Recent developments in geosciences and engineering software, and 3-Dvisualization in particular, ease construction and utilization of detailedgeologic descriptions in reservoir simulation. This ease enhances theinteraction among team members such as geophysicists, geologists, and reservoirengineers. The ability to model accurately subsurface reservoir detail andthose features that are likely to control or influence the displacementprocesses associated with hydrocarbon production, is essential if we are toimprove our ability to predict field performance through reservoirsimulation.
This paper outlines a methodology for integrating a variety of data types atdifferent scales into a comprehensive description of the subsurface that can berapidly updated as new data become available. The methodology makes use ofrecent developments in artificial neural network techniques coupled with highresolution 3-D seismic to extrapolate reservoir properties away from theborehole.
The NE Rabbit Hills Field is used to demonstrate the benefits and theinteractive nature of this integrated study. Results of performing historymatching using the neural network-derived rock properties are alsopresented.
The petroleum industry has recognized the need for a cross-disciplinary teamapproach to reservoir characterization and simulation studies. A reservoirsimulation study combines data from many different sources and processes thedata through complex nonlinear systems of equations to generate the reservoirproduction forecast required for economic analysis. These data types typicallyinclude geologic, seismic, petro physical, well log, and production data. Asignificant problem with these simulations is that data generated fromdifferent sources are measured at different scales and on different rockvolumes.
The recent intense focus on reservoir characterization as a production tool hasoccurred largely because of the enhanced computer capability now available tointegrate and visualize large volumes of data in a three-dimensional format. Along with the advances in integration and visualization capabilities, a newmodeling method has been introduced that is capable of making predictions ofphysical parameters in the absence of well defined analytic relationships. Thistechnique, artificial neural networks (ANN), is becoming increasingly popularin the areas of pattern recognition and nonlinear problem solving.
Petroleum reservoirs are, in large part, characterized by the porosity andpermeability of the hydrocarbon-bearing formation. Accurate estimates of theseparameters are important for predicting reserves and developing oil extractiontechniques. At present, to obtain porosity or permeability distributions, values are interpolated between existing wells using conventional griddingtechniques. This technique can account for large-scale variations in reservoirparameters but is unable to provide the level of resolution required to produceaccurate production estimates or to determine injection locations if thereservoir is located in a zone of complex geology. A neural network approachcan provide a method for extrapolating well log data guided by another type ofinformation, such as seismic data. Neural networks are thus being used toaddress the challenge posed by the considerable difference in physical scale ofthe different data sets that are to be integrated, for example, outcrop andcore data with surface and crosswell seismic data.