论文部分内容阅读
阐述了递阶辨识原理 ,提出了传递函数阵模型参数的递阶随机梯度 (HSG)辨识方法 .在递阶辨识中 ,系统参数被分解为参数向量和参数矩阵 .前者是由系统的特征多项式的系数构成的 ,后者是由传递函数矩阵分子多项式的系数构成的 .借助于鞅超收敛定理的收敛性分析表明 ,HSG算法的参数估计误差一致有界 ;当持续激励条件成立时 ,参数估计误差一致收敛于零 .递阶辨识方法具有计算量小和容易实现等特点
In this paper, the principle of hierarchical identification is expounded, and a hierarchical stochastic gradient (HSG) identification method of transfer function matrix model parameters is proposed. In hierarchical identification, the system parameters are decomposed into parameter vectors and parameter matrices. The former is composed of system characteristic polynomials The latter is composed of the coefficients of the polynomial of the transfer function matrix. By means of the convergence analysis of martingale super convergence theorem, it is shown that the parameter estimation error of the HSG algorithm is uniformly bounded. When the continuous excitation condition is established, the parameter estimation error Uniformly converge to zero.Hierarchy identification has the characteristics of low computational complexity and easy implementation