Research On Condition Monitoring Of DC-link Capacitor For BEV Inverter

DC-link capacitor is one of the key devices in the inverters.It can balance the instantaneous power difference between the input source and the output load in the converter,so as to minimize the DC-link voltage change and filter the ripple.In electric drive system of electric vehicle(EV),DC-link capacitor bears more severe working conditions and more complex and changeable stress.The performance degradation of DC-link capacitors will apply more power load on the input of the inverter easily,even destroy the reliability and safety of the system.With the rapid development of EV,higher demands are put forward on the reliability of their matching electric driving systems and components.The condition monitoring of DC-link capacitor is helpful to optimize the control strategy of electric vehicle control system,reduce the electrothermal stress of power devices and electric driving system,and improve the stability of EV control system and even the reliability of EV operation.Based on the double closed-loop electric drive system of battery electric vehicle(BEV),online condition monitoring methods of DC-link capacitor under different working conditions are studied in this paper.Firstly,the simplified equivalent circuit of metalized film capacitor(MFC)is established,and the frequency characteristic test of MFC parameters is carried out.The failure modes of MFC under working conditions are analyzed,and the monitoring parameters of capacitor condition monitoring are determined.Based on the physical structure of MFC,the thermal characteristic is analyzed,the thermal distribution law is studied,and the thermal stability test is carried out to verify the theoretical results.Secondly,based on the double closed-loop electric driving system model controlled by SVPWM which is built in MATLAB,a current reconstruction method is designed to monitor DC-link capacitor.The ripple current can be obtained by threephase output current,and the theoretical results are verified by simulation.Based on genetic factor least square method,a DC-link capacitor condition-monitoring scheme is designed,and simulation results show that the least square method based on genetic factor is better than the recursive least square method.Thirdly,aiming at the condition monitoring method of capacitor based on BP neural network,the BP neural network optimization algorithm based on genetic algorithm(GABP)and the BP neural network optimization algorithm based on particle swarm optimization(PSOBP)are presented.The effects of the optimization algorithms are compared by simulation.The test scheme of on-line monitoring of DC-link capacitor based on working condition is designed,and the experimental results show that the PSOBP performs better on online condition monitoring of DC-link capacitor,for it can obtain more accurate results.Finally,the condition monitoring of DC-Link capacitors under open-circuit faults of inverters is carried out.The common open-circuit faults of inverters and the corresponding three-phase current characteristics are analyzed.The variation characteristics of ripple current of DC-link capacitor in time and frequency domains are analyzed under various fault modes.The correctness of the theory is verified by MATLAB simulation and result show that the ripple current is mostly affected by the simultaneous open fault of power transistors on the same side of the bridge arm in time domain and by the simultaneous opening of two power transistors on the same arm in frequency domain.Besides,the maximum monitoring error of capacitor is less than 3%.