Multiparameter Model Order Reduction for Thermal Modeling of Power Electronics

Abstract

In this article, a multiparameter model order reduction (MOR) is applied to the thermal modeling and simulation of power electronics modules and air cooling systems. Although widely employed, simulation tools based on finite elements (FE) or finite difference methods result in computationally expensive models that hamper the analysis in studies where one or more parameters are changed. MOR techniques can be applied to reduce the computational complexity. However, standard reduction techniques cannot easily consider parameters variability and need to be reapplied for each parameter value. This article proposes a multiparameter order reduction technique that can significantly improve the thermal simulation efficiency without having a significant impact on the prediction accuracy. The method is applied to multichip SiC power module mounted on a forced air-cooled finned heatsink with variable air flow.