To solve the problem of parameter mismatch in model predictive control (MPC), this paper presents a robust model predictive control method based on a fixed window optimization (FWO) algorithm for a three-level voltage source inverter that only needs to sample the current value. When compared with the traditional observer-based model predictive control (such as the Luenberger observer, sliding mode observer (SMO) and Kalman filter (KMF)), the proposed method does not require an observer with a complicated design, and its algorithm is simple and easy to understand. Meanwhile, high current sampling accuracy is not needed in the proposed method. However, it is necessary in some types of model-free predictive control. In addition, low switching frequency operation and delay compensation are also considered in this paper. In general, the proposed method is simple to implement and does not have high requirements in terms of the accuracy of its current sensor. Experimental results show that the proposed method can accurately estimate parameter values and improve the parameter robustness of MPC.

为了解决模型预测控制（MPC）中参数不匹配的问题，本文提出了一种基于固定窗口优化（FWO）算法的鲁棒模型预测控制方法，该方法适用于仅需要对电流值进行采样的三电平电压源逆变器。与传统的基于观测器的模型预测控制（如Luenberger观测器，滑模观测器（SMO）和卡尔曼滤波器（KMF））相比，该方法不需要设计复杂的观测器，并且算法简单并易于理解。同时，该方法不需要高电流采样精度。但是，在某些类型的无模型预测控制中，这是必需的。此外，本文还考虑了低开关频率操作和延迟补偿。一般来说，所提出的方法易于实现，并且对电流传感器的精度要求不高。实验结果表明，该方法能够准确估计参数值，提高了MPC的参数鲁棒性。