基于相关向量机样本选择的钢铁企业副产煤气系统预测  

Prediction of by-product gas system in iron and steel enterprises based on sample selection of relevance vector machine

作　　者：闫亚亮 陈龙[1,2] 赵珺[1,2] 王伟[1,2] YAN Yaliang;CHEN Long;ZHAO Jun;WANG Wei(School of Control Science and Engineering,Dalian University of Technology,Dalian 116024,China;Key Laboratory of Intelligent Control and Optimization for Industrial Equipment,Ministry of Education,Dalian 116024,China)

机构地区：[1]大连理工大学控制科学与工程学院,辽宁大连116024 [2]工业装备智能控制与优化教育部重点实验室,辽宁大连116024

出　　处：《冶金自动化》2023年第3期35-43,共9页Metallurgical Industry Automation

基　　金：国家重点研发计划项目(2017YFA0700300);国家自然科学基金项目(62125302,62003072,61833003,U1908218)。

摘　　要：针对副产煤气系统的运行数据冗余度高、噪声强等特点,提出了一种基于相关向量机算法(relevance vector machine,RVM)的训练样本选择的副产煤气系统预测算法。鉴于RVM算法具有相关样本自动选择的特点,提出采用此算法对原始训练集数据进行训练,以获取的相关向量作为基本训练集;之后利用K近邻算法(K nearest neighbor,KNN)实现对基本训练集合的样本增强,并以此作为新的训练集,从而实现样本的去冗余,提高训练样本质量,提升算法效率与预测准确度。采用国内某钢铁厂高炉煤气数据进行试验,试验效果表明,本文所提的方法可有效针对高炉煤气数据进行样本选择,并以较快的模型训练效率获得较高的煤气柜柜容预测精度,预测结果可为钢铁煤气系统的优化调度工作提供基础。In view of the features of high redundancy and strong noise in the operation data of byproduct gas system,a prediction algorithm based on relevance vector machine(RVM)was proposed to select training samples for by-product gas system.Since the RVM algorithm has the feature of automatic selection of relevant samples,this algorithm was proposed to train the original training set data,and the obtained relevance vector was used as the basic training set.Then,K nearest neighbor(KNN)algorithm was used to enhance the samples of the basic training set and serve as the new training set,so as to eliminate the redundancy of samples,improve the quality of training samples,and enhance the efficiency and prediction accuracy of the algorithm.The experiment was conducted using blast furnace gas data from a domestic steel plant and the results show that the method proposed in this paper can effectively select samples for blast furnace gas data,and obtain higher prediction accuracy of gas holder capacity with faster model training efficiency.The prediction results can provide a basis for the optimization of steel gas system scheduling.

关 键 词：样本选择 相关向量机 副产煤气系统 预测 

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]