Phase Stability of CH₄ and CO₂ Hydrates under Confinement Predicted by Machine Learning

Area of Science:

• Geochemistry and Earth Science
• Computational Chemistry and Materials Science

Background:

• Phase stability of gas hydrates under confinement is critical for understanding geological stability.
• Gas hydrates, particularly methane (CH4) and carbon dioxide (CO2) hydrates, play a significant role in Earth's systems.

Purpose of the Study:

• To predict the phase stability of confined CH4 and CO2 hydrates using machine learning.
• To evaluate and compare the performance of different machine learning models for this prediction task.

Main Methods:

• Development and application of three machine learning models: support vector machine, random forest, and gradient boosting decision tree.
• Training and validation of models using a dataset for predicting confined gas hydrate phase stability.
• Selection of the best-performing model based on prediction accuracy.

Main Results:

• Support vector machine (SVM) model achieved the highest prediction accuracy for confined gas hydrate phase stability.
• Random forest model exhibited the lowest prediction accuracy among the tested models.
• The chosen SVM model, with a 0.7 training set fraction, demonstrated over 90% average accuracy for predicting unknown phase stability of CH4 and CO2 hydrates.

Conclusions:

• Machine learning, particularly SVM, provides a highly accurate and efficient method for determining the phase stability of confined CH4 and CO2 hydrates.
• The developed models can be valuable tools for rapid and precise phase stability assessments in future applications.
• This approach enhances our understanding of gas hydrate behavior under geological confinement.