Application of Recurrent Neural Network and Matrix Pencil Method in Predicting Blast Furnace Thermal State Indicators

The article considers the problem of automation of operational control of blast furnace thermal state. The main focus is on predicting of blast furnace thermal state indicators. Two approaches are suggested for predicting the blast furnace thermal state. The prediction is based on the matrix pencil method, and the Elman's recurrent neural network model for the blast-furnace process is considered. These approaches are investigated on experimental data. The analysis of the results of the forecast showed that these approaches of predicting the blast furnace thermal state indicators are suitable for the expert systems as an advisor for the operator. Thus, a technique, that allows one to build a forecast, using software tools for the in-depth analysis of operation statistics, was developed. The approach for creating the expert system, that allows the blast furnace foreman to analyse the blast furnace thermal state condition and to decide on its adjustment, was suggested. The results of the work are useful for further research in the field of the ferrous industry.