Constrained long-horizon direct model predictive control for power electronics

Abstract

The direct model predictive control (MPC) problem for linear systems with integer inputs, such as many power electronic systems, can be formulated as an integer least-squares (ILS) optimization problem. However, solving this problem when state and/or output constraints are explicitly included is challenging. In this paper, a method that allows to effectively use the sphere decoder - even in the presence of the aforementioned constraints - is proposed. This is done by computing a new hypersphere based on the feasible control input set, as defined by the imposed state/output constraints. A variable speed drive system with a three-level voltage source inverter serves as an illustrative example to demonstrate the performance of the proposed algorithm.