Identification for Polynomial Wiener Model Using Improved Back Propagation Method

Abstract

A new method is introduced for the identification of nonlinear dynamic system described by Wiener model, which consists of a linear dynamic block followed by a static non-linearity. Firstly, it is assumed that static non-linear part is invertible and its inverse characteristics can be expressed or approximated by a polynomial of known orders. Secondly, based on these assumptions, a novel neural network structure is designed, the weights in which are corresponding with the parameters of polynomial Wiener model. Finally, to solve the problem of non-convergence of conventional back propagation iterative, the improved one is derived, through which the non-linear and linear dynamic part can be optimized and the coefficients of the polynomial Wiener model are gotten with a higher convergence rate. A numerical example is included to show the effectiveness and the practical feasibility of the presented approach.