System, method and program product for automatically matching new members of a population with analogous members

• US 9,417,256 B2
• Filed: 12/12/2012
• Issued: 08/16/2016
• Est. Priority Date: 12/12/2012
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation, Repsol, S. A.
• Original Assignee: International Business Machines Corporation, Repsol, S. A.
• Inventors: Rodriguez, Hilario Martin, Hegazy, Mohamed Ahmed, Embid Droz, Sonia Mariette, Flach, Bruno Da Costa, Valladao, Davi Michel, Zadrozny, Bianca
• Primary Examiner(s): Toatley, Gregory J
• Assistant Examiner(s): Alkafawi, Eman
• Application Number:
  US13/711,616
• Publication Number:
  US 20140163901A1
• Time in Patent Office: 1,343 Days
• Field of Search: 702/22, 702/6, 702/13, 702/11, 702/9, 702/34, 706/52, 706/46, 340/610
• US Class Current: 1/1
• CPC Class Codes: G01N 37/00 Details not covered by any other group of this subclass
• CPC Class Codes: G01V 20/00 Geomodelling in general
• CPC Class Codes: G01V 2210/624 Reservoir parameters
• CPC Class Codes: G01V 2210/665 using geostatistical modeling

Abstract

A population comparison system, method and a computer program product therefor. A stored list of population members, e.g., hydrocarbon reservoirs, characteristics and analogous members is partitioned into lists for each member. A weighting system automatically uses the partitions to determine a weight set (w*) for population member characteristic and a similarity function. The weighting system may include an objective model that iteratively, blindly identifies analogous members for each population member until the identified analogous members match the listed analogous members. An analogous member selector uses the weights set (w*) and similarity function to automatically select analogous listed members for each new population member.

Claims

1. A population comparison system comprising:

a population storage storing a list of population members, each list entry including one or more characteristics for a respective population member and an analogous member list listing one or more other population members that have been previously identified as analogous to said respective population member;

an automatic weighting system automatically extracting weights (w*) for population member characteristics, said automatically extracted weights (w*) determining a similarity function for identifying analogous population members from said list, such that any error between the identified analogous members and said analogous member list is minimized, said automatic weighting unit comprising;

cross validation pre-processor partitions said population into a plurality of partitions, wherein said cross validation pre-processor partitions said plurality (I) into one partition for each member (i, where iε

I), each partition including two sub-groups, one sub-group including said each member (i) and the other sub-group including all other members remaining after excluding the respective said each member (Ii=I\{i}), and

an objective model automatically selecting property weights to select from other said listed population members for each listed population member the same members as said more previously identified population members;

a new population member input receiving new population member descriptions including characteristics for each respective new population member;

an analogous member selector selecting a subset of listed population members as analogous members for each new population member responsive to said automatically extracted weights (w*) applied to said similarity function; and

means for providing a value for said new population member estimated from the analogous members subset and using the estimated value to decide how to proceed with said new population member.

2. A population comparison system as in claim 1, wherein said population members are reservoirs (R) with properties (pε

P), each property has a known value (θ

_i^p), and said objective model compares said known value for a target property (η

) against a property value estimate (v_iε

R;

individual iε

I estimate of property η

ε

P) for said target property to arrive at property weights (w_pε

[1,0]).

3. A population comparison system as in claim 2, wherein said objective model minimizes the comparison difference and has the form:

4. A population comparison system as in claim 2, wherein two reservoirs (i and j) are separated by a 2-norm univariate distance (δ

_ij^p), and inputs to said objective model include;

θ

_i^pε

[1,0];

(known) value of property pε

P of individual iε

I; and

δ

_ij^pε

[1,0];

distance along property pε

P between reservoirs iε

I and jε

I.

5. A population comparison system as in claim 4, wherein

reservoirs are separated by a distance, d_ijε

R;

iε

I and jε

I;

each reservoir has K neighboring reservoirs within a neighborhood threshold, t_iε

R;

iε

I; and

said objective model optimizes subject to property estimate constraints, neighboring distance constraints, weight constraints, and neighboring member constraints.

6. A population comparison system as in claim 5, wherein

property estimate constraints have the form

7. A population comparison system as in claim 1, said analogous member selector comprising:

a member selection unit applying said automatically extracted weights (w*) to new population member characteristics and automatically selecting listed members satisfying a threshold as members of said subset; and

an estimator determining said value for the new population member from said subset.

8. A population comparison system as in claim 7, wherein said threshold is a preselected number, said member selection unit selecting said preselected number of listed members with weighted properties closest to said the new population member.

9. A reservoir (R) valuation system comprising:

a reservoir storage storing a reservoir database listing known reservoirs, each list entry including characteristics, and an analogous reservoir list listing one or more other reservoirs that are previously identified as analogous to the respective known reservoir;

an automatic weighting system automatically extracting weights (w*) for known reservoir characteristics, said automatically extracted weights (w*) determining a similarity function identifying analogous reservoirs from said list, such that any error between the identified analogous reservoirs and said analogous reservoir list is minimized;

a new reservoir input receiving new reservoir descriptions including characteristics for each respective said new reservoir;

an analogous reservoirs selection unit applying said automatically extracting weights (w*) in said similarity function to said new reservoir characteristics to automatically select a subset of known reservoirs satisfying a threshold;

said automatic weighting unit comprises;

cross validation pre-processor partitions said population into a plurality of partitions, wherein said cross validation pre-processor partitions said plurality (I) into one partition for each member (i, where iε

I), each partition including two sub-groups, one sub-group including said each member (i) and the other sub-group including all other members remaining after excluding the respective said each member (Ii=I\{i}), and

an objective model automatically selecting property weights to select from other said listed population members for each listed population member the same members as said more previously identified analogous reservoir;

an estimator determining a value for the said new reservoir from said subset; and

one or more user terminals providing the estimated value, wherein the estimated value provides a decision of whether to pass or exploit said new reservoir.

10. A reservoir valuation system as in claim 9, wherein said reservoirs are hydrocarbon reservoirs with properties (pε

P), each property has a known value (θ

_i^p), and said objective model compares said known value for a target property (η

) against a property value estimate (v_iε

R;

individual iε

I estimate of property η

ε

P) for said target property to arrive at property weights (w_pε

[1,0]).

11. A reservoir valuation system as in claim 10, wherein two reservoirs (i and j) are separated by a 2-norm univariate distance (δ

_ij^p), said objective model minimizes the comparison difference and has the form;

12. A reservoir valuation system as in claim 11, wherein

said hydrocarbon reservoirs are separated by a distance, d_ijε

R;

iε

I and jε

I;

each said hydrocarbon reservoir has K neighboring reservoirs within a neighborhood threshold, t_iε

R;

iε

I; and

said objective model optimizes subject to property estimate constraints, neighboring distance constraints, weight constraints, and neighboring member constraints.

13. A reservoir valuation system as in claim 12, wherein

said property estimate constraints have the form

14. A computer program product for comparing members of a population, said computer program product comprising a computer usable medium having computer readable program code stored thereon, said computer readable program code comprising:

computer readable program code means for storing a list of known population members, each list entry including one or more characteristics for a respective population member and an analogous member list listing one or more other population members that are previously identified as analogous to said respective population member, and further indicating each one or more previously identified analogous population members for each listed population member;

computer readable program code means for automatically extracting weights (w*) for population member characteristics, said automatically extracted weights (w*) determining a similarity function for identifying analogous population members from said list, such that any error between the identified analogous members and said analogous member list is minimized, said computer readable program code means for automatically extracting weights comprising;

computer readable program code means for partitioning said population (I) into a a plurality of partitions, one partition for each member (i, where iε

I), each partition including two sub-groups, one sub-group including said each member (i) and the other sub-group including all other members remaining after excluding the respective said each member (Ii=I\{i}), and

computer readable program code means for automatically selecting property weights to select from other said listed population members for each listed population member the same members as said more previously identified population members;

computer readable program code means for receiving new population member descriptions including characteristics for each respective new population member; and

computer readable program code means for selecting a subset of listed population members as analogous members for each new population member responsive to said automatically extracted determined weights (w*) applied to said similarity function, said subset being used to estimate a value for said new population member, the estimated value being used to decide how to proceed with said new population member.

15. A computer program product for comparing members of a population as in claim 14, wherein said population members are hydrocarbon reservoirs (R), computer readable program code means for receiving new population member descriptions comprises computer readable program code means for receiving description information directly from remotely connected sensors, automatically, or manually entered into one or more of the networked computers, and said computer readable program code means for automatically selecting comprises computer readable program code means for comparing a known value for a target property (η

) against a property value estimate (v_iε

R;

individual iε

I estimate of property η

ε

P) for said target property to arrive at property weights (w_pε

[1,0]).

16. A computer program product for comparing members of a population as in claim 15, wherein said computer readable program code means for automatically selecting further comprises computer readable program code means for minimizing the comparison difference and has the form:

17. A computer program product for comparing members of a population as in claim 16, wherein computer readable program code means for automatically selecting optimizes subject to property estimate constraints, neighboring distance constraints, weight constraints, and neighboring member constraints, where

property estimate constraints have the form

18. A computer program product for valuating reservoirs (R), said computer program product comprising a computer usable medium having computer readable program code stored thereon, said computer readable program code comprising:

computer readable program code means for storing a reservoir database with a list of a population of known reservoirs, each reservoir list entry including;

characteristics for a respective known reservoir and an identification of one or more other listed said known reservoirs previously identified as analogous reservoirs;

computer readable program code means for partitioning said population (I) into a plurality of partitions, one partition for each member (i, where iε

I), each partition including two sub-groups, one sub-group including said each member (i) and the other sub-group including all other members remaining after excluding the respective said each member (Ii=I{i});

computer readable program code means for selecting property weights to select in each said partition analogous reservoirs from the second sub-group for the reservoir in the first group, such that the selected reservoirs are the same as the previously identified analogous reservoirs;

computer readable program code means for determining a set of weights (w*) for known reservoir characteristics and a similarity function independently identifying for said analogous reservoirs for said each known reservoir;

computer readable program code means for receiving new reservoir descriptions including characteristics for each respective new reservoir;

computer readable program code means for applying determined said weights (w*) to new reservoir characteristics and automatically selecting a subset of known reservoirs satisfying a threshold;

computer readable program code means for determining a value for said each respective new reservoir from said subset; and

computer readable program code means for providing the determined value, wherein the determined value provides a decision of whether to pass or exploit said new reservoir.

19. A computer program product for comparing members of a population as in claim 18, wherein said population members are hydrocarbon reservoirs with properties (pε

P), each property has a known value (θ

_i^p), and said computer readable program code means for determining comprises;

computer readable program code means for comparing said known value for a target property (η

) against a property value estimate (v_iε

R;

individual iε

I estimate of property η

ε

P) for said target property to arrive at property weights (w_pε

[1,0]); and

computer readable program code means for constraining said computer readable program code means for comparing, neighboring distance constraints, weight constraints, and neighboring member constraints, where

whether reservoir j is a nearest neighbor to reservoir i is indicated by y_ij;

property estimate constraints have the form

20. A computer program product for comparing members of a population as in claim 19, wherein said computer readable program code means for automatically selecting further comprises computer readable program code means for minimizing the comparison difference and has the form: