电工技术学报  2023, Vol. 38 Issue (22): 6027-6038    DOI: 10.19595/j.cnki.1000-6753.tces.230159
“电动汽车驱动电机系统”专题(特约主编:温旭辉 研究员) |
基于时间序列的永磁同步电机连续控制集无模型预测电流控制
魏尧1, 柯栋梁1, 黄东晓1, 汪凤翔1, 康劲松2
1.电机驱动与功率电子国家地方联合研究中心(中国科学院海西研究院泉州装备制造研究中心)晋江 362216;
2.同济大学铁道与城市轨道交通研究院 上海 201804
A Continuous-Control-Set Type Model-Free Predictive Current Control Based on Time-Series for Permanent Magnet Synchronous Motor Drives
Wei Yao1, Ke Dongliang1, Huang Dongxiao1, Wang Fengxiang1, Kang Jinsong2
1. National and Local Joint Engineering Research Center for Electrical Drives and Power Electronics Quanzhou Institute of Equipment Manufacturing Haixi Institutes Chinese Academy of Science Jinjiang 362216 China;
2. Institute of Rail Transit Tongji University Shanghai 201804 China
全文: PDF (6567 KB)   HTML
输出: BibTeX | EndNote (RIS)      
摘要 时间序列数据驱动模型通过采样输入输出数据将被控对象在线拟合为离散传递函数,但在连续控制集(CCS)预测控制中直接应用存在困难。为了解决这个问题,该文结合最小二乘法,提出一种基于时间序列的永磁同步电机(PMSM)连续控制集无模型预测电流控制方法。该方法通过拉格朗日法合理设计回归矢量,在线估算模型待定系数,并建立数据驱动模型预测所需变量。不仅从根本上消除了模型预测控制(MPC)中先验模型对被控对象时变物理参数的依赖,而且所得模型符合电机运动特性,有更高的模型精度和良好控制性能。仿真和实验结果验证了提出方法的有效性,以及在动态性能、电流质量和系统噪声方面的优势。
服务
把本文推荐给朋友
加入我的书架
加入引用管理器
E-mail Alert
RSS
作者相关文章
魏尧
柯栋梁
黄东晓
汪凤翔
康劲松
关键词 无模型预测控制时间序列模型最小二乘估计算法电流预测    
Abstract:The finite-control-set type (FCS-type) predictive control method is not ideal for electric vehicles because of its limited control accuracy and high harmonic content. In contrast, the continuous-control-set type (CCS-type) model-free predictive current control (MF-PCC) is more suitable. The model-free predictive current control (MF-PCC) uses a time-series model as a discrete transfer function, which is more compliant with the motion characteristics of the motor system. However, calculating the inverse of the model in the digital processor makes it challenging to apply in the CCS-type. Therefore, a time-series-based CCS-type MF-PCC strategy is proposed in this paper for a permanent magnet synchronous motor (PMSM) driving system. The plant is accurately expressed, and the control strategy is easily realized in the CCS-type by the online building and updating the time-series model based on the sampled data.
Firstly, this approach establishes a time-series model and updates the regressive vector summarizing input and output signals based on sampled data. Secondly, all undetermined coefficients in the model are estimated through the recursive least square (RLS) algorithm. Herein, the current operating state is described as a discrete- time transfer function within the model. Finally, according to predictive reference by the Lagrange algorithm, the regressive vector is updated to predict the output signal and generate the control law. This time-series model is easily realized in the CCS type with good accuracy, addresses the problem of the complex calculation process, and eliminates all time-varying physical parameters and their influences in the a priori model of the plant.
Simulation and experimental results on a PMSM driving system show that the proposed method resists disturbances and tracks the reference successfully to suit electric vehicle driving operations. The disturbances mainly include changed parameter mismatches, load torque, and DC voltage. Continuous Fourier analysis and accumulated error comparisons between different control strategies demonstrate that the proposed method achieves a range of total harmonic distortion (THD) between 3.3%~3.9%, with an average value of 3.53%. Compared to conventional strategies, the proposed method has the minimum ascending slope of accumulated error of current. Additionally, to analyze the impact of system noise on driver perception, the proposed method is tested under different speed references and load torques. Finally, experimental results are obtained with different parameter mismatches of typical physical parameters, including stator resistance, stator inductance, and magnet flux linkage, and compared using continuous Fourier analysis.
The following conclusions can be drawn. (1) The proposed method represents the current operating state of the motor driving system as a time-series model in the CCS-type. Compared with the conventional MF-PCC strategy using ultra-local, it formulates the system as a group of discrete-time transfer functions. (2) The obtained current quality, dynamics, and system noises are improved due to the good accuracy of the time-series model. (3) Based on the designed estimation algorithm and sampled data, the time-series model includes multiple time-varying physical parameters of the system. (4) The orders of the model should be selected comprehensively, considering the permitted calculation time and performances to prevent overrun error or insufficient accuracy.
Key wordsModel-free predictive control    time series model    least square estimation algorithm    current prediction   
收稿日期: 2023-02-14     
PACS: TM341  
基金资助:国家自然科学基金(52277070)和福建省科技计划(2022T3070)资助项目
通讯作者: 汪凤翔 男,1982年生,博士,研究员,博士生导师,研究方向为电力电子与电力传动。E-mail: fengxiang.wang@fjirsm.ac.cn   
作者简介: 魏 尧 男,1993年生,博士,研究方向为新能源汽车电控系统、交流电机伺服系统及其先进控制。E-mail: yao.wei@fjirsm.ac.cn
引用本文:   
魏尧, 柯栋梁, 黄东晓, 汪凤翔, 康劲松. 基于时间序列的永磁同步电机连续控制集无模型预测电流控制[J]. 电工技术学报, 2023, 38(22): 6027-6038. Wei Yao1, Ke Dongliang1, Huang Dongxiao1, Wang Fengxiang1, Kang Jinsong2. A Continuous-Control-Set Type Model-Free Predictive Current Control Based on Time-Series for Permanent Magnet Synchronous Motor Drives. Transactions of China Electrotechnical Society, 2023, 38(22): 6027-6038.
链接本文:  
https://dgjsxb.ces-transaction.com/CN/10.19595/j.cnki.1000-6753.tces.230159          https://dgjsxb.ces-transaction.com/CN/Y2023/V38/I22/6027