Large volumes of gas can be produced at high rates with conventional horizontal- or vertical-well configurations for long periods of time from some methane-hydrate accumulations by means of depressurization-induced dissociation. However, most assessments of hydrate production use simplified or reduced-scale 3D or 2D production simulations. This study used a message-passing-interface parallel code to make the first field-scale assessment of a large deep-ocean hydrate reservoir. Systems of up to 2.5 million gridblocks, running on thousands of supercomputing nodes, were required to simulate such a large system at the highest level of detail. The simulations begin to reveal the challenges inherent in producing from deep relatively cold systems with extensive water-bearing channels and connectivity to large aquifers. The main difficulties are water production and achieving depressurization.
Introduction
Gas hydrates are solid crystalline compounds in which gas molecules occupy the lattices of ice-like crystal structures called hosts. Gas hydrates occur in the permafrost and in deep-ocean sediments, where the necessary conditions of low temperature and high pressure exist for their formation and stability. Most naturally occurring gas hydrates contain methane in abundance. Not all hydrates are desirable targets for production. Of the three possible methods of hydrate dissociation for gas production—depressurization, thermal stimulation, and use of inhibitors—depressurization appears to be the most efficient. Recent studies have indicated that, under certain conditions, gas can be produced from natural hydrate deposits at high rates over long periods of time by use of conventional technology. Earlier work focused on production from vertical wells, but more-recent studies show that horizontal wells are more productive and are easier to manage, if the technology is available.
Study
The objective of this study was to use real geophysical data at the field scale to simulate a realistic 3D gas-hydrate reservoir. Previous studies focused on simple 2D modeling, 2D modeling with artificial heterogeneity, or extremely simplified (coarse-scale) 3D modeling. Through collaboration, Statoil provided real data on the geometry and geology of a known oceanic-hydrate system. The deposit is a layered system, approximately 7×5.5 km and 350 m thick. The hydrate is arranged in a series of high-permeability channels bounded by slightly-lower-permeability levees. The system is probably impermeable at the top and bottom boundaries, but may connect to an aquifer along the x–z face.
Constructing 3D volume meshes for systems of this scale is well beyond the means of standard routines. To build a horizontal-well system, custom tools for manipulating the raw list of element centroids and creating new mesh-element configurations (i.e., cylinders) were combined with a meshing toolkit to create a new meshing tool that is capable of generating a full 3D-Voronoi mesh for any valid configuration of cell centers. The system keeps track of all relevant cell and interface properties and makes them available for manipulation. This feature enabled creating highly flexible and dynamically refined meshes. The domain was trimmed such that regions of the mesh that do not represent permeable reservoir rocks were removed to promote computational efficiency. All subsequent meshing begins with this reduced mesh, which contains 1,663,900 elements.
For the horizontal well, the geophysical data were analyzed to find the longest continuous horizontal layers of hydrate that are more than 60 m thick. A region of roughly 65-m thickness was found near the core of the system, accommodating a horizontal well 1900 m in length, with no intersection with nonhydrate-bearing channels or levees. A cylindrical mesh was constructed around the well to a radius of 250 m, and this radial mesh was inserted into the rectilinear mesh derived from the geophysical data (Fig. 1). Element properties (i.e., hydrate saturation) were interpolated from the rectilinear data onto the new mixed mesh. The resulting mesh contained 2,264,000 elements, requiring the simultaneous solution of more than 9,000,000 equations.
The new meshing tool also provides geometric information required for ongoing research into the visualization of Voronoi meshes and the data sets generated by simulations of such unstructured systems. This visualization involves tessellation of the unstructured finite-volume meshes, creation of visualization meshes of 10 million to 100 million volume elements, and then rendering these tessellated volumes into 3D images by use of a visualization cluster. This visualization problem has been difficult for reservoir simulators (because most visualization tools have strict data-set-size limits or are limited to meshes of certain geometries), and these tools are helping to create new possibilities for visualizing large-scale problems.
First, the system was brought to hydrostatic equilibrium at the known depth and then to thermal equilibrium by use of the known geothermal gradient. Because of the size and extent of the system, this process consumed significant computing time. Hydrate was added to the system according to the distribution derived from the geophysical data. The system then was allowed to reach full thermal, hydrological, and chemical equilibrium at stated conditions before any simulations were performed.
Results
Preliminary simulation work by Statoil, using commercial software with gas-hydrate add-ons, suggested that the system could be productive if vertical wells were drilled into the lower part of the formation near areas of high hydrate saturation. However, these simulations used a 2D slice of the full system, and the initial configuration of the reservoir assumed that significant free gas existed. For Class-2 and Class-3 hydrate deposits, it has been demonstrated that properly placed horizontal wells are a more-effective production strategy. Therefore, production from the horizontal well was simulated, as shown in Fig. 1. Because of the nature of the Voronoi grid (i.e., on average, more connections per element than a rectilinear grid), the computational requirements for this simulation were approximately an order of magnitude greater than anything previously attempted with this parallel code. Initial equilibration was performed with a 220-processor in-house cluster, but production was scaled up to the National Energy Research Scientific Computing Center Hopper supercomputer for multi-day runs using 960 and 1,920 processors. At the time this paper was written, the authors had been able to simulate 135 days of production time, with simulations ongoing.
Evolution of the rate of hydrate dissociation/methane release and the rate of methane production at the well is shown in Fig. 2. Release exceeds production from the beginning, which is necessary to generate free gas in the reservoir for production later. Water production is large but steady; however, after less than 5 months of production, both release and production ceased to increase with time, while water production remained approximately constant. This trend suggests that effective depressurization is hindered by the inflow of large quantities of water from the surrounding formation. Because economically viable production rates for offshore wells must be in the millions of cubic feet per day, this system is evolving toward unsustainability.
Cumulative quantities of methane, released and produced, are shown in Fig. 3. More significant, however, is the water/gas ratio (WGR), shown in green. While the WGR drops rapidly as production proceeds, the curve is heading toward asymptotic behavior at values of 700 to 800 kg/m3, which suggests that methane transport occurs entirely in the aqueous phase, and even then, the quantity of water produced greatly exceeds the quantity of water needed to dissolve and transport the methane observed at the well. Therefore, it is likely that the hydrate zone around the well has been perforated by dissociation, connecting the well to the massive reservoir of mobile water surrounding the hydrate layers and hindering depressurization after a few months of production.
Visualization Challenges
For simpler simulations, a key part of the analysis is to make detailed 2D illustrations of the state of the simulated systems at various times, ideally creating a movie or time sequence of property evolution. This visualization puts the derived properties of the simulation (e.g., dissociation rate and production rate) into context, and highlights the physical processes that lead to success or failure of a particular production strategy. However, for large 3D-volume meshes, simple scripts or off-the-shelf visualization tools rapidly become unworkable. The codes used in this study use finite-volume meshes that are, in essence, nonpositional. The x/y/z coordinates of gridblocks are retained for post-simulation analysis, but the mesh used by the code sees each gridblock only as a volume connected to other gridblocks by connections of known area and length. The shape of the gridblock (i.e., the corners and edges used in finite-element meshes) is not part of the mesh itself. For a simple rectangular mesh, it is simple to convert a series of element-centroid locations into an equivalent finite-element-style mesh for visualization with several commercial or freeware visualization programs. For more-complicated meshes in two dimensions, a representative visualization mesh can be created through triangularization, and the properties of the elements can be interpolated onto this mesh with minimal loss of information.
Such simplifications are not possible for this field-scale Voronoi mesh. The difference in scale between the largest gridblocks (50-m length) and the smallest (0.2 m) prevents interpolation onto a uniform mesh without either a severe loss of information (i.e., reducing all near-well gridblocks to a set of 5000-m3 averaged cells) or the creation of a mesh too large to render (approximately 1016 cells). The alternative is to create the visualization mesh through tetrahedralization, analogous to the 2D triangularization used for simpler problems. The element properties then can be interpolated onto this tetrahedral mesh without loss of information.
Conclusions
While the simulations used the most extensive, realistic, and data-based model available, the insights found here strongly relate to earlier studies on the productivity and nonproductivity of various configurations of hydrates. Preliminary results for a horizontal-well configuration showed that large quantities of water are able to reach the well early in the depressurization process, reducing depressurization effectiveness. Whether this is caused by rapid dissociation and breakthough, or other effects (e.g., greatly reduced but still finite effective permeability of hydrate-bearing media), remains to be proved (ongoing simulations and advanced visualization will produce detailed 3D images of system evolution to evaluate this hypothesis). As a result, it is becoming clear that systems with large quantities of water in communication with the hydrate reservoir, even if not connected to aquifers beyond reservoir boundaries, are particularly challenging production targets. Permeability of the bounding media, distance from the hydrate-stability boundary, and configuration of the hydrate into large contiguous masses must be considered in reservoir evaluation.
This article, written by Senior Technology Editor Dennis Denney, contains highlights of paper IPTC 17026, “Massively Parallel Simulation of Production From Oceanic-Gas-Hydrate Deposits,” by Matthew T. Reagan, SPE, George J. Moridis, SPE, Craig M. Freeman, SPE, Katie L. Boyle, and Noel D. Keen, Lawrence Berkeley National Laboratory, prepared for the 2013 International Petroleum Technology Conference, Beijing, 26–28 March. The paper has not been peer reviewed. Copyright 2013 International Petroleum Technology Conference. Reproduced by permission.