长短时记忆网络在除尘内模控制中的应用  

Application of Long Short-Term Memory Network in Internal Model Control of Dust-removal System

作　　者：朱明山 刘剑敏[1] 胡小虎[1] 余数[1] 酆强胜 王华秋 ZHU Ming-shan;LIU Jian-min;HU Xiao-hu;YU Shu;FENG Qiang-sheng;WANG Hua-qiu(Ningbo Cigarette Factory,China Tobacco Zhejiang Industrial Co.,Ltd.,Ningbo 315504,China;Liangjiang School of Artificial Intelligence,Chongqing University of Technology,Chongqing 401135,China)

机构地区：[1]浙江中烟工业有限责任公司宁波卷烟厂,浙江宁波315504 [2]重庆理工大学两江人工智能学院,重庆401135

出　　处：《控制工程》2021年第2期251-257,共7页Control Engineering of China

基　　金：浙江中烟科技项目:除尘房烟末回收关键技术研究与应用(ZJZY2018D013)。

摘　　要：卷接机组风力除尘系统受到多种因素影响会造成负荷波动,而且风压和风量之间存在耦合关系,传统的控制策略一般将控制参数固定或者缺乏预测机制,无法很好地解决生产工艺的变化带来的除尘系统不稳定问题。提出了用Goldstein搜索算法对长短时记忆网络(LSTM)的隐层记忆单元个数进行调优,构建了基于长短时记忆网络的卷接机组风力除尘系统内模控制,能够自动调整变频器频率和补风阀开度,实现了对除尘系统的恒压恒流控制。通过仿真实验,采取的内模控制系统具有良好的跟踪性和抗扰动性,解决了除尘系统在受到干扰时能快速稳定的问题。The air supplying and dust-removal system of cigarette making machine is affected by many factors to cause load fluctuation.And there is a coupling relationship between wind pressure and air volume.Traditional control strategies usually fix control parameters or lack predictive mechanism.The problem of dust-removal system instability caused by production process change can’t be solved by traditional control strategies.In this paper,Goldstein search algorithm is presented to optimize the number of hidden memory cells in long short-term memory(LSTM)network.And an internal model control(IMC)based on long short-term memory network is designed for air supplying and dust-removal system.The controller can automatically adjust the frequency of inverter and the opening degree of supplementary air valve.The constant pressure and flow control of air supplying and dust-removal system is realized by the IMC controller.Through simulation experiments,the presented internal model control system has good tracking and anti-disturbance performance.The rapid stability problem of dust-removal system is solved when the system is disturbed.

关 键 词：风力除尘系统 长短时记忆网络 Goldstein搜索算法 内模控制 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置] TP183[自动化与计算机技术—控制科学与工程]