一类神经网络整定PID参数的非线性系统解耦方法

对具体非线性系统有不同的解耦方法，如模糊解耦、神经网络解耦等。因神经网络有自学习功能以及对参数摄动的不敏感，故对某些特殊的非线性系统采用神经网络解耦作为补偿器进行输入输出解耦，能够获得较强的鲁棒性。首先介绍了神经网络解耦的一般结构，然后给出了递归多层神经网络的学习方法。针对一类非线性系统设计了一类基于递归多层神经网络整定PID参数的多变量非线性系统的解耦控制器。从仿真可以看出，对于文中所给出的非线性系统，这种递归多层神经网络的解耦控制器对于输入输出解耦是可行的。

Method Using Neural Network to Adjust PID Parameters for Decoupling Nonlinear System

For the decoupling of nonlinear system,a decoupling method has been put forward which is aimed at nonlinear system,such as fuzzy control decoupling and neural network decoupling.Because of the function of self-learning and non-sensitive of perturbation of parameters,neural network is used for input and output decoupling of some special nonlinear systems.In this way,it can gain much better characteristic of robust.Firstly,the common structures of decoupling of neural network are introduced.Secondly,the research methods of Diagonal Recurrent Neural Network (DRNN) are introduced.A PID decoupling method of multi-variable and nonlinear system using Diagonal Recurrent Neural Network is advanced.The emulative results indicate that the decoupling method is feasible and effective for the given system.