Thermodynamic Parameter Regression Using Open Source Optimizers

Parameterization of activity models and equations of state for multicomponent systems is a complicated endeavor when the number of system components is high. Even when restricting interaction parameters to the binary level, the inherent non-linearity of thermodynamic models often leads to local solutions that are not globally optimal. In this talk, we examine the quality of open source optimizers when applied to thermodynamic parameter regression tasks, using the Python programming environment as a framework to drive Aspen Properties calculations. In particular, we consider techniques designed for discovering global optima (e.g., Particle Swarm Optimization). Finally, we provide practical examples relevant to Acid Gas Treating, such as multicomponent aqueous amine solvent systems.