Optimization-Based Estimation and Model Predictive Control for High Performance, Low Cost Software-Defined Power Electronics

Abstract

A combined optimization-based estimation (OBE)-model predictive control (MPC) technique is developed to improve the dynamic performance of a reconfigurable LC-based power module with lower cost and less noise/oscillation. The developed OBE-MPC technique and the corresponding power module are based on a software-defined power electronics concept that can be reconstructed as different topologies and applied to various load/source applications, e.g., dc/dc converters, dc/ac single/three-phase grid-connected inverters, and ac/ac motor traction inverter to improve the energy conversion performance. The software-defined power electronics are designed in a generalized way by manipulating a different number of OBE-MPC power modules with holistic high-level control functions for wide applications. Symmetrically mirrored to the MPC, the OBE is configured as a constrained finite time optimal estimation (CFTOE) problem to solve the quadratic cost function based on the past sampling information. With the designed OBE, the sensor count is reduced with less noise/oscillation. And the highly accurate OBE contributes to the correction of possible modeling parametric or sampling errors. The integration of OBE-MPC algorithms improves both the steady-state and dynamic performances with less noise/oscillation, more robust transient behavior, and higher control bandwidth. The explicit design of OBE-MPC algorithms makes it possible to implement the functions on a low-cost DSP. Also, the state-space modeling of OBE-MPC for the LC-based power module is immune to the output side unknown inductance, which further improves the parametric accuracy. The proposed methods have been validated experimentally.