Stratigraphic forward modeling platform and methods of use

Patent metadata

Field	Value
Publication number	US-12242013-B2
Application number	US-202318467046-A
Country	US
Kind code	B2
Filing date	Sep 14, 2023
Priority date	Sep 14, 2022
Publication date	Mar 4, 2025
Grant date	Mar 4, 2025

Abstract

A process mimicking forward modeler with deposition and erosion at each specific geological time step. The 3D derived properties are high resolution depositional environments and rock properties that are used to generate multiscale labelled synthetic data. These synthetic data can range from 1D logs such as grain size, gamma ray, density, and velocity, to 3D synthetic seismic, and are used directly as training data for various AIML applications.

First claim

What is claimed is:

1. A method of using a stratigraphic forward modeling platform, comprising:
generating at least one synthetic stratigraphic models at the reservoir scale using at least one process mimicking algorithms by generating multiple different depositional facies using a time stepping approach;
defining an initial topography and an initial rates of subsidence and an initial rate of uplift over an area region of interest;
computing deposition of depofacies during each time step over a predefined map area, the predefined map area comprising at least a portion of the areal region of interest;
generating petrophysical and rock physics properties of the depofacies over the predefined map area;
generating multiple realizations of each stratigraphic model based on random seed and different hyperparameters;
storing each of the multiple model realizations and indexing the multiple model realizations in geological time; and
utilizing the stored multiple model realization data sets from the model to train an artificial intelligence and machine learning (AIML) application.

2. The method according to claim 1 wherein generating multiple different depositional facies comprises generating at least one of channel lag and overbank deposits, point bars, levees, oxbow lakes, and crevasse splays.

3. The method according to claim 1 wherein using a time stepping approach comprises utilizing a plurality of time steps in a geological time.

4. The method of claim 3 , wherein the plurality of time steps comprise at least one of years (a for annum), kilo-years (Ka), or mega-years (Ma).

5. The method according to claim 3 wherein computing deposition comprises computing a net thickness of each depofacies during each time step over the predefined map area.

6. The method according to claim 3 wherein computing deposition comprises computing a total thickness of deposition during each time step over the predefined map area.

7. The method according to claim 3 wherein computing deposition comprising computing an erosion, compaction and diagenesis of underlying sediment during each time step over the predefined map area.

8. The method according to claim 3 wherein computing deposition comprises computing 3D petrophysical and rock properties such as depofacies classification, porosity, grain size and grain sorting for each depofacies as a function of spatial location (latitude, longitude, elevation) and relative position to sources of active deposition.

9. The method according to claim 1 wherein generating multiple realizations of each stratigraphic model is based on random seed and different hyperparameters such as channel width and depth, deposition rates and properties of each depofacies such as characteristic levee width and variability.

10. The method according to claim 1 wherein storing comprises storing the model realizations in at least one of a local location or a cloud location.

11. The method according to claim 1 further comprising converting the geological time-based model to true vertical depth (TVD) for consumption by modeling and interpretation applications.

12. The method according to claim 11 wherein geological age is a primary index of the time-based model, but when converted to TVD index, the geological age is an additional property of the model.

13. The method according to claim 1 wherein generating petrophysical and rock physics properties based on spatial distribution, grain size and sorting.

14. The method according to claim 1 wherein the AIML application comprises Generative Adversarial Network (GAN) based well conditioning.

15. The method according to claim 1 wherein training comprises training deep learning-based architectures such as Convolutional Neural Network (CNN), UNet, ResNet or Long Short-Term Memory (LSTM) for multiscale automatic stratigraphy interpretation along and among wells.

Assignees

• Schlumberger Technology Corp

Inventors

• Tilke Peter
• Etchebes Marie
• Lefranc Marie Emeline Cecile
• Zhu Lingchen
• Lis Michael
• Sabathier Remy

Classifications

• G06F30/27

  using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model · CPC title

• G01V20/00 Primary

  Geomodelling in general · CPC title