COREX冷煤气成分预测的二步建模方法  

Two-stage modeling method for predicting COREX cold gas content

作　　者：赵晨熙[1] 张群[2] 徐金梧[1] 黎敏[1] 

机构地区：[1]北京科技大学机械工程学院,北京100083 [2]宝钢股份研究院,上海201900

出　　处：《北京科技大学学报》2012年第10期1184-1189,共6页Journal of University of Science and Technology Beijing

基　　金：国家自然科学基金资助项目(50934007;51004013);中央高校基本科研业务费专项(FRF-MP-09-009B;FRF-AS-09-008B)

摘　　要：针对熔融气化炉冷煤气成分含量,提出了基于熵权模糊C均值聚类和偏最小二乘的COREX冷煤气成分预测方法.建模过程中首先根据料单中各种原料的单耗量,利用熵权模糊C均值聚类的方法将料单聚类成若干种料单类别,然后针对不同的料单类别,利用偏最小二乘法分别建立冷煤气成分预测模型.对宝钢COREX-1#炉实际生产数据验证结果表明:该方法可以有效地建立COREX冷煤气成分预测模型,具有较好的预测精度.A method for predicting cold gas content in a melter-gasifier was proposed based on entropy-weighted fuzzy C-means clustering and partial least squares（PLS）.In the modeling process,an entropy-weighted fuzzy C-means clustering algorithm is used to get the clustering result of burden calculation reports according to the consumption of raw materials at first.Then,different prediction models are built based on a PLS algorithm for various cluster types.The real field data of cold gas content from Baosteel COREX-1# were used for verification.It is shown that the method can build the prediction model of COREX cold gas content effectively,and has an advantage in prediction accuracy.

关 键 词：熔炼 铁矿石还原 煤气 预测 模糊聚类 偏最小二乘法 

分 类 号：TF557[冶金工程—钢铁冶金] O212.4[理学—概率论与数理统计]