间歇过程最优迭代学习控制的发展:从基于模型到数据驱动

发布时间：2018-01-23 03:15

本文关键词： 间歇过程基于模型的最优迭代学习控制迭代动态线性化数据驱动的最优迭代学习控制　出处：《自动化学报》2017年06期 　论文类型：期刊论文

【摘要】：本文综述了间歇过程的基于模型的和数据驱动的最优迭代学习控制方法.基于模型的最优迭代学习控制方法需要已知被控对象精确的线性模型,其研究较为成熟和完善,有着系统的设计方法和分析工具.数据驱动的最优迭代学习控制系统设计和分析的关键是非线性重复系统的迭代动态线性化.本文简要综述了基于模型的最优迭代学习控制的研究进展,详细回顾了数据驱动的迭代动态线性化方法,包括其详细的推导过程和突出的特点.回顾和讨论了广义的数据驱动最优迭代学习控制方法,包括完整轨迹跟踪的数据驱动最优迭代学习控制方法,提出和讨论了多中间点跟踪的数据驱动最优点到点迭代学习控制方法,和终端输出跟踪的数据驱动最优终端迭代学习控制方法.进一步,迭代学习控制研究中的关键问题,如随机迭代变化初始条件、迭代变化参考轨迹、输入输出约束、高阶学习控制律、计算复杂性等.本文突出强调了基于模型的和数据驱动的最优迭代学习控制方法各自的特点与区别联系,以方便读者理解.最后,本文提出数据驱动的迭代学习控制方法已成为越来越复杂间歇过程控制发展的未来方向,一些开放的具有挑战性的问题还有待于进一步研究.
[Abstract]:This paper reviews the model-based and data-driven optimal iterative learning control methods for batch processes. A model-based optimal iterative learning control method requires a linear model with known accuracy of the plant under control. Its research is more mature and perfect. The key to the design and analysis of data-driven optimal iterative learning control systems is the iterative dynamic linearization of nonlinear repetitive systems. This paper briefly reviews the optimal iteration based on the model. The research progress of generation learning control. The data-driven iterative dynamic linearization method, including its detailed derivation process and outstanding characteristics, is reviewed in detail. The generalized data-driven optimal iterative learning control method is reviewed and discussed. A data-driven optimal iterative learning control method for complete trajectory tracking is proposed and discussed. A data-driven optimal iterative learning control method for multi-intermediate point tracking is proposed and discussed. And terminal output tracking data driven optimal terminal iterative learning control method. Further, the key issues in iterative learning control research, such as random iterative change of initial conditions, iterative change reference trajectory. Input and output constraints, higher-order learning control laws, computational complexity, etc. This paper highlights the characteristics and differences of model-based and data-driven optimal iterative learning control methods. Finally, this paper proposes a data-driven iterative learning control method, which has become the future direction of the development of more and more complex batch process control. Some open and challenging issues need to be further studied.
【作者单位】：青岛科技大学自动化与电子工程学院;北京交通大学先进控制系统研究所;阿尔伯塔大学化学与材料工程学院;
【基金】：国家自然科学基金(61374102,61433002) 山东省泰山学者工程资助~~
【分类号】：+
【正文快照】： 引用格式池荣虎,侯忠生,黄彪.间歇过程最优迭代学习控制的发展:从基于模型到数据驱动.自动化学报,2017,43(6):917-932Optimal Iterative Learning Control of Batch Processes:From Model-based to Data-drivenCHI Rong-Hu1HOU Zhong-Sheng2交通大学先进控制系统研究所北京100

【相似文献】