Source determination of produced water from oilfields with artificial intelligence techniques

1 . A method, comprising:

collecting a first geochemical data set comprising geochemical data for a first plurality of produced water samples;

collecting a second plurality of produced water samples from an oil and gas field;

performing a geochemical analysis on the second plurality of produced water samples to form a second geochemical data set, wherein the second geochemical data set comprises geochemical data for the second plurality of produced water samples;

combining the first geochemical data set and the second geochemical data set into a database;

determining, by a subject matter expert, a water type for each produced water sample in the database;

training a machine-learned model on at least a portion of the database to predict the water type of a produced water sample given its geochemical data;

collecting a third plurality of produced water samples from a well of the oil and gas field that penetrates a subsurface of the oil and gas field;

performing the geochemical analysis on the third plurality of produced water samples to form a third geochemical data set, wherein the third geochemical data set comprises geochemical data for the third plurality of produced water samples;

determining, with the trained machine-learned model, the water type for each produced water sample in the third plurality of produced water samples using the third geochemical data set;

generating a regional salinity map for the subsurface based on the determined water type for each produced water sample in the third plurality of produced water samples; and

adjust a production operation of the well based on the regional salinity map.

2 . The method of claim 1 , further comprising:

augmenting geochemical data with derived data; and

pre-processing the geochemical data.

3 . The method of claim 1 , wherein the machine-learned model is a supervised gradient boosting classifier.

4 . The method of claim 1 , wherein the machine-learned model is a semi-supervised gradient boosting classifier.

5 . The method of claim 1 , further comprising:

selecting a machine-learned model type and hyperparameters;

evaluating the trained machine-learned model;

adjusting the machine-learned model hyperparameters; and

re-training the trained machine learned model with the adjusted machine-learned model hyperparameters.

6 . The method of claim 1 , further comprising estimating the generalization error of the machine-learned model.

7 . The method of claim 1 , wherein the water types comprise: brine, formation water, mixed, mud filtrate, condensate water, and supply water.

8 . The method of claim 1 , the well having been previously drilled using a drilling system, wherein drilling the well comprised injecting a drilling fluid into the subsurface.

9 . A non-transitory computer readable medium storing instructions executable by a computer processor, the instruction comprising functionality for:

receiving a first geochemical data set comprising geochemical data for a first plurality of produced water samples;

receiving a second geochemical data set for a second plurality of produced water samples from an oil and gas field, wherein the second geochemical data set comprises geochemical data for the second plurality of produced water samples;

combining the first geochemical data set and the second geochemical data set into a database;

receiving a water type for each produced water sample in the database, wherein the water type is determined by a subject matter expert;

training a machine-learned model on at least a portion of the database to predict the water type of a produced water sample given its geochemical data;

receiving a third geochemical data set for a third plurality of produced water samples from a well of the oil and gas field that penetrates a subsurface of the oil and gas field wherein the third geochemical data set comprises geochemical data for the third plurality of produced water samples;

determining, with the trained machine-learned model, the water type for each produced water sample in the third plurality of produced water samples using the third geochemical data set;

generating a regional salinity map for the subsurface based on the determined water type for each produced water sample in the third plurality of produced water samples; and

commanding an adjustment to a production operation of the well based on the regional salinity map.

10 . The non-transitory computer readable medium of claim 9 , the instructions further comprising functionality for:

augmenting geochemical data with derived data; and

pre-processing the geochemical data.

11 . The non-transitory computer readable medium of claim 9 , wherein the machine-learned model is a supervised gradient boosting classifier.

12 . The non-transitory computer readable medium of claim 9 , wherein the machine-learned model is a semi-supervised gradient boosting classifier.

13 . The non-transitory computer readable medium of claim 9 , the instructions further comprising functionality for:

selecting a machine-learned model type and architecture;

evaluating the trained machine-learned model;

adjusting the machine-learned model architecture; and

re-training the trained machine learned model with the adjusted machine-learned model architecture.

14 . The non-transitory computer readable medium of claim 9 , wherein the machine-learned model determines that the water type is one of: brine, formation water, mixed, mud filtrate, condensate water, and supply water.

15 . A system, comprising:

a first geochemical data set comprising geochemical data for a first plurality of produced water samples;

a second plurality of produced water samples from an oil and gas field;

a second geochemical data set, wherein the second geochemical data set comprises geochemical data for the second plurality of produced water samples;

a third plurality of produced water samples having been acquired from a well of the oil and gas field that penetrates a subsurface of the oil and gas field;

a third geochemical data set, wherein the third geochemical data set comprises geochemical data for the third plurality of produced water samples;

a plurality of water types, wherein the plurality of water types comprises a water type for each produced water sample in both the first and second pluralities of produced water samples, and wherein the water type for each produced water sample was determined by a subject matter expert;

a trained machine-learned model; and

a computer, the computer comprising:

one or more computer processors, and

a non-transitory computer readable medium storing instructions executable by a

computer processor, the instructions comprising functionality for:

processing, with the trained machine-learned model, the third geochemical data set to determine a water type for each produced water sample in the third plurality of produced water samples,

generating a regional salinity map for the subsurface based on the determined water type for each produced water sample in the third plurality of produced water samples, and

commanding an adjustment to a production operation of the well based on the regional salinity map.

16 . The system of claim 15 , wherein the instructions further comprising functionality for:

augmenting geochemical data with derived data; and

pre-processing the geochemical data.

17 . The system of claim 15 , wherein the trained machine-learned model is a supervised gradient boosting classifier.

18 . The system of claim 15 , wherein the trained machine-learned model is a semi-supervised gradient boosting classifier.

19 . The system of claim 15 , wherein the trained machine-learned model determines that the water type is one of: brine, formation water, mixed, mud filtrate, condensate water, and supply water.

20 . The system of claim 15 , further comprising a drilling system, wherein the well having been previously drilled using the drilling system, wherein drilling the well comprised injecting a drilling fluid into the subsurface.