基于AdaBoost-HKSVR算法的烧结固体燃耗预测模型  

Sintering solid fuel consumption prediction model based on AdaBoost-HKSVR algorithm

作　　者：许戈钦 黄晓贤[1] 范晓慧[1] 向家发 周茂军[2] 陈许玲[1] XU Geqin;HUANG Xiaoxian;FAN Xiaohui;XIANG Jiafa;ZHOU Maojun;CHEN Xuling(School of Minerals Processing and Bioengineering,Central South University,Changsha 410083,Hunan,China;Iron-making Plant,Baoshan Iron and Steel Co.,Ltd.,Shanghai 200941,China)

机构地区：[1]中南大学资源加工与生物工程学院,湖南长沙410083 [2]宝山钢铁股份有限公司炼铁厂,上海200941

出　　处：《钢铁研究学报》2025年第1期24-32,共9页Journal of Iron and Steel Research

基　　金：国家自然科学基金资助项目(52274344)。

摘　　要：烧结是钢铁生产中能耗第二高的工序,其中固体燃耗占烧结工序能耗70%以上,对固体燃耗进行提前预测可以为现场操作人员提供生产参数调控依据,对烧结工序的节能减排具有重要意义。针对目前固体燃耗预测模型泛化能力不佳、预测精度受限的问题,提出了一种基于混合核支持向量机(HKSVR)与AdaBoost集成学习算法的烧结固体燃耗预测模型,并采用贝叶斯算法优化模型参数。使用现场生产数据对模型进行训练和测试,结果表明基于多项式核函数与拉普拉斯核函数构建的AdaBoost-HKSVR模型具有较好的预测性能,预测结果的MAE、RMSE、决定系数R^(2)为0.14、0.19、0.99,可为烧结工序智能控制与工艺参数优化调控提供有力支持。Sintering is the second most energy-consuming process in steel production,with solid fuel consumption accounting for more than 70%of the energy consumption in the sintering process.Early prediction of solid fuel consumption can provide a basis for adjusting production parameters for on-site operators and is of great significance for energy saving and emission reduction in the sintering process.In response to the current problems of limited prediction accuracy and poor generalization ability of solid fuel consumption prediction models,a sintering solid fuel consumption prediction model is proposed based on the Hybrid Kernel Support Vector Machine(HKSVR)and AdaBoost ensemble learning algorithm,and its model parameters are optimized using the Bayesian algorithm.The model is trained and tested using on-site production data.The results show that the AdaBoost-HKSVR model built with polynomial kernel functions and Laplace kernel functions has good predictive performance,with MAE,RMSE,and R^(2) of the prediction results being 0.14,0.19,and 0.99,respectively,which can provide strong support for intelligent control and process parameter optimization and adjustment of the sintering process.

关 键 词：烧结工序 固体燃耗 集成学习 混合核函数 支持向量回归 

分 类 号：TF325.1[冶金工程—冶金机械及自动化]