基于双阶段网络模型的改进多变量广义预测控制  

An Improved Multivariable Generalized Predictive Control with A Two-Stage Network Model

作　　者：丁海丽 陈宽文 刘朋远 胡婷婷 梁飞 杨杨 DING Haili;CHEN Kuanwen;LIU Pengyuan;HU Tingting;LIANG Fei;YANG Yang(State Grid Ningxia Marketing Service Center(State Grid Ningxia Metrology Center),Yinchuan 750002,China;College of Automation&College of Artificial Intelligence,Nanjing University of Posts and Telecommunications,Nanjing 210023,China)

机构地区：[1]国网宁夏电力有限公司营销服务中心(国网宁夏电力有限公司计量中心),银川750002 [2]南京邮电大学自动化学院、人工智能学院,南京210023

出　　处：《计算机测量与控制》2022年第7期141-147,共7页Computer Measurement &Control

基　　金：国家自然科学基金项目(61873130);江苏省自然科学基金(BK20191377)。

摘　　要：针对多变量非线性时滞系统存在多变量间复杂的耦合情况,多输入多输出系统转化为多个多输入单输出系统,并构建多变量双阶段神经网络时滞预测模型;在考虑耦合关系的基础上,将改进比例性能指标型广义预测控制器引入到多变量系统中;该控制器含有预测控制增量表征系统未来变化趋势,将其作为当前控制量的补偿,优化控制性能;通过300 MW单元机W型火焰直吹式燃煤锅炉系统的仿真研究验证了控制方案的有效性。Aimimg at the coupling relationships in the multi-variable nonlinear time-delay system,the multi-input multi-output(MIMO)systems are transformed into the multiple multi-input single-output(MISO)systems,and a predictive model for the multi-variable double-stage network is constructed.On the basis of the coupling relationship,a generalized predictive controller with an improved proportional performance index is introduced into the multi-variable system.This controller contains the predictive control increment which represents the future trends of the system,which is taken as the compensation of the current control increment and optimizes the control performance.The effectiveness of the controller is verified by the simulation of the system with a 300MW unit W-type flame direct-fired coal-fired boiler.

关 键 词：多变量系统 广义预测控制 双阶段模型 非线性系统 神经网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]