操作变量与前馈变量转换对模型预测控制可行域影响机制  

Influential mechanism of switch between manipulated variables and feed-forward variables on feasible region of model predictive control

作　　者：罗雄麟 张佳 许锋 许鋆 

机构地区：[1]中国石油大学自动化系,北京102249

出　　处：《大连理工大学学报》2017年第3期308-315,共8页Journal of Dalian University of Technology

基　　金：国家自然科学基金资助项目(21676295);"九七三"国家重点基础研究发展计划资助项目(2012CB720500)

摘　　要：模型预测控制因其解耦性和强鲁棒性得以在过程控制中广泛使用.在实际生产过程中,因操作需要常常放开控制器对部分操作变量的控制,此部分操作变量则转化为前馈变量.变量之间的转换将使系统的结构在方系统、胖系统和瘦系统之间发生转换.对于瘦系统,操作变量的维数少于被控变量,其控制效果往往低于其他结构的系统.同时,原操作变量不被控制器控制后,转换为前馈变量,也会影响对系统的控制效果.可行域能直接反映系统控制效果.在对状态空间模型进行多步预测推导的基础上,将部分控制变量与前馈变量进行转换,用空间几何映射表示变量转换,并分析转换对系统可行域的影响.最后,针对实际生产过程中放开部分控制变量的情况,提出预判机制,以保证系统可控.Model predictive control is widely used in process control because of its decoupling ability and strong robustness.In the actual industrial process,considering operation requirements,apart of manipulated variables are released by controller and transformed into feed-forward variables at the same time.The transformation between manipulated variables and feed-forward variables results in structure changing among different systems,including square system,fat system and thin system.For the thin system,in which the dimension of manipulated variables is less than that of the controlled variables,the control result is usually less effective than the others.Besides,the partial original manipulated variables,released by controller,transformed into feed-forward variables,also affect the control performance.Feasible region can reflect the effect of the control system directly.On the basis of the multi-step prediction formula derivation for state space model,the method of spatial geometry mapping is used to express the transformation between manipulated variables and feed-forward variables.In addition,it is also used to analyze the influence on the change of feasible region caused by the transformation.Finally,aiming at the situation for releasing partial manipulated variables,prejudgment method is proposed to ensure the stability.

关 键 词：过程控制 模型预测控制 前馈变量 可行域 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]