New insights into forced and free fall gravity drainage performance in a fractured physical model

Abstract

Gravity is one of the major driving forces for fluid flow in naturally fractured reservoirs. Gravity drainage appears to be a vital recovery mechanism over the oil production process in a majority of fractured reserves. In this paper, free-fall gravity drainage (FFGD) and forced gravity drainage (FGD) production approaches are investigated through employing a new fractured sand pack with a rectangular geometry. Distilled water and condensate are used as the wetting phase. Nitrogen, carbon dioxide, and air are also utilized as the non-wetting phase. A systematic sensitivity analysis is conducted to find the effects of the key factors such as fracture blockage, injection rate, dip angle, and type of wetting phase and injection fluid on the performance of FFGD and FGD processes. Also, a numerical simulation is performed to simulate the experiments. Results reveal that the FGD has a better performance under controlled process conditions, compared to the FFGD so that gas injection with a controlled rate improves the recovery. Moreover, it was found that increasing the gas injection rate improves the recovery factor (RF) of wetting-phase when the up and bottom sides of the fractures are blocked; the RF is decreased in the case with open fracture sides. The recovery of the wetting phase decreases in both FFGD and FGD modes when the porous system has a deviation from the vertical orientation. According to the study outcomes, the immiscible gas-assisted gravity drainage offers a higher condensate recovery, compared to water recovery, due to lower viscosity and interfacial tension (IFT) of the condensate phase. The comparison of the experimental data and simulation results implies that the simulation approach can satisfactorily predict the performance of the FGD process. Based on the error analysis, there is a good match between the simulation results and the experimental measurements; the fluid pair of CO2-condensate exhibits the highest accuracy with an R-squared of 0.9961 and an average relative error (ARE) of 0.0054.