引用本文:孙浩杰,邹涛,李丽娟.双层结构模型预测控制算法性能分析[J].控制理论与应用,2020,37(1):13~22.[点击复制]
SUN Hao-jie,ZOU Tao,LI Li-juan.Performance analysis of double-layered model predictive control algorithm[J].Control Theory and Technology,2020,37(1):13~22.[点击复制]
双层结构模型预测控制算法性能分析
Performance analysis of double-layered model predictive control algorithm
摘要点击 2582  全文点击 1082  投稿时间：2018-08-22  修订日期：2019-03-23
查看全文  查看/发表评论  下载PDF阅读器
DOI编号  10.7641/CTA.2019.80630
  2020,37(1):13-22
中文关键词  双层结构  模型预测控制  控制性能；经济性能  鲁棒性  
英文关键词  double-layered  model predictive control（MPC）  steady-state target calculation (SSTC)  control performance  economic performance  robustness
基金项目  国家重点基础研究发展计划（2017YFB0603703）;,国家自然科学基金(61773366,61533015,61873121);辽宁省博士启动基金(20180540066)
作者单位E-mail
孙浩杰 中国科学院沈阳自动化研究所 sunhaojie@sia.cn 
邹涛* 中国科学院沈阳自动化研究所 zoutao@sia.cn 
李丽娟 南京工业大学  
中文摘要
      双层结构模型预测控制策略被当前主流工业流程控制软件所采用（如DMCplus、RMPCT等），但是在理论层面尚缺少深入研究。本文着重于分析稳态目标计算层对常规单层模型预测控制策略在控制性能、经济性能、及鲁棒性方面的影响。阐明稳态目标计算是在动态不确定性下的最优决策，在当前工况环境下给出实时最优设定值，并证明在恒值扰动下可实现无静差控制；此外稳态目标计算层可在保证对重要变量跟踪控制的同时利用操作自由度进行经济性优化，不仅有助于系统经济性能提升，而且使得系统对于特定方向的扰动和通道失配具有更强的鲁棒性。最后通过实验仿真验证了所做分析的正确性。
英文摘要
      Double-layered structure model predictive control strategy adopted by the current mainstream industrial process predictive control software (eg: DMCplus, RMPCT. et al), but lack of deep research in theoretical level. This paper focuses on the analysis of the effects on the control performance, economic performance and robustness of adding a steady-state target calculation layer on conventional single model predictive control strategies. It is clarified that the steady-state target calculation is the optimal decision under dynamic uncertainty, which can give the real-time optimal setting value under the current working condition and prove that the offset-free control can be realized under constant disturbance. In addition, the steady-state target calculation layer can make use of the degree of freedom to optimize the economy while ensuring the tracking control of important variables, which not only helps to improve the economic performance of the system, but also helps to make the system more robust to perturbations and channel mismatches in specific directions. Finally, the correctness of the analysis is verified by simulation.