高炉硅含量预测控制的时间序列混合建模  

Hybrid time series predictive control model for silicon content in blast furnace hot metal

作　　者：贺诗波[1] 刘祥官[1] 郜传厚[1] 黄丽英[2] 黄雅彬[2] 

机构地区：[1]浙江大学系统优化技术研究所,浙江杭州310027 [2]包头钢铁公司炼铁厂,内蒙古自治区包头014100

出　　处：《浙江大学学报（工学版）》2007年第10期1739-1742,共4页Journal of Zhejiang University：Engineering Science

基　　金：国家发改委高技术产业化示范工程项目

摘　　要：根据时间序列原理,对铁水硅质量分数w(Si)序列的动态特性进行了分析,找出了其主体成分.综合炼铁工艺理论和6号高炉炼铁现场的实际情况,对影响铁水硅质量分数w(Si)的各个因素进行了典型相关分析,确定了w(Si)的主要影响因素为:料速、风量、喷煤、透气性、富氧率、铁水物理温度和渣碱度.建立了由AR(2)、状态变量和控制变量混合构成的w(Si)预测控制的传递函数模型.使用包钢6号高炉生产数据验证表明,此模型克服了以往炉温控制模型把炼铁理论和统计知识分离的缺点,炉温预报的命中率高达84%,预测控制取得了较好的效果.Based on principle of time series analysis, the dynamic characteristic of silicon content series in molten iron was discussed and the main components were discovered. Integrated the theories of iron-making technique and the situation of Blast Furnace No. 6 on-the-spot, canonical correlation analysis （CCA） was used to decide the main influencing parameters of silicon content, eg. lowering speed, blast volume, pulverized coal, permeability, oxygen-enriched, physical temperature of molten iron and basicity. Then a hybrid transfer function model composed of AR （2）, state variables and control variables was proposed for predictive control of silicon content. The model was tested using datasets from Blast Furnace No. 6 of Baotou Steel Corporation. Simulation showed that the hit rate was over 84%. Thus the predictive control model is effective and economic, and avoids the traditional model＇s shortcoming of separating the theories of iron-making and statistics.

关 键 词：时间序列 统计分析 典型相关分析 传递函数模型 

分 类 号：TG250.2[金属学及工艺—铸造]