Model predictive control (MPC) has garnered increasing scholarly interest attributable to its straightforward implementation, its excellent efficacy in managing multiple control objectives, and its fast dynamic response. However, MPC requires accurate mathematical models and precise physical parameters to ensure optimal performance. Furthermore, the high switching frequency associated with conventional MPC approaches often results in increased undesirable losses. To solve the above issues, a two-step prediction horizon-based predictor neural network model-free predictive control (TSP-PNN) methodology is proposed in this paper. In the proposed method, the unknown non-linear components of the model and the uncertainties in the system are estimated through the employment of a predictor-based neural network (PNN). Simultaneously, the introduction of a two-step prediction horizon framework extends the optimization landscape to encompass two operational cycles, which enhances the likelihood of selecting identical voltage vectors, thereby effectuating a reduction in the switching frequency. The primary contribution of the proposed approach is its ability to maintain a lower switching frequency while exhibiting strong robustness. Finally, simulations are used to validate the robustness of the proposed method as well as its efficacy in maintaining low switching frequencies.

Finite control-set model predictive control,predictor-based neural network control,robust control,switching frequency,Model-free predictive control (MFPC)