有色噪声中非线性动态系统的BP网络辨识

前馈多层神经网络为复杂的非线性系统提供了一种极具吸引力的模型结构。本文不利用仅含一个隐层的前馈多层神经网络来拟合离散时间非线性动态系统的问题进行了探讨。由于有色噪声的存在会导致网络模型偏差产生，文中引入了一种对噪声建模的方案。借助于非线性模型检验技术，本文给出了在有色噪声存在的情况下，利用ＢＰ网络辨识离散时间非线性动态系统的一般方法，仿真结果亦表明该方法行之有效。

Identification of Non-Linear Dynamic Systems in Coloured Noise Using Neural Networks

Feed forward multi-layered neural networks offer an attractive alternative of modellingcomplex non-linear systems. This paper inverstigates the identification of discrete-take non-lineardynamic systems using neural networks with a single hidden layer. As the coloured noise can result in severely biased models,a scheme for modelling the noise is introduced. With the aid of thenon-linear model test technique, a general approach to modelling non-linear dynamical systemscorrupted by coloured noise using BP networks is pressented. Simulation results of an exampleshow the effectiveness of the approach.