Prediction and Regulation of Blast Furnace Pressure Drop Using an Error-Weighted SSA-BP-XGBoost Model

The pressure drop in a blast furnace significantly impacts its stable operation and smelting efficiency. Excessive pressure drop can cause material suspension, disrupting the smooth operation of the furnace, while insufficient pressure drop can lead to over-blowing or even gas channeling within the furnace. Therefore, maintaining the pressure drop within an appropriate range is a prerequisite for ensuring stable furnace conditions. Accurate prediction of blast furnace pressure drop enables operators to monitor furnace conditions in real time, optimize operational parameters, and prevent adverse fluctuations. However, conventional predictive models often struggle with limited accuracy due to the complex nonlinear and time-varying nature of the furnace system. To address this issue, this study develops a hybrid prediction model that integrates the Sparrow Search Algorithm (SSA), Extreme Gradient Boosting (XGBoost), and the Back Propagation Neural Network (BP) based on real blast furnace operation data from a domestic steel enterprise. SSA optimizes the individual models, and prediction errors are minimized using a weighted averaging approach based on inverse error. Results show that the SSA-optimized XGBoost-BP hybrid model outperforms other models in prediction accuracy. Finally, an MPC controller was designed based on the proposed hybrid model to regulate the pressure drop by adjusting operational parameters, providing a theoretical foundation for closed-loop control of blast furnace pressure drop.