Energy-efficient dynamic control of carbon capture for hot blast stoves: A physics-constrained and surrogate-based framework

• Dynamic HBS carbon capture boosts steel industry decarbonization. • Particle swarm reconstruction mapping enhances signal quality (R 2 > 0.95). • Physics-constrained UKF ensures precise gas variable prediction (R 2 > 0.9). • Surrogate model-based dynamic predictive control saves 9.08 GJ per cycle. The inherent non-stationarity and measurement uncertainties of hot blast stove (HBS) flue gas severely hinder efficient waste heat utilization. Driven by the imperative of carbon neutrality, integrating carbon capture and solvent regeneration with HBS waste heat offers a critical pathway to reduce the energy penalty of industrial decarbonization. Consequently, this study proposes a hybrid predictive control framework to synchronize these coupled energy systems. To mitigate the impact of combustion dynamics on process monitoring, a particle swarm reconstruction mapping method is introduced, extracting high-fidelity process data with an R 2 > 0.95. Building upon this foundation, a physics-constrained Unscented Kalman Filter (UKF) is employed to forecast transient thermodynamic parameters, effectively incorporating mass and energy constraints to limit anomaly influence and ensure a prediction accuracy of R 2 > 0.90. Through this robust predictive capability, an online control paradigm based on an Aspen surrogate model is established to regulate the dynamic coupling process, yielding significant energy savings of 9.08 GJ per cycle. This research provides a solution for managing the non-linear emission characteristics of HBS and similar periodically fluctuating industrial systems, bridging the gap between signal processing and process intensification for low-carbon steel manufacturing.