Machine Learning-Based Research on Reserve Prediction of Natural-Gas-Hydrates

Currently, there are challenges in reserve prediction of natural-gas-hydrates, including difficulties in obtaining parameters, limited calculation accuracy, and complex processes. Machine learning is a data-driven predictive method with strong generalization capabilities, capable of abstracting and inducing knowledge and rules from basic data witch has made significant advancements in various fields and has been preliminarily applied in fields of earth sciences and energy exploration and development. Various machine learning algorithms for determining the parameters of gas hydrate reserves were introduced in this paper. The support vector machine algorithm was employed to calculate effective area and thickness, while neural network algorithm was introduced to calculate porosity, saturation, and decomposition coefficient for the determination of parameters of natural-gas-hydrates reserves in Shenhu area of China's offshore, and then the feasibility and accuracy of the method was verified. Compared to evaluation results based on trial mining or numerical simulation, the machine learning-based reserve prediction method demonstrates high predictive accuracy, clear processes, and simple structure, making it highly valuable for future research on storage and transportation of natural-gas-hydrates and provides foundation for future industrial applications.