Research On Temperature Prediction Model Of Molten Iron Transportation Based On Data Analysis

In the iron and steel industry,molten iron transportation refers to the entire process of transferring molten iron from blast furnaces to a steel-making plant through a torpedo car.In this process,if the drop of molten iron temperature is too large,it will not only affect the steelmaking efficiency,but also increase the energy consumption of the steelmaking process.Therefore,the research on temperature prediction of molten iron transportation has practical significance.Accurate prediction can help technicians to dynamically set the logistics schedule of torpedo car according to the real-time temperature changes so as to reduce the temperature-dropping and energy saving.The real-time measurement of temperature in the process of molten iron transportation costs too much,and many factors would affect the temperature-dropping of the process,so it is very difficult to describe the process by single model accurately.In order to solve the above problems,this thesis uses the actual production data to establish a corresponding data analytical model,and studies the prediction of molten iron real-time temperature and pour-in temperature.At the same time,a molten iron temperature prediction system is developed.This thesis includes:1)Aimed at the real-time temperature prediction of molten iron,considering the flaw of a single model,a temperature hybrid prediction method of molten iron transportation based on Kalman filter is proposed.The predicted value of the mechanism of temperature-dropping in molten iron is taken as a priori estimate in Kalman filter and the predicted value of piecewise linear model is used as a measure to realize the hybrid model under Kalman filter theory.Thus,the problems of error divergence caused by the lack of accuracy of the mechanism model and the inaccurate description of the details of the transportation process by the piecewise linear model are solved.By calculating the experiment,the prediction results are compared with the mechanism and the linear model to verify the effectiveness of the hybrid model.2)In order to predict pour-in temperature of molten iron,a pour-in temperature forecasting method based on multi-model predictive data fusion is proposed.To solve the noise problem,the outlier was eliminated by the idea of hierarchical clustering.The LSSVM(least squares support vector machine)model is based on differential evolution algorithm(DE-LSSVM).Based on Kalman filter theory,the multiple linear regression model and the prediction of molten iron pouring temperature by DE-LSSVM are merged to realize the high precision during the prediction of molten iron pour-in temperature.3)Based on the characteristics of molten iron transportation process,a molten iron temperature prediction system is designed.The main functions of the system include real-time molten iron temperature prediction,pour-in temperature prediction,abnormal data prediction and statistical analysis of prediction results.The prediction model developed above is embedded in the system,to predict the temperature of molten iron transportation.