基于 Hammerstein 型神经网络的非线性动态系统辨识

摘要

Hammerstein model is widely applied to the identification of nonlinear systems,which consists of a nonlinear static gain part in cascade with a linear dynamic part.We propose a Hammerstein-type neural network (HNN)to simulate the conventional Hammerstein model,and apply it in the identification of nonlinear dynamic systems.The Lipschitz entropy is employed to determine the order of HNN,and the back-propagation (BP)algorithm is used for training the network weights.Simulation results show that HNN has satisfied identification performance on nonlinear dynamic systems.%Hammerstein 模型广泛应用于非线性系统的辨识中，其结构是由非线性静态增益部分和一个线性动态部分串联。提出一种 Hammerstein 型神经网络用来模拟传统的 Hammerstein 模型，并将其应用于非线性动态系统的辨识中。由 Lipschitz 熵来确定 Ham-merstein 型神经网络的阶次，并利用反向传播算法对网络权值的进行训练。仿真结果表明，Hammerstein 型神经网络具有较好的非线性动态系统辨识性能。