Hierarchical Surrogate Modeling and Optimization of the SO2 Recovery–Water Loss Trade-Off in Basic Aluminum Sulfate Regeneration

Vacuum regeneration can close the absorbent loop in basic aluminum sulfate (BAS) wet flue-gas desulfurization and SO2 capture, but SO2 release occurs together with energy-intensive water evaporation. This study developed a response-specific framework to predict and balance these coupled rates using existing vacuum-regeneration data. The SO2 model included absolute pressure, temperature, aluminum concentration, basicity, and sulfite concentration, whereas the H2O model used pressure and temperature. Candidate regressors were assessed by cross-validation, error statistics, and physical-trend consistency. Gaussian process regression with a Matérn-5/2 automatic-relevance-determination kernel best predicted SO2 desorption (RMSE = 0.0631; R2 = 0.9746), while an automatic-relevance-determination squared-exponential kernel best predicted H2O evaporation (RMSE = 0.0793; R2 = 0.9928). Sulfite concentration dominated the SO2 response; pressure and temperature controlled both rates. Lower pressure and higher temperature improved SO2 recovery but increased water loss. Pareto analysis identified a model-supported compromise at approximately 26–30 kPa and 340–342 K. A 120 min recursive simulation showed declining instantaneous SO2 release as sulfite was depleted, while water evaporation continued. Within the experimental domain, the framework supports joint selection of pressure, temperature, and operating time for BAS regeneration.