基于数据驱动的纸浆洗涤过程优化控制  

Data-driven Optimal Control for Pulp Washing Process

作　　者：单文娟 汤伟[2] 刘炳 SHAN Wen-juan;TANG Wei;LIU Bing(College of Light Industry Science and Engineering,Shaanxi University of Science & Technology,Xi'an,Shaanxi Province,710021;College of Electrical and Information Engineering,Shaanxi University of Science & Technology,Xi'an,Shaanxi Province,710021;Huawei Technology Co.,Ltd.,Xi an,Shaanxi Province,710077)

机构地区：[1]陕西科技大学轻工科学与工程学院,陕西西安710021 [2]陕西科技大学电气与信息工程学院,陕西西安710021 [3]华为技术有限公司,陕西西安710077

出　　处：《中国造纸学报》2018年第4期44-49,共6页Transactions of China Pulp and Paper

基　　金：陕西省重点科技创新团队计划项目(2014KCT-15)

摘　　要：针对纸浆洗涤过程的残碱和黑液波美度不能直接在线测量、控制回路的动态特性难以用数学模型描述的问题,通过研究数据驱动操作模式的优化思想,提出了基于数据驱动的纸浆洗涤过程综合优化的方法。基于PCA-BP神经网络法和多元回归分析建立了残碱和黑液波美度的预测模型及工况综合评价模型。基于大量工业运行数据和工况评价模型对纸浆洗涤过程的操作模式进行优化,构建优化操作模式库。以高产、低成本、低耗为目标对优化模式库寻优,找出最优操作模式。通过实际应用,证实该方法能准确预测残碱和黑液波美度,并在满足洗涤质量的同时,使出浆量提高,清水加入量减少,达到优化生产的效果。Considering the difficulties of modeling,online-measurement of residual soda and Baume degree indexes,and optimal control in pulp washing process,a data-driven operational-pattern optimization method for pulp washing process was proposed in this paper.Firstly,it described the basic concepts about data-driven operational-pattern for pulp washing process.Secondly,it established the data-driven prediction models of residual soda,Baume degree and operation performance integrated evaluation by PCA-BP neural network and multivariate logistic regression.Operational-pattern database was conducted based on large number of industrial operation data and the optimized operation patterns in washing process by the operation performance evaluation model.Thirdly,under the target of high production,low cost and low water consumption,the optimal operational-pattern was obtained by ant colony optimization algorithm from the optimized operational-pattern database.The practical results showed this method was effective in pulp washing process.

关 键 词：纸浆洗涤过程 数据驱动 神经网络 动态建模 操作模式 

分 类 号：TS736[轻工技术与工程—制浆造纸工程]