Pruned Bagging Aggregated Hybrid Prediction Models for Forecasting the Steel Temperature in Ladle Furnace

Abstract

For accurate control of the steel temperature during ladle furnace (LF) refining, in this study, a novel ensemble based steel temperature prediction model is presented. The base learning algorithm of this ensemble predictor is developed by hybrid modeling of the LF thermal system, where first principle method and data‐driven modeling algorithm are combined for obtaining a precise prediction. Thereafter, for addressing the over‐fitting problem and thereby further boosting the prediction performance, an ensemble based prediction model is presented on the basis of the proposed pruned Bagging method. The pruned Bagging uses negative correlation learning to prune the original Bagging ensemble for resulting in a robust and efficient ensemble which is able to improve the ensemble efficiency without deteriorating the prediction performance. As a result, the proposed ensemble based steel temperature predictor is able to provide a precise prediction with high reliability. The effectiveness of the presented steel temperature prediction model is validated by the practical data.