Gas-oil gravity drainage mechanism in fractured oil reservoirs: surrogate model development and sensitivity analysis

In this study, a surrogate model is developed based on single porosity modelling approach to predict the gas-oil gravity drainage recovery curves. The single porosity model is validated against experimental data associated with gravity drainage available for an actual core sample gathered from a naturally fractured reservoir. Then, using the single porosity model and a vast databank of rock and fluid data, the gravity drainage recovery curves are generated. Two empirical functions, namely Lambert and Aronofsky are then utilized to match the generated recovery curves. An exact graphical approach based on saturation profiles is put forth to compute ultimate oil recovery and an artificial neural network is developed to predict the convergence constant. The findings imply that unlike Lambert, Aronofsky is more accurate in capturing the simulation-generated recovery curves. The Lambert function overpredicts the early-time and underpredicts late-time oil recovery in most cases. The statistical parameters and plotted graphs indicate that the developed surrogate model is robust and precise in predicting the Aronofsky function convergence constant. In addition, it is found that unlike absolute permeability and maximum oil relative permeability, other parameters including block height, oil viscosity, and porosity directly correlate with the time required to reach ultimate oil recovery. The presented graphical approach indicates that initial water saturation, density difference, block height, and capillary pressure curve are the only parameters that could affect the ultimate oil recovery. The sensitivity analysis aided by the developed surrogate model and Monte Carlo algorithm affirms that absolute permeability and density difference are the parameters that have the maximum and minimum impact, respectively, on the time required to reach the ultimate recovery.