Machine learning assisted reservoir simulation

1. An apparatus for controlling wellhead equipment in a fluid reservoir, comprising:

a memory; and

at least one processor coupled with the memory, the processor operative to implement at least one machine learning classifier, the machine classifier being operative to:

obtain one or more recent results from at least one geomechanical simulation;

compare the recent results to stored historical data;

based on the comparing, decide at least one reservoir model for use by at least one reservoir simulation; and

control the wellhead equipment using the results from the at least one geomechanical simulation and from the at least one reservoir simulation, wherein:

the one or more recent results from the at least one geomechanical simulation comprise at least one point, each point comprising at least one strain value and at least one stress value for at least a given geomaterial; and

the stored historical data comprises at least one stress-strain curve for at least the given geomaterial;

wherein:

the at least one point in the one or more recent results from the at least one geomechanical simulation comprises a current point for at least a given geomaterial;

the current point for the given geomaterial comprises a current stress value and a current strain value; and

the machine classifier is operative to decide by at least one of:

selecting at least one single porosity model if the current point is within a pre-peak area of the at least one stress-strain curve for at least the given geomaterial;

selecting at least one dual porosity model if the current point is within a peak area of the at least one stress-strain curve for at least the given geomaterial; and

selecting at least one dual permeability model if the current point is within a post-peak area of the at least one stress-strain curve for at least the given geomaterial.

2. The apparatus of claim 1 , wherein the stored historical data comprises a plurality of stress-strain curves for the given geomaterial.

3. The apparatus of claim 2 , wherein the stored historical data further comprises:

an average stress-strain curve for the given geomaterial;

at least one stress-strain curve based at least in part on:

the average stress-strain curve for the given geomaterial; and

a standard deviation of the plurality of stress-strain curves for the given geomaterial.

4. The apparatus of claim 1 , wherein:

the stored historical data comprises a plurality of stress-strain curves comprising at least one stress-strain curve for each of a plurality of geomaterials comprising at least the given geomaterial.

5. An apparatus for controlling wellhead equipment in a fluid reservoir, comprising:

a memory; and

at least one processor coupled with the memory, the processor operative to implement at least one machine learning classifier, the machine classifier being operative to:

obtain one or more recent results from at least one geomechanical simulation;

compare the recent results to stored historical data;

based on the comparing, decide at least one reservoir model for use by at least one reservoir simulation; and

control the wellhead equipment using the results from the at least one geomechanical simulation and from the at least one reservoir simulation, wherein:

the one or more recent results from the at least one geomechanical simulation comprise at least one point, each point comprising at least one strain value and at least one stress value for at least a given geomaterial; and

the stored historical data comprises at least one stress-strain curve for at least the given geomaterial;

wherein:

the at least one point in the one or more recent results from the at least one geomechanical simulation comprises a plurality of points for at least the given geomaterial from the at least one geomechanical simulation;

each of the plurality of points for the given geomaterial comprises a respective stress value and a respective strain value; and

the machine classifier is operative to compare by computing a proportion of the plurality of points which are within a pre-peak area of the at least one stress-strain curve for at least the given geomaterial.

6. The apparatus of claim 5 , wherein the plurality of points for at least the given geomaterial from the at least one geomechanical simulation a current point for at least a given geomaterial comprises a current point and at least one prior point.

7. The apparatus of claim 5 , wherein:

the machine classifier is further operative to compare by determining at least one characteristic of at least one fracture in at least the given geomaterial based at least in part on the computed proportion; and

the machine classifier is operative to decide by deciding the at least one reservoir model based at least in part on the determined characteristic of the fracture.

8. The apparatus of claim 5 , wherein the machine classifier is operative to decide by at least one of:

selecting at least one single porosity model if the proportion is less than a first value;

selecting at least one dual porosity model if the proportion is between the first value and a second value, the second value being greater than the first value; and

selecting at least one dual permeability model if the proportion is greater than the second value.

9. The apparatus of claim 8 , wherein the first value is about forty percent, and wherein the second value is about sixty percent.

10. An apparatus for controlling wellhead equipment in a fluid reservoir, comprising:

a memory; and

at least one processor coupled with the memory, the processor operative to implement at least one machine learning classifier, the machine classifier being operative to:

obtain one or more recent results from at least one geomechanical simulation;

compare the recent results to stored historical data;

based on the comparing, decide at least one reservoir model for use by at least one reservoir simulation; and

control the wellhead equipment using the results from the at least one geomechanical simulation and from the at least one reservoir simulation, wherein:

the machine classifier is operative to compare by determining at least one characteristic of at least one fracture in at least one geomaterial; and

the machine classifier is operative to decide on the at least one reservoir model based at least in part on the determined characteristic of the fracture.

11. The apparatus of claim 10 , wherein the characteristic of the fracture comprise at least one of presence, movement, and connectedness.

12. The apparatus of claim 11 , wherein the machine classifier is operative to decide by at least one of:

selecting at least one single porosity model if the fracture is not present or is not moving;

selecting at least one dual porosity model if the fracture is moving and connected; and

selecting at least one dual permeability model if the fracture is moving and not connected.

13. A non-transitory computer readable medium comprising computer executable instructions which when executed by a computer cause the computer to perform a method for controlling wellhead equipment in a fluid reservoir, the method comprising:

implementing at least one machine learning classifier;

obtaining, with the implemented machine learning classifier, one or more recent results from at least one geomechanical simulation;

comparing, with the implemented machine learning classifier, the recent results to stored historical data;

based on the comparing, deciding, with the implemented machine learning classifier, at least one reservoir model for use by at least one reservoir simulation; and

controlling, with the implemented machine learning classifier, the wellhead equipment using the results from the at least one geomechanical simulation and from the at least one reservoir simulation, wherein:

the one or more recent results from the at least one geomechanical simulation comprise at least one point, each point comprising at least one strain value and at least one stress value for at least a given geomaterial; and

the stored historical data comprises at least one stress-strain curve for at least the given geomaterial;

wherein:

the at least one point in the one or more recent results from the at least one geomechanical simulation comprises a current point for at least a given geomaterial;

the current point for the given geomaterial comprises a current stress value and a current strain value; and

the machine learning classifier is operative to decide by at least one of:

selecting at least one single porosity model if the current point is within a pre-peak area of the at least one stress-strain curve for at least the given geomaterial;

selecting at least one dual porosity model if the current point is within a peak area of the at least one stress-strain curve for at least the given geomaterial; and

selecting at least one dual permeability model if the current point is within a post-peak area of the at least one stress-strain curve for at least the given geomaterial.

14. A non-transitory computer readable medium comprising computer executable instructions which when executed by a computer cause the computer to perform a method for controlling wellhead equipment in a fluid reservoir, the method comprising:

implementing at least one machine learning classifier;

obtaining, with the implemented machine learning classifier, one or more recent results from at least one geomechanical simulation;

comparing, with the implemented machine learning classifier, the recent results to stored historical data;

based on the comparing, deciding, with the implemented machine learning classifier, at least one reservoir model for use by at least one reservoir simulation; and

controlling, with the implemented machine learning classifier, the wellhead equipment using the results from the at least one geomechanical simulation and from the at least one reservoir simulation, wherein:

the one or more recent results from the at least one geomechanical simulation comprise at least one point, each point comprising at least one strain value and at least one stress value for at least a given geomaterial; and

the stored historical data comprises at least one stress-strain curve for at least the given geomaterial;

wherein:

the at least one point in the one or more recent results from the at least one geomechanical simulation comprises a plurality of points for at least the given geomaterial from the at least one geomechanical simulation;

each of the plurality of points for the given geomaterial comprises a respective stress value and a respective strain value; and

the machine learning classifier is operative to compare by computing a proportion of the plurality of points which are within a pre-peak area of the at least one stress-strain curve for at least the given geomaterial.

15. The non-transitory computer readable medium of claim 14 , wherein the plurality of points for at least the given geomaterial from the at least one geomechanical simulation a current point for at least a given geomaterial comprises a current point and at least one prior point.

16. The non-transitory computer readable medium of claim 14 , wherein:

the machine learning classifier is further operative to compare by determining at least one characteristic of at least one fracture in at least the given geomaterial based at least in part on the computed proportion; and

the machine learning classifier is operative to decide by deciding the at least one reservoir model based at least in part on the determined characteristic of the fracture.