MRST: Public data sets
Good data sets are essential to enable tests of new computational methods in a realistic and relevant setting. Such data sets arehard to come by, and when making MRST we have made an e ort to provide simple access to a number of public data sets that can be freely downloaded

Downloading the data sets

To simplify dataset management, MRST offers a graphical user interface, mrstDatasetGUI() that lists all public data sets known MRST, gives a short description of each, and can download and unpack most of the data sets to the correct sub-folder of the standard path. A few data sets require you to register your email address or fill in a license form, and in this case we provide a link to the correct webpage. TheGUI will also list all files that are part of the data set and list the examples in various modules of MRST that utilize each particular data set.

Short description of each data set

SPE 1:
This is the first SPE benchmark. It is a live oil/dry gas black-oil model with nearly immobile water. There is a single producer controlled by bottom-hole pressure, and a single injector injecting gas into the reservoir initially filled with undersaturated oil. Setup: three-phase, black-oil with gas in oil phase, Cartesian grid model with 300 cells Model contains: grid, rock and fluid.
SPE 3:
This is a modified wet-gas black-oil version of the third SPE benchmark which originally described a compositional model for a condensate reservoir. Setup: three-phase, black-oil with oil component in gas phase, Cartesian grid model with 324 cells. Model contains: grid, rock and fluid.
SPE 9:
This is the ninth SPE benchmark, which consists of dipping reservoir produced by a single water injector and 25 producers. Water breakthrough and schedule dynamics result in a large amount of control switching between oil-rate and bottom-hole pressure. Setup: three-phase, black-oil with gas in oil phase, dipping Cartesian grid model with 9000 cells Model contains: grid, rock and fluid.
SPE 10:
The tenth SPE benchmark was initially designed as a difficult upscaling test. The benchmark contains two models. MRST offers access to both through a dedicated module spe10. Model 2 contains a highly heterogeneous set of petrophysical parameters that have been used extensively, both in their original form or in the form of various subsamples, to create standard benchmark tests in a large number of publications. Setup (Model 2): two-phase fluid, Cartesian grid model with 1122000 cells Model contains: grid, rock and fluid.
BedModels 1:
Benchmark models for local upscaling created by Lars Vingli Odsaeter in his master thesis "Numerical Aspects of Flow Based Local Upscaling" at the Norwegian University of Science and Technology (2013). Setup: Corner point models. Models contain: six different grid models, some with rock parameters.
BedModel 2:
The model represents a 30^3 cm model of a sedimentary bed that contains six different rock types. The model is a good example of a corner-point grid having a large number of inactive cells and cells with degenerate geometry. Setup: corner-point model with 91831 cells. Model contains: grid and rock properties.
The "Egg Model" is a synthetic reservoir model consisting of an ensemble of 101 relatively small three-dimensional realizations of a channelized reservoir produced under water flooding conditions with eight water injectors and four producers. This data set represents a "standard version" of the Egg Model which is meant to serve as a standard test case.
CO2 Storage Atlas:
The CO2 Atlas is a set of map files produced by the Norwegian Petroleum Directorate to aid the evaluation of possible CO2 storage sites on the Norwegian Continental Shelf. The dataset includes thickness and top surface maps of a wide variety of aquifers and rock formations. The data sets are used extensively in MRST-co2lab, see e.g., here.
Simulation model of the Norne field in the Norwegian Sea operated by Statoil. First made available to researchers via NTNU. Later, published as an open data set by the Open Porous Media (OPM) initiative. Run scripts makeNorneSubsetAvailable() and makeNorneGRDECL() to set up the Eclipse input files. Variants of this data set has been used with MRST as a primary example of a realistic reservoir geometry and petrophysical properties from the very start of the software. Setup: three-phase, black-oil, corner-point model with 44420 cells. Model contains: Grid, rock and fluid.
The SAIGUP (Sensitivity Analysis of the Impact of Geological Uncertainties on Production) project generated a large number of realistic geological models representing shallow marine environments. Unfortunately, this large ensemble of models is not generally available to the public. However, the developers of MRST have been allowed to publish one specific realization that has faults, inactive cells, disconnected components, but no pinchouts. Setup: corner-point model with 78720 cells. Model contains: Grid and rock.
The Johansen formation was for a period considered as a candidate site for large-scale CO2 storage offshore the south-west coast of Norway. The data set contains four different models: one model of the whole field and three sector models with different vertical resolution. The data sets are published courtesy of the MatMoRA project funded by the Climit program at the Research Council of Norway. Setup: corner-point model. Model contains: Grid and rock.
Case B4:
The dataset contains four realizations of two intersecting faults modelled with a stair-step and a pillar grid with 36x48 or 72x96 cells in the lateral resolution. Model contains: Grid.

Published December 20, 2016