The Sphere Decoding with Probabilistic Constraint for Multi-Step MPC of PMSM

Although multi-step model predictive control (MPC) can effectively enhance the performance of motor drive systems and reduce current harmonic distortion, the exhaustive search leads to an exponential increase in computational burden. To overcome this problem, this paper proposes a novel sphere decoding algorithm incorporating probabilistic constraints. The algorithm adopts a breadth-first search strategy, where full enumeration is performed in the first three layers to ensure optimality, while probabilistic constraints are applied in the subsequent layers to reduce the number of candidates. This hybrid approach balances computational efficiency and control accuracy. Simulation results confirm that the proposed method significantly alleviates the computational load while maintaining desirable control performance.