基于递归小波神经网络的非线性动态系统仿真  

Nonlinear Dynamical System Simulation Based on Recurrent Wavelet Neural Network

作　　者：赵凤遥[1] 马震岳[1] 

机构地区：[1]大连理工大学土木水利学院

出　　处：《系统仿真学报》2007年第7期1453-1455,1539,共4页Journal of System Simulation

基　　金：国家自然科学基金(50279003)

摘　　要：为提高动态递归神经网络的动态系统仿真能力,在Elman神经网络的基础上,提出动态递归小波神经网络(RWNN),给出了其动态梯度下降算法,并将其成功应用于非线性动态系统仿真。仿真算例表明,该网络具有收敛快,精度高等优点,仿真效果很好,同时具有较好的泛化性能,具有广阔的应用前景。For the aim of improving the dynamical system simulation ability of recurrent neural network, Based on Elman network, the recurrent wavelet neural network （RWNN ） was proposed in the paper, and the dynamic gradient descent algorithm of RWNN was given. The RWNN could be used in the nonlinear dynamical system simulation successfully. Simulation example shows that RWNN has a faster convergence speed and a better precision in calculation, and a good result on the nonlinear dynamical system simulation is obtained. At the same time, the network also has better generalization ability, which means it has a broad prospect on application.

关 键 词：ELMAN神经网络 递归小波神经网络(RWNN) 梯度下降算法 非线性动态系统 仿真 

分 类 号：TP393.9[自动化与计算机技术—计算机应用技术]