US 10,519,766 B2 · Reservoir modelling with multiple point statistics from a non-stationary training image

Abstract

A multiple point simulation technique for generating a model realization of a subterranean formation having different facies is described which uses a non-stationary training image which reflects facies spatial trends across the formation. The realization is formed by sequentially populating each cell; a facies pattern of neighboring cells is identified for each cell in the model grid, and then corresponding facies patterns identified in the training image. The probability of a target cell in the model grid having a given facies is calculated based on the proportion of occurrences of the corresponding facies pattern where the central cell has that facies. The contribution of each corresponding facies pattern occurrence in the training image to this proportion or probability is weighted according to the distance between its central cell and the training image cell corresponding in location to the target cell in the model grid.

Claims4

Dependent claims

A computer implemented method of generating a realistic model of a subterranean formation, the method comprising:

(a) obtaining a training image of a subterranean formation having a number of grid nodes corresponding to a number of simulation points u in the model;
(b) at each said simulation point u:
(i) identifying a data event B consisting of a facies pattern in a neighborhood surrounding said simulation point u;
(ii) identifying all grid nodes in said training image whose neighboring facies patterns are compatible with data event B;
(iii) computing a conditional probability distribution P(Ai|B,u) based on possible facies events Ai at said simulation point u based on the number of said compatible grid nodes, weighted according to the distance between a respective compatible grid node and a grid node corresponding in location to the simulation point u;
(iv) drawing a facies event for said simulation point u according to the conditional probability distribution P(Ai|B,u);
and
(c) repeating step (b) for each next simulation point until all facies events for all simulation points u are drawn, thereby generating the realistic model of said subterranean formation.

5 more elements

A computer implemented multiple point statistics method of generating a model realization of a subterranean formation having different facies, the method comprising sequentially populating and drawing cells in a simulation grid, wherein the method comprises the steps of:

(a) obtaining a training image representative of geometric patterns and/or features of said different facies and their spatial distribution in the subterranean formation, the training image comprising a plurality of training image cells and having equivalent dimensions to said model realization;
(b) for a target cell in the simulation grid to be populated, identifying a facies pattern of neighboring cells in the simulation grid and then identifying occurrences of a corresponding facies pattern of neighboring training image cells in the training image;
(c) assigning a weighting factor to each said corresponding facies pattern occurrence, the weighting factor reflecting the distance of the respective facies pattern occurrence from a training image cell corresponding in location to said target cell;
(d) calculating a probability of the target cell having certain facies from steps (b) and (c) and drawing said target cell, wherein said probability is defined as P(A|B,u) where:
A stands for the event of observing a certain facies at any location in the simulation grid, B stands for the event of observing a certain facies pattern in the neighborhood of any location in the simulation grid, and u stands for the event of being at a certain location u in the simulation grid, and wherein the calculation of the probability involves the calculation of probability density functions based on the location of events in the training image;
(e) generating said model realization by repeating steps (b) to (d) sequentially until all target cells of the simulation grid are populated and drawn, wherein the model realization represents geological and petrophysical heterogeneities of a geological structure of the subterranean formation.

3 more elements

Dependencies are inferred from the claim text.

Reservoir modelling with multiple point statistics from a non-stationary training image

Life of this patent

Reservoir modelling with multiple point statistics from a non-stationary training image

Life of this patent

Abstract

Claims4