Thermal Analysis of Power Electronic Modules with Parametric Model Order Reduction

This study presents a new approach for conducting parametric analyses via exploring model parameters of a power electronics module (PEM) thermal model, integrating finite element analysis (FEA) with a parametric model order reduction (pMOR) method. A three-dimensional model has been utilised for thermal analyses using a pMOR method, parametrising the heat-source parameter, i.e., the internal heat generation (IHG) rate, and the convection parameter, i.e., the film coefficient of the coolant's channel surface, of the PEM. This approach allows for the exploration of various cyclic thermal loads observed in power cycle experiments. The ‘pMOR by matrix interpolation’ method has been employed for the global parametric reducedorder modelling study, and PRIMA, a Krylov subspace and Arnoldi process-based model order reduction (MOR) technique, for local model order reductions and the generalised trapezoidal rule (GTR) for numerical analysis procedure, have been utilised within the modelling procedure. The local models of the parametric reduced-order model (pROM) have a total degree of freedom (DOF) of just 20, in contrast to the local models of the parametric full-order model's (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{p F O M}$</tex>) total DOF of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{1 8 5, 4 8 9}$</tex>. The pROM solutions have been validated by comparing corresponding pFOM solutions and experimental data, which demonstrate that the proposed parametric reduced-order modelling approach provides valid solutions and significantly reduces computational requirements, e.g., time, by over 99 % for the case presented. The developed parametric reduced-order modelling approach for thermal analyses, addressing the parametric (global) analysis of power electronic components, has been implemented via MATLAB programming (.m) environments.