TGARCH模型预测高炉铁水硅质量分数  

Prediction of silicon content in blast furnace hot metal based on TGARCH model

作　　者：潘伟[1] 刘祥官[1] 曾九孙[1] 

机构地区：[1]浙江大学数学系,浙江杭州310027

出　　处：《浙江大学学报（工学版）》2010年第4期696-699,731,共5页Journal of Zhejiang University：Engineering Science

基　　金：国家科技部重点推广计划资助项目(2005EC000166);国家发改委高技术产业化示范工程资助项目(发改高技[2004]2080号);国家自然科学基金资助项目(10826100);浙江省自然科学基金资助项目(Y107110);高等学校博士点科研基金资助项目(20070335161);浙江大学控制技术国家重点实验室开放课题资助项目(ICT0904)

摘　　要：为了更好地反映高炉铁水硅质量分数序列的高波动特性,利用门限广义自回归条件异方差(TGARCH)模型对硅质量分数序列进行预测.应用Portmantea Q检验、拉格朗日乘子检验以及非对称项系数显著性检验,验证了高炉铁水硅质量分数序列存在异方差性和非对称性.在此基础上将TGARCH模型应用于高炉铁水硅质量分数预测,采用极大似然估计法确定参数,建立TGARCH(1,1,1)预测模型,并采用命中率和误差率2种评价准则对预测结果进行分析.这种方法克服了以往模型没有考虑序列非对称性影响的缺陷,更加适合于高炉铁水硅质量分数的预测.将预测模型应用于包钢6号高炉,取得了较好的预测效果.In order to account for the high volatility of silicon content in blast furnace hot metal,threshold generalized autoregressive conditional heteroskedastic （TGARCH） model was used to construct a predictive model for the time series of silicon content in hot metal.Portmantea Q test,Lagrange multiplier （LM） test and asymmetric test were applied to the silicon content series.It was proved that silicon content series exhibits ARCH and asymmetric effect,which validated the application of TGARCH model.A TGARCH （1,1,1） model was then adopted and the coefficients were estimated by maximum likelihood method.Two criteria,i.e.hit rate and error rate,were used to evaluate the performance of the model.This approach takes the asymmetric effect as well as ARCH effect of silicon content into account and is more appropriate for prediction of silicon content in hot metal.Data collected from No.6 blast furnace of Baotou Iron Steel Corporation were used to test the identified model and good results were achieved.

关 键 词：异方差效应 非对称效应 TGARCH 硅含量 时间序列 

分 类 号：TG250.2[金属学及工艺—铸造]