Process metallurgy and data-driven prediction and feedback of blast furnace heat indicators

The prediction and control of furnace heat indicators are of great importance for improving the heat levels and conditions of the complex and difficult-to-operate hour-class delay blast furnace (BF) system. In this work, a prediction and feedback model of furnace heat indicators based on the fusion of data-driven and BF ironmaking processes was proposed. The data on raw and fuel materials, process operation, smelting state, and slag and iron discharge during the whole BF process comprised 171 variables with 9223 groups of data and were comprehensively analyzed. A novel method for the delay analysis of furnace heat indicators was established. The extracted delay variables were found to play an important role in modeling. The method that combined the genetic algorithm and stacking efficiently improved performance compared with the traditional machine learning algorithm in improving the hit ratio of the furnace heat prediction model. The hit ratio for predicting the temperature of hot metal in the error range of ±10°C was 92.4%, and that for the chemical heat of hot metal in the error range of ±0.1wt% was 93.3%. On the basis of the furnace heat prediction model and expert experience, a feedback model of furnace heat operation was established to obtain quantitative operation suggestions for stabilizing BF heat levels. These suggestions were highly accepted by BF operators. Finally, the comprehensive and dynamic model proposed in this work was successfully applied in a practical BF system. It improved the BF temperature level remarkably, increasing the furnace temperature stability rate from 54.9% to 84.9%. This improvement achieved considerable economic benefits.