A variable-structure variable-order simulation paradigm for power electronic circuits

Solid-state power converters are used in a rapidly growing number of applications including variable-speed motor drives for hybrid electric vehicles and industrial applications, battery energy storage systems, and for interfacing renewable energy sources and controlling power flow in electric power systems. The desire for higher power densities and improved efficiencies necessitates the accurate prediction of switching transients and losses that, historically, have been categorized as conduction and switching losses. In the vast majority of analyses, the power semiconductors (diodes, transistors) are represented using simplified or empirical models. Conduction losses are calculated as the product of circuit-dependent currents and on-state voltage drops. Switching losses are estimated using approximate voltage-current waveforms with empirically derived turn-on and turn-off times. With recent increases in switching speeds, these approximations are no longer valid in many applications. Although it is possible to simulate power converters using physics-based models of power semiconductors based upon coupled drift, diffusion, continuity (CDDC) equations, such simulations are generally prohibitively slow. In this thesis, a variable-structure variable-order simulation paradigm is set forth in which the detailed CCDC-based models are used to calculate the switching transients and corresponding losses. As devices (e.g., diodes) become active or inactive, the structure and order of the simulation is dynamically changed without sacrificing accuracy. A time-step control algorithm is devised such that the overall simulation captures only the relevant transients. Finally, parallelization strategies are identified that can produce a 483 % improvement in simulation speed compared with a conventional solution of the CDDC equations