基于改进K近邻和最小二乘支持向量机的钢铁蒸汽管网压力预测  

Improved KNN and LSSVM-based pressure prediction for steel steam networks

作　　者：赵烽敏 徐晓东 李峥 姚文贵 陈龙[2] ZHAO Fengmin;XU Xiaodong;LI Zheng;YAO Wengui;CHEN Long(Engineering Department,Shanghai Baoneng Information and Technology Co.,Ltd.,Shanghai 201900,China;School of Control Science and Engineering,Dalian University of Technology,Dalian 116081,China;Integrated Department,Shanghai Baoneng Information and Technology Co.,Ltd.,Shanghai 201900,China;Research and Development Department,Shanghai Baoneng Information and Technology Co.,Ltd.,Shanghai 201900,China)

机构地区：[1]上海宝能信息科技有限公司工程部,上海201900 [2]大连理工大学控制科学与工程学院,辽宁大连116081 [3]上海宝能信息科技有限公司综合部,上海201900 [4]上海宝能信息科技有限公司研发部,上海201900

出　　处：《冶金自动化》2022年第4期64-71,共8页Metallurgical Industry Automation

摘　　要：针对钢铁企业蒸汽管网压力变化波动大且存在流量数据随机缺失和异常等因素导致的难以对蒸汽系统进行实时有效调度的问题,提出了一种基于数据填补-最小二乘支持向量机(least squares support vector machine,LSSVM)的蒸汽管网压力预测方法。首先利用基于近邻噪声处理的K近邻缺失数据填补算法(eliminate neighbor noise K-nearest neighbor,ENN-KNN)对蒸汽流量原始数据进行异常点和缺失点填补,以降低数据中的异常点干扰;然后通过选取影响蒸汽管网压力的主要蒸汽发生和使用用户,包括煤精、海水淡化、烧结、干熄焦和炼钢等,利用LSSVM算法建模,对蒸汽管网压力进行预测。试验结果表明,基于本文所提方法的钢铁企业蒸汽管网压力的预测精度高、建模速度快,具有良好的抗干扰性和实用性,可以为蒸汽的合理调度提供科学的理论依据。Aiming at the problem of difficult real-time and effective scheduling of the steam system caused by factors such as large fluctuations in the pressure of the steam networks of the iron and steel enterprise and random missing and abnormal flow data,a least squares support vector machine(LSSVM) based data imputation method was proposed to predict the pressure of steam networks in steel enterprises.First,the missing data imputation algorithm(eliminate neighbor noise K-nearest neighbor,ENN-KNN) was used to fill in the anomalies and missing points in the raw steam flow data to reduce the interference of anomalies in the data.Then,the main steam generating and using users that affect the steam networks pressure,including coal refining,seawater desalination,sintering,dry predict the steam networks pressure.The experimental results show that the prediction of steam networks pressure in iron and steel enterprises based on the method proposed in this paper is highly accurate and fast,with good anti-interference and practicality,and can provide a scientific and theoretical basis for rational steam dispatching.

关 键 词：压力预测 蒸汽管网 数据缺失 数据填补 最小二乘支持向量机 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程] TF089[自动化与计算机技术—控制科学与工程]