Prediction model for hot metal temperature based on MI-PSO-LSTM

With the development of artificial intelligence, digital empowerment to help the intelligent transformation of blast furnaces and achieve the "30-60" target has become a major research objective. As one of the important indicators of blast furnace production, the reasonable prediction for hot metal temperature could help to master the operating status of furnace conditions and reduce the energy consumption in time. Therefore, a molten iron temperature model based on PSO-LSTM is proposed in this paper. Firstly, for the characteristics of blast furnace data such as long time lag and the abnormal values difficultly judge, the handling for sample data is completed by using the big data method. Secondly, the selection for parameters based on the complementary of process and big data techniques is used to improve the interpretation of model input variables. Finally, a PSO-LSTM prediction model was established by the LSTM algorithm, and the prediction capability is analysed from four perspectives include mean absolute error (MAE) and root mean square error (RMSE) and so on. The results show that the prediction accuracy of hot metal temperature is up to 95.6%, which has certain implications for production guidance.