Abstract: The theory and method of dynamic system identification by dynamic recurrent neural network are studied.An improved Elman neural network is successfully used to identify the nonlinear dynamic system even though without any prior information of identified system.Simulation results show that the Elman neural network has higher learning speed and better generalization ability than the feedforward neural network,and that it is suitable for the nonlinear dynamic system identification．

摘要： 研究了应用动态递归神经网络实现动态系统辨识的原理和方法,在没有被辨识对象的先验知识情况下,通过改进的Elman网络实现了非线性动态系统的辨识。仿真结果表明,与前馈网络相比,Elman网络具有学习速度快、泛化能力强的特点,可用较小的网络结构实现高阶系统的辨识,适用于具有本质非线性动态系统的辨识。