多变量系统的广义预测控制解耦设计  

Decoupling Design of Generalized Predictive Control for Multivariable Systems

作　　者：薛美盛[1] 樊弟[1] 魏衡华[1] 

机构地区：[1]中国科学技术大学自动化系,安徽合肥230026

出　　处：《控制工程》2011年第1期39-42,共4页Control Engineering of China

基　　金：国家高技术研究发展计划项目(2007AA04Z195);国家水体污染控制与治理科技重大专项课题(2008ZX07010-003)

摘　　要：针对在系统关联严重的情况下,通过选择合适的权矩阵,预测控制可以实现间接解耦控制,但是系统的跟踪性能将大大降低的问题,提出了一种带设定值观测器的多变量广义预测解耦控制算法,结合前馈解耦控制策略与广义预测控制律,将其他通道对当前通道的影响作为扰动,从而实现近似解耦。在目标函数中用控制量的增量代替控制量,使控制增量变化不至于过于剧烈。并且增加了一种新的设定值观测器,根据设定值与实际值的偏差实时改变误差权矩阵,进一步减弱了耦合。最后,磨煤机的仿真结果验证了该控制算法的优越性。The decoupling problem of the multiple-input multiple-output coupling systems is discussed,the generalized model predictive decoupling control.A multivariable generalized predictive control algorithm with reference observer is provided,which combines the feed-forward decoupling with the generalized predictive control.The other channel impact on the researching channel is taken as disturbs,and the system is docoupled approximately.The increments of control variables are introduced in the cost function to weaken the great changes in the increments of control variables.And a reference observer is added,which changes the error weight matrices according to the errors between references and system outputs.It can get a further weakening of the coupling.The simulations on a ball mill show the superiority of the proposed method.

关 键 词：多变量系统 解耦 广义预测控制 观测器 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]