摘要
本文提出两种基于神经网络的多变量解耦控制方法。方法1通过设计神经网络补偿装置,使得包括补偿神经网络在内的广义对象的 Bristol 第一系数矩阵为对角阵;方法2首先定义了神经网络的串联、并联和反馈运算,然后在此基础上设计一个神经网络补偿装置,使得包括补偿神经网络在内的广义对象矩阵为对角阵。将其用于某二元精馏塔的塔顶和塔底组分控制,仿真结果证实了本文方法的有效性。
Two novel multivariable decoupling control methods using neural nets are presented. In one method, a neural net interaction compensator (NNIC) is designed to make the first Bristol's coefficient matrix including the NNIC of generalized plant become a diagonal matrix. In the other method , the operating of neural net' s cascade, parallel and feedback is defined and then a NNIC is constructed to make the generalized plant matrix including NNIC become a diagonal matrix. The two methods are applied to overhead and bottom composition control of a binary distillation column. Simulation results show that both methods are efficient.
出处
《控制与决策》
EI
CSCD
北大核心
1993年第6期409-413,共5页
Control and Decision
基金
国家教委博士点基金
关键词
神经网络
解耦控制
多变量
neural networks, multivariable systems, decoupling control , learning control, distillation control
