基于固定尺度欧氏距离的变因子递推子空间辨识算法  

Variable Factor Recursive Subspace Identification Algorithm Based on Fixed Scale Euclidean-distance

作　　者：黄金峰[1] 张合新[1] 张植 

机构地区：[1]第二炮兵工程学院自动控制系,陕西西安710025 [2]第二炮兵驻二00厂军代室,北京100854

出　　处：《测试技术学报》2012年第1期51-56,共6页Journal of Test and Measurement Technology

摘　　要：针对传统的子空间辨识算法采用固定遗忘因子出现跟踪能力不足或效果易受噪声影响等问题,提出了一种新的变因子递推子空间辨识算法.该算法分为3个步骤:首先引入变因子构造与更新Hankel矩阵和观测向量;其次为保证广义能观测阵的列向量收敛于主子空间的正交基上,采用OPAST算法递推估计广义能观测矩阵,并由广义能观测矩阵估计系统参数矩阵;最后用A的特征值空间距离信息实现变因子,因此,算法具有自适应能力.应用于一类时变系统,仿真结果表明改进算法具有较好的快速跟踪能力和跟踪效果.A new recursive subspace identification algorithm is proposed aiming at the problems existed in traditional algorithms using fixed factor, such as easily disturbed by noise and low tracking capability. It has three steps： firstly, Hankel matrices and observation vectors are updated with variable factor; secondly, in or- der to insure that the column vector of extended observation matrix is convergent on principal-subspace-or- thogonal-basis, the orthogonal PAST algorithm is used for recursive estimation of extended observation ma- trix, which is used to estimate system matrices; finally, with the help of eigenvalues＇ Euclidean-distance of matrix A, the approach of changing forgetting factor is realized, which ensures the self-adaptation of the modified algorithm. The results of simulations show that tracking performance of the new algorithm is faster and better than that of the traditional algorithms.

关 键 词：子空间辨识 递推算法 欧氏距离准则 正交化PAST跟踪算法 变因子 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]