摘要
建立RH精炼终点钢液温度预报模型,有效控制RH终点钢液温度,有利于降低钢铁生产成本,提高钢铁生产的质量和效率。采集某钢厂RH精炼实际生产数据,对生产数据进行预处理,利用递归特征消除法选择对RH钢液温度影响度高的生产工艺关键参数作为特征集,利用BO和PSO算法优化MLP模型提高了钢水终点温度预测精度和鲁棒性。研究结果表明,基于PSO-MLP的RH精炼终点钢液预测模型的平均误差和均方根误差分别为1.14和1.67,误差绝对值≤3℃的命中率为94%;模型现场应用过程中误差绝对值≤3℃的命中率≥96.86%。该模型的应用为RH生产过程中的钢液温度控制提供准确可靠的支撑,有助于优化工艺参数,提高产品质量,降低生产成本。
The forecast model of the end point temperature of RH refining has been established for effectively controlling the RH endpoint steel temperature,which is beneficial for reducing steel production costs and improving the quality and efficiency of steel production.This article first preprocesses the actual production data of RH refining in a steel plant,then selects the key production process parameters that have a high impact on RH steel temperature as the feature set and uses the recursive feature elimination method.Finally,the BO and PSO algorithms are used to optimize the MLP model to improving the accuracy and robustness of predicting the final temperature of steel.The research results indicate that the MAE and RMSE of the RH refining endpoint steel prediction model based on PSO-MLP are 1.14 and 1.67,respectively,with a prediction accuracy of 94%within±3℃.The application of this model provides accurate and reliable support for steel temperature control in RH production process,helps optimize process parameters,improve product quality,and reduce production costs.
作者
施树蓉
韩啸
胡倩倩
何志军
辛宇
Shi Shurong;Han Xiao;Hu Qianqian;He Zhijun;Xin Yu(University of Science and Technology Liaoning;Green Low-Carbon and Intelligent Metallurgy)
出处
《冶金能源》
北大核心
2025年第2期60-65,共6页
Energy For Metallurgical Industry
基金
国家自然科学基金资助项目(52374339)。
关键词
RH精炼
MLP模型
PSO优化算法
预测精度
RH refining
multi-layer perceptron
particle swarm optimization
forecast accuracy
