基于小批量随机梯度下降的 EV 电机模型预测电流控制

Model Predictive Current Control for Electric Vehicle Motors Based on Mini-Batch Stochastic Gradient Descent

摘要

为了克服电动汽车电机模型预测电流控制系统中的参数敏感性问题，并兼顾系统稳、动态性能，提出了一种基于小批量随机梯度下降(MSGD)的模型预测误差补偿算法．首先，构建了一种易于求解的线性辨识模型，通过辨识参数失配信息实现预测误差在线补偿．其次，提出了一种基于MSGD的线性回归求解方法．与递归算法和单采样点梯度下降算法相比，MSGD 利用滑动窗口存储时序数据，形成小批量样本集进行迭代优化．理论分析表明，所提算法优化了辨识收敛过程中的电流环零极点分布，同时减少了对单个数据点的依赖，避免了过拟合问题．此外，该算法引入随机性，能够平滑处理测量噪声和电机状态量脉动，提升了参数估计的鲁棒性．最后，设计了内嵌均方根传播(RMSProp)算法的自适应迭代步长更新机制，可快速收敛至最优解，提高了在线计算适应性．通过仿真和实验验证，在工况突变下，所提算法能够在准确补偿预测误差的同时，保证动态响应快速性，削弱收敛过程引入的暂态脉动．

Abstract

To overcome the parameter sensitivity issue in the model predictive current control system for electric vehicle motors and ensure steady and dynamic system performance，a prediction error compensation algorithm based on mini-batch stochastic gradient descent(MSGD)was developed. First，an easily solvable linear identification model was constructed to achieve online prediction error compensation through parameter mismatch identification. Second，a method for solving linear regression based on MSGD was introduced. Compared with recursive algorithms and single-sample-point gradient descent algorithm，MSGD utilized a sliding window to store time-series data，forming mini-batch sample sets for iterative optimization. Theoretical analysis demonstrated that the proposed algorithm optimized the zero-pole distribution of the current loop during identification convergence，reduced reliance on individual data points，and mitigated overfitting. Additionally，this algorithm introduced randomness，which enabled to smooth measurement noise and motor state pulsations，thereby enhancing the robustness of parameter estimation. Finally，an adaptive iterative step-size update mechanism embedded with root mean square propagation(RMSProp)algorithm was designed，enabling fast convergence to the optimal solution and improved online computational adaptability. Simulations and experiments confirmed that upon sudden changes in operating conditions，the proposed algorithm achieved accurate prediction error compensation while maintaining fast dynamic response and suppressing transient oscillations during convergence.

关键词

模型预测电流控制 / 电动汽车 / 参数失配 / 小批量随机梯度下降

Key words

model predictive current control(MPCC) / electric vehicle(EV) / parameter mismatch / mini-batch stochastic gradient descent(MSGD)

参考文献

[1] Xu B, Jiang Q, Ji W, et al. An improved three—vector—based model predictive current control method for surface—mounted PMSM drives[J]. IEEE Transactions on Transportation Electrification, 2022, 8(4):4418-4430.

[2] Wang X Q, Wang Z, Xu Z X, et al. Optimization of torque tracking performance for direct—torque—controlled PMSM drives with composite torque regulator[J]. IEEE Transactions on Industrial Electronics, 2020, 67(12):10095-10108.

[3] 李桂丹, 凌雨璐, 李斌, 等. 考虑相位延时的三矢量非线性模型预测控制[J]. 天津大学学报(自然科学与工程技术版), 2025, 58(11):1145-1154.

[4] Li Guidan, Ling Yulu, Li Bin, et al. Three—vector nonlinear model predictive control considering phase delays[J]. Journal of Tianjin University(Science and Technology), 2025, 58(11):1145-1154(in Chinese).

[5] Vazquez S, Rodriguez J, Rivera M, et al. Model predictive control for power converters and drives：Advances and trends[J]. IEEE Transactions on Industrial Electronics, 2017, 64(2):935-947.

[6] 齐昕, 苏涛, 周珂, 等. 交流电机模型预测控制策略发展概述[J]. 中国电机工程学报, 2021, 41(18):6408-6419.

[7] Qi Xin, Su Tao, Zhou Ke, et al. Development of AC motor model predictive control strategy：An overview[J]. Proceedings of the CSEE, 2021, 41(18):6408-6419(in Chinese).

[8] Belda K, Vosmik D. Explicit generalized predictive control of speed and position of PMSM drives[J]. IEEE Transactions on Industrial Electronics, 2016, 63(6):3889-3896.

[9] Rodriguez J, Garcia C, Mora A, et al. Latest advances of model predictive control in electrical drives—Part I：Basic concepts and advanced strategies[J]. IEEE Transactions on Power Electronics, 2022, 37(4):3927-3942.

[10] Liu X D, Zhang Q. Robust current predictive control—based equivalent input disturbance approach for PMSM drive[J]. Electronics, 2019, 8(9):1034.

[11] Liu K, Zhu Z Q. Mechanical parameter estimation of permanent—magnet synchronous machines with aiding from estimation of rotor PM flux linkage[J]. IEEE Transactions on Industry Applications, 2015, 51(4):3115-3125.

[12] Wang Z G, Zheng Z D, Li Y D, et al. An offset—free robust model predictive control with incremental model and improved current observer for induction motor[J]. IEEE Transactions of Electrical Energy System, 2019, 29(12):1-18.

[13] Zhang X G, Cheng Y, Zhang L. Disturbance deadbeat inductance observer—based current predictive control for surface—mounted permanent magnet synchronous motors drives[J]. IET Power Electronics, 2020, 13(6):1172-1180.

[14] Wei Y, Ke D L, Yu X H, et al. Adaptive inertia observer—based model—free predictive current control for PMSM driving system of electrical vehicle[C]// 2023 26th International Conference on Electrical Machines and Systems. Zhuhai，China， 2023:1206-1211.

[15] Wang J X, Wang F X, Wang G L, et al. Generalized proportional integral observer based robust finite control set predictive current control for induction motor systems with time—varying disturbances[J]. IEEE Transactions on Industrial Informatics, 2018, 14(9):4159-4168.

[16] 高仁璟, 王赓. 基于自适应扰动观测器的电磁直线驱动器改进滑模控制[J]. 电机与控制学报, 2025, 29(5):64-75.

[17] Gao Renjing, Wang Geng. Improved sliding mode control for electromagnetic linear actuator based on adaptive disturbance observer[J]. Electric Machines and Control, 2025, 29(5):64-75(in Chinese).

[18] 赵凯辉, 戴旺坷, 周瑞睿, 等. 基于扩展滑模扰动观测器的永磁同步电机新型无模型滑模控制[J]. 中国电机工程学报, 2022, 42(6):2375-2386.

[19] Zhao Kaihui, Dai Wangke, Zhou Ruirui, et al. Novel model—free sliding mode control of permanent magnet synchronous motor based on extended sliding mode disturbance observer[J]. Proceedings of the CSEE, 2022, 42(6):2375-2386(in Chinese).

[20] Wu T, Lyu M C, Wu X, et al. Enhanced model predictive control for PMSM based on reference voltage predictive model[J]. IEEE Journal of Emerging and Selected Topics in Power Electronics, 2023, 11(5):5290-5300.

[21] An X K, Liu G H, Chen Q, et al. Adjustable model predictive control for IPMSM drives based on online stator inductance identification[J]. IEEE Transactions on Industrial Electronics, 2022, 69(4):3368-3381.

[22] Kallio S, Karttunen J, Peltoniemi P, et al. Online estimation of double—star IPM machine parameters using RLS algorithm[J]. IEEE Transactions on Industrial Electronics, 2014, 61(9):4519-4530.

[23] 李洪凤, 徐浩博, 徐越. 扩展卡尔曼滤波参数辨识下永磁同步电机模型预测转矩控制[J]. 电机与控制学报, 2023, 27(9):19-30.

[24] Li Hongfeng, Xu Haobo, Xu Yue. Model prediction torque control of PMSM based on extended Kalman filter parameter identification[J]. Electric Machines and Control, 2023, 27(9):19-30(in Chinese).

[25] Yuan X, Zhang S, Zhang C N. Improved model predictive current control for SPMSM drives with parameter mismatch[J]. IEEE Transactions on Industrial Electronics, 2020, 67(2):852-862.

[26] Zhou Y, Zhang S, Zhang C N, et al. Current prediction error based parameter identification method for SPMSM with deadbeat predictive current control[J]. IEEE Transactions on Energy Conversion, 2021, 36(3):1700-1710.

[27] Li H F, Shao J Y, Liu Z Y. Incremental model predictive current control for PMSM with online compensation for parameter mismatch[J]. IEEE Transactions on Energy Conversion, 2023, 38(2):1050-1059.

[28] Siami M, Khaburi D A, Rodríguez J. Torque ripple reduction of predictive torque control for PMSM drives with parameter mismatch[J]. IEEE Transactions on Power Electronics, 2017, 32(9):7160-7168.

[29] Gohain P B, Jansson M. Robust information criterion for model selection in sparse high—dimensional linear regression models[J]. IEEE Transactions on Signal Processing, 2023, 71:2251-2266.

基金资助

国家自然科学基金资助项目(52377065)