基于神经网络补偿的广义离散模型参考自适应控制系统设计  

A Design Method for Generalized Discrete Model Reference Adaptive Control System Based on the Compensation via a Neural Network

作　　者：张秀玲[1] 

机构地区：[1]燕山大学电气工程学院自动化系,秦皇岛066004

出　　处：《系统仿真学报》2002年第1期125-127,共3页Journal of System Simulation

摘　　要：用一个常规线性模型对被控对象进行辨识,线性模型辨识的余差用一个神经网络进行补偿,线性模型和神经网络共同构成对象的辨识模型。利用参考模型的输出状态和被控对象的预测值,提出了适于任何形式被控对象的广义离散MRACS设计方法,适用于线性、非线性、最小相位和非最小相位系统。仿真结果表明,系统响应速度快,能够跟踪任意给定的参考序列,具有较强的鲁棒性和较好的控制精度,从而拓宽了模型参考自适应控制的应用范围。The plant is identified by a normal linear model, and the deviation identified by the linear model is compensated via a neutral network. The identification model consists a linear model and a neutral network. By using the state of model抯 output and plant抯 forecast, a design method is presented for generalized discrete MRACS (Model Reference Adaptive Control System). It applies to linear, nonlinear, minimum phase and non-minimum phase system.. The simulation shows that this plant has fast response speed and can follow the arbitrary given reference sequence. It has stronger robustness and better control precision.

关 键 词：神经网络 离散模型参考自适应控制系统(MRACS) 补偿 辨识 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]