Experimental Program Investigates Miscible CO2 WAG Injection in Carbonate Reservoirs

You have access to this full article to experience the outstanding content available to SPE members and JPT subscribers.

To ensure continued access to JPT's content, please Sign In, JOIN SPE, or Subscribe to JPT

Prediction of miscible water-alternating-gas (WAG) injection performance relies on proper calibration of thermodynamical and petrophysical models. Swelling, miscibility, and stripping phenomena must be captured in the equation of state (EOS), and the oscillations of gas and water saturations require using history-dependent relative permeabilities. This paper provides a robust methodology for miscible CO2 WAG experimental-data acquisition and history matching.


Miscible CO2 injection in oil reservoirs leads to low residual oil saturations in the swept areas. However, macroscopic sweeping can be poor because of the high mobility of CO2. One way of improving macroscopic sweep is to inject CO2 in the presence of mobile water to reduce its mobility (in tertiary or WAG modes).

The prediction of WAG efficiency and the sizing of surface installations rely partly on the ability of the three-phase relative-permeability model to calculate proper mobility for each phase in any part of the reservoir. Numerous three-phase relative-permeability models have been proposed in the literature. Among these models, one proposed by Larsen and Skauge (1998) is considered by the authors as a starting point for simulation work. The experimental program discussed in this paper was launched in order to clarify several points, including the existence and the amplitude of residual oil saturation in miscible flooding and the validity of existing three-phase models.

Experimental-Program Definition for WAG Study

The experimental program is ­composed of several coreflooding experiments performed at reservoir conditions of 260 bar and 94°C. Materials and methods are discussed in detail in the ­complete paper.

Flow Regions and Saturation Paths. WAG mechanisms involve alternating two fluids of different density and mobility properties at the injection well. These differences lead to gas migration toward the top of the reservoir. The consequence is a large variability of flow sequences in space as illustrated in Fig. 1, which shows a cross section between two vertical wells from a WAG phenomenological simulation. Several saturations paths were extracted to identify six flow regions that can be grouped into three categories:

  • Two-phase flow with monotonous saturations variations (Flow Regions 1 and 2)
  • Two-phase flow with oscillating saturations (Flow Regions 3 and 4)
  • Three-phase flow with oscillating saturations (Flow Regions 5 and 6)
Fig. 1—Cross section between two vertical wells from a numerical simulation of secondary immiscible WAG injection in a homogeneous 3-m-thick layer. Green=oil; red=gas; blue=water.


Depending on reservoir properties and the development plan, the proportions between these flow regions can vary significantly and should be assessed for a given reservoir before the experimental-program definition. A reliable three-phase relative-permeability model should be valid over each flow region, or at least over the most significant one for a given case.

The experimental program discussed in this paper covers this variability of saturations paths, including four WAG coreflooding experiments. The first slugs from the long-slug experiments described Flow Regions 1 and 2. The following slugs enabled calibration of the three-phase relative-permeability model required to describe flow in three-phase zones (Flow Regions 5 and 6). Flow Regions 3 and 4 were not studied in this experimental program.

Experimental Results and History Matching: Three-Phase Flow

Experiments dedicated to secondary waterflooding and secondary miscible CO2 flooding are detailed in the complete paper. An attempt was made to match three-phase slugs W2 and G3 from Experiment 1 and slugs G2 and W3 from Experiment 4 using the Larsen and Skauge three-phase hysteresis model.

The observations to be matched in the optimization process included the following:

  • Oil production at surface conditions
  • Condensate production at surface conditions
  • Differential pressure
  • Average water saturation

The authors claim that, when history matched, these properties were sufficient to guarantee that recovery mechanisms were properly captured in the simulations.

Incremental Oil During Three-Phase Flow. The first challenge was to properly represent oil and condensate production during three-phase flows. This problem was reduced to slug G2 from Experiment 4, the only three-phase slug to provide incremental oil compared with a two-phase slug. In this case, CO2 displaced both water and oil. For all other slugs, the problem was reduced to two-phase flow in the presence of a third ­immobile phase.

Additional oil and condensate production from G2 (Experiment 4) could be matched using the same EOS, alpha factors, and two-phase relative permeabilities obtained during history matching of G1 (Experiment 1). The function was chosen linearly between residual oil-to-water saturation and residual oil-to-gas saturation. Consequently, it can be concluded that three-phase oil recovery during G2 was from a microscopic point of view identical to that seen in recovery during secondary CO2 flooding.

Gas and Water Relative Permeabilities During Three-Phase Flow. Water Injections (W2 and W3). Water slugs did not recover additional oil. The flow problem was reduced for all three-phase water slugs to a water-displacing-gas problem in the presence of the immobile remaining oil fraction. Gas displacement by water was stable because of a favorable viscosity ratio, regardless of pore-space occupancy. The authors confirmed that gas was microscopically trapped and not retained because of capillary pressure discontinuity at the outlet (the capillary end effect).

Gas Injections (G2 and G3). The main interest of research, apart from incremental oil production, during three-phase gas injection was to determine if the water cluster was fragmented during gas injection. Trapped-water-saturation assessment was not as straightforward as it is for determination of trapped-gas saturation because of the unstable nature of the displacement. Large volumes were required to approach residual saturations, which were never reached in most of the core.

Three-phase gas injections require more caution in terms of experimental procedures. Trapped water should be obtained by history matching and should not be read directly from experimental data. In addition, to be constrained properly, very large amounts of gas should be injected to reach residual water saturation to gas. The authors concluded that water retention in the core was not caused by the large macroscopic capillary end effect.


Reliability of Obtained Three-Phase Relative Permeabilities For Field-Scale Evaluations: Saturation Paths and Simultaneous History Matching. A reliable three-phase relative-permeability model should be valid over every saturation path encountered in the reservoir (i.e., obtained by simultaneous history match of several experiments). The authors claim that, in a configuration such as the one shown in Fig. 1, if the three-phase relative permeability model is valid for Saturation Paths 1, 2, and 5, flow will be captured in an acceptable way in large-scale simulations. Simultaneous history matching was properly achieved following this work flow. Every point in the three-phase zone of the reservoir undergoes a different saturation path when performing WAG injection. The validity of the three-phase relative-permeability model cannot be constrained experimentally over every saturation path encountered in the reservoir.

Any change in the cell dimensions in the simulations from microscopic scale to larger scale should be accompanied by a change in relative permeabilities and alpha factors. Although upscaling was not addressed in this study, the authors’ comparison of 2D and 1D simulation provides an indication of the effect of gridding in Z direction at core scale.

Recovery-Mechanism Efficiency.

Recovery-mechanism efficiency is a combination of microscopic and macroscopic efficiencies. Macroscopic efficiency depends on flow stability, heterogeneity, and gravity segregation. Coreflood experiments investigate microscopic effects. However, the authors’ experiments underwent macroscopic gravity effects. Results at core scale were found to originate from a combination of microscopic efficiency and macroscopic vertical efficiency.

The history matching enabled uncorrelating of microscopic and macroscopic effects. Waterflooding efficiency was not good because of flow instability. Remaining oil distribution was controlled by gas-gravity segregation with better sweep in the upper part of the core and close to the inlet. Recovery in tertiary CO2 mode combined good microscopic efficiency (the same as seen in secondary CO2 flooding) with a much better vertical sweep efficiency. Residual oil saturation of 7% was observed almost everywhere in the core. The recovery improvement in WAG compared with that in secondary CO2 was explained by improved vertical conformance at core scale because of reduced CO2 permeability in the presence of mobile water.


Experiments were performed at complete reservoir conditions for proper investigation of gas/oil exchanges during miscible injections. Heavy protocols and monitoring techniques were used to observe recovery mechanisms, capillary end effects, and gravity-override effects.

2D compositional simulations with three-phase hysteresis were performed, matching the observed thermodynamical and petrophysical observations. A classical three-phase hysteresis model was used and captured satisfactorily the observations except water saturation during gas injections. A new model was implemented, allowing water trapping that improved the history match. This new model did not increase simulation time.

CO2 injection performed better than waterflooding at core scale, whatever the injection mode. However, recovery by CO2 injection was not 100% at microscopic scale because a residual oil saturation of 7% was observed. The best performance at core scale for CO2 injection was achieved in tertiary mode, in the presence of mobile water. This better recovery was explained by an increased vertical sweep efficiency compared with that of the secondary CO2 flood.

Hysteresis was observed on both phases. A model allowing water trapping in addition to gas trapping was proposed and helped to improve the history match. However, the assessment of water trapping appeared to be much more difficult because of the unstable nature of water displacement by gas and the potential capillary end effects.


Larsen, J.A., and Skauge, A. 1998. Methodology for Numerical Simulation With Cycle-Dependent Relative Permeabilities. SPE J. 3 (2): 163–173. SPE 38546.

For a limited time, the complete paper SPE 188471 is free to SPE members.

This article, written by JPT Technology Editor Chris Carpenter, contains highlights of paper SPE 188471, “Experimental Data Acquisition and Modeling for the Study of Miscible CO2 WAG in Carbonate Reservoirs Under Oil-Wet Conditions,” by T. Joubert, S. Duchenne, M. Bourgeois, SPE, R. de Loubens, SPE, and M. Petitfrere, Total, prepared for the 2017 Abu Dhabi International Petroleum Exhibition and Conference, Abu Dhabi, 13–16 November. The paper has not been peer reviewed.

Experimental Program Investigates Miscible CO2 WAG Injection in Carbonate Reservoirs

01 January 2019

Volume: 71 | Issue: 1


Don't miss out on the latest technology delivered to your email weekly.  Sign up for the JPT newsletter.  If you are not logged in, you will receive a confirmation email that you will need to click on to confirm you want to receive the newsletter.