Comprehensive Review of the Impact of Thermodynamic Inhibitors and the Predictive Power of Machine Learning Models on Hydrate Formation Pressure and Temperature

Gas hydrate formation presents challenges in the petroleum and gas industry, such as pipeline blockages. This study evaluates thermodynamic inhibitors, including amino acids, ionic liquids, salts, and commercial inhibitors, using 213 data entries covering a range of gases and inhibitors over pressures from 0.13 to 200 MPa and temperatures from 238.15 to 333.15 K. Glycine is identified as the most effective amino acid inhibitor, especially when combined with methanol. The inhibition efficiency of ionic liquids depends on functional groups (e.g., OH, NH₂) and side chain lengths, while salts like MgCl₂ perform well due to high ionic charge densities. Methanol and monoethylene glycol remain effective in high-flow systems. Machine learning models, including random forest (RF), support vector machines (SVM), deep neural networks (DNN), and convolutional neural networks (CNN), were applied to predict hydrate formation conditions. The RF model showed the best accuracy with an R² of 0.96 and a root-mean-square error (RMSE) of 1.51 MPa for pressure, and an R² of 0.92 and an RMSE of 2.66 K for temperature. Compared to physically based models, these machine learning methods demonstrated better generalization across varied compositions and inhibitor types, particularly in cases involving complex nonlinear interactions, offering a powerful approach to optimize hydrate control strategies in operations.