Sphere Decoding Based Long-Horizon Predictive Control of Three-level NPC Back-to-Back PMSG Wind Turbine Systems

We propose a computationally efficient control scheme for model predictive control of a three level-neutral point clamped (NPC) inverter permanent magnet synchronous generator (PMSG) back-to-back system. In order to increase the performance a multistep prediction of the system behaviour is applied. Therefore, multistep prediction models for the machine side as well as the grid side are introduced. Since the computational complexity would be growing exponentially with increasing horizon length we apply a sphere decoding algorithm to reduce the computational complexity. The sphere decoding algorithm can only be used for linear systems while the 3L-NPC BTB system is nonlinear. Therefore different simplification methods are studied in this paper.