Data processing: measuring – calibrating – or testing – Measurement system in a specific environment – Earth science
Reexamination Certificate
2001-03-02
2004-11-02
Barlow, John (Department: 2863)
Data processing: measuring, calibrating, or testing
Measurement system in a specific environment
Earth science
C703S002000
Reexamination Certificate
active
06813565
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a method of gradual deformation of representations or realizations, generated by sequential simulation, of a model of a heterogeneous medium which is not limited to a Gaussian stochastic model, based on a gradual deformation algorithm of Gaussian stochastic models.
2. Description of the Prior Art
In French patent application 98/09,018 a method is described which gradually deforms a stochastic (Gaussian type or similar) model of a heterogeneous medium such as an underground zone, constrained by a set of parameters relative to the structure of the medium. This method comprises drawing a number p (p=2. for example) of realizations (or representations) independent of the model or of at least part of the selected model of the medium from all the possible realizations and one or more iterative stages of gradual deformation of the model by performing one or more successive linear combinations of p independent initial realizations and then composite realizations are successively obtained possibly with new draws, etc., the coefficients of this combination being such that the sum of their squares is 1.
Gaussian or similar models are well-suited for modelling continuous quantity fields but are therefore ill-suited for modelling zones crossed by fracture networks or channel systems for example.
The most commonly used geostatistical simulation algorithms are those referred to as sequential simulation algorithms. Although sequential simulation algorithms are particularly well-suited for simulation of Gaussian models, these algorithms are not in principle limited to this type of model.
A geostatistical representation of an underground zone is formed for example by subdividing thereof by a network with N meshes and by determining a random vector with N dimensions Z=(Z
1
, Z
2
, . . . Z
N
) best corresponding to measurements or observations obtained on the zone. As shown for example by Johnson, M. E.; in “Multivariate Statistical Simulation”; Wiley & Sons, New York, 1987, this approach reduces the problem of the creation of an N-dimensional vector to a series of N one-dimensional problems. Such a random vector is neither necessarily multi-Gaussian nor stationary. Sequential simulation of Z first involves the definition of an order according to which the N elements (Z
1
, Z
2
, . . . Z
N
) of vector Z are generated one after the other. Apart from any particular case, it is assumed that the N elements of Z are generated in sequence from Z
1
to Z
N
. To determine a value of each element Z
1
, (i=1, . . . , N), the following operations have to be carried out:
a) building the distribution of Z
i
conditioned by (Z
1
, Z
2
. . . Z
i-1
) F
c
(Z
i
)=P(Z
i
≦/Z
1
, Z
2
, . . . Z
i-1
); and
b) determining a value of Zi from distribution F
c
(Z
i
).
In geostatistical practice, sequential simulation is frequently used to generate multi-Gaussian vectors and non-Gaussian indicator vectors. The main function of sequential simulation is to determine conditional distributions F
c
(Z
i
) (i=1, . . . , N). Algorithms and softwares for estimating these distributions are for example described in:
Deutsch, C. V. et al. “GSLIB (Geostatistical Software Library) and Useres Guide”; Oxford University Press, New York, Oxford 1992.
Conceming determining the values from distribution F
c
(Z
i
), there also is a wide set of known algorithms.
The inverse distribution method is considered by means of which a realization of Z
i
;
i
=F
c
−1
(u
l
) is obtained, where u
1
is taken from a uniform distribution between 0 and 1. A realization of vector Z therefore corresponds to a realization of vector U whose elements U
1
, U
2
, . . . , U
N
, are mutually independent and evenly distributed between 0 and 1.
It can be seen that a sequential simulation is an operation S which converts a uniform vector U=(U
1
, U
2
, . . . U
N
) to a structured vector Z=(Z
1
, Z
2
, . . . , Z
N
):
Z=S
(
U
) (1).
The problem of the constraint of a vector Z to various types of data can be solved by constraining conditional distributions F
c
(Z
i
) (i=1, . . . , N) and/or uniform vector U=(U
1
, U
2
, . . . , U
N
).
Recent work on the sequential algorithm was focused on improving the estimation of conditional distributions F
c
(z
i
) by geologic data and seismic data integration. An article by Zhu, H. et al: “Formatting and Integrating Soft Data: Stochastic imaging via the Markov-Bayes Algorithm” in Soares, A., Ed. GeostaUstics Troia 92, vol.l: KiuwerAcad. Pubi., Dordrecht, The Netherlands, pp.1-12, 1993 is an example.
However, this approach cannot be extended to integration of non-linear data such as pressures from well tests and production records, unless a severe linearization is imposed. Furthermore, since any combination of uniform vectors U does not give a uniform vector, the method for gradual deformation of a stochastic model developed in the aforementioned patent application cannot be directly applied within the scope of the sequential technique described above.
The method according to the invention thus allows making the two approaches compatible, that is to extend the formalism developed in the aforementioned patent application to gradual deformation of realizations, generated by sequential simulation, of a model which is not limited to a Gaussian stochastic model.
SUMMARY OF THE INVENTION
The method allows gradual deformation of a representation or realization, generated by sequential simulation, of a model which is not limited to a Gaussian stochastic model of a physical quantity z in a heterogeneous medium such as an underground zone, in order to constrain the model to a set of data collected in the medium by previous measurements and observations relative to the state or the structure thereof.
The method of the invention comprises applying an algorithm of gradual deformation of a stochastic model to a Gaussian vector (Y) having a number N of mutually independent variables that is connected to a uniform vector (U) with N mutually independent uniform variables by a Gaussian distribution function (G), so as to define a chain of realizations u(t) of vector (U), and using these realizations u(t) to generate realizations z(t) of this physical quantity that are adjusted in relation to the (non-linear) data.
According to a first embodiment, the chain of realizations u(t) of uniform vector (U) is defined from a linear combination of realizations of Gaussian vector (Y) whose combination coefficients are such that the sum of their squares is one.
According to another embodiment, gradual deformation of a number n of parts of the model representative of the heterogeneous model is performed while preserving the continuity between these n parts of the model by subdividing uniform vector (U) into a number n of mutually independent subvectors.
The method of the invention finds applications in modeling underground zones which generates representations showing how a certain physical quantity is distributed in an underground zone (permeability z for example) and which is compatible in the best manner with observed or measured data: geologic data, seismic records, measurements obtained in wells, notably measurements of the variation with time of the pressure and of the flow rate of fluids from a reservoir, etc.
REFERENCES:
patent: 5416750 (1995-05-01), Doyen et al.
patent: 5764515 (1998-06-01), Guerillot et al.
patent: 5838634 (1998-11-01), Jones et al.
patent: 6618695 (2003-09-01), Hu et al.
Hegstad et al., Uncertainty Assessment in History Matching and Forecasting, 1996, International Geostatistical Congress, Wollongong.*
Caers, Direct Sequential Indicator Simulation, Stanford University, Depart. of Petroleum Engineering, date is unknown.
Hu Lin-Ying
Noetinger Benoît
Antonelli Terry Stout & Kraus LLP
Barlow John
Institut Francais du Pe'trole
Le Toan M.
LandOfFree
Method for gradually deforming sequential simulations of a... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method for gradually deforming sequential simulations of a..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method for gradually deforming sequential simulations of a... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3301530