多变量系统子子模型辨识

目前,对多变量系统辨识的研究已经越来越多,但是,基本上都属于子模型的辨识问题,不便于实际应用.本文采用子模型与子子模型技术,将一个有R个输入,m个输出的多变量系统分解为M个带R个输入的子模型,而每个子模型又可分解为R个子子模型(单输入单输出系统).在多变量系统中,只要做一次试验(同时加入月个输入信号),使用月个输入信号与M个输出信号的数据,可以自动辨识子模型的阶次与参数,再从子模型的有关数据,最终辨识出子子模型的阶次与参数.本文采用分块矩阵求逆公式,提出一种快速的自动定阶与估计参数的递推算法.

Subsubmodel Identification of Multivariable System

In recent years, multivariable system identification is discussed in many papers. However, only submodels are used in most of the identificaion metheds, and are not saetisfactory for engineering applications. The identification problems of submodels and subsubmodels are discussed. A main model with R inputs and M outputs is decomposed into M submodels. And then a submodel with R inputs is decomposed into R subsubmodels (SISO-system). All the inputs excite all the outputs in a experiment simultaneously. The order and Parameter of the submodels and subsubmodels can be identified separately and automatically using all the data. A fast recursive algorithm of cestimating the order and parameters using the property of partitioned matrix inverse is presented in this paper.