Predicting Gas Hydrate Equilibria in Multicomponent Systems with a Machine Learning Approach

Natural gas production and transportation can be threatened by gas hydrate plugging, especially in offshore environments with low temperatures and high pressures. This can cause pipeline blockages, increased back pressure, production stoppage, and pipeline ruptures. Machine learning (ML) models were created to predict gas hydrate formation in multicomponent systems with and without inhibitors. The models utilized input parameters such as gas-mixture specific gravity, pressure, and inhibitor concentrations, which were fed into five supervised ML algorithms. The accuracy of the models' predictions was compared, and three models had accuracy greater than 90%, while two had accuracies between 80 and 90%.

Automated summary

Natural gas production and transportation in offshore environments can be threatened by gas hydrate plugging, leading to various problems. Machine learning models were developed to predict gas hydrate formation with and without inhibitors, using input parameters such as specific gravity, pressure, and inhibitor concentrations. The accuracy of the models' predictions was evaluated, with three models achieving accuracy greater than 90% and two models achieving accuracies between 80 and 90%.