摘要
本研究建立了一种结合T-PLS(全潜结构投影)和GRA(灰色关联度分析)的造纸干燥过程能耗非优原因追溯模型。该模型首先基于机理知识和方差特性,去除造纸干燥过程生产数据中的非核心特征变量,并通过3σ原则和箱形图剔除异常值;然后使用施胶前定量与卷取车速数据,结合K-Means聚类算法,实现不同生产状态的分类;最后针对不同的生产状态,对比T-PLS和PLS建立的经济指标计算模型,选用基于T-PLS-GRA算法,构建造纸干燥过程能效非优原因追溯模型。结合国内某造纸厂实时生产数据对该模型进行了验证。结果表明,该模型基于经济指标判断工业生产状态过程,对非优过程预测精准率为77.7%,可较好地跟踪造纸过程设备运行状态的变化过程;并且能追溯非优状态原因及整个工况下,非优生产状态中最大贡献变量出现频次,为企业改进工艺流程及节能优化提供了参考依据。
In this study,an energy consumption non-optimal cause identification model combining T-PLS total latent structural projection and GRA grey correlation analysis in paper drying process was established.The model firstly removed the non-core characteristic variables of production data in paper drying process based on the mechanism knowledge and variance characteristics,and eliminated the outliers through the 3σprinciple and box plots;then the model used the data of pre-sizing quantitatively and winding speed,and realized the classification of different production states by combining with the K-Means clustering algorithm;finally,in view of the different production states,the model compared the economic indexes calculation models established by T-PLS and PLS,and choosed energy efficiency non-optimal reason tracing model based on the T-PLS-GRA in paper drying process.The model was verified with the real-time production data of a paper mill in China,and the results showed that the model judged the industrial production state process based on the economic indexes,and the prediction accuracy rate of the non-optimal process was 77.7%,which can better track the change process of the running state of the equipment in the papermaking process.The model can trace the reasons for the non-optimal state and the frequency of occurrence of the largest contributing variable in the non-optimal production state during the whole working process,which provides a reference basis for the enterprise to improve the process and optimize the energy saving.
作者
戴景波
陈晓彬
方子言
郑启富
张垚
张敏
董云渊
廖建明
DAI Jingbo;CHEN Xiaobin;FANG Ziyan;ZHENG Qifu;ZHANG Yao;ZHANG Min;DONG Yunyuan;LIAO Jianming(College of Chemical Engineering,Zhejiang University of Technology,Hangzhou,Zhejiang Province,310014;College of Chemical and Material Engineering,Quzhou University,Quzhou,Zhejiang Province,324000)
出处
《中国造纸学报》
CAS
CSCD
北大核心
2024年第1期91-99,共9页
Transactions of China Pulp and Paper
基金
国家自然科学基金(62303265)
浙江省重点研发计划(2024C03120)
浙江省基础公益研究计划(LTGN24B060001,LGN21C030001)
衢州市科技计划项目(2023K230)。
关键词
造纸干燥过程
能效
非优原因追溯
机器学习
paper drying process
energy efficiency
non-optimal cause identification
machine learning
