Experimental analysis and machine learning modeling of CO2 and CH4 gas hydrate phase equilibria with quaternary ammonium salts

The study investigates the hydrate liquid vapor equilibrium HLVE experimentally at concentrations of 1, 5, and 10 wt % for tetramethylammonium acetate, tetraethylammonium acetate, and their mixtures with the commercially used thermodynamic hydrate inhibitor, monoethylene glycol (MEG). The experimental pressure range for CO 2 hydrates is from 2.0 to 3.50 MPa, while the pressure range for CH 4 hydrates varies between the values of 3.40 to 8.30 MPa. A Gaussian Process Regression (GPR) based machine learning model has also been developed to predict the hydrate liquid vapor equilibrium (HLVE) of CO 2 and CH 4 hydrate formation for quaternary ammonium salts (QAS). This analysis suggests kernel optimization of the GPR. With an R 2 value of 0.9 and an RMSE value of 0.031, the optimized model has strong predictability using the concentration of QAS, pressure, and temperature as inputs. The results of the model are correlated with the experimental outcome, and the predicted results are in fair accordance with the results of the experiment.