A Systematic Evaluation of SAFT-Type Equations of State for CO2–Polar Aprotic Solvent Systems: Model Selection and Optimization Strategies

Accurate vapor–liquid equilibrium (VLE) modeling of CO2 in polar aprotic solvents (PASs) is critical for solvent screening and the design of carbon-capture processes. This study systematically evaluates three SAFT-type equations of state (EoSs), namely CPA, PC-SAFT, and SAFT-VR Mie, under a unified parameter optimization framework. For pure-component saturation properties, SAFT-VR Mie yielded the lowest mean deviations across all investigated substances, with average AARD values of 0.066% for Psat and 0.127% for ρsat. However, the most accurate EoS for an individual substance depended on the property evaluated, and PC-SAFT produced lower deviations for several substance–property combinations. For binary CO2 + PAS systems, all three EoSs exhibited substantial pressure deviations when the binary interaction parameter (kij) was set to zero. Optimizing kij reduced the pressure AARD for every model–solvent combination, indicating that the default combining rules did not adequately represent the CO2–solvent cross-interactions. Notably, the specific improvement in pressure AARD% after introducing kij for each model revealed CPA’s superior sensitivity (e.g., for the CO2 + MEK system, pressure AARD% decreased from 14.43% to 4.07%, a reduction of over 70%). Furthermore, a comparative analysis of single-temperature (ST) and multi-temperature (MT) optimization strategies reveals that the ST optimization reduces the average pressure AARD% from 4.34% (MT) to 3.05%, whereas the MT-optimized kij values offer superior transferability across a broader temperature range. These findings elucidate the different applicable scenarios of model architecture and kij optimization strategies, providing actionable guidance for selecting appropriate thermodynamic models to support the simulation and optimization of carbon-capture systems.