基于机器学习方法的天然气水合物稳定带厚度计算

随着互联网大数据发展，人工智能算法逐渐能自动学习数据特征和挖掘大数据隐藏的信息，且其预测结果具有极高的准确度和可靠性。机器学习是人工智能的核心算法，目前已应用于多个领域，在地球科学领域的应用也已兴起。天然气水合物稳定带是评估水合物资源潜力的关键参数，其准确性直接影响水合物勘探进程和结果。前人在计算水合物稳定带厚度时往往采用较为简单的模型，忽略气体组分、热导率等因素的影响，计算结果存在较大偏差。基于不同海水盐度和气体组分条件下的天然气水合物实验数据构建机器学习模型，利用机器学习算法预测天然气水合物的相平衡条件；进而结合南海北部的气体组分、热流、热导率等数据，计算得到南海北部水合物稳定带厚度。结果分析表明机器学习模型预测的水合物相平衡曲线与实验数据高度吻合，决定系数高达0.997。计算的南海北部水合物稳定带厚度与水合物钻井和地震资料揭示的结果基本一致。 本研究提供了一个机器学习算法在水合物稳定带厚度估算中的应用实例，表明人工智能算法在未来天然气水合物资源预测和潜力评价中具有较大的应用前景。

Thickness of gas hydrate stability zone calculated by machine learning method

With the development of big data processing techniques，artificial intelligence algorithm is emerging to automatically dig out the features and hidden information from large datasets，which performs with high accuracy and reliability.Machine learning is one of the core algorithms of artificial intelligence，which has been applied to a variety of fields，including geoscience research.The thickness of gas hydrate stability zone is one of the key parameters to assess the resource potential of gas hydrate，and its accuracy directly affects the process and result of gas hydrate exploration.Previous studies to calculate the thickness of gas hydrate stability zone，usually utilize simple gas hydrate phase equations，which ignore many effects，such as gas component，thermal conductivity，etc.，so that the results contain large bias.This paper predicted the phase condition of gas hydrate using machine learning，based on the experimental hydrate formation data of different gas components and seawater salinities.Then combining with heat flow，water depth and thermal conductivity of northern South China Sea，the thickness of gas hydrate stability zone was obtained by machine learning model.The predicted result is highly consistent with the experimental data and the coefficient of determination is up to 0.997.The computed thickness of gas hydrate stability agrees well with drilling data and seismic profiles.This paper provides an applied example for machine learning to calculate the thickness of gas hydrate in the northern South China Sea，which indicates that artificial intelligence has great application prospect in the potential evaluation of gas hydrate resource in the feature.