Multistep Finite Control Set Model Predictive Control for Power Electronics

Abstract

For direct model predictive control with reference tracking of the converter current, we derive an efficient optimization algorithm that allows us to solve the control problem for very long prediction horizons. This is achieved by adapting sphere decoding principles to the underlying optimization problem. The proposed algorithm requires only few computations and directly provides the optimal switch positions. Since the computational burden of our algorithm is effectively independent of the number of converter output levels, the concept is particularly suitable for multi-level topologies with a large number of voltage levels. Our method is illustrated for the case of a variable speed drive system with a three-level voltage source converter.