Temperature Prediction for Finish Entry of Hot Strip Mill Based on Data-driven

摘要

In the production and control of hot-rolled strip, the prediction accuracy of the finish rolling inlet temperature directly affects the temperature control of the subsequent finish rolling outlet temperature and coiling temperature. The heat exchange during the cooling process of the strip is a very complicated nonlinear process, which is difficult to accurately express with mathematical models. However, data-driven models can represent nonlinear processes very well. This paper uses data-driven temperature prediction of the finishing rolling inlet for the full length of the slab surface, and uses the factors related to the full length temperature of the rolled slab surface in the hot continuous rolling line as training data to predict the next full surface temperature at the finishing rolling inlet. Aiming at the problems of traditional prediction models that have low accuracy, many input parameters, and difficult prediction, a gray wolf algorithm optimized BP neural network (GWO-BP) prediction model is proposed. This model is suitable for prediction of the entrance temperature of the final rolling. The experimental results were verified that the GWO-BP prediction model is suitable for the prediction of the entrance temperature of finishing rolling compared with Particle Swarm Optimization Support Vector Machine (PSO-SVR) and Genetic Algorithm Optimized Support Vector Machine (GA-SVR). Finally, the GWO-BP model was further improved to obtain and verify the AGWO-BP model. The experimental results show that the data-driven finish rolling temperature prediction model built in this paper has high prediction accuracy and prediction ability, which is of great significance for practical applications.