Geologic drivers affecting buoyant plume migration patterns in small-scale heterogeneous media : characterizing capillary channels of sequestered CO₂

CO₂ sequestration aims for the most efficient utilization of reservoir pore volume and for maximizing security of storage. For typical field conditions and injection rates, buoyancy and capillary forces grow dominant over viscous forces within hundreds of meters of the injection wells as the pressure gradient from injection becomes less influential on flow processes. Flow regimes ranging from compact flow to capillary channel flow or secondary accumulation beneath a seal are possible through time as the CO₂ plume travels through the storage reservoir. Here we model the range of possible migration behavior in the capillary channel regime in small-scale domains whose heterogeneity has been resolved at depositional (sub-millimeter) scale. Two types of model domains have been studied in this work: domains with depositional fabric from real, naturally-occurring geologic samples and geostatistically generated synthetic model fabrics. The real domains come from quasi-2D physical geologic samples (peel # 1: ~1 m × 0.5 m sample and peel # 2: ~0.4 m × 0.6 m sample) that are vertically oriented relief peels of fluvial sediment extracted from the Brazos River, Texas. Peel # 1 is oriented perpendicular to dominant depositional flow while peel # 2 is a flow-parallel specimen. The various depositional fabrics represent definite correlation lengths of threshold pressures in the horizontal and vertical directions which can be extracted. High-resolution (~2 million element model) laser scanning of the samples provided detailed topography which is the result of nearly linear corresponding changes in measured grain size (normal distribution) and sorting. We model the basic physics of buoyant migration in heterogeneous domain using commercial software which applies the principle of invasion percolation (IP). The criterion for governing drainage at the pore scale is that the capillary pressure of the fluid needs to be greater than or equal to the threshold pressure of the pore throat it is trying to enter for the interface to advance into the pore. Here we employ the extension of this concept to flows at larger scales, which replaces the pore throat with a volume of rock with a characteristic value of capillary entry pressure. The fluid capillary pressure is proportional to the height of continuous column of the buoyant phase. The effects of (i) threshold pressure range, i.e. difference between the maximum and minimum threshold pressures in the domain; and (ii) the density difference between CO₂ and connate water on capillary channels of CO₂ were studied on the various sedimentologic fabrics. As the rock and fluid properties varied for different model domains, ₂ migration patterns varied between predominantly fingering and predominantly back-filling structures. Sufficiently heterogeneous media (threshold pressures varying by a factor of 10 or more) and media with depositional fabrics having high ratios of horizontal and vertical correlation lengths of capillary entry pressures in the domain yield back-filling pattern, resulting in a significantly large storage capacity. Invasion percolation simulation models give qualitatively similar CO₂ migration patterns compared to full-physics simulators in small-scale but high resolution domains which are sufficiently heterogeneous. On the other hand, we find the invasion percolation simulations predicting disperse capillary fingering pattern in relatively homogeneous media (threshold pressures varying by less than a factor of 10) while the full-physics simulations reveal a very compact CO₂ front in the same media. This stark difference needs to be investigated to understand the governing flow physics in these domains. Fingering flow pattern in the capillary channel regime would clearly result in the estimated storage capacity being much less than the nominal value (the pore volume of the rock) as the rock-fluid contact is minimal. The importance of this work lies in the verification that a relatively simple model (invasion percolation), which runs in a very small fraction of the time required by full-physics simulators, can be used to study buoyant migration in rocks at the micro-scale. Understanding migration behavior at the small-scale can help us approach the problem of upscaling better and hence define the complex plume dynamics at the reservoir scale more realistically. Knowledge of the correlation structure of the sedimentologic fabric (ratio of correlation lengths of threshold pressures in horizontal and vertical directions) and the threshold pressure distribution (permeability distribution) for any given reservoir rock could help evaluate amount of CO₂ that can be stored per unit volume of rock (storage potential) for a reservoir in the migration phase of sequestration. The possibility of predictive ability for expected capillary channel flow patterns kindles the prospect of enabling an engineered storage strategy that drives the behavior toward the desired flow patterns in the subsurface.