Application Of Time Series Method Predicting Hot Metal Silicon For No.6 BF In Limfen Steel Works

For clarify the complex of ironmaking, BF's (blast furnace) ironmaking process anddifficulties of predicting hot metal silicon content were described in this paper. By means of timeseries method, determing the order and identificating parameters on silicon time series weredone .From the above, the first-step, second-step ahead predicting models were built, and hittingratio of all predicting models were calculated. To increase the hitting ratio of first-step headforecasts models, and to use the information of other time series sufficiently, more than 20 first-step predicting models were built using "Leading indictor", and "Multivariate time series".Predictors were modified using two parameters: ventilation, hot air volume and temperature whichcan increase hitting ratio for 4-5%. In order to apply these predicting models on line, "RecursiveFixed Memory method" was used to identify parameters which were optimized by Newtonmethod and modified. The calculations got to the hitting ratio level off line.