Accelerated Simulation Across Multiple Resolutions for Power Electronic Circuits

An automated multi-resolution simulation (MRS) framework based on model order reduction (MOR) for accelerating simulation of high-fidelity power electronic circuits (PECs) is presented in this paper. The wide span of eigenvalues of PECs makes their simulation prohibitively slow. This problem is tackled using the proposed approach. Singular perturbation approximation (SPA) is used to derive the reduced order models. The transient contribution of the non-dominant eigenvalues is ignored, retaining the steady-state contribution of all the eigenvalues. Reduced models of different orders or resolutions, each corresponding to a particular level of accuracy are derived. Then, the simulation engine is set such that it dynamically changes the resolution of the simulation. Initially, the original high-resolution model is simulated. As the transients of the fast-states die down, the simulation shifts to a lower resolution. The approach has an advantage that instead of simulating the original system over the entire time-range, a combination of the full and reduced models are simulated. The high-fidelity model is simulated for only a fraction of the simulation cycle. This accelerates the simulation considerably and at the same time ensures that the complete response of the system is obtained with negligible approximation error.