改进的TSVR转炉终点预测模型研究  

Study on Improved TSVR Prediction Model for Controlling End-point Situation of Converter

作　　者：汪淼 高闯[2] 艾新港[1] 李胜利[1] WANG Miao;GAO Chuang;AI Xingang;LI Shengli(School of Materials and Metallurgy,University of Science and Technology Liaoning,Anshan 114051,Liaoning,China;School of Electronic and Information Engineering,University of Science and Technology Liaoning,Anshan 114051,Liaoning,China)

机构地区：[1]辽宁科技大学材料与冶金学院,辽宁鞍山114051 [2]辽宁科技大学电子与信息工程学院,辽宁鞍山114051

出　　处：《鞍钢技术》2022年第6期29-32,共4页Angang Technology

基　　金：“十四五”国家重点研发计划(2021YFB3702005);2020年度辽宁省科技重大专项计划(2020JH1/10100001);辽宁省科技厅博士启动基金项目(No.2021-BS-244)。

摘　　要：准确的预测转炉炼钢终点碳含量和终点温度,对于提高钢的生产效率具有重要意义。建立了一种基于加权拉格朗日ε-孪生支持向量回归机(WL-ε-TSVR)的转炉终点预测模型。根据采集到的某钢厂260 t转炉的实际生产数据,将450组样本用于模型的训练,50组样本用于测试模型的准确性。仿真结果表明,预测误差碳含量控制在±0.005%范围,终点温度控制在±10℃范围,模型的命中率分别达到88%和92%;另外,双命中率达80%。所建模型具有较高的精度,能够更好的满足钢厂的生产需求。Accurately predicting the content of end-point carbon in a converter and end-point temperature of molten steel was of great significance for improving the production efficiency of molten steel by converter.The converter end-point prediction model based on weighted Lagrangeε-Twin Support Vector Regression Machine(WL-ε-TSVR)was established for discussing.According to the actual production data on the 260 t converter of a steel plant,450 sets of samples were used for model training and 50 sets of samples were used for testing the accuracy of the model.Then the simulation testing results showed that the prediction errors were controlled in the range of±0.005%for content of carbon and±10℃for end point temperature,and the hit rates for the model were 88%and 92%respectively.In addition,the double hit rate for the model was 80%.So the established model had high precision,being well capable of meeting the requirements in production asked by steel mill.

关 键 词：转炉炼钢 预测模型 孪生支持向量回归机 

分 类 号：TP39[自动化与计算机技术—计算机应用技术] TF71[自动化与计算机技术—计算机科学与技术]