Singular perturbation‐based model reduction of power electronic circuits

Abstract

Here, the authors present a model order reduction (MOR) framework based on singular perturbation approximation to accelerate the simulation of high-fidelity power electronic converters. The problem of slow simulation speeds caused due to the wide span of the eigenvalues is mitigated by implementing the proposed framework. The dynamics of the original stiff system is approximated by neglecting the transient contribution of the non-dominant eigenvalues and retaining the steady-state contribution of all the eigenvalues of the system. The model reduction problem has been reformulated to fit the switched nature of these circuits. An error bound for the approximation method has been derived. The method is demonstrated on a DC–DC boost converter and a Class-E amplifier. Significant improvement in speed and reduction in the size of the solution arrays is achieved. It is seen that the reduced-order models are able to replicate the response of the original models and the approximation error is within acceptable limits.