基于机器学习的造纸用能负荷特征日获取模型  

A Model for Obtaining the Representative Day of Papermaking Energy Load Based on Machine Learning

作　　者：刘昌 何正磊 朱小林[1] 满奕[1,2] LIU Chang;HE Zhenglei;ZHU Xiaolin;MAN Yi(State Key Lab of Pulp and Paper Engineering,Guangzhou 510640,China;Pazhou Lab,Guangzhou 510335,China)

机构地区：[1]华南理工大学浆造纸工程国家重点实验室,广东广州510640 [2]琶洲实验室,广东广州510335

出　　处：《造纸科学与技术》2023年第2期6-12,共7页Paper Science & Technology

摘　　要：造纸的用电负荷及用热负荷数据是反映造纸企业运行效益和工艺合理性的重要指标,对于企业生产来说,在大量的历史用能负荷数据中,挖掘出企业用能的分布情况对分析生产耗能情况,优化设备运行具有重要的意义。传统的用能分析方法存在着滞后性,无法保证结果的准确等问题。随着工业大数据的发展,基于机器学习的数据挖掘方法为用能负荷数据分析提供了有效的途径。通过相关性分析和降维算法,剔除数据冗余的信息。在此基础上,通结合聚类算法获取全年用能负荷特征日。结果表明,采用使用K均值与主成分分析法相结合能有效捕捉输入数据分布特征的同时获取代表全年用能负荷的数据。Electricity load and heat load data of paper manufacturing are important indicators reflecting the operational efficiency and process rationality of paper manufacturing enterprises.For enterprise production,in a large number of historical energy load data,mining the distribution of enterprise energy use is important to analyze the production energy consumption and optimize equipment operation.Traditional energy analysis methods have problems such as lagging and inability to guarantee the accuracy of results.With the development of industrial big data,the data mining method based on machine learning provides an effective way to analyze energy consumption load data.Through correlation analysis and dimensionality reduction algorithm,the redundant information of data is eliminated.On this basis,the annual energy use load characteristic days are obtained by combining with clustering algorithm.The results show that the combination of Kmeans and principal component analysis can effectively capture the characteristics of the input data distribution and obtain the data representing the annual energy load.

关 键 词：用能负荷 机器学习 相关性分析 聚类算法 

分 类 号：TS71[轻工技术与工程—制浆造纸工程]