【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 .更多还原

【Abstract】 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.更多还原