摘要
针对工业生产过程中的多变量耦合系统采用传统控制方法不能达到满意的效果,提出了一种基于神经网络的PID解耦控制方案。在实验研究中,采用改进型动态BRF神经网络辨识器,在线辨识多变量系统的非线性时变模型,同时自动调整PID控制器各项参数,最终实现对系统的智能化解耦控制。给出了BRF神经网络的拓扑结构和算法,并对一组二变量强耦合时变系统的控制过程进行了计算机仿真,结果表明:基于BRF神经网络的PID控制不仅超调量小、响应速度快、控制精度高,而且具有很强的鲁棒性和自适应能力。该设计方案使得解耦后的多变量系统具备了良好的动、静态特性。
In industrial process control fields, the traditional control methods can＇t acquire satisfying result for multivariate coupling system. A design scheme of decoupling PID control was proposed for multivariable system based on RBF neural network, which identified the non-linear and time-varying system model on line by means of an improved dynamic RBF neural network identifier, adjusted automatically each parameter of the PID controller and achieved intelligence decoupling control of the system. Both neural network structure and algorithm were given and a double variable and strong coupled time-varying control system was real-timely simulated in computer. The simulation results demonstrate that this control system works well with quick response, good robustness and perfect self adaptation. This decoupling design is effective to improve the dynamic behavior and static characteristic of the decoupled system.
出处
《系统仿真学报》
EI
CAS
CSCD
北大核心
2008年第3期627-630,634,共5页
Journal of System Simulation
基金
国家科技部"十五"攻关项目(2001BA204B01-03)
作者简介
焦竹青（1983-），男，山东烟台人，研究方向为智能控制与系统仿真；
屈百达（1956-），男，辽宁北镇人，满族，博士，教授，研究方向为现代控制技术与应用；
徐保国（1950-），男，江苏淮阴人，博导，教授，研究方向为过程控制、智能仪表及现场总线等。
