Geological property modeling with neural network representations

1. A method comprising:

retrieving, via a computerized geological database and one or more processors, original primary data and original secondary data obtained for at least a first subset of a first plurality of locations in a first geological formation;

training, via the one or more processors, a first neural network to generate estimated secondary data at the first plurality of locations and at a plurality of unknown locations in the first geological formation;

replacing the original secondary data with the first trained neural network;

generating estimated primary data using the original primary data and the first trained neural network; and

guiding an oil and gas operation, based, at least in part, on the estimated primary data.

2. The method of claim 1 , further comprising:

replacing first geological property values of the original secondary data with the first trained neural network;

storing the first trained neural network in memory;

associating, in memory, the first trained neural network with the first plurality of locations in the first geological formation;

deleting the first geological property values from memory; and

responding, via the one or more processors, to a query to the computerized geological database for the first geological property values obtained for the first plurality of locations with output from the first trained neural network.

3. The method of claim 2 , wherein the memory comprises memory in cloud storage.

4. The method of claim 2 , further comprising,

retrieving the first trained neural network based, at least in part, on the query, the query indicating one or more of the first plurality of locations in the first geological formation; and

generating first estimated geological property values at a second subset of the first plurality of locations in the first geological formation based, at least in part, on inputting indications of the second subset of the first plurality of locations into the first trained neural network.

5. The method of claim 4 , wherein the indications of the second subset of the first plurality of locations comprises coordinates for the second subset of the first plurality of locations.

6. The method of claim 4 , wherein the query further indicates one or more network characteristics of the first trained neural network.

7. The method of claim 2 , further comprising:

retrieving, via the computerized geological database, indications of the first plurality of locations in the first geological formation and first geological property values,

wherein the training uses the retrieved indications of the first plurality of locations in the first geological formation and the first geological property values obtained for the first plurality of locations, and

wherein replacing the first geological property values obtained for the first plurality of locations with the first trained neural network comprises associating the first trained neural network with the indications of the first plurality of locations in the first geological formation.

8. The method of claim 1 , further comprising training a second neural network to estimate geological property values at a second plurality of locations in a second geological formation using second geological property values obtained for at least a subset of the second plurality of locations in the second geological formation.

9. The method of claim 8 , wherein training the second neural network to estimate the geological property values at the second plurality of locations in the second geological formation is based, at least in part, on a query indicating one or more of the second plurality of locations in the second geological formation.

10. One or more non-transitory machine-readable media comprising program code to:

retrieve, via a computerized geological database and one or more processors, original primary data and original secondary data obtained for at least a first subset of a first plurality of locations in a first geological formation;

train, via the one or more processors, a first neural network to generate estimated secondary data at the first plurality of locations and at a plurality of unknown locations in the first geological formation;

replace the original secondary data with the first trained neural network;

generate estimated primary data using the original primary data and the first trained neural network; and

guide an oil and gas operation based, at least in part, on the estimated primary data.

11. The non-transitory machine-readable media of claim 10 , further comprising program code to:

replace first geological property values of the original secondary data with the first trained neural network;

store the first trained neural network in memory;

associate, in memory, the first trained neural network with the first plurality of locations in the first geological formation;

delete the first geological property values from memory; and

respond, via the one or more processors, to a query to the computerized geological database for the first geological property values obtained for the first plurality of locations with output from the first trained neural network.

12. The non-transitory machine-readable media of claim 11 , wherein the memory comprises memory in cloud storage.

13. The non-transitory machine-readable media of claim 1 , further comprising program code to:

retrieve the first trained neural network based, at least in part, on the query, the query indicating one or more of the first plurality of locations in the first geological formation; and

generate first estimated geological property values at a second subset of the first plurality of locations in the first geological formation based, at least in part, on inputting indications of the second subset of the first plurality of locations into the first trained neural network, wherein the indications of the second subset of the first plurality of locations comprises coordinates for the second subset of the first plurality of locations.

14. The non-transitory machine-readable media of claim 13 , wherein the query further indicates one or more network characteristics of the first trained neural network.

15. The non-transitory machine-readable media of claim 10 , further comprising program code to train a second neural network to estimate geological property values at a second plurality of locations in a second geological formation using second geological property values obtained for at least a subset of the second plurality of locations in the second geological formation.

16. The non-transitory machine-readable media of claim 15 , wherein the program code to train the second neural network to estimate the geological property values at the second plurality of locations in the second geological formation is based, at least in part, on a query indicating one or more of the second plurality of locations in the second geological formation.

17. The non-transitory machine-readable media of claim 10 , further comprising program code to:

retrieve, via the computerized geological database, indications of the first plurality of locations in the first geological formation and first geological property values,

wherein the program code configured to train the first neural network comprises program code to use the retrieved indications of the first plurality of locations in the first geological formation and the first geological property values obtained for the first plurality of locations, and

wherein the program code to replace the first geological property values obtained for the first plurality of locations with the first trained neural network comprises program code to associate the first trained neural network with the indications of the first plurality of locations in the first geological formation.

18. An apparatus comprising:

one or more processors; and

a machine-readable medium having program code executable by the processor to cause the apparatus to, retrieve, via a computerized geological database and the one or more processors, original primary data and original secondary data obtained for at least a first subset of a first plurality of locations in a first geological formation;

train, via the one or more processors, a first neural network to generate estimated secondary data at the first plurality of locations and at a plurality of unknown locations in the first geological formation;

replace the original secondary data with the first trained neural network;

generate estimated primary data using the original primary data and the first trained neural network; and

guide an oil and gas operation based, at least in part, on the estimated primary data.

19. The apparatus of claim 18 , further comprising program code executable by the one or more processors to cause the apparatus to:

process, via the computerized geological database, a query indicating at least the first plurality of locations in the first geological formation at which to estimate geological property values;

retrieve, based on the query, the first trained neural network that was trained to estimate geological property values at a second plurality of locations in the first geological formation, wherein the second plurality of locations at least include one or more of the first plurality of locations; and

generate, via the computerized geological database, first estimated geological property values at the first plurality of locations in the first geological formation based, at least in part, on inputting indications of the first plurality of locations into the first trained neural network,

wherein the query further indicates one or more network characteristics of the first trained neural network.

20. The apparatus of claim 19 , further comprising program code executable by the one or more processors to cause the apparatus to train a second neural network to estimate geological property values at the first plurality of locations in the first geological formation based, at least in part, on an absence of a trained neural network satisfying the query.