Reservoir simulation

Experimental and Numerical Studies of CO2 EOR in Unconventional Reservoirs

This study explores the mechanisms contributing to oil recovery with numerical modeling of experimental work and investigates the effects of various parameters on oil recovery.

Abstract image
Getty Images

This study investigates oil-production mechanisms from the matrix into the fracture by simulating two laboratory experiments as well as several field-scale studies and evaluates the potential of using carbon dioxide (CO2) huff ’n’ puff to enhance oil recovery in unconventional liquid reservoirs (ULRs) with nanodarcy-range matrix permeability and complex natural-fracture networks. This study explores the mechanisms contributing to oil recovery with numerical modeling of experimental work and investigates the effects of various parameters on oil recovery.


CO2 injection has been recognized as one of the more important and successfully applied enhanced-oil-recovery (EOR) processes in the US since the start of the first commercial CO2-injection project in 1972. Statistics show that, out of the 153 active CO2 EOR projects worldwide, 139 of them are in the US. CO2 flooding is the only EOR method in the US that has been consistent and economical since the drop in oil prices in the 1980s. One of the main advantages of CO2 injection is that CO2 can achieve miscibility with the resident hydrocarbon when minimum miscibility pressure (MMP) is attained.

Although CO2 flooding has demonstrated the potential of increasing oil production in conventional reservoirs, if this EOR process were to be applied to ULRs, the oil-recovery mechanism could not be considered the same as that during a conventional CO2 flooding because of the differences in reservoir petrophysical properties, fluid-phase behavior, and the fundamental mass-transfer mechanisms.

In general, diffusion during gas injection has been recognized as a critical mechanism that affects the oil recovery in fractured reservoirs. If the diffusive flow in the matrix is being neglected during simulation, the calculated result will underestimate productivity.

Furthermore, complex fracture geometry has been proposed for hydraulic fracturing because of the intersection between hydraulic fractures and in-situ natural fractures. Numerous simulation and experimental studies have evaluated production performance of complex fracture networks. However, there is no extensive work regarding the potential of CO2 huff ’n’ puff in complex fracture geometries.

Core-Scale Laboratory Experiment

Two experiments were performed using two sets of preserved sidewall core from the same well in a ULR. The petrophysical properties and the saturations of the cores were not measured before the experiments, to preserve the core original conditions. However, because the cores are not stored in a pressurized environment, the fluids saturating the cores can be assumed to be dead oil and water. To simulate the presence of a highly permeable fracture around the core, glass beads were packed outside the cores to allow CO2 to have direct contact with the cores. Two Berea sandstone disks were placed on each end of the setup to block the glass beads from entering the production tube. The core holder was then placed inside a water bath that circulated hot water to simulate the reservoir temperature. The assembly was mounted in a computed-tomography (CT) scanner, and the cores were kept horizontal during the experiment.

The experiments were performed under constant-pressure conditions where viscous displacement was absent. The first experiment (i.e., Experiment 1 as seen in Fig. 1) was performed at 3,000 psi at 150°F, and the second experiment (i.e., Experiment 2 as seen in Fig. 2) was performed at 1,600 psi and 150°F. Periodic scanning of the cores with the CT scanner was performed throughout the experiment. The experiments typically last for 2 to 3 days, and production was allowed twice a day on average. The CT images of the core revealed constant density/saturation/composition changes of the resident fluid during both experiments, which indicated that CO2 was constantly penetrating into the core throughout the process. Final oil-production volumes of 0.4 mL were recorded for both experiments. On the basis of the estimated original oil in place, the experiments yielded high oil recovery.

Fig. 1: Saturation variation of two CT images with time for Experiment 1.
Fig. 2: Saturation variation of two CT images with time for Experiment 2.

Core-Scale Numerical Modeling

The core-sale modeling followed the ­experimental work described in the preceding section. The petrophysical properties of the cores and the fluid properties of the dead oil were evaluated after the experiments.

Core-Scale Experiment Simulation. To save computational time, a horizontal slice of the Cartesian core model was taken for investigation.

For the base case, the conditions are 3,000 psi and 150°F, which are the same as for Experiment 1. Injectors were placed in the fracture region for pressurizing and pressure maintenance. Producers were not introduced at this point.

The recovery factor is not sensitive to the matrix and fracture permeability. The higher the fracture and matrix porosity, the larger the diffusion coefficient and the better the recovery factor. In addition, because the MMP of the fluid system was determined to be 1,727 psi, three system pressures were considered—3,000 psi (miscible as Experiment 1), 1,600 psi (immiscible as Experiment 2), and 1,800 psi (near miscible). The main mechanism for mass transfer occurring between the matrix and fracture is considered to be diffusion. For a diffusion case, higher pressure would result in better CO2 solubility in oil, which contributes to the increase in recovery. Furthermore, for the capillary pressure, the base case is under miscible conditions with zero capillary pressure. Water saturation slightly affects the recovery factor, which might be because of differences in the compressibility of water and oil and CO2 being soluble in water.

Field-Scale Numerical Modeling

A synthetic field case was developed to study oil-recovery mechanisms on a larger scale. To save computational time, a single stage with one fracture on a horizontal well was modeled. The main focus of the field study was to use the properties from the core model to predict and optimize the performance of CO2 injection in various cases. In this study, the field-scale input parameters were obtained from the previous core-scale simulation.

The fluid model for the experimental simulation was built on the basis of the dead-oil data from the reservoir. To model the fluid in the reservoir with little knowledge of the live-oil properties, methane was combined with the dead-oil components to create a live-oil model.

Base cases with different production pressures were considered for primary depletion and for huff ’n’ puff. The bottomhole flowing pressures for the different cases are 2,000 psi; 1,550 psi, which is slightly above the MMP; 1,300 psi, which is below the MMP but above bubblepoint pressure; and 1,000 psi, which is below the bubblepoint pressure. After 30 days of injection and 15 days of soaking, production took place at the same pressure as the corresponding primary-depletion bottomhole pressure.

The oil incremental percentage for the cases is modest, as is expected for huff ’n’ puff. In reality, natural fractures may exist in these reservoirs so that production should be higher than the calculated value in this model, depending on the ­intensity of the natural fractures.

Summary of Sensitivity Parameters. The effect of several parameters was investigated on both cumulative oil production and oil incremental percentage. Among all the parameters, the matrix/fracture porosity and permeability are the most significant, and the higher the matrix/fracture porosity and permeability, the higher the incremental oil. On the other hand, diffusion, capillary pressure, and soaking time have almost no effect on the field model. For the field model, convection from pressure drawdown can be considered to be the dominating mechanism instead of diffusion, which is different from the previous core-scale simulations where diffusion is the dominant mechanism for oil recovery. Furthermore, regarding injection operating parameters, the later the initial injection time is, the larger the injection rate is, the higher the injection pressure is, and the longer the injection length is; the more cycles there are, the more chances there are to repressurize the reservoir and provide gas drive to assist production under immiscible conditions.


  • Core-scale experiments show that CO2 huff ’n’ puff recovers oil from ULRs.
  • Core-scale simulations show that diffusion plays a key role in oil recovery. The recovery factor is sensitive to matrix porosity, fracture porosity, saturation, and diffusion coefficient, and it is not sensitive to matrix permeability, fracture permeability, relative permeability, or capillary pressure. Improved oil recovery was observed by changing the porosity and the diffusion coefficient.
  • From field-scale simulation, it was observed that gas expansion is an important mechanism during depletion; thus, huff ’n’ puff is most effective below bubblepoint pressure.
  • Field-scale simulation shows that the incremental oil is sensitive to matrix porosity, matrix permeability, fracture porosity, fracture permeability, time of first injection, cycling length, injection rate, injection pressure, and number of cycles, and it is not sensitive to capillary pressure, diffusion coefficient, or length of soaking time.

This article, written by Special Publications Editor Adam Wilson, contains highlights of paper SPE 179634, “Experimental and Numerical Studies of CO2 EOR in Unconventional Liquid Reservoirs With Complex Fracture Networks,” by Jianlei Sun, SPE, Amy Zou, and David Schechter, SPE, Texas A&M University, prepared for the 2016 SPE Improved Oil Recovery Conference, Tulsa, 11–13 April. The paper has not been peer reviewed.