基于滞环反馈时域鲁棒辨识的控制器自整定  

Robust Time-domain Identification Based on Relay Feedback Test and Its Applications to Controller Auto-tuning

作　　者：刘清源 韩昱城 尚超[1] 黄德先[1] LIU Qing-yuan;HAN Yu-cheng;SHANG Chao;HUANG De-xian(Department of Automation,Tsinghua University,Beijing 100084,China)

机构地区：[1]清华大学自动化系,北京100084

出　　处：《控制工程》2022年第4期684-691,716,共9页Control Engineering of China

基　　金：国家自然科学基金资助项目(62003187)。

摘　　要：现有的滞环反馈自动整定方法大多仅使用频域信息,包含较多近似,因此在处理质量较差的工业数据时精度与鲁棒性较差;此外,现有方法依赖稳定震荡的极限环数据,需要较长的滞环反馈测试时间。针对这些问题,提出一种基于滞环反馈测试的时域鲁棒辨识与控制器自动整定方法。首先,利用低阶延时模型时域响应Hankel矩阵的低秩特性,提出使用秩约束优化对滞环反馈时域辨识问题进行建模,并使用基于交替方向乘子法的求解方法;然后,给出了算法中超参数和优化初始值的选择方案,并讨论了使用非零初始状态滞环反馈数据进行辨识的方法;最后,提出了基于辨识结果的控制器整定方案。仿真实验结果表明,所提方法与现有方法相比具有更高的辨识精度与鲁棒性,并能取得更好的控制器整定效果。Current relay feedback auto-tuning methods are mainly based on frequency-domain information and involve massive approximation, thus show poor performance in the accuracy and robustness of identification when dealing with poor-quality industrial data. In addition, current relay feedback auto-tuning methods strongly rely on limit cycle data of stable oscillation for identification, which requires longer relay feedback test time. Aiming at these problems, a robust time-domain identification algorithm based on relay feedback test is proposed in this paper, as well as its applications to controller auto-tuning. Firstly, by exploiting the inherent low-rank structure of the Hankel matrix of time-domain response of the low-order plus time delay model, we proposed a novel method for the modeling of relay feedback time-domain identification problem with rank-constrained optimization, and a solution based on the alternating direction multiplier method. Then, the selection criterion of hyperparameters and initial value of the optimization problem is given, and the case of non-zero state relay feedback test is discussed. Finally, the controller auto-tuning method based on the identification result is proposed. The numerical example shows that, in comparison with current relay feedback auto-tuning methods, the proposed method has higher accuracy and robustness of identification, which secures better controller tuning effect.

关 键 词：滞环反馈自动整定 系统辨识 滞环反馈测试 时域辨识 秩约束优化 

分 类 号：TP13[自动化与计算机技术—控制理论与控制工程]