Hammerstein系统自适应神经网络控制算法的收敛性分析

摘要

An adaptive control algorithm for deterministic hammerstein system based on neural networks is presented. Considering the nonlinearity and the adaptive learning ability of the neural network, two neural networks are used as estimator and controller of the control system respectively. The train of the networks is used Widrow-Hopff rule. The convergence of the control algorithm is analyzed and the results are shown that the system is global convergence and has bounds for the input and output.%对一类确定性Hammerstein系统，给出了基于神经网络的自适应控制算法。考虑到神经网络的非线性特点，特别是其自适应学习能力，控制系统采用两个神经网络分别作为估计器和控制器，通过在线训练网络的权重来获得模型参数和控制输入。神经网络的训练用Widrow-Hoff学习规则。对算法的全局收敛性进行分析表明系统具有总体收敛性，输入输出有界。