Prediction of gas hydrate formation in the Shenhu sea area based on machine learning

A reliable method for evaluating hydrate resources is essential for accurately predicting the distribution of natural gas hydrates, supporting industrialization, and achieving carbon neutrality goals. By integrating marine geological survey data from the northern South China Sea and using the KNN algorithm, a feature dataset of seabed characteristics, regional organic carbon content, geothermal gradient, sediment density, seabed temperature, and sedimentation rate was constructed. The contribution of pyrolysis gas was introduced to optimize the hydrate formation model, improve the prediction accuracy of marine hydrate saturation, and the Pearson correlation coefficient was used to perform sensitivity analysis on the data.. Simulation results indicate that the predicted average hydrate saturation in typical hydrate drilling areas is reliable and showed robust model generalization capabilities. In addition, sensitivity analysis revealed that organic carbon content and seabed temperature correlated significantly with natural gas hydrate formation. The research results will provide effective guidance for the optimization of characteristic parameters of machine learning for marine natural gas hydrates in the region and even the world, which will help to improve the accuracy of natural gas hydrate resource evaluation, improve global carbon estimation, and enhance awareness of potential seabed hazards..