Simplified Predictive Modeling of CO2 Geologic Sequestration in Saline Formations: Insights

Geologic sequestration of CO2 in deep saline formations has been recognized as an attractive proposition to control greenhouse gas emissions.  Understanding the nature of pressure buildup and plume movement as injected CO2 displaces the native brine is key to ensuring sound storage strategy.  Detailed numerical simulation of such dynamic processes generally requires significant reservoir characterization effort and computational burden.  Simplified models, when properly constructed and benchmarked, can be useful alternatives for rapid evaluation of CO2 sequestration projects.  The overall objective of this work is to develop simplified 2-D models to characterize CO2 plume movement in saline aquifers based on the most relevant physical processes. To this end, we use a set of well-designed full-physics compositional simulations using CMG-GEM(R) to understand key processes and parameters affecting buoyant plume movement and pressure propagation when supercritical CO2 is injected into a saline aquifer.

The computational model consists of a single vertical well injecting supercritical CO2 into a 2-D layered reservoir bounded by a less permeable caprock.  Adding the caprock to the model along with vertical layering (as compared to the simplified 1-D model of Oruganti and Mishra, IJGGC, 2013) introduces buoyancy effects for plume migration as well as pressure attenuation effects.  Simulations are run for an injection period of 30 years to observe CO2 displacement characteristics in a closed system – as would be the case in a network of injection wells.  Key performance metrics of interest are: magnitude of pressure buildup in the reservoir and caprock, average pressure in the reservoir, radial extent of the CO2 plume and pressure fronts, average CO2 saturation behind the front, and volumetric sweep efficiency.  The independent variables of interest for this sensitivity analysis are thickness of reservoir and caprock, their respective porosities and permeabilities, permeability anisotropy ratio (ratio of vertical to horizontal permeability) in the reservoir, spatial arrangement of the heterogeneous permeability layers in the reservoir, capillary entry pressure of the caprock, and relative permeability curves for the reservoir. We investigate system behavior for high and low variants from a reference case for each of these variables and seek to quantify their effect on each of the dependent performance metrics.

In all of the simulations, the pressure at the injection well quickly jumps to a quasi-steady value and remains relatively stable thereafter during the early transient period before boundary effects come into play.  One useful representation of such behavior can be captured via the injectivity index, calculated as the ratio of the injection rate to the quasi-steady pressure buildup at the injection well.  This index correlates very well with the permeability-thickness product of the reservoir – thus providing a robust predictive tool.  Using a steady-state version of Darcy’s law as a starting point, we develop an empirical equation for injectivity index that can predict the magnitude of injection well pressure buildup given a target injection rate, or the injection rate corresponding to a target pressure differential.

The pressure buildup in the reservoir is affected primarily by the permeability-thickness product of the reservoir, the injection rate, and the relative permeability model for the reservoir.  The corresponding pressure buildup in the caprock is additionally affected by the permeability-thickness and the porosity of the caprock.  We are currently exploring the appropriate form of a simplified expression that relates the incremental pressure buildup in the caprock to the hydraulic properties of the caprock.   

Once the boundary effects are felt, the system tends to behave like a closed tank – whereupon the average pressure buildup in the reservoir can be related to the reservoir pore volume and total compressibility.  We find that any attenuation of pressure from this linear relationship because of the presence of a caprock (which acts to “bleed off” reservoir pressure rise) can be described using a correction term related to the relative thickness of the caprock in the stratigraphic column.  Such simplified models can be useful for predicting the magnitude of pressure dissipation for various postulated caprock characteristics in the absence of detailed site characterization data.

When CO2 is injected into the reservoir, it displaces the resident brine and advances into the reservoir as a two-phase front.  The efficiency of this CO2-brine displacement process reflects the ability to effectively sequester CO2 in that reservoir.  It is a product of two factors: (a) volumetric sweep efficiency reflecting the fraction of the total pore volume contacted by CO2, and (b) displacement efficiency within the pore volume contacted by CO2. As the initial gas saturation in the aquifer is zero, the average CO2 saturation behind the front gives the displacement efficiency.  The volumetric sweep efficiency is calculated from the ratio of the actual pore volume contacted by CO2 to the pore volume corresponding to the maximum extent of the CO2 saturation front. 

Based on our suite of sensitivity simulations, the average CO2 saturation behind the front is only weakly affected by all parameters excepting the relative permeability model.  This is consistent with the standard two-phase Buckley-Leverett immiscible displacement theory, in which the average displacing phase saturation is proportional to the slope of the fractional flow curve.  We have developed a simple relationship that relates the average saturation to a parameter group characterizing the fractional flow curve shape.  Similarly, the volumetric sweep efficiency is related to a gravity number group which includes the vertical permeability, injection rate, reservoir thickness and system dimensions.  These simplified correlations can be useful in predicting how plume migration is affected by buoyancy.

In summary, the results of this study will provide a suite of very useful simplified relationships and correlations for characterizing CO2 plume movement and pressure propagation in a 2-D layered aquifer bounded by a caprock. 

*This work was supported by U.S. Department of Energy National Energy Technology Laboratory award DE-FE0009051 and Ohio Department of Development grant D-13-02.