基于随机搜索算法和AdaBoost模型预测LF精炼过程脱硫率  

Predicting desulfurization ratio during LF refining process basedon Random Research and AdaBoost model

作　　者：严旭梅 陈超[1,2] 王楠 陈敏 Yan Xumei;Chen Chao;Wang Nan;Chen Min(Northeastern University Key Laboratory for Ecological Metallurgy of Multimetallic Mineral(Ministry of Education),Shenyang 110819,China;Northeastern UniversitySchool of Metallurgy,Shenyang 110819,China)

机构地区：[1]东北大学多金属共生矿生态化冶金教育部重点实验室,沈阳110819 [2]东北大学冶金学院,沈阳110819

出　　处：《材料与冶金学报》2023年第5期430-436,443,共8页Journal of Materials and Metallurgy

基　　金：国家自然科学基金项目(52174301,52074077,51974080);中央高校基本科研业务费由教育部资助项目(N2125018)。

摘　　要：脱硫是LF精炼过程的主要任务之一.为达到稳定脱硫率的目的,利用AdaBoost模型对某钢铁厂LF精炼过程的实际生产数据进行建模,通过对比网格搜索算法、随机搜索算法及贝叶斯优化算法对AdaBoost模型超参数的优化效果和优化时间的影响,讨论了实际应用中AdaBoost模型的超参数优化方案.此外,根据实验对比结果及实际应用情况,确定了基于随机搜索算法和AdaBoost模型的LF精炼过程脱硫率预测模型.结果表明:该模型可以实现脱硫率误差在±0.07,±0.06和±0.05时,准确度分别为95.3%,93.0%和86.0%.Desulfurization is one of the main tasks in the LF refining process.In order to achieve the purpose of stabilizing the desulfurization rate,the AdaBoost model was used to model the actual production data of the LF refining process of a steel plant.The grid search algorithm,random search algorithm,and Bayesian optimization algorithm were used to compare the optimization effects and optimization time of the AdaBoost model hyperparameters,and the hyperparameter optimization scheme of the AdaBoost model in practical applications was discussed.In addition,based on the experimental comparison results and practical application,a prediction model for the desulfurization rate of the LF refining process based on the random search algorithm and the AdaBoost model was determined.The results show that the hit rates of the desulfurization ratio can achieve 953%,930%,and 860%within the prediction error of±007,±006,and±005 respectively.

关 键 词：脱硫率 LF精炼过程 预测模型 AdaBoost模型 超参数优化 

分 类 号：TF721[冶金工程—钢铁冶金]