线性离散时间大系统次优分散控制器设计

THE DESIGN OF NEAR-OPTIMUM DECENTRALIZED CONTROLLERS FOR LINEAR DISCRETE-TIME LARGESCALE SYSTEMS

在大系统分散控制器设计中,如果完全忽略各子系统之间的联系,控制性能往往较差,有时甚至整个系统都不能稳定;如果完全考虑各子系统之间的联系,则通常需要求解一个高维问题(其维数等于系统总阶数)。为此,本文提出了一种简化设计方法,在求解某个子系统控制器时,将该子系统以外的系统其余部分降阶,得到一个低阶近似系统,再求解由它们构成的分散控制器设计问题,从而得到该子系统的控制器。该方法可以大大降低分散控制器设计问题的复杂性,且控制性能甚好。

In the design of decentralized controllers for large seale systems, the system performance is usually not good enough. Sometimes the whole system may be unstable if all the interactions between the subsystems are neglected, while a problem of high demension (equals to the order of the whole system) has to be solved if the interactions are fully considered, which is not realizable in many cases. A simplified approach is proposed here. The decentralized controller of a subsystem is designed using the reduced order model for the rest of the system. In other words. the subsystem and the order reduced model are taken to compose an approximated low order system, upon which the decentralized controller for the subsystem is designed. And it is shown that the approach could greatly simplify the design of decentralized controllers for high demension systems and achieve high performance