一种记忆元网络自适应控制方法

本文提出了一种神经网络自适应方法。该方法采用记忆元网络采用记忆元神经网络进行对象模型辩识，用单个神经元实现了自适应ＰＩＤ 控制器。被控对象输出误差经记忆元辩识网络反传后得到控制器的输出误差，以此修正控制器网络权值，由于记忆元网络无需引入延迟算子，能够逼近任意阶线性动态，保证了模型辩识的精度和误差返传的精度，神经元ＰＩＤ 控制器具有极为简单的结构与算法，保证了自适应控制的实时性，大量仿真结果表明该方法可以有效地应付非线性对象。

A Adaptive Control Method Using Memory Neuron Network

In this paper, a neuron network adaptive control method is presented. In this method, Memory Neuron Network(MNN) is used for identificating plant model and single neuron is used as PID controller. The output error of the controller is obtained by back propagating the output error of the plant through the MNN. Weights of the controller are updated on line according to it. Because the MNN is able to approximate arbitrary order non linear dynamicas without using the tapped delay line,it can ensure the identification precision and that of propagated error.The structure and algorithm of the controller are simple and the real time control is ensured. Simulation results show that this method can effectively treat with some non linear objects.