基于BP神经网络的多变量PID解耦控制.pdfVIP

论述与研究 基于BP神经网络的多变量PID解耦控制 Multivariate Decoupling Control Based on Back Propagation Neural Netw ork 薛昊洋 刘红军 （华北电力大学自动化系，保定市 071003） Xue Haoyang Liu Hongj un (Automatic Department of North China Electric Power University , Baoding 07 1003) 【摘 要】基于神经网络实现智能PID控制的策略，它以经典的PID控制为基础,通过神经网络参数整定 实现,进而进行自学，用于多变量系统的解耦控制。本文给出了网络的结构和算法，对一组二 变量强耦合时变系统进行了仿真。通过计算机仿真证明了基于神经网络的PID控制器网络结构 简单规范具有良好的自学习和自适应解耦控制能力。系统易于实现，融解耦器与控制器于一 体，适用于非线性多变量系统的解耦控制。能够使解耦后的系统具有良好的动态和静态性能， 特别是依据BP控制规律来确定网络连接权的初值，还具有参数快速收敛的优点。 【关键词】PID控制 神经网络 多变量系统 解耦控制 Abstract: Intelligent PID control strategy , which is based on neural network, is according to classical PID control, realized through neural network parameter setting which has self study function for multivariate decoupling control. The structure and the algorithm were given and th e real-time simulation results o f doub le variable and strong- coupled time-varying sy stem were shown in the pap er. It proved th at PID control based on neural network has preferably self study and self adapting decoupling control ability through simulation . The sy stem, which inosculates the decoupler and controler, is easy to implement and applicable for multivariate decoupling control. It makes the decoupled sy stem have better dynamic behavior and static characteristic. Esp ecially , it makes parameters astringe fast when determing initial value of network according to BP control law . Key words: PID Control Neural Network Multivariable Sy stem Decoupling Control 引言 制算法得到相应各变量的控制量参与控制，并能在 多变量控制系统 （MIMO）属于高级而又复杂的 该过程中实现解耦控制，而不用给定样本信号进行 过程控制系统，从广度和深度上都超过了单