维纳模型的最小二乘支持向量机辨识

Identification of Wiener Models Based on Least Square Support Vector Machine

针对传统维纳模型辨识方法存在算法复杂、精度低的问题,通过对最小二乘支持向量机建模原理和维纳模型结构特点的分析,提出一种基于最小二乘支持向量机的维纳模型辨识新方法。该方法充分利用了维纳模型中具有线性环节这一先验知识,实现了线性和非线性环节参数的同时辨识。对于多变量维纳模型,该方法同样适用。给出并证明了该方法存在唯一解的约束条件—参数部分列满秩。仿真实验表明了该方法的有效性,与标准最小二乘支持向量机辨识方法相比,该方法具有更高的精度。

Considering the shortcomings of Wiener model＇s identification,this paper proposes a new identification method based on least squares support vector machine（LS-SVM）.This method allows for the estimation of the model parameters of the linear part as well as the determination of the nonlinearity by the prior information of Wiener models.Then the MIMO identification case is also elaborated for the MIMO Wiener models.This paper also proofs the constraint condition of the method.An unique solution exists when the parametric part has full column rank.The experiments show the effectiveness of the presented method.This method has higher precision than the standard LS-SVM.