用神经网络进行非线性离散动态系统辨识的可行性  

The Feasibility of Using Neural Networks for the Identification of Nonlinear Discrete Dynamic Systems

作　　者：刘延年[1] 冯纯伯[1] 

机构地区：[1]东南大学自动化所

出　　处：《控制理论与应用》1994年第4期413-420,共8页Control Theory & Applications

摘　　要：本文通过证明由神经网络组成的动态系统具有可以逼近满足一定条件的非线性离散动态系统到任意精度的能力，以及讨论将其作为辨识模型对非线性离散动态系统进行实时辨识时系统的稳定性，对用神经网络进行非线性离激动态系统辨识的可行性进行了探讨。In this paper, by proving that dynamic systems constituted by neural networks can approximete a kind of nonlinear discrete dynamic systems with arbitrary accuracy, and analyzing the stability of applying neuralnetworks to the real-time identification of the nonlinear discrete dynamic systems, we discuss th feasibility of us ing neural networks for the identification of nonlinear discrete dynamic systems. We also address the problems remained to be solved.

关 键 词：神经网络 离散动态系统 系统辨识 

分 类 号：TP271.8[自动化与计算机技术—检测技术与自动化装置]