Modeling of Temperature Drop Prediction of Hot Metal Based on Heat-Transfer Mechanism and Machine Learning

Predicting the temperature drop of hot metalHot metal is greatly significant for the decision-making of scrap steel amount in basic oxygen furnace (BOF). In addition, accurate prediction of temperature drop can guide the scheduling of hot metalHot metal, keeping the steelmaking process stable. In this paper, a prediction model of temperature drop was developed by integrating heat-transfer mechanismsHeat-transfer mechanism and machine learningMachine learning models. Extreme learning machineExtreme learning machine (ELM) was applied to establish the machine learningMachine learning model (MLM). Further, the performance of MLM was optimized by introducing regularization and particle swarm optimization (PSO). Based on actual data from a steelmaking plant in China, the above models were trained and verified. The results show that the hit ratio of the integration model is 88.53% when the prediction error is within 10 ℃, which is higher than those of the heat-transfer mechanismHeat-transfer mechanism and the optimized machine learningMachine learning models. Meanwhile, the robustness of the integration model is also optimal.