基于动态神经网络非线性结构辨识的研究

The Study of Non-Linear Structure’s Identification using Dynamic Neural Network

利用 Elman动态递归神经网络 ,对非线性结构进行黑箱辨识 ,建立了它的非线性状态方程 ;提出了加快网络收敛速度的自适应学习算法。辨识结果表明 ,动态递归网络模型优于传统辨识模型 ,适于非线性、不确定结构的辨识。

In this paper, the Elman’s recurrent neural network is developed to identify non linear structure in block box, and a nonlinear state space model of structure is determined. An accelerated adaptive learning rate algorithm is proposed. Results of identification show that the Elman’s recurrent model is superior to the traditional model. It is adaptive to the identification of the non linear and uncertain structure.