Application of Machine Learning in Predicting Formation Condition of Multi-Gas Hydrate

Thermodynamic models are usually employed to predict formation condition of hydrates. However, these thermodynamic require a large amount calculations approach phase equilibrium. Additionally, parameters included in the model need be calibrated based on experimental data, which leads high uncertainties predicted results. With rapid development artificial intelligence (AI), machine learning as one sub-discipline has been developed and widely applied various research area. In this work, was innovatively natural gas hydrates overcome computation cost low accuracy. Three data-driven models, Random Forest (RF), Naive Bayes (NB), Support Vector Regression (SVR) were tentatively used determine hydrate formed by pure mixed gases. Experimental data reported previous work taken train test models. As representative Chen–Guo (C-G) analyze computational efficiency accuracy The comparison results C-G with indicated that RF performed better than NB SVR both speed According average AADP calculated is 7.62 times model. Meanwhile, time costed 75.65 Compared other two expected predicting under field conditions.