Improving equilibrium based CO2-PCGLY process simulations via neural network-optimized Murphree efficiency: Accuracy and energy insights

Abstract

Potassium carbonate promoted with glycine (PCGLY) has emerged as a promising solvent for CO₂ removal from high CO₂ content natural gas, offering enhanced absorption kinetics and reduced regeneration energy. While steady-state behaviour of CO₂-PCGLY systems is well documented, their dynamic performance remains underexplored. This posing a challenge for accurate simulation and control design. This study addresses the limitations of rate-based models in Aspen Dynamics by proposing an innovative equilibrium-based simulation framework in Aspen Plus V12.1 that powered by Artificial Neural Networks (ANNs) to dynamically optimize Murphree efficiency. The ANN model enables rapid recalibration across varying operational conditions, achieving predictive deviations of less than 5% compared to rate-based benchmarks while significantly reducing computational load. At operating conditions of 15 wt% K₂CO₃+ 3 wt% glycine, and 40 bar, the system attains 75% CO₂ removal efficiency. Additionally, energy analysis through conceptual stripper design and process flow modifications reveals that solvent pre-heating delivers a 20.09% improvement in overall energy efficiency with minimal impact on separation performance. This integrated approach offers a powerful tool for advancing both simulation fidelity and sustainable process design in CO₂ capture applications using PCGLY systems.