Layer | Fill | Outline |
---|
Map layers
Theme | Visible | Selectable | Appearance | Zoom Range (now: 0) |
---|
Fill | Stroke |
---|---|
Collaborating Authors
Results
Influence of Micro-Computed Tomography Image Resolution on Petrophysical Properties
Alyafei, Nayef (Imperial College London) | Gharbi, Oussama (Imperial College London) | Qaseminejad Raeini, Ali (Imperial College London) | Yang, Jianhui (Imperial College London) | Iglauer, Stefan (Curtin University) | Blunt, Martin J. (Imperial College)
Abstract Micro-CT scanning is a non-destructive technique that can provide three-dimensional images of rock pore space at a resolution of a few microns. However, these grey scale images cannot be directly input into simulators to predict flow properties; they require image processing to segment the solid and void space in the rock. Dynamic and static single phase properties can then be computed using the images directly or on extracted equivalent network models. In this paper, we study the effect of imaging resolution (five different voxel sizes ranging from 6โ20 ฮผm) of Clashach and Doddington sandstone on predicted single phase properties (porosity and absolute permeability) and network properties. Experimental data is used to validate the predictions. The results suggest that the computed porosity was largely independent of resolution and in good agreement with the measured value, while image resolutions of a few microns are sufficient to determine the permeability of a high-permeability rock such as Doddington but may not be sufficient for lower permeability samples. The topologically representative networks are sensitive to resolution, adding additional smaller pores and throats as the resolution is increased. This latter reason was confirmed by a network extraction analysis that indicated the average throat radius was 6 ยตm, similar to the highest resolution used and insufficient to image all important features of the pore space properly.
- North America > United States (0.68)
- Asia > Middle East (0.47)
- Research Report > New Finding (0.48)
- Research Report > Experimental Study (0.34)
- Geology > Geological Subdiscipline > Geomechanics (0.46)
- Geology > Rock Type > Sedimentary Rock (0.39)
Summary We use pore-scale network modeling to simulate imbibition in fractures and the matrix/fracture interaction accounting for viscous forces and flow in wetting layers. We represent the fracture as a 2D lattice of conceptual pores and throats connected to a 3D network that models the rock matrix. We find that the matrix adjacent to the fracture plays an important role in controlling matrix/ fracture transfer. Introduction Field-scale numerical modeling of naturally fractured reservoirs commonly is conducted through the use of dual-porosity or dual-permeability simulators. These simulators discretize the reservoir into grid blocks, which can contain a large number of individual or connected sets of fractures. Both formulations of the flow equations assume that the fractures form one continuum and the matrix forms another. Both also require that all the matrix and fracture properties within a grid block be described by single values of the properties for the entire block, no matter how complicated the matrix or fracture systems are within the block. Physically, the flow of fluids within a fracture and the transfer between the fracture and the matrix are controlled by the structure of the porous medium, the flow rate, and capillary forces. In a water-wet medium, capillary pressure will drive water from a fracture to the matrix, while viscous forces favor flow through the high-permeability fracture. The ratio of viscous to capillary forces, which controls the matrix/fracture transfer, is called the capillary number. One common definition for the capillary number isEquation 1 where q=the volumetric flow rate per unit area, ยต=the fluid viscosity and ?=the interfacial tension between the fluids in the pore space. Typical reservoir values for NCa are 10 and lower. Flow rates in fractures can be much higher than in granular media. Fracture capillary numbers may be as high as 10 away from wells, with values approximately 10 near wells. It is known from experiments in normal granular media that viscous effects are important for NCa greater than approximately 10. The effect of flow rate in fractured media, one of the main topics of this paper, has been less well-studied, but may be significant, because high values of NCa are achieved readily. Typically, matrix/fracture transfer is incorporated into simulators using shape-based functions whereby the saturation, flow rate, and history dependence of the transfer are all provided by effective relative permeabilities. Modified transfer functions by de Swaan, Kazemi et al., and Gupta and Civan take advantage of the exponential-like behavior of immersion experiments but have had difficulty modeling rate-dependent experiments. Beckner et al., Dutra and Aziz, and Chen et al. have postulated transfer functions that are better able to capture these rate effects. The appropriate form for fracture relative permeabilities, particularly at high flow rates, is also poorly understood. Traditionally, linear-relative permeabilities have been assumed with endpoint values at 0 and 1. This was based on experimental work by Romm for flow between glass plates. Romm's work, however, was designed to obtain a large degree of phase segregation. He lined the glass plates with waxed paper and added waxed paper strips running through the plates as spacers. Merrill and Pieters and Graves attempted to duplicate Romm's experiments without the use of the waxed paper strips. Merrill found that the wetting-phase saturations were scattered around a value of 0.72 for nearly the entire range of flow rates and fractional flows. Deviations from this behavior only occurred when both the total flow rate and the fractional flow were high. This behavior was seen also when Merrill performed experiments on Berea sandstone blocks which had been sealed with epoxy. The only difference was that the wetting-phase saturation value where multiphase flow occurred was approximately 0.62. The Pieters and Graves experiments also showed deviations from straight-line relative permeabilities. The experiments above were all performed on artificial fractures with a constant aperture. However, many authors have measured variable aperture distributions in more realistic fractures. New theoretical evaluations of fracture relative permeability have been proposed based on these findings. Pyrak-Nolte et al. numerically generated a spatially correlated, log-normal aperture distribution. Drainage relative-permeability curves were generated for different stress levels using a percolation process. This model found that the relative permeabilities were non linear functions of saturation and that the relative-permeability curve crossover point was essentially invariant of stress. Pyrak-Nolte et al. modified the displacement model to be one involving invasion percolation with trapping, which was then used for imbibition. They found a strong shift in the crossover point towards lower relative-permeability values and lower wetting-phase saturation. Pruess and Tsang used a general-purpose simulator to solve for the effective phase permeabilities on a 20ร20 grid. They used a log-normal aperture distribution. It was assumed that gridblock occupancy was dependent only on the capillary pressure of the block (accessibility was ignored). This model is then, essentially, a percolation model. For short-range correlations, very little multi phase flow occurred. This result concurs with that of Wilkinson and Willemsen, who have shown that for random percolation models, it is not topologically possible to have continuous pathways of two phases across a 2D medium. Nonwetting-phase relative permeability fell as wetting-phase saturation, Sw, increased (wetting-phase relative permeability was 0). Once the saturation was above a percolation threshold, the nonwetting-phase relative permeability was 0 and wetting-phase relative permeability rose. For correlations that were much stronger in the flow direction, Pruess and Tsang found that wetting-phase relative permeability curves could be approximated as a power-law function of saturation (Corey-like). Nonwetting-phase relative permeabilities started on a similar Corey-like path, but fell to 0 once a critical gridblock was filled. Rossen and Kumar used an effective medium approach on the Pruess and Tsang aperture distribution. They found relative permeabilities nearly identical to those found by Pruess and Tsang with less computational expense. Rossen and Kumar evaluated the effects of the variance of the fracture aperture distribution, gravity, and wetting-layer flow on relative permeability.
- Geology > Geological Subdiscipline > Geomechanics (0.49)
- Geology > Rock Type (0.34)
We use pore-scale network modeling to simulate imbibition in fractures and the matrix/fracture interaction. We represent the fracture as a two-dimensional lattice of conceptual pores and throats. We allow flow in connected wetting layers that occupy roughness and crevices in the pore space. We model piston-like advance with a capillary pressure that accounts for the curvature of the meniscus due to the fracture aperture, as well as the curvature of the wetting front in the fracture plane. We show that the model gives results that are insensitive to the resolution or pore spacing of the network. To account for viscous forces, the wetting phase pressure is computed assuming a fixed conductance in wetting layers. This pressure, in combination with the local capillary pressure, is used to determine the displacement sequence. A matrix is incorporated by surrounding the two-dimensional fracture plane by a three-dimensional network of pores and throats. We model multiphase flow in a real fracture using an aperture distribution obtained from CT scanning. The simulated saturation distributions agree with those measured using in situ imaging. We also study the matrix/fracture transfer in a large three-dimensional network.
- Europe (0.68)
- North America > United States > Texas > Harris County > Houston (0.28)
Abstract This paper presents the extension of the streamline approach to full-field, three-dimensional (3D) compositional simulation. The streamline technique decomposes a heterogeneous 3D domain into a number of one-dimensional (1D) streamlines along which all fluid flow calculations are done. Streamlines represent a natural, dynamically changing grid for modeling fluid flow. We use a 1D compositional finite-difference simulator to move components numerically along streamlines, and then map the 1D solutions back onto an underlying Cartesian grid to obtain a full 3D compositional solution at a new time level. Because of the natural decomposition of the 3D domain into a number of 1D problems, the streamline approach offers substantial computational efficiency and minimizes numerical diffusion compared to traditional finite-difference methods. We compare our three and four component solutions with solutions from two finite difference codes, UTCOMP and Eclipse 300 (E300). These examples show that our streamline solutions are in agreement with the finite-difference solutions, are able to minimize the impact of numerical diffusion, are faster by orders of magnitude. Numerical diffusion in finite-difference formulations can interact with reservoir heterogeneity to substantially mitigate mobility differences and lead to optimistic recovery predictions. We demonstrate the efficiency and usefulness of the streamline-based simulator on a 518,400 gridblock, 3D, heterogeneous, 36-well problem for a condensing-vaporizing gas drive with four components. We can simulate this problem on an average-size workstation in three CPU days. It takes approximately the same amount of time to simulate the upscaled 28,800 gridblock version of the problem using finite-differences. We conclude with a qualitative discussion explaining the near-linear scaling of the streamline approach with the number of gridblocks and the cubic and higher scaling exhibited by one of the finite-difference codes. Introduction The use of streamlines and streamtubes to model convective displacements in heterogeneous media has been presented repeatedly since the early work by Muskat, Fay and Prats, and Higgins and Leighton. Important subsequent contributions are due to Parsons, Martin and Wegner, Bommer and Schechter, Lake et al., Mathews et al., Emanuel et al., Renard, and Hewett and Behrens. Recently, streamline methods have received renewed attention by several groups as a viable alternative to traditional finite-difference (FD) methods for large, heterogeneous, multiwell, multiphase simulations, which are particularly difficult for FD simulators to model adequately. Large speed-up factors compared to traditional FD solutions, minimization of numerical diffusion and grid orientation effects, and the inherent simplicity of the approach offer unique opportunities for integration with modern reservoir characterization methods. Examples include ranking of equiprobable earth models, estimation of the uncertainty in production forecasts due to the uncertainty in the geological description, rapid assessment of production strategies such as infill drilling patterns and miscible gas injection. In addition, streamlines may offer an attractive alternative to well-known problems with upscaling of absolute and pseudorelative permeabilities by allowing larger geological models and requiring upscaling across a smaller range of scales. Our streamline approach for reservoir simulation hinges on two important extensions to past streamline/streamtube methods:the use of true 3D streamlines and and numerical solutions of the transport equations along periodically changing streamlines. With these extensions we have been able to simulate realistic fluid flow in detailed, heterogeneous, 3D reservoir models much more efficiently than FD methods. We emphasize that reservoir simulation using streamlines is not a minor modification of current FD approaches, but instead represents a significant shift in methodology. P. 471^
- Reservoir Description and Dynamics > Reservoir Simulation > Scaling methods (1.00)
- Reservoir Description and Dynamics > Reservoir Fluid Dynamics > Flow in porous media (1.00)
- Reservoir Description and Dynamics > Reservoir Characterization (1.00)
- Reservoir Description and Dynamics > Improved and Enhanced Recovery > Gas-injection methods (1.00)