论文部分内容阅读
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。
This paper presents a neural network adaptive method. In the method, the memory element network is adopted to identify the object model using the memory element neural network, and an adaptive PID controller is implemented with a single neuron. The output error of the controlled object gets the output error of the controller after it is backtracked by the memory element identification network so as to correct the network weight of the controller. Since the memory element network does not need to introduce the delay operator, the linear dynamic can be approximated to any order, The accuracies and accuracy of error feedback are analyzed. The neuron PID controller has a very simple structure and algorithm to ensure the real-time performance of adaptive control. A large number of simulation results show that this method can effectively deal with nonlinear objects.