基于最小二乘改进算法的时变系统参数辨识

在系统辨识领域,常规最小二乘法是一种最基本的辨识方法之一.然而,随着观测数据的不断增加,会出现“数据饱和”的现象,造成新观测数据对估计值起不到修正的作用.由于新观测值对未知参数估计的影响要比旧观测值大,采用了渐消记忆和限定记忆最小二乘改进算法,来实现时变过程的参数辨识,并进行了仿真实验.仿真结果表明,它们能够克服“数据饱和”现象,从而改善参数辨识结果

Parameter identification of time- varying systems based on the least- squares modified algorithm

In the system identification domain, the least-squares are one of the most fundamental methods. However, as the observed data are increased, data saturation comes about, which leads to a phenomenon that the estimated values can't be modified by new observed data. Because the new observed data have a more remarkable effect on parameter estimation than the old observed data, this paper introduces the least-squares modified algorithm to identify the parameter of time varying process. Finally, the simulated experiment is proposed. The result indicates that both regression memory and restriction memory method can overcome the data saturation phenomenon, and improve the precision of parameter identification.