Reduced space clustering representatives and its application to long term prediction
In one embodiment, a technique is provided to cluster model prior realizations into flow-equivalent sets. Thereby, far fewer representatives are identified from each cluster that can still predict the same range of future production reliably. Only representatives of each realization cluster that indicate unique forecasts will undergo the computationally expensive process of history matching.
1. A method for estimating variability of future recovery of a natural resource, the method comprising:
receiving, from a database, a plurality of model realizations of the natural resource, the plurality of model realizations representing uncertainty in determining model properties of the natural resource;
determining, for each of the plurality of model realizations, respective principal flow pattern coefficients representing a reduced model space;
identifying similar principal flow patterns amongst the received plurality of model realizations using the coefficients representing the reduced model space;
clustering the received plurality of model realizations according to the identified similar principal flow patterns in the reduced model space, each cluster having a similar principal flow pattern and representing a prior input to a history matching process;
determining a set of history matched models by applying the history matching process on a set of selected model prior realizations, wherein the selected model prior realizations are a subset of representative realizations selected from the clustered realizations; and
determining variability in natural resource production by utilizing the set of history matched models as inputs to a prediction simulation tool, wherein the prediction simulation tool is configured to run simulations based on the set of history matched models to estimate the future production scenarios of the natural resource;
wherein a processor connected to a memory device is configured to perform: the receiving the plurality of model realizations, the determining the respective principal flow pattern coefficients, the identifying similar principal flow patterns, the clustering the received plurality of model realizations, the determining the set of history matched models, the determining the variability, and running the prediction simulation tool.
2. The method according to claim 1 , wherein the process of identifying similar principal flow patterns comprises:
determining a distance between coefficients of the principal flow patterns of the received plurality of model realizations; and
determining model realization cluster association based upon whether each determined distance is less than a pre-determined threshold.
3. The method according to claim 1 , wherein each cluster represents a prior input to the history matching process, of which subsequent future simulation is performed to estimate a future production scenario of the natural resource.
4. The method according to claim 1 , further comprising:
selecting a representative of each cluster to generate the selected model prior realizations.
5. The method according to claim 4 , further comprising performing a singular value decomposition upon the flow indicators to enable reduced representation of respective associated flow patterns related to the dynamic behavior of each model realization.
6. The method according to claim 5 , further comprising:
representing each flow pattern in a reduced space using a respective singular vector obtained from the singular value decomposition.
7. The method according to claim 1 , further comprising:
hierarchically organizing each cluster, each cluster aggregating a set of one or more of the received model realizations.
8. A computer program product for estimating variability of future recovery of a natural resource, the program code of the computer program product executable by a computer to perform a method comprising:
receiving, from a database, a plurality of model realizations of the natural resource, the plurality of model realizations representing uncertainty in determining model properties of the natural resource;
determining, for each of the plurality of model realizations, respective principal flow pattern coefficients representing a reduced model space;
identifying similar principal flow patterns amongst the received plurality of model realizations using the coefficients representing the reduced model space;
clustering the received plurality of model realizations according to the identified similar principal flow patterns in the reduced model space, each cluster having a similar principal flow pattern and representing a prior input to a history matching process;
determining a set of history matched models by applying the history matching process on a set of selected model prior realizations, wherein the selected model prior realizations are a subset of representative realizations selected from the clustered realizations; and
determining variability in natural resource production by utilizing the set of history matched models as inputs to a prediction simulation tool, wherein the prediction simulation tool is configured to run simulations based on the set of history matched models to estimate the future production scenarios of the natural resource.
9. The computer program product according to claim 8 , wherein the process of identifying similar principal flow patterns comprises:
determining a distance between coefficients of the principal flow patterns of the received plurality of model realizations; and
determining model realization cluster association based upon whether each determined distance is less than a pre-determined threshold.
10. The computer program product according to claim 8 , wherein each cluster represents a prior input to the history matching process, of which subsequent future simulation is performed to estimate a future production scenario of the natural resource.
11. The computer program product according to claim 8 , the method performed further comprising:
selecting a representative of each cluster to generate the selected model prior realizations.
12. The computer program product according to claim 11 , the method performed further comprising performing a singular value decomposition upon the flow indicators to enable reduced representation of respective associated flow patterns related to the dynamic behavior of each model realization.
13. The computer program product according to claim 12 , the method performed further comprising:
representing each flow pattern in a reduced space using a respective singular vector obtained from the singular value decomposition.
14. The computer program product according to claim 8 , the method performed further comprising:
hierarchically organizing each cluster, each cluster aggregating a set of one or more of the received model realizations.
15. A computer-implemented system for estimating variability of future recovery of a natural resource, the system comprising:
a processor; and
a memory storing computer-readable instructions which, when executed by the processor, implement:
a receiving element configured to receive, from a database, a plurality of model realizations of the natural resource, the plurality of model realizations representing uncertainty in determining model properties of the natural resource;
a first determining element configured to determine, for each of the plurality of model realizations, respective principal flow pattern coefficients representing a reduced model space;
an identifying element configured to identify similar principal flow patterns amongst the received plurality of model realizations using the coefficients representing the reduced model space;
a clustering element configured to cluster the received plurality of model realizations according to the identified similar principal flow patterns in the reduced model space, each cluster having a similar principal flow pattern and representing a prior input to a history matching process;
a utilizing element configured to determine a set of history matched models by applying the history matching process on a set of selected model prior realizations, wherein the selected model prior realizations are a subset of representative realizations selected from the clustered realizations; and
a prediction element configured to determine variability in natural resource production by utilizing the set of history matched models as inputs to a prediction simulation tool, wherein the prediction simulation tool is configured to run simulations based on the set of history matched models to estimate the future production scenarios of the natural resource.
16. The system according to claim 15 , wherein the process of identifying similar principal flow patterns by the identifying element comprises:
determining a distance between coefficients of the principal flow patterns of the received plurality of model realizations; and
determining model realization cluster association based upon whether each determined distance is less than a pre-determined threshold.
17. The system according to claim 15 , wherein each cluster represents a prior input to the history matching process, of which subsequent future simulation is performed to estimate a future production scenario of the natural resource.
18. The system according to claim 15 , further comprising:
a selecting element configured to select a representative of each cluster to generate the selected model prior realizations.
19. The system according to claim 18 , further comprising a singular value decomposition element configured to perform singular value decomposition upon the flow indicators to enable reduced representation of respective associated flow patterns related to the dynamic behavior of each model realization.
20. The system according to claim 15 , further comprising:
an organizing element configured to hierarchically organize each cluster, each cluster aggregating a set of one or more of the received model realizations.