未知非线性动态系统神经网络模型在线识别

神经网络为未知非线性动态系统的建模提供了一条新途径。本文探讨了只用单个隐含层的前向神经网络对未知非线性动态系统的识别。只要动态系统输入输出可测量，未知非线性动态系统就能在线识别。在线识别算法依据性能函数在输出空间最优原则导出，有别于常规的BP算法。求解速度快，适合于在线识别。仿真实例进一步表明，采用神经网络建立未知非线性动态系统的在线模型具有可行性。

Unknown Nonlinear Dynamical System Identification and Online Algorithm Using Multilayered Feedforward Neural Network

Multilayered neural network offers a new exciting alternative for modelling unknown nonlinear dynamical system. This paper investigates the identification of unknown nonlinear dynamical system using multilayered feedforward neural network with a single hidden layer. The online identifiaction of the unknown nonlinear dynamical system can be implemented when only the inputs and outputs are accessible for measurement. A novel parameter learning algorithm is derived for the multilayered feedforward neural network model based on gradient descent in the output space. The simulation results are presented to demonstrate that the model of an unknown nonlinear dynamical system is built with the multilayered feedforward neural network model.