Michele Hirsch

Real-time thermal monitoring of power semiconductors in power electronics using linear parameter-varying models for variable coolant flow situations

Real-time thermal monitoring of power-semiconductor temperatures is an important task to establish thermal protection within a water-cooled power-electronic inverter unit. A model-based method is presented, which estimates the junction temperatures. Via feedback of measured temperature signals, the observer system is also able to estimate the coolant temperature. However, if the module is cooled by variable coolant flow rates, the heat transfer coefficient from heat sink to coolant changes. Consequently, the estimated coolant temperature of the observer system based on a static thermal model becomes inaccurate. This paper proposes a method to adapt the thermal model dependent on the current coolant flow rate using a parameter-varying system design.

Real-time coolant temperature monitoring in power electronics using linear parameter-varying models for variable coolant flow situations

In power converters of (hybrid) electric vehicles, the power losses within the semiconductors (diodes and IGBTs) lead to strong self-heating. Therefore, methods for online monitoring of the relevant temperatures are a very important topic regarding thermal self-protection. In order to reduce the number of sensors (and costs) it can be advantageous to use fewer sensors than the number of relevant temperatures and to use a model-based method to reconstruct/estimate the non-measureable temperatures. However, the accuracy of such models in water cooled systems is affected if the coolant flow rate deviates from nominal (constant) flow rate. This paper presents a method to estimate the junction temperatures of the semiconductors and the coolant inlet and outlet temperature of a 3 phase voltage inverter for automotive usage. It proposes a linear parameter-varying model to cope with variable coolant flow situations. Eventually, the method is verified by measurements.

Enhanced online thermal modeling for power electronic temperatures in (hybrid) electric vehicles

In electric and hybrid electric vehicles, the power losses of the semiconductors within power modules lead to strong self-heating. The resulting thermal stresses are major factors regarding critical overload and lifetime reliability. The relevant temperatures and temperature profiles are important design parameters. Additionally, it is necessary to monitor these temperatures during operation of the vehicle. However, if the number of relevant temperatures is larger than the number of sensors or the point of interest is not accessible for sensor placement, it is advantageous to use a model-based method to estimate these temperatures and use the given sensor information to correct the model. A so-called feedback observer can be used to reconstruct the non-measureable temperature. The capability of such an observer is based on the size and structure of the implemented model. In this paper multiple models and model structures are presented, which can be used to meet different requirements: Asymmetrical load distribution within the modules; coolant inlet and outlet estimation and/or the adaption of those models for the situation if the coolant flow rate of the inverter is variable. However, every extension of the model results in an increase of complexity and therefore in an increase of computational effort. Hence, to model as compact as possible is another requirement that has to be met. The result is a systematic approach which allows to combine those modeling methods within a modular construction system for thermal modeling of power converters. The presented models are compared regarding capability and computational effort.