Climate-adaptive amine sweetening optimization via the critical reversal temperature for seasonal mercaptan-accumulation and dehydration control

Seasonal temperature variability can fundamentally alter the limiting mechanism in industrial amine gas sweetening, creating a trade-off between low-temperature methyl-mercaptan accumulation and high-temperature dehydration demand that cannot be adequately addressed by a single annual operating point. Here, we develop a climate-adaptive, cost-aware optimization framework for an industrial methyl diethanolamine-piperazine (MDEA-PZ) sweetening system exposed to ambient temperatures from -4 to 46°C. A calibrated reduced-order model was constructed from eleven plant-validated seasonal operating states and evaluated using leave-one-season-out (LOSO) validation, yielding mean absolute percentage errors of 0.040% for reboiler duty, 2.34% for methyl mercaptan concentration, and 9.46% for sweet-gas water content. A polynomial Ridge surrogate trained over a 1400-point Latin hypercube design subsequently achieved a mean R² of 0.9782, a minimum R² of 0.9381 across all outputs, and a maximum MAPE of 2.05%. The central contribution is the identification of a Critical Reversal Temperature (CRT), defined here as the ambient temperature at which the dominant normalized operating burden switches from a methyl-mercaptan accumulation proxy to sweet-gas dehydration; it is not presented as a direct corrosion-rate threshold. For the nominal benchmark composition and solvent strength, the CRT was 19.61°C, and water-content, normalization-bound, and TOPSIS-weight sensitivity tests consistently retained the thermal-policy transition near 20°C. In contrast, exploratory off-design composition/solvent-strength scenarios shifted the modeled CRT over a wider 8.36-24.22°C range and are explicitly treated as scenario exploration rather than independently plant-validated predictions. Relative to the reference seasonal operation, the model-based CRT-adaptive screening strategy reduced annual OPEX from approximately 11.5 to 9.8 MUSD yr⁻¹, corresponding to about 1.8 MUSD yr⁻¹ in annual savings; the 15-year incremental NPV was approximately 16.2 MUSD with a simple payback of about 0.5 years. Because off-design circulation response is semi-empirical, the adaptive setpoints and associated economic outputs should be interpreted as screening predictions pending high-fidelity or plant re-validation. These results establish CRT as a physically interpretable regime-transition descriptor conditional on the benchmark model and provide a control-oriented framework for climate-responsive operation.