Based on an analysis on the modeling principles of nonlinear system, the identification of a nonlinear system was realized with Diagonal Recurrent Neural Networks (DRNN). Serial-parallel identification architecture was applied in the modeling. Time variation was taken into account in the adjustment algorithm of weights. Compared with static neural network, the method based on DRNN displays better ability to deal with a dynamic system, due to its advantages such as without the need of system order number, a smaller neural network structure and a faster convergence. Simulation results testified the feasibility and validity of the proposed method.
Narendra K S, Parthasarathy K. Identification and control of dynamical systems using neural networks[J].IEEE Tran on Neural Networks.1990,1(1):4-27
Ku Chao-Chee, Lee K Y. Diagonal recurrent neural networks for dynamic system control[J].IEEE Tran on Neural Networks.1995,6(1):144-156
Liang X, Lee K Y. Comments on "Diagonal recurrent neural networks for dynamic system control" Reproof of theorems 2 and 4 and reply[J].IEEE Tran on Neural Networks.1997,8(3):811-814
����Ң,��ƽ��.��������ģ������[M].����:�廪��ѧ������,1998 Zhang Naiyao,Yan Pingfan. Neural network and fuzzy control[M]. Beijing:Tsinghua University Press, 1998(in Chinese)
Chen Ping, Qiu Lihua. System Identification in hydraulic servo system with diagonal recurrent neural networks . In:Lu Yongxiang, Chen Ying, Xu Li,eds. Proceedings of the Fifth International Conference on Fluid Power Transmission and Control (ICFP'2001) . Beijing:Beijing World Publishing Corporation, 2001. 499~503
������ ,Ϳ ��.��Ԫ�������[M].����:��е��ҵ������, 1998 Wang Yongji, Tu Jian. Neuron network control[M].Beijing:China Machine Press,1998(in Chinese)