IEEE Transactions on Power Electronics. Dec2009, Vol. 24 Issue 12, p2833-2846. 14p.
FINITE element method, ELECTRIC current converters, EIGENVALUES, REDUCED-order models, and LEAST squares
The virtual prototyping of power electronic converters requires electrothermal models with various abstraction levels and easy identification. Numerous methods for the construction of compact thermal models have been presented in this paper. Few of them propose state-space models, where the model order can be controlled according to the necessity of the virtual prototyping analyses. Moreover, the model reduction methods require the experience of the engineer and previous calibration. Diffusive representation (DR) is proposed here as an original and efficient method to build compact thermal models as state-space models. The model reduction is obtained through the model parameter identification and/or the time horizon of the measurement data provided for the identification. Instead of eigenvalue elimination, the method enables to specify adequately inside the model the frequency domain wished for the virtual analysis at hand. The proposed method is particularly dedicated to the system optimization phases. Experimental and simulation results are in good agreement. The advantages and limitations of the DR are discussed in comparison to published methods. [ABSTRACT FROM AUTHOR]
Evans, Paul L., Castellazzi, Alberto, and Johnson, C. Mark
IEEE Transactions on Power Electronics. Oct2013, Vol. 28 Issue 10, p4791-4802. 12p.
SYSTEMS design, APPROXIMATION theory, POWER electronics, FINITE difference method, EIGENVALUES, and SPARSE matrices
Virtual prototyping of power electronic modules aims to allow rapid evaluation of potential designs without the need to resort to building and testing physical prototypes. A key requirement for this process is the ability to quickly generate small, compact models describing the thermal performance of a potential design and this study presents a novel approach for this model generation process. The approach starts with a finite-difference mesh of the proposed design and applies a fast sparse matrix solver (GMRES) to determine the steady-state response to a particular power input. In doing this, approximations for the eigenvalues of the system can also be obtained from the same algorithm. It is shown that these two results can then be used to create small compact models describing the dynamic thermal properties within the design. The method is validated against an analytical solution for 1-D heat conduction and against experimental results for a simple power module. This process can be automated and it is shown that compact models can be generated in around 12 s per power input from the finite-difference mesh of the power module containing 14 623 nodes on a standard desktop PC. [ABSTRACT FROM AUTHOR]