Model parameter reductions and model parameter selection to optimize execution time of reservoir management workflows
An apparatus for generating forecasts from a high-dimensional parameter data space comprising a reservoir model, a model order reduction module, and an assisted history matching module. The reservoir model having input variables, output variables, and an algorithmic model. The input variables, output variables, and the algorithmic model are generated by a flow simulator module and from a formation and reservoir properties database and a field production database. The model order reduction module generates a subset of the original or transformed input variables. The assisted history matching module adjust values of the output variables based on a difference between the at least one of the output variables and dynamic field production data to improve model accuracy.
1. An apparatus for generating forecasts from an input high-dimensional parameter data space, the apparatus comprising:
a reservoir model that includes a plurality of input variables, a plurality of output variables, and at least one algorithmic model, wherein the plurality of input variables, the plurality of output variables, and the at least one algorithmic model are generated by a flow simulator module based on variables from a formation and reservoir properties database and a field database;
a model order reduction module configured to generate a subset of the plurality of input variables, the subset having a reduced parameter space relative to the input high-dimensional parameter data space by identifying input variables within the plurality of input variables useful to approximate at least one output variable of the plurality of output variables, wherein the model order reduction module uses a function decomposition algorithm to identify the input variables within the plurality of input variables useful to approximate the at least one output variable to generate the subset of the plurality of input variables with a reduced a number of input variables such that the subset of input variables of the plurality of input variables most effectively approximates the at least one output variable; and
an assisted history matching module configured to adjust values of at least one of the output variables based on a difference between the at least one of the output variables and dynamic field production data.
2. The apparatus of claim 1 , wherein the flow simulator module generates the reservoir model using at least one of a full-physics flow simulator and a proxy flow simulator.
3. The apparatus of claim 1 , wherein the model order reduction module additionally uses a design of experiments algorithm to identify input variables that are useful to approximate at least one output variable.
4. The apparatus of claim 1 , wherein the plurality of input variables include reservoir parameters represented by a first coefficient and the subset is a subset of the reservoir parameters represented by a second coefficient having a value less than that of the first coefficient.
5. The apparatus of claim 4 , wherein the second coefficient is derived by applying a series expansion algorithm to the plurality of input variables and the plurality of output variables.
6. The apparatus of claim 5 , further comprising a sensitivity analysis module configured to measure sensitivity of at least one output variable of the plurality of output variables based on change in values of the subset of the reservoir parameters.
7. The apparatus of claim 1 , wherein the assisted history matching module uses Bayesian optimization to adjust values of at least one of the plurality of output variables based on a difference between the at least one of the output variables and dynamic field production data.
8. A system for generating forecasts from an input high-dimensional parameter data space, the system comprising:
at least one flow simulator configured to generate a reservoir model including a plurality of input variables, a plurality of output variables, and at least one algorithmic model, wherein the at least one flow simulator includes at least one of a full-physics flow simulator module and a proxy flow simulator module,
wherein the full-physics flow simulator module generates the reservoir model using variables from at least one of a formation and reservoir properties database and a field database,
wherein the proxy flow simulator module generates the reservoir model using a machine learning algorithm, an assisted history matching module and/or field development optimization module, and variables from at least one of the formation and reservoir properties database and the field database;
a model order reduction module configured to generate a subset of the plurality of input variables, the subset having a reduced parameter space relative to the input high-dimensional parameter data space by identifying input variables within the plurality of input variables useful to approximate at least one output variable of the plurality of output variables, wherein the model order reduction module uses a function decomposition algorithm to identify the input variables within the plurality of input variables useful that can be used to approximate the at least one output variable to generate the subset of the plurality of input variables with a reduced number of input variables such that the subset of input variables of the plurality of input variables most effectively approximates the at least one output variable; and
an assisted history matching module configured to adjust values of at least one of the output variables based on a difference between the at least one of the output variables and dynamic field production data.
9. The system of claim 8 , wherein the model order reduction module additionally uses a design of experiments algorithm to identify input variables that are useful to approximate at least one output variable.
10. The system of claim 8 , wherein the plurality of input variables include reservoir parameters represented by a first coefficient and the subset is a subset of the reservoir parameters represented by a second coefficient having a value less than that of the first coefficient.
11. The system of claim 10 , wherein the second coefficient is derived by applying a series expansion algorithm to the plurality of input variables and the plurality of output variables.
12. The system of claim 11 , further comprising a sensitivity analysis module configured to measure sensitivity of at least one output variable of the plurality of output variables based on change in values of the subset of the reservoir parameters.
13. The system of claim 8 , wherein the assisted history matching module uses Bayesian optimization to adjust values of at least one of the plurality of output variables based on a difference between the at least one of the output variables and dynamic field production data.
14. A method for generating forecasts from an input high-dimensional parameter data space, the method comprising:
generating a reservoir model that includes a plurality of input variables, a plurality of output variables, and at least one algorithmic model, wherein the generation of the reservoir model is performed using a flow simulator and variables from at least one of a formation and reservoir properties database and a field database;
generating, using a model order reduction module, a subset of the plurality of input variables, the subset having a reduced parameter space relative to the input high-dimensional parameter data space by identifying input variables within the plurality of input variables useful to approximate at least one output variable of the plurality of output variables, wherein the model order reduction module uses a function decomposition algorithm to identify input variables within the plurality of input variables useful to approximate the at least one output variable to generate the subset of the plurality of input variables with a reduced number of input variables such that the subset of input variables of the plurality of input variables most effectively approximates the at least one output variable; and
adjusting, using an assisted history matching module, values of at least one of the output variables based on a difference between the at least one of the output variables and dynamic field production data.
15. The method of claim 14 , wherein the flow simulator used for the generation of the reservoir model is at least one of a full-physics flow simulator and a proxy flow simulator.
16. The method of claim 14 , wherein the generation of the subset of input variables is also performed using a design of experiments algorithm, input variables that that are useful to approximate at least one output variable.
17. The method of claim 14 , wherein the plurality of input variables include reservoir parameters represented by a first coefficient and subset is a subset of the reservoir parameters represented by a second coefficient having a value less than that of the first coefficient.
18. The method of claim 17 , further comprising deriving the second coefficient by applying a series expansion algorithm to the input variables and the output variables, and measuring sensitivity of at least one output variable based on change in values of the subset of the reservoir parameters.
19. The method of claim 14 , further comprising applying Bayesian optimization, using assisted history matching, to adjust values of at least one of the output variables based on a difference between the at least one of the plurality of output variables and dynamic field production data.