Prediction of Silicon Content in Blast Furnace Hot Metal Based on TGARCH Model

潘伟,刘祥官,曾九孙
DOI: https://doi.org/10.3785/j.issn.1008-973x.2010.04.013
2010-01-01
Abstract: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.
What problem does this paper attempt to address?