结合电力电子系统有限开关状态特性，近年来有限状态集模型预测控制(FCS-MPC)利用被控对象离散预测模型在同步磁阻电机(SynRM)能效提升系统中得到广泛应用。然而，FCS-MPC建模过程严重依赖电机数学模型，电机参数精确性会直接影响系统控制效果。为解决上述问题，提出一种数据驱动型模型预测控制(DD-MPC)方法。该方法不利用任何电机数学模型信息，仅基于电机运行时的输入输出数据关系，通过对数据的处理分析实现“边建模、边控制”。此外，为保证DD-MPC系统的收敛性和稳定性，设计了一种DD-MPC输入输出数据关系的高更新率实现方法，利用最近输出的三电压矢量对应电流变化信息，推演并更新全局电压矢量对应电流变化关系。最后，基于25 kW样机对DD-MPC控制效果进行测试与分析。与传统FCS-MPC方法对比，所提方法在保留FCS-MPC快速性和灵活性等优势的基础上，可有效提升系统的鲁棒性和稳定性。

In recent years, using the discrete model of the object and considering the finite switching state characteristics of power electronic systems, finite control set model predictive control (FCS-MPC) has been widely used in synchronous reluctance motor (SynRM) drive system to promote the energy efficiency. However, the FCS-MPC modeling process heavily depends on the SynRM mathematical model, and the accuracy of motor parameters directly affects the control effect. To solve the above problems, a data-driven MPC (DD-MPC) method is proposed. This method does not need the SynRM mathematical model information. It only utilizes the input-output data relationship of the SynRM. So the “synchronization of modeling and control” is realized. In addition, in order to ensure the convergence and stability of DD-MPC system, a DD-MPC data relation with high update rate is designed. By analyzing the current change information corresponding to the recent three-voltage-vector output, the corresponding current change relation of the global voltage vectors is deduced and updated. Finally, the control effect of DD-MPC is tested and analyzed based on a 25 kW prototype. Compared with the traditional FCS-MPC, the proposed method can effectively improve the robustness and stability, while maintaining the rapidity and flexibility of FCS-MPC.

国家重点研发计划“科技助力2020”项目（SQ2020YFF0426410）；国网江苏省电力有限公司科技项目（J2021057）