Scholarlink

As for the steel rolling reheating furnace(SRRF) in steel production, the prediction of heat required for slab heating is of great significance to the use of slab mixed gas supply and the quality of slab under different heating conditions. In this paper, a two stage heat prediction model for slab heating is proposed by combining the multivariate linear-regression variable parameter spatio-temporal zoning model(MLR-VPST) and the Bayesian optimization gradient boosting decision tree(BO-GBDT). In the first stage, since there are different heating zones in a reheating furnace, the real-time slab temperature at the outlet of each zone is predicted by using the MLR-VPST. Then, on the basis of the temperature prediction results, the GBDT is used to predict the required heat in the second stage, where the model parameters are optimized by using the Bayesian optimization. Real operational data of SRRF of a steel plant in China are employed. The experimental results show that the method can accurately estimate the heat required for slab heating and help to effectively arrange the consumption of mixed gas.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Modeling and Prediction of Gas Consumption for Slab Heating in Steel Rolling Reheating Furnace Based on Gradient Boosting Decision Tree with Bayesian Optimization

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Chong Tian,
  • Long Chen,
  • Jun Zhao,
  • Wei Wang

摘要

As for the steel rolling reheating furnace(SRRF) in steel production, the prediction of heat required for slab heating is of great significance to the use of slab mixed gas supply and the quality of slab under different heating conditions. In this paper, a two stage heat prediction model for slab heating is proposed by combining the multivariate linear-regression variable parameter spatio-temporal zoning model(MLR-VPST) and the Bayesian optimization gradient boosting decision tree(BO-GBDT). In the first stage, since there are different heating zones in a reheating furnace, the real-time slab temperature at the outlet of each zone is predicted by using the MLR-VPST. Then, on the basis of the temperature prediction results, the GBDT is used to predict the required heat in the second stage, where the model parameters are optimized by using the Bayesian optimization. Real operational data of SRRF of a steel plant in China are employed. The experimental results show that the method can accurately estimate the heat required for slab heating and help to effectively arrange the consumption of mixed gas.