Batteries are widely applied to the energy storage and power supply in portable electronics, transportation, power systems, communication networks, and so forth. They are particularly demanded in the emerging technologies of vehicle electrification and renewable energy integration for a green and sustainable society. To meet various voltage, power, and energy requirements in large-scale applications, multiple battery cells have to be connected in series and/or parallel. While battery technology has advanced significantly during the past decade, existing battery management systems (BMSs) mainly focus on the state monitoring and control of battery systems packed in fixed configurations. In fixed configurations, though, battery system performance is, in principle, limited by the weakest cells, which can leave large parts severely underutilized.

This paper deals with the detection of single battery or inverter switch faults during operation of a common and an active three-level neutral-point-clamped (NPC) inverter with a connected neutral point. Here, a main focus lies on the detection and localization of open circuit faults of the inverter’s switches. Therefore, a fault detection algorithm, using a current estimator, and two fault localization algorithms, a pulse pattern injection principle and an online adaption of the space vector modulation, are investigated and verified through simulations and experiments. Also investigated is how the power-train can be operated under fault condition, so that the vehicle can drive with a limited maximum power using an adapted space vector modulation, referred to as “limp home“ mode, to the next service station. It is shown, that an active neutral-point-clamped inverter can cope with any single short or open circuit fault of the inverter’s switches without bringing the vehicle to standstill, whereas a generic neutral-point-clamped inverter loses controllability, if an open circuit fault at an inner switch occurs. Furthermore, both inverter types are able to be operated just with half of the DC link voltage in case of a failure in one part of the battery.

In this paper, two types of filter capacitors of varying capacity were connected to the battery packs of a cascaded H-Bridge single-star multilevel vehicle traction inverter, and their influence on the battery losses has been analyzed. The battery and capacitor simulation models used are experimentally verified in a downscaled system. Different capacitor configurations were simulated for four-drive cycle scenarios to determine the potentials for the mitigation of current pulse stresses and battery loss reduction with respect to the added weight. By adding capacitors corresponding to a weight of 4% of the initial battery storage, the peak current is reduced by 5%-20%, depending on the operating point from dc to a few kilohertz, and the battery losses are reduced by 10%. In comparison, it is demonstrated that adding supercapacitors is more beneficial for lower output frequencies, while adding electrolytic capacitors is better for higher output frequencies. Furthermore, the low-order voltage harmonics of the dc-rails between the converter and battery were reduced by 10%-30% for frequencies above 9 kHz, which decreases the potential of electromagnetic disturbances. In addition, during cold battery temperatures, when it is very important to avoid heavy cyclings, the loss reduction using the capacitors was 2.5 times larger than for nominal temperature.

The reverse conduction capability of MOSFETs is beneficial for the efficiency of a three-phase inverter. In this paper analytical expressions in closed form are presented which allow to quickly evaluate the conduction losses, considering the effect of the reverse conduction and blanking time for both sinusoidal PWM operation with and without third harmonic injection. The losses of a three-phase SiC MOSFET inverter suitable for traction applications are estimated with the proposed method and show good agreement of about 98.5 % with measurements, performed with a calorimetric setup.

This paper deals with the modeling and the parameterization of high-power, LiFePO4 battery packs used in cascaded H-bridge multilevel propulsion inverters. Since the battery packs are intermittently conducting the motor currents, the battery cells are stressed with a dynamic current containing a substantial amount of low-order harmonic components up to a couple of kHz, which is a major difference in comparison to a traditional two-level inverter drive. Different models, such as pure resistive and dynamic RC-networks, are considered to model the energy losses for different operating points and driving cycles. Using a small-scale setup, the models' parameters are extracted using both a low-frequency, pulsed current and an electrochemical impedance spectroscopy sweep. The models are compared against measurements conducted on the small-scale setup at different operating points. Additionally, a drive cycle loss comparison is simulated. The simple resistive model overestimates the losses by about 20% and is, thus, not suitable. The dynamic three-time constant model, parameterized by a pulsed current, complies with the measurements for all analyzed operating points, especially at low speed, with a maximum deviation of 3.8%. Extracting the parameters using an electrochemical impedance spectroscopy seems suitable for higher speeds, though the losses for the chosen operating points are underestimated by 1.5% to 7.9%.