Simulation stochastiques de faciès par la méthode des membership functions

Laurent Labat. ( 2004 )
Institut National Polytechnique de Lorraine

Abstract

An oil reservoir is a complex physical system whose challenge is to estimate as best as possible the heterogeneities from a set of heterogeneous available data. This estimation can be carried out using a deterministic approach (accounting for all available data integrated into a unique model) or geostatistical simulation methods : these methods do not aim at providing a single model of the under- ground, but rather to generate a large set of alternative, equiprobable models representing the heterogeneities expected into the reservoir. The main goal of this work is to propose a new method for generating al- ternative and equiprobable models, accounting for a large set of data. The first part is a review of commonly used geostatistical methods. In a second part, the theory of membership functions is introduced, and the integration of several data types (well logs, proportion map and curve, seismic data, anisotropy) into the DSI interpolator is detailed, and demonstrate the ability of the DSI Interpolator to integrate a large set of constraints. The next part discusses the specific problem of the interpolation of facies occurrence probabilities, and presents different ways to compute a proportion cube(computing an initial solution, using a multigrid algorithm or using the DSI algorithm). The last part presents three alternative methods to generate stochastic models from a proportion cube, and the results obtained are discussed.

Download / Links

BibTeX Reference

@PHDTHESIS{Labat2004a,
    author = { Labat, Laurent },
     title = { Simulation stochastiques de faciès par la méthode des membership functions },
      year = { 2004 },
    school = { Institut National Polytechnique de Lorraine },
  abstract = { An oil reservoir is a complex physical system whose challenge is to estimate
as best as possible the heterogeneities from a set of heterogeneous available data.
This estimation can be carried out using a deterministic approach (accounting
for all available data integrated into a unique model) or geostatistical simulation
methods : these methods do not aim at providing a single model of the under-
ground, but rather to generate a large set of alternative, equiprobable models
representing the heterogeneities expected into the reservoir.
The main goal of this work is to propose a new method for generating al-
ternative and equiprobable models, accounting for a large set of data. The first
part is a review of commonly used geostatistical methods. In a second part, the
theory of membership functions is introduced, and the integration of several data
types (well logs, proportion map and curve, seismic data, anisotropy) into the DSI
interpolator is detailed, and demonstrate the ability of the DSI Interpolator to
integrate a large set of constraints. The next part discusses the specific problem
of the interpolation of facies occurrence probabilities, and presents different ways
to compute a proportion cube(computing an initial solution, using a multigrid
algorithm or using the DSI algorithm). The last part presents three alternative
methods to generate stochastic models from a proportion cube, and the results
obtained are discussed. }
}