Investigation of Fluid Phase Behavior in Shale Reservoirs Using Equation of State, Molecular Simulation and Machine Learning
Abstract
Fluid phase behavior in shale reservoirs differs significantly from phase behavior in conventional reservoirs due to the strong interactions between fluid and boundary in nanopores. In this study, we applied equation-of-state (EOS) modeling, machine learning (ML) technique and molecular simulation to investigate fluid phase behavior in shale reservoirs.
One common issue observed in liquid-rich shale (LRS) production is that oil recovery of LRS reservoirs is much lower compared to oil recovery from a conventional reservoir with the same drawdown. To understand this phenomenon, EOS modeling is developed to analyze the fluid compositions in the bulk and confined regions. Our simulation results indicate that hydrocarbons distribute heterogeneously with respect to pore size on a nanoscale. The leaner bulk composition leads to the reduction in oil recovery from LRS reservoirs. Although EOS modeling can accurately simulate fluid phase behavior in shale reservoirs, the required simulation time is much longer than that for models of conventional reservoirs. To solve this problem, ML techniques were applied to accelerate the phase-equilibrium calculations in the EOS modeling. In contrast to previous models designed for a specific type of hydrocarbon, we have developed a generalized, ML-assisted phase-equilibrium calculation model that is suitable for shale reservoirs. In total, the average CPU time required for phase-equilibrium calculation using the generalized ML-assisted phase-equilibrium model was reduced by more than two orders of magnitude while maintaining an accuracy of 97%. With the development of shale oil and gas, depleted shale gas reservoirs may be attractive candidates for hydrogen (H2) storage. Molecular simulation was used to investigate the potential for H2 storage in depleted shale gas reservoirs. The results of the simulation suggest that a higher proportion of H2 exists in the bulk region. Because fluid is mainly produced from the bulk region, the high percentage of H2 in bulk fluid would lead to high purity of H2 during the recovery process. This work contributes to the understanding and application of fluid phase behavior in shale reservoirs.
Citation
Chen, Fangxuan (2023). Investigation of Fluid Phase Behavior in Shale Reservoirs Using Equation of State, Molecular Simulation and Machine Learning. Doctoral dissertation, Texas A&M University. Available electronically from https : / /hdl .handle .net /1969 .1 /198938.