Automated Extraction of Low-Order Thermal Model With Controllable Error Bounds for SiC MOSFET Power Modules
Cameron Entzminger, Wei Qiao, Liyan Qu, J.L. Hudgins
Abstract
This article explores modeling the thermal process of a Silicon carbide (SiC) metal-oxide-semiconductor field-effect transistor ( <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">mosfet</small> ) power module through a finite element analysis (FEA) based full-order thermal model (FOM) and then reducing the order of the FEA thermal model using a hybrid model order reduction (MOR) method. This hybrid MOR method takes advantage of Krylov subspace projection method's ability to be applied to higher order systems and the controllable error bound of the Hankel singular value based MOR method using a pure mathematical approximation process without any heuristic assumption. The resulting reduced-order thermal model has a significantly reduced computation cost compared to the FEA model while preserving the accuracy of the FEA model with controllable error bounds. The proposed method is applied to a SiC <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">mosfet</small> power module to generate reduced-order thermal models, which are validated by computer simulation with respect to the FEA thermal model and are compared with a state-of-the-art three-dimensional thermal equivalent circuit model and the reduced-order thermal models generated by using a Krylov subspace projection method. Experimental validation of the thermal models with respect to the measured SiC <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">mosfet</small> junction temperature is provided. The results demonstrate higher accuracy and controllable error bound of the proposed method.