大时滞生产过程的鲁棒抗扰控制与批次运行优化

To cope with the problems of poor anti-disturbance control performance and inconsistence of batch running associated with manufacturing processes with large time delay, this project investigates active disturbance rejection control (ADRC) methods based on delay-free response prediction and real-time disturbance estimation, and robust anti-disturbance batch control & running optimization methods based on historical batch data and iterative learning strategies. For the presence of delay uncertainty and process dynamic uncertainties, robust anti-disturbance control methods will be established to stabilize these uncertainties, and correspondingly, sufficient conditions for holding robust stability of the closed-loop system will be analyzed. For batch processes with delay response, robust anti-disturbance iterative learning control (ILC) methods will be developed based on historical batch information, to overcome adverse effect of repeatable load disturbance, such that perfect tracking of the desired output trajectory could be realized. Moreover, non-repeatable load disturbance, delay uncertainty and time-varying process uncertainties could be stabilized. Concerning batch processes with time delay that are difficult to model the dynamic response characteristics, robust data-driven ILC methods based on measured input/output data will be explored, and meanwhile, sufficient conditions for robustly tracking the desired output trajectory will be analyzed. The project results will give some new theory and application methods of robust anti-disturbance control and batch running optimization for manufacturing processes with large time delay, and therefore have reference significance in aspects of theoretical guidance and engineering application for improving industrial process control technologies.

针对大时滞生产过程存在抗扰性能较差和批次运行质量不一致的问题，研究基于无时滞响应预估和实时扰动估计的主动抗扰控制方法、基于历史批次数据和迭代学习策略的鲁棒抗扰批次控制与运行优化方法。对于存在时滞不确定性和过程动态不确定性的情况，建立能镇定这些不确定性的鲁棒抗扰控制方法，分析闭环控制系统保证鲁棒稳定性的充分条件。针对带有时滞响应的批次生产过程，发展基于历史批次信息的鲁棒抗扰迭代学习控制方法，以克服重复性负载干扰的不利影响，实现完全跟踪期望输出轨迹，而且能镇定非重复性负载干扰、时滞不确定性以及过程时变不确定性。对于难以建模动态响应特性的时滞响应批次生产过程，研究基于过程输入和输出测量数据的鲁棒数据驱动迭代学习控制方法，论证鲁棒跟踪期望输出轨迹的充分条件。本项目成果将给出大时滞生产过程鲁棒抗扰控制与批次运行优化的一些新理论与应用方法，对提高工业过程控制技术水平具有理论指导和工程应用参考意义。

工业生产过程中广泛存在着由物料传送、信号处理、长距离信号传输等引起的时滞现象。此外，实际生产过程中还存在由添加原理、产物输出、以及负荷变化等引起的干扰。这些不利因素往往会导致系统性能大幅下降，甚至造成闭环控制系统不稳定。如何针对时滞系统设计抗扰控制方法是过程控制领域具有挑战性的问题。对于批次生产过程，如何克服非重复过程不确定性以及时滞响应，实现鲁棒批次运行优化，也是过程控制领域亟待解决的难题。围绕这些难题和挑战，本项目取得了一批具有原创性的研究成果，主要包括：(i) 针对存在非对称饱和约束的输出时滞温度控制系统，提出基于无时滞输出预测的反饱和主动抗扰控制方法，显著提高大时滞系统的抗扰性能；(ii) 针对存在输入时滞的工业生产过程，提出基于预估器的主动抗扰控制方法；(iii) 针对存在时变和批次变化不确定性的批次生产过程，提出基于扩张状态观测器的间接型ILC方法，提高了被控过程的跟踪和抗扰性能；(iv) 针对存在非重复性负载干扰的批次过程，提出基于静态输出反馈的有限频域两维H无穷性能规范的ILC方法；(v)针对存在输出时滞响应的批次生产过程，提出基于预测状态观测器的间接型设定点学习控制方法；(vi) 针对存在非重复性负载干扰未知非线性批次生产过程，提出基于扰动估计和补偿的数据驱动迭代学习控制方法，提高了现有数据驱动迭代学习控制方法的抗扰性能。共计发表学术论文17篇，其中7篇发表于控制领域国际重要SCI期刊，申请授权国家发明专利2项，合作出版中文著作1部。