基于高阶神经网络扩展卡尔曼滤波器算法的非线性动态系统辨识  

Identification of nonlinear dynamic system based on high-order neural networks and extended kalman filter algorithm

作　　者：刘春梅[1] 沈毅[1] 胡恒章[1] 葛升民[1] 

机构地区：[1]哈尔滨工业大学控制工程系,黑龙江哈尔滨150001

出　　处：《哈尔滨工业大学学报》2000年第2期107-110,113,共5页Journal of Harbin Institute of Technology

摘　　要：针对非线性动态系统辨识 ,采用高阶神经网络和径向基函数网络相结合的方法 ,神经网络的连接权值可作为系统的未知参数 ,用扩展卡尔曼滤波器 (EKF)算法来估计 ,确保了该方法的快速收敛 .具体模型的仿真结果表明该方法能快速收敛 ,并能方便的用于在线辨识 .High order neural networks and radial basis function networks are combined to identify nonlinear dynamic system. The linkweights of neural networks can be taken as the unknown parameters of system and can be estimated by EKF algorithm so that the fast convergence is assured. The simulations of specific model imply that this method is of fast convergence and can be used in on-line identification expediently.

关 键 词：高阶神经网络 卡尔曼滤波器 动态系统辨识 

分 类 号：TN713.7[电子电信—电路与系统] TP13[自动化与计算机技术—控制理论与控制工程]