流量(数学)
材料科学
化学
热力学
碳纤维
环境科学
一氧化碳
工作(物理)
矿物学
化学工程
流动条件
燃烧
作者
Rongchen TONG,Sida Jia,Quan Gan,Zihao Wang,Lulu Xu,Cunbao Li,Junlei WANG,Zhenxue Dai
出处
期刊:Energy & Fuels
[American Chemical Society]
日期:2026-06-25
卷期号:40 (27): 14500-14515
标识
DOI:10.1021/acs.energyfuels.6c00995
摘要
Carbon dioxide (CO2) injection for methane (CH4) displacement has become a pivotal technology for synergizing enhanced gas recovery with carbon storage. However, the absence of a systematic framework to decipher the complex interplay between pore structure and flow dynamics remains a critical barrier to optimizing displacement efficiency. This study systematically investigates the combined effects of pore structure and flow conditions on CO2-driven CH4 displacement through 2000 numerical simulations. A total of 19 input parameters, including pore-scale variables, flow condition parameters, and parameters associated with the quartet structure generation set method, were used to predict breakthrough time (Tb), sweep efficiency (Es30, Es430), fingering coefficient (Fv), and residual amount of CH4 (Rm430). This integrated framework combines 2D pore-scale modeling, numerical simulation, and machine learning to provide interpretable insights, although the use of 2D pore structures may underestimate connectivity-related descriptors such as coordination number (Nc) and tortuosity (τ). Key quantitative results reveal that injection pressure (Pi) is the dominant control (total-effect index ST up to 0.77), while the characteristic length ratio (Lhv) and porosity (ϕ) primarily govern fingering (e.g., Lhv first-order index S1 = 0.28 for Fv). The optimized XGBoost model achieved high predictive accuracy, with R2 values of 0.91 for Tb and 0.90 for Es430 on the test set. Interpretability analysis further uncovered critical trade-offs, such as high Lhv accelerating early displacement at the cost of intensified fingering. In conclusion, this work not only quantifies the key controls on CO2–CH4 displacement but also provides an interpretable, data-driven methodology to translate simulation data into actionable insights for optimizing injection strategies, thereby bridging the gap between mechanistic understanding and field application.
科研通智能强力驱动
Strongly Powered by AbleSci AI