基于机器学习的海底地质封存CO_(2)快速泄漏评估  

Machine Learning-Based Assessment of CO_(2) Rapid Leakage from Offshore Geological Storage

作　　者：张海滨 卢迪 王永昌 高凤溥 宋学行 孙楠楠 ZHANG Haibin;LU Di;WANG Yongchang;GAO Fengpu;SONG Xuehang;SUN Nannan(CNOOC Energy Conservation and Emission Reduction Center,Tianjin 300452,China;CAS Key Laboratory of Low-carbon Conversion Science and Engineering,Shanghai Advanced Research Institute,Chinese Academy of Sciences,Shanghai 201210,China)

机构地区：[1]海油总节能减排监测中心有限公司,天津300452 [2]中国科学院低碳转化科学与工程重点实验室,中国科学院上海高等研究院,上海201210

出　　处：《力学季刊》2025年第1期99-107,共9页Chinese Quarterly of Mechanics

基　　金：中海油能源发展股份有限公司重大科技专项与中国科学院青促会项目(2021286)。

摘　　要：鉴于我国东部沿海地区碳排放量高且陆上封存容量有限的现状,CO_(2)海底地质封存具有显著的应用潜力.尽管海底地质封存过程中CO_(2)泄漏的风险极低,但准确评估快速泄漏情景下的海水中CO_(2)扩散规律对于确保封存安全至关重要.本研究基于典型海洋环境数据,采用VolumeofFluid(VOF)模型模拟CO_(2)气泡动态特性,并用欧拉模型模拟CO_(2)集群在海水中的扩散特征,从而完成流体动力学的综合仿真分析.在此基础上,开发基于机器学习的海底CO_(2)泄漏量评估方法.构建神经网络模型框架,并据此建立反问题求解模型,解析不同海洋环境条件下CO_(2)泄漏时的形态演变与分布范围,反演出导致此泄漏情景的裂缝宽度及泄漏速度等关键参数,进而精确计算出泄漏量.该模型预测精度可达95%以上,证明其在处理CO_(2)泄漏反问题上的有效性和适用性.本研究不仅为海底地质封存CO_(2)快速泄漏评估提供了一种创新性的方法路径,也为未来相关领域的风险评估与安全管理提供了重要的科学依据和技术支撑.In view of the high carbon emissions and limited land storage capacity in the eastern coastal areas of China,CO_(2) offshore geological storage holds significant application potential.Although the risk of CO_(2) leakage during offshore geological storage is extremely low,accurate assessment of CO_(2) diffusion in seawater under rapid leak scenarios is critical to ensure the safety of storage.Based on typical marine environment data,this study employs the Volume of Fluid(VOF)model to simulate the dynamic characteristics of CO₂bubbles and utilizes the Euler model to simulate the diffusion characteristics of CO_(2) clusters in seawater,thereby completing a comprehensive simulation analysis of fluid dynamics.On this basis,a machine learning-based assessment method for subsea CO_(2) leakage is developed.The neural network model framework is constructed,and an inverseproblem solving model is established accordingly to analyze the morphological evolution and distribution range of CO_(2) leakage under different marine environmental conditions.Additionally,key parameters such as crack width and leakage velocity resulting from the leakage scenario are analyzed,enabling accurate calculation of the leakage amount.The prediction accuracy of this model can reach 95%,demonstrating its effectiveness and applicability in addressing CO₂leakage.This study not only offers an innovative method for rapid leakage assessment of CO 2 offshore geological storage but also provides an important scientific basis and technical support for risk assessment and safety management in related fields in the future.

关 键 词：机器学习 CO_(2) 海底地质封存 快速泄漏 

分 类 号：P744.5[天文地球—海洋科学]