一种时变参数多项式扩展递推最小二乘法PRLS  

A Recursive Least Square Identification Method with Time-Varying Parameter Polynomial Expansion

作　　者：赵正义[1] 宋文忠[1] 

机构地区：[1]东南大学自动化研究所,南京210018

出　　处：《控制与决策》1992年第3期194-198,共5页Control and Decision

摘　　要：本文对带有方程噪声的时变模型提出了一种新的参数辨识算法——PRLS,并克服了时间观测变量的无限增长问题,从而能长时期地跟踪辨识快速时变参数,并且效果明显优于时变遗忘因子和常值遗忘因子最小二乘法。A new method-PRLS-for the identification of linear timevarying system models with equation noises is presented and a way to prevent the infinite increase of the observed time variable, t, is found. So, this method can track and identify fast time-varying parameters for any long time. The effect is better than both RLS with constant forgetting factor and RLS with variable forgetting factor obviously.

关 键 词：系统辨识 辨识算法 最小二乘法 

分 类 号：TP11[自动化与计算机技术—控制理论与控制工程]