacitors in traction drive system,and discusses the influence of sampling frequency,measurement error and noise level about parameter identification accuracy,which focuses on the identification effect and practicability of capacitor parameters in each scheme.It is necessary to reconstruct DC-link current because of immeasurability of DC-link current.In order to avoid the influence of switching state and load changing and obtain DC-link current accurately,capacitor parameter identification is selected in pre-charge stage.The capacitor parameter identification method based on least square method is studied firstly,which is applied to DC-link without resonant branch and with resonant branch respectively.The capacitor can be identified accuracy by using the classical least square method when DC-link without resonant branch,and can be improved to recursive form which avoids complex matrix calculation.However,this method can’t be applied to DC-link with the resonant branch.The reason for the inaccurate identification is further explained by mathematical derivation.In order to realize accurate capacitance identification in DC-link with resonant branch,the capacitor parameter identification method based on state observer is studied.The parameters of Luenberger observer and sliding mode observer are designed respectively,and the accuracy of identification in different factors are studied deeply.The simulation results show that the identification error exceeds the expected accuracy when noise is existed in the system,In order to improve the robustness under noise,a capacitor parameter identification algorithm based on Kalman filter is studied.Simulation results show that The capacitor can be identified accuracy by using this method when the system contains noise,which the physical meaning of filter parameters is clear,and the design process is relatively simple.In addition,this method can also be used for capacitor parameter identification of DC-link without resonant branch.Therefore,this method can be used for parameter identification of DC-link without resonant branch and with resonant branch.Finally,an 1.5MW real-time simulation platform and a 7.5k W experimental prototype are carried out to validate the above method.The results show that all the above parameter identification methods can identify the capacitance of DC-link without resonant branch accurately,while only the parameter identification method based on Kalman filter can identify the capacitance of DC-link with resonant branch accurately,which dentification error of capacitance is less than 2%.And is insensitive to the change of system parameters and measurement noise.In addition,extra hardware circuitry and injection of external signals,this method exhibits high accuracy and cost-effectiveness.