Prediction Of Hot Metal Temperature On Ironmaking And Steelmaking Interface Based On Data-driven Model

As the basic raw material for steel making,the physical heat and chemical heat of hot metal are the basic heat sources of converter steelmaking.Higher hot metal temperature is helpful to the stable and smooth production of converter steelmaking process and the reduction of energy consumption.Low temperature affects steelmaking production,increases interface energy consumption,and may even affect the quality of steel and produce safety problems.However,due to the hot metal transportation at the ironmaking and steelmaking interface is a long process and multi-process production process,there are many factors affecting the temperature of hot metal,and the coupling is strong,which leads to the uncontrollable temperature of hot metal reaching the converter.Therefore,the prediction of the temperature of the hot metal at the ironmaking and steelmaking interface is of great significance.With the continuous progress of informatization in the iron and steel industry,a large number of valuable production historical data are stored in the enterprise,which also makes it possible to predict the hot metal temperature at the ironmaking and steelmaking interface based on data-driven model.In this paper,the ironmaking and steelmaking interface in an iron and steel enterprise were taken as the research object.Through combing the process flow,the prediction model of hot metal temperature including pretreatment process and without pretreatment process were established for the unit 1 and 2 steelmaking plant.The production history data comes from January 2016 to June 2018 in the L3 level database of the EMS system.The different process data was correlated to form a complete data chain,and then the data was preprocessed and analyzed,The proportion of different operation processes was calculated.For example,the secondary tapping rate is 18.68%;the average proportion of pretreatment processes in the unit 1 and 2 steelmaking plant is 63.21%.Analyzed the data characteristics of different parameters at the interface.For example,the average of heavy tank time is177min;the average of empty tank time is 177 min.Using the processed data,combined with the heat dissipation process of hot metal and correlation analysis,six variables were selected as the input variables of the prediction model,including hot metal temperature of tapping,tapping time,heavy tank time,empty tank time,the weight of hot metal and pretreatment time.The prediction model of hot metal temperature was established based on linear regression,support vector machine regression,BP neural network and BP neural network optimized by GA algorithm.The performance of different models was evaluated by using test sample data.Among the above four kinds of prediction models of hot metal temperature,the GA-BP neural network model has the best prediction effect.The predicted hit rate of the GA-BP neural network model within the absolute error of 10? between the predicted value and the actual value reaches 87%;the average of mean absolute percentage error is 0.44;the average of root mean square error is 6.96.The comparison shows that the GA-BP neural network model has certain advantages.Based on the hot metal temperature prediction model,three kinds of hot metal temperature control methods at the ironmaking and steelmaking interface were given,and good application results have been obtained,such as reducing the weight threshold of hot metal by secondary tapping from 200 t to 140 t,the change for temperature of hot metal at the interface decreases from 160? to 135?.