Condition Monitoring of DC-Link Capacitor in Traction Inverters

Abstract

Abstract With the increasing use of battery electric vehicles (BEVs) in the road transport sector for driving the decarbonization of industry poses additional challenge for reliability. The DC-link capacitor which links the power source, typically a battery, with the load - an electric machine, performs a critical function of reducing the voltage and current ripple induced mainly by the switching action of IGBTs in inverter module. Metallized polypropylene film(MPPF) capacitor are used in traction application which carries out its operation by providing an instaneous power by acting as a reservoir and limiting the charge and discharge cycle stress on the main power source. This continuous usage induces high stress mainly due to over voltage or surge current eventually leading to wearout of dielectric medium. These factors make capacitors the weakest link in the drive system which alone contributes to about 30% of all the failures. Reliability of these components is thus crucial for safe operation of the drive system. Condition monitoring (CM) helps to address these challenges by providing a means of continuous monitoring of some of the health indicators like capacitance, equivalent series resistance (ESR) and dissipation factor (DF). AsMPPF capacitors have relatively small ESR and DF values, the thesis will be focused on only utilizing capacitance as the estimation parameter. Although many literature study have proposed the offline monitoring technique, it requries a direct access of capacitor terminals which in traction application demands the converter module to be dismantled. As this is a major inconvenience, only online CM methods are considered for the comprehensive study in this thesis. The DC-link capacitor has a wide range of application ranging from adjustable speed drives (ASD) to DC/DC convereters and power factor correction (PFC), photovoltaic (PV) grid. As most of the methods presented in literature depend on certain specific parameter or hardware present in the topology, only ASD systems were examined as the topology closely relates to that of traction EVs. Whilstmany of the approaches proposed in ASD system are online, they still rely on additional hardware like capacitor current sensor or measurement circuit to accurately estimate the health of the capacitor. Applying these techniques directly to traction application will induce additional stray inductance and cost overhead. Two methods were found suitable, one of which is completely online that leverages the zero state working of inverters and estimate the capacitance by change in voltage to current accumulated during which the capacitor is charged. The current is the same as the battery current and can be accurately measured with an existing sensor. Another method is a quasi-online approach is implemented when the motor is stationary and the capacitor is already charged upto some level. A unipolar pulse widthmodulation strategy is utilized to partially discharge the capacitor through any of the two stator windings and capacitance is estimated by measuring the change in DC-link voltage and current discharged through windings. Both the methods were modelled and evaulated in PLECS/Simulink environment and results found that the discharge method was more accurate at < 0.05% error in estimation than its counterpart at < 0.4%.