The global decline of conventional oil and gas resources has positioned enhanced oil recovery (EOR) as a key strategy for strengthening energy security through improved recovery from unconventional and residual reserves. However, widespread EOR implementation faces significant challenges, including strong reservoir heterogeneity, high prediction uncertainty, substantial costs, and the limited efficiency of the traditional numerical simulation. Machine learning (ML), with its advanced data mining and nonlinear modeling capabilities, offers a promising technological pathway to overcoming these barriers. This review systematically examines ML applications across four major EOR categories: chemical flooding (such as polymer, surfactant, and nanofluid flooding), gas injection (including nitrogen, hydrocarbon gas, and CO2 flooding), thermal recovery (covering steam flooding, in situ combustion, and steam-assisted gravity drainage), and emerging microbial flooding. The study further identifies four core issues in current ML-based EOR research: interdisciplinary collaboration, multiscale intelligent modeling, model interpretability, and data quality and fusion. Based on these findings, six forward-looking recommendations are proposed to guide future developments. By providing a comprehensive and integrated overview of ML applications throughout the EOR process chain, this work not only fills a gap in the existing literature but also supports the intelligent and precise transformation of EOR technology, offering practical value for both engineering and sustainable oilfield development.
Jing et al. (Wed,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: