利用基序列逼近的一类非线性时变系统辨识的实用方法

A Practical Method for Identification Nonlinear Time-Varying Systems Using Basis Sequences

提出一种针对一类非线性时变系统 (时变的 Hammerstein级数 )辨识的实用方法。为减少时变系统建模所需参数个数 ,利用已知基序列的线性组合来逼近系统的时变动态特性 ,并采用递推最小二乘来估计模型的参数 ,克服了以往基序列逼近用于时变系统辨识方法中离线最小二乘计算效率不高的缺陷。仿真结果表明 ,本文提出的方法能经济有效地对一大类时变非线性系统进行较好的辨识。

In the paper a new identification method for nonlinear time varying models, called the time varying Hammerstein series is presented. Basis sequences is introduced to approximate the temporal variation of each time varying kernels in order to reduce the number of coefficients required in modeling. A significant advantage of the method is that only a single input output record is required to obtain recursive least squares estimates of the model parameters. RLS estimates can avoid lower computational efficiency of off line LS which have been used to estimate the parameters of basis sequences approximation to time varying system identification.Simulated results are presented to indicate that the presented method can identify a wide class of time varying and nonlinear systems in a parsimonious and computationally attractive manner.