SPE Journal
Volume 13, Number 4, December 2008, pp. 447-454

Abstract

We have previously proposed the "inject low and let rise" strategy of storing CO₂ in deep saline aquifers. The idea is to maximize the amount of CO₂ stored in immobile forms by letting CO₂ rise toward the top seal of the aquifer but not reach it. The distance that the CO₂ rises depends on the uniformity of the displacement front. In this paper, we address the question of whether the intrinsic instability of a buoyancy-driven immiscible displacement leads to fingering. Fingers could reach the top seal of the aquifer, leading to an accumulation of CO₂ at large saturations. We study the mechanisms governing this type of displacement in a series of fine-grid numerical simulations. Each simulation begins with a finite volume of CO₂ placed at large saturation at the bottom of a 2D aquifer. Only buoyancy forces drive the displacement. Boundaries are closed, so CO₂ rises and brine falls as the simulation proceeds. Several fine-scale geostatistical realizations of permeability are considered, and the effects of capillary pressure, anisotropy, and dip angle are examined. In these simulations, buoyant instability has very little effect on the uniformity of the displacement front. Instead, the CO₂ rises along preferential flow paths that are the consequence of spatially heterogeneous rock properties (permeability, drainage capillary pressure curve, and anisotropy). Capillary pressure broadens the lateral extent of the flow paths. If the formation beds are not horizontal, capillary pressure and anisotropy can cause the CO₂ to move predominantly along the bedding plane rather than vertically. Accurate assessment of CO₂ migration after injection ends will therefore require accurate characterization of the spatial correlation of permeability in the target formation and of the capillary pressure and relative permeability curves.

Introduction

Storing CO₂ in deep saline aquifers will be a key technology if society elects to limit the amount of greenhouse gases entering the atmosphere. Large-scale (10⁶ tonnes of CO₂ per year) examples of this type of storage are underway at Sleipner and In Salah, and more are planned (IPCC 2008). Effective mitigation of CO₂ emissions will require many more projects of this type, storing on the order of 10⁹ tonnes per year (Pacala and Socolow 2004). In terms of volumetric flow rates through wellbores, this rate of storage is of the same magnitude as the current global rate of oil production. Thus inexpensive, reliable methods of ensuring that stored CO₂ remains in place will be essential.

CO2 can be stored in an aquifer in four modes: as a bulk phase within a structural trap, as a residual phase trapped by capillary forces, as aqueous species dissolved in brine, and as a precipitated mineral. The latter three forms of storage are permanent in the sense that the CO₂ will remain in the aquifer at least as long as the residence time of water in the aquifer. On the other hand, CO₂ held in a structural trap at large saturations (above residual) is potentially mobile in that it will remain trapped only as long as the seal remains intact. Storage methods that reduce the amount of potentially mobile CO₂ correspondingly reduce the risk of leakage over the long term.

The inject-low-and-let-rise strategy is one such method (Kumar et al. 2005; Ozah et al. 2005). Under typical storage conditions, CO₂ is less dense than brine. If CO₂ is injected only into the lower part of an aquifer, then, after injection ends, the CO₂ will continue to migrate, driven only by buoyancy. As CO₂ rises into the upper part of the aquifer, it will leave behind a residual phase trapped by capillary forces. The permanency of residual phase trapping is the main motivation for this approach, but an additional benefit is that vertical movement toward the top seal is also retarded. By choosing the volume injected, one can, in principle, prevent the CO₂ from reaching the top of the aquifer. The distance that the CO₂ rises depends on the uniformity of the displacement front and on the saturation of CO₂ behind the front. In this paper, we discuss factors that control the former feature. We will report on the latter in future publications.

Coarse-grid simulations suggest that CO₂ will rise in a compact plume having a smooth outline. As the grid is refined, the shape of plume becomes more uneven. Can this loss of uniformity be attributed to the intrinsically unstable character of buoyancy-driven immiscible flow? In analogy with immiscible displacements that exhibit viscous instability, we might anticipate the emergence of fingers as the CO₂ rises. Such fingers conceivably could reach the top seal of the aquifer quickly, even when the volume of stored CO₂ is insufficient to allow a uniform displacement to reach the top. This could lead to an accumulation of potentially mobile CO₂, the very situation the inject-low-and-let-rise strategy seeks to avoid. Thus, it is important to assess the extent to which gravity fingers develop under typical storage conditions for a range of target formations.

Some aspects of this problem are familiar from the long experience of gas and CO₂ injection into oil reservoirs (Stalkup 1983). In gas-injection processes, the competition between viscous forces and buoyancy leads to gravity override. The larger mobility of the gas phase also leads to viscous fingering. We will see that some factors that govern gas-injection displacements also influence the situation of interest here--that is, when injection has ended and the only driving force is buoyancy. On one hand, this is not surprising. On the other, it should not be taken for granted because there has been relatively little examination of the buoyancy-dominated dynamics. The key question is whether the absence of competing forces allows the intrinsic instability of a buoyant displacement to dominate the shape of the plume.

The idealized initial condition for our simulations is an approximation of the situation commonly observed at the end of the injection period in simulations of the inject-low-and-let-rise strategy. The simplification allows us to attribute differences in behaviors unequivocally to differences in petrophysical properties and to the physics of buoyant flow. The understanding thus obtained will provide insight into the post-injection behavior when the injection period is simulated more realistically.

© 2008. Society of Petroleum Engineers

View full textPDF ( 7,450 KB )

History

• Original manuscript received: 17 February 2006
• Meeting paper published: 22 April 2006
• Revised manuscript received: 11 September 2007
• Manuscript approved: 2 April 2008
• Version of record: 15 December 2008