Methods and Systems For Machine - Learning Based Simulation of Flow

• US 20130118736A1
• Filed: 05/19/2011
• Published: 05/16/2013
• Est. Priority Date: 07/29/2010
• Status: Active Grant

• Current Assignee: Adam Usadi, Dachang Li, Rossen Parashkevov, Sergey A. Terekhov, Xiaohui Wu, Yahan Yang
• Original Assignee: Adam Usadi, Dachang Li, Rossen Parashkevov, Sergey A. Terekhov, Xiaohui Wu, Yahan Yang
• Inventors: Usadi, Adam, Li, Dachang, Parashkevov, Rossen, Terekhov, Sergey A., Yang, Yahan, Wu, Xiaohui
• Granted Patent: US 9,187,984 B2
• Time in Patent Office: Days
• US Class Current: 166/268
• CPC Class Codes: E21B 43/00 Methods or apparatus for obtaining oil, gas, water, soluble or meltable materials or a slurry of minerals from wells applicable only to water E03B
• CPC Class Codes: G01V 20/00 Geomodelling in general
• CPC Class Codes: G06F 2111/10 Numerical modelling
• CPC Class Codes: G06F 2113/08 Fluids
• CPC Class Codes: G06F 30/20 Design optimisation, verification or simulation optimisation, verification or simulation of circuit designs G06F30/30
• CPC Class Codes: G06F 30/23 using finite element methods [FEM] or finite difference methods [FDM]
• CPC Class Codes: G06F 30/27 using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
• CPC Class Codes: G06F 30/28 using fluid dynamics, e.g. using Navier-Stokes equations or computational fluid dynamics [CFD]
• CPC Class Codes: G06N 3/042 Knowledge-based neural networks; Logical representations of neural networks

Abstract

There is provided a method for modeling a hydrocarbon reservoir that includes generating a reservoir model that has a plurality of coarse grid cells. A plurality of fine grid models is generated, wherein each fine grid model corresponds to one of the plurality of coarse grid cells that surround a flux interface. The method also includes simulating the plurality of fine grid models using a training simulation to obtain a set of training parameters, including a potential at each coarse grid cell surrounding the flux interface and a flux across the flux interface. A machine learning algorithm is used to generate a constitutive relationship that provides a solution to fluid flow through the flux interface. The method also includes simulating the hydrocarbon reservoir using the constitutive relationship and generating a data representation of a physical hydrocarbon reservoir in a non-transitory, computer-readable medium based on the results of the simulation.

Claims

1. A method for modeling a hydrocarbon reservoir, comprising:

generating a reservoir model comprising a plurality of coarse grid cells;

generating a plurality of fine grid models, each fine grid model corresponding to one of the plurality of coarse grid cells that surround a flux interface;

simulating the plurality of fine grid models using a training simulation to obtain a set of training parameters comprising a potential at each coarse grid cell surrounding the flux interface and a flux across the flux interface;

using a machine learning algorithm to generate a constitutive relationship that provides a solution to fluid flow through the flux interface;

simulating the hydrocarbon reservoir using the constitutive relationship; and

generating a data representation of a physical hydrocarbon reservoir in a non-transitory, computer-readable medium based, at least in part, on the results of the simulation.

2. The method of claim 1, wherein simulating the plurality of fine grid models using the training simulation comprises specifying a set of physical, geometrical, or numerical parameters for each one of the plurality of fine grid models, wherein the set of physical, geometrical, or numerical parameters comprises potential, phase permeability, phase saturation, geological characteristics, and combinations thereof.

3. The method of claim 1, comprising storing the constitutive relationship generated for the flux interface and physical, geometrical, or numerical parameters corresponding to the plurality of coarse grid cells that surround the flux interface to a database of surrogate solutions for use in subsequent reservoir simulations.

4. The method of claim 1, wherein the constitutive relationship generated for the flux interface is re-used for a second flux interface based on a comparison of a set of physical, geometrical, or numerical parameters corresponding to the flux interface and a new set of physical, geometrical, or numerical parameters that characterize the second flux interface.

5. The method of claim 1, wherein using the machine learning algorithm to generate the constitutive relationship comprises training a neural net using the training parameters, wherein the potential at each coarse gird cell surrounding the flux interface is used as an input to the neural net and the flux across the flux interface is used as a desired output.

6. The method of claim 5, wherein another input to the neural net comprises geometric parameters of the coarse grid cells surrounding the flux interface.

7. The method of claim 1, wherein generating a plurality of fine grid models comprises generating two or more fine grid models of varying fine grid cell sizes for each corresponding coarse grid cell.

8. A method for producing a hydrocarbon from a hydrocarbon reservoir, comprising:

generating a reservoir model comprising a plurality of coarse grid cells;

generating a plurality of fine grid models, each fine grid model corresponding to one of the plurality of coarse grid cells that surround a flux interface;

simulating the plurality of fine grid models using a training simulation to obtain a set of training parameters comprising a potential at each coarse grid cell surrounding the flux interface and a flux across the flux interface;

using a machine learning algorithm to generate a constitutive relationship that provides a solution to fluid flow through the flux interface;

simulating the hydrocarbon reservoir using the constitutive relationship; and

producing a hydrocarbon from the hydrocarbon reservoir based, at least in part, upon the results of the simulation.

9. The method of claim 8, wherein producing the hydrocarbon comprises:

drilling one or more wells to the hydrocarbon reservoir, wherein the wells comprise production wells, injection wells, or both;

setting production rates from the hydrocarbon reservoir;

or

any combinations thereof.

10. A system for modelling reservoir properties, comprising:

a processor;

a non-transitory machine readable medium comprising code configured to direct the processor to;

generate a reservoir model comprising a plurality of coarse grid cells;

generate a plurality of fine grid models, each fine grid model corresponding to one of the plurality of coarse grid cells that surround a flux interface;

simulate the plurality of fine grid models using a training simulation to obtain a set of training parameters comprising a potential at each coarse grid cell surrounding the flux interface and a flux across the flux interface;

use a machine learning algorithm to generate a constitutive relationship that provides a solution to fluid flow through the flux interface;

simulate the reservoir using the constitutive relationship.

11. The system for claim 10, wherein the machine readable medium comprises code configured to direct the processor to receive a set of physical, geometrical, or numerical parameters for each one of the plurality of fine grid models, the physical, geometrical, or numerical parameters used for simulating each one of the plurality of fine grid models using the training simulation, wherein the set of physical, geometrical, or numerical parameters comprises at least one of rock porosity, phase permeability, and geological characteristics.

12. The system for claim 10, wherein the machine readable medium comprises code configured to direct the processor to store the constitutive relationship generated for the flux interface and physical, geometrical, or numerical parameters corresponding to the plurality of coarse grid cells that surround the flux interface to a database of surrogate solutions for use in subsequent reservoir simulations.

13. The system for claim 10, wherein the machine readable medium comprises code configured to direct the processor to re-use the constitutive relationship generated for the flux interface for a second flux interface based on a comparison of a set of physical, geometrical, or numerical parameters corresponding to the flux interface and a new set of physical, geometrical, or numerical parameters that characterize the second flux interface.

14. The system of claim 11, comprising a neural net, wherein the machine readable medium comprises code configured to direct the processor to train the neural net using the training parameters, wherein the potential at each coarse gird cell surrounding the flux interface is used as an input to the neural net and the flux across the flux interface is used as a desired output.

15. The system of claim 14, wherein another input to the neural net comprises geometric parameters of the coarse grid cells surrounding the flux interface.

16. The system of claim 11, wherein generating the plurality of fine grid models comprises generating two or more fine grid models of varying fine grid cell sizes for each corresponding coarse grid cell.

17. A non-transitory, computer readable medium comprising code configured to direct a processor to:

generate a reservoir model comprising a plurality of coarse grid cells;

generate a plurality of fine grid models, each fine grid model corresponding to one of the plurality of coarse grid cells that surround a flux interface;

simulate the plurality of fine grid models using a training simulation to obtain a set of training parameters comprising a potential at each coarse grid cell surrounding the flux interface and a flux across the flux interface;

use a machine learning algorithm to generate a constitutive relationship that provides a solution to fluid flow through the flux interface;

simulate the reservoir model using the constitutive relationship.

18. The non-transitory, computer readable medium of claim 17 comprising code configured to store the constitutive relationship generated for the flux interface and physical, geometrical, or numerical parameters corresponding to the plurality of coarse grid cells that surround the flux interface to a database of surrogate solutions for use in the simulation model or a different simulation model.

19. The non-transitory, computer readable medium of claim 17 comprising code configured to direct the processor to generate a neural net and train the neural net using the training parameters, wherein the potential at each coarse gird cell surrounding the flux interface is used as an input to the neural net and the flux across the flux interface is used as a desired output.

20. The non-transitory, computer readable medium of claim 19, wherein another input to the neural net comprises geometric parameters of the coarse grid cells surrounding the flux interface.