Effect of guest-dependent reference hydrate vapor pressure in thermodynamic modeling of gas hydrate phase equilibria, with various combinations of equations of state and activity coefficient models

Abstract

Using the van der Waals solid solution theory and the fugacity-based approach of Klauda and Sandler, a thermodynamic model to predict the phase equilibria of gas hydrates using an Equation of State (EOS) and an Activity Coefficient Model (ACM) is presented. Here, we have studied five EOS models, that are, Peng-Robinson-Stryjek-Vera (PRSV), Patel-Teja (PT), Soave-Redlich-Kwong (SRK), Peng-Robinson (PR), and Redlich-Kwong (RK) to calculate the gas phase fugacity and four ACMs, that are, modified UNIFAC (mUNIFAC), UNIQUAC, Wilson and NRTL to model the liquid phase non-ideality. The Wilson and NRTL activity coefficient models have previously been used for systems containing additional chemicals in the liquid phase but not for pure water in coexistence with gas hydrates, to the best of our knowledge, and this has been explored in this work. By making the empty hydrate vapor pressure parameters guest dependent, the model improves upon the predictions from our previous model for sI and sII hydrates significantly and has drastically reduced the errors especially in modeling sII hydrates. The best combinations of EOS and ACM in L-H-V equilibrium and best EOS for I-H-V equilibrium, respectively, for each compound, to be used in the guest dependent model from the 100 combinations of EOS and ACM (for L-H-V equilibrium) and 25 possible EOS options (for I-H-V equilibrium) examined are: RK–mUNIFAC and SRK for CH4, RK–mUNIFAC and PRSV for CO2, RK–mUNIFAC and PR for C2H6, SRK–mUNIFAC and SRK for N2, and PR–NRTL and RK for C3H8. While different EOS work well for CH4, CO2, C2H6 and C3H8 hydrates in L-H-V and I-H-V equilibria, the SRK EOS works well for N2 hydrates in both L-H-V and I-H-V equilibria. We have also reported the solubilities of guest species in liquid water in equilibrium with hydrates and the hydrate cage occupancies from the models producing the least error as data on these facets is limited.