Computational tools for reservoir modelling play a critical role in the development of strategies for optimal recovery of hydrocarbon resources by providing a means of forecasting recovery given a set of data, assumptions, and operating constraints. Reservoir modelling is crucial to the much more general goal of quantifying what is known about a particular reservoir as well as what is uncertain. Traditional reservoir simulation is computationally demanding and a single simulation on a full reservoir model may require from a few tens of minutes to hours or even days. In many geomodelling workflows, it is impractical or impossible to wait for full-featured flow simulations.
Flow diagnostics are based on controlled numerical experiments that yield quantitative information about the flow behaviour of a reservoir model in settings simpler than would be encountered in the actual field. Using flow diagnostic tools, information about volumetric connections and timelines and heterogeneity measures for pressure and displacement fronts can be obtained within seconds, even for models with multimillion cells. These tools are therefore ideal for interactive visualization and inspection of large reservoir models, and can be used to evaluate, rank and compare realizations or production strategies, or to control the quality of model upscaling. In particular, we have shown how flow diagnostics can be used as a computationally inexpensive substitute for full-featured simulations when optimizing rates, placement, and drilling sequence of wells.
The new diagnostic tools have been implemented as an open-source module in the Matlab Reservoir Simulation Toolbox (MRST), and can be freely downloaded and used under the GNU GPLv3 license, see http://www.sintef.no/Projectweb/MRST/Modules/diagnostics/
MRST has many hundred users worldwide.
The project proposes to develop and mature computational tools for flow diagnostics that are specifically designed to quantify uncertainty in reservoir characterization. Flow diagnostic tools are based on controlled numerical flow experiments that yield q uantitative information regarding the flow behavior of a model in settings much simpler than would be encountered in the actual field. While full-featured simulators are capable of making these predictions, they generally cannot do so in a computationally efficient manner unless an unacceptably large degree of upscaling is applied. Industrial applications require fast tools that can be applied directly to high resolution reservoir models.
The main R&D challenges are to develop appropriate numerical formu lations and implement a set of prototypes that can be used for comparing, ranking, and clustering high resolution reservoir models. The tools will also be applied to suggest appropriate model updates during data integration and optimization.
Flow diagnos tics will be tested for use in evaluating upscaling errors, assessing discretization errors, ranking earth models, and clustering reservoir models based on flow information. The primary application that will initially be targeted is waterflood optimizatio n for mature fields, for which simpler modeling tools are often favored over full-featured simulation to generate multiple history matched models for optimization. Flow diagnostics on fine-scale models have the potential for improved estimation of sweep e fficiency, which is the basis for many optimization strategies. They are also of interest for unconventional resources, such as tight gas/oil and shale gas, for which unstructured grids are already being used. In general, the tools have the potential for widespread use within earth sciences and will help with overall integration of reservoir modeling workflows.