Prediction of desulfurizer addition based on heterogeneous stack ensemble learning
KR mixing method is one of the important processes for pre-desulfurization of hot metal,but the amount of desulfurizer added in the process mainly depends on manual experience control,which leads to low hit rate of desulfurization and unstable quality of the billet.In order to accurately control the amount of desulfurizer added and improve the hit rate of primary desulfurization,a method for predicting the amount of desulfurizer added based on heterogeneous stack ensemble learning was proposed.First of all,the data was cleaned and the outliers were eliminated using LOF algorithm combined with expert experience.Secondly,the maximum information coefficient combined with Spearman rank correlation coefficient was used for feature screening.Finally,the stack integration algorithm was introduced to select different base learners of the model based on the evaluation index and error correlation analysis,and Optuna algorithm was used to find the optimal parameter combination for the base learner,and the heterogeneous stack integration prediction model was established.The results show that the coefficient of determination(R2)is 91.6％,the root mean square error(RMSE)is 197.79,the mean absolute error(MAE)is 117.14,and the mean absolute percentage error(MAPE)is 6.95％.
KR mixing methodamount of desulfurizer addedfeature screeningOptunastacking integration
万方数据
维普
