Robust machine learning models for predicting methane hydrate formation conditions in the presence of brine

Methane hydrates, crystalline compounds of methane and water form under high pressure and low temperatures, presenting opportunities as an energy resource and challenges like pipeline blockages. Accurate prediction of hydrate equilibrium conditions is crucial for optimizing energy extraction and ensuring pipeline safety. In this study, machine learning models were developed to predict hydrate equilibrium temperatures in various brine solutions, using a dataset of 1039 data points. Eleven models were tested, with each evaluated using 10-fold cross-validation to ensure accuracy and robustness. Extreme Gradient Boosting (XGBoost) emerged as the most accurate model, achieving the lowest error rates and highest R² values. Sensitivity analysis identified pressure as the most significant factor influencing hydrate formation, followed by specific ions in the brines. This research highlights the effectiveness of machine learning, particularly XGBoost, in predicting methane hydrate formation, offering valuable insights for industrial applications and advancing hydrate management in energy processes.