Granted Patent B2

US 12,529,690 · App. 18/597,525 · Granted Jan 20, 2026

Inventors: Nazli Demirer (Tomball, TX); Conrad Teran (Tomball, TX); Alejandra Inmaculada Lopez Trosell (Cypress, TX); Chunli Li (The Woodlands, TX); Robert P. Darbe (Tomball, TX)

Assignee: Halliburton Energy Services, Inc.

G01N33/2823G01N33/2835G06N20/00

Abstract

Claims (64)

1 . A method comprising:

generating first conditions for a first bottle test using a trained machine-learning model;

performing the first bottle test using the first conditions to obtain a first result including one or more key performance indicators;

feeding the first result to the trained machine-learning model to generate additional conditions for an additional bottle test, wherein the additional conditions are different from the first conditions;

performing the additional bottle test, using the additional conditions to obtain an additional result including one or more key performance indicators;

determining that the one or more key performance indicators of the additional result are within a threshold limit of one or more reference key performance indicators;

outputting the additional conditions used to obtain the additional result;

performing a second bottle test with an incumbent emulsion breaker formulation using the additional conditions to obtain a second result including one or more key performance indicators;

feeding the second result to a second trained machine-learning model to generate a revised emulsion breaker formulation for a new bottle test, wherein the revised emulsion breaker formulation is different from the incumbent emulsion breaker formulation;

performing the new bottle test with a revised emulsion breaker formulation using the additional conditions to obtain a new result including one or more key performance indicators;

determining that the one or more key performance indicators of the new result are within a threshold limit of one or more incumbent key performance indicators; and

outputting the revised emulsion breaker formulation.

2 . The method of claim 1 , further comprising:

repeating feeding results to the trained machine-learning model and performing additional bottle tests until the key performance indicators are within the threshold limit of the one or more reference key performance indicators.

3 . The method of claim 1 , wherein the key performance indicators include a water drop rate, an interface quality, a water clarity, a basic sediments and water metric, an unresolved emulsion metric, or any combination of these.

4 . The method of claim 1 , wherein the first conditions and the additional conditions include a concentration of demulsifier, a bottle test temperature, a bottle test agitation energy, a bottle test mechanical energy, an agitation energy, a mechanical energy, a volume of an emulsion sample used in the bottle test, a volume of free water used in the bottle test, and any combination of these.

5 . The method of claim 1 , wherein the threshold limit corresponds to values of 10% different or less to the one or more reference key performance indicators.

6 . The method of claim 1 , wherein the trained machine-learning model performs an optimization algorithm that emulates bottle test procedures to predict desired key performance indicators based on a set of predictors that describe emulsion characteristics, test conditions, and emulsion breaker formulation and dosage.

7 . The method of claim 1 , further comprising:

modifying a production emulsion breaker formulation based on the revised emulsion breaker formulation to generate a modified production emulsion breaker formulation; and

using the modified production emulsion breaker formulation for demulsifying produced crude oil.

8 . The method of claim 1 , further comprising:

feeding a set of constraints to the second trained-machine learning model, wherein the constraints include at least one of reducing a cost of an emulsion breaker formulation, limiting maximum concentrations of one or more chemicals in the emulsion breaker formulation, requiring minimum concentrations of one or more chemicals in the emulsion breaker formulation, and wherein the second trained machine-learning model uses the set of constraints when generating emulsion breaker formulations.

9 . The method of claim 1 , wherein the second trained machine-learning model performs an optimization algorithm that emulates bottle test procedures to predict desired key performance indicators based on a set of predictors that describe emulsion characteristics, test conditions, and emulsion breaker formulation and dosage.

10 . The method of claim 1 , wherein the trained machine-learning model and the second trained machine-learning model are a same machine-learning model configured to operate to generate bottle test conditions and emulsion breaker formulations based on inputs including one or more of target key performance indicators, one or more constraints, a crude oil or emulsion characteristics, or any combination of these.

11 . A system comprising:

a processor; and

a non-transitory computer-readable medium comprising instructions that are executable by the processor to cause the processor to perform operations including:

receiving input corresponding to one or more reference key performance indicators;

generating first conditions for a first bottle test using a trained machine-learning model;

outputting the first conditions for the first bottle test for use in generating a first result including one or more key performance indicators;

receiving input corresponding to the first result;

generating additional conditions for an additional bottle test using the trained machine-learning model and the first result;

outputting the additional conditions for the additional bottle test for use in generating an additional result including one or more key performance indicators, wherein the additional conditions are different from the first conditions;

receiving input corresponding to the additional result;

determining that the one or more key performance indicators of the additional result are within a threshold limit of one or more reference key performance indicators;

outputting an indicator that the additional conditions used to obtain the additional result are within a threshold limit of one or more reference key performance indicators;

receiving input corresponding to second results including one or more key performance indicators, wherein the second results are for a second bottle test performed using an incumbent emulsion breaker formulation and the additional conditions;

generating a revised emulsion breaker formulation using a second trained machine-learning model, wherein the revised emulsion breaker formulation is different from the incumbent emulsion breaker formulation;

receiving input corresponding to new results including one or more key performance indicators for a new bottle test performed using the revised emulsion breaker formulation and the additional conditions;

determining that the one or more key performance indicators of the new result are within a threshold limit of one or more incumbent key performance indicators; and

outputting the revised emulsion breaker formulation.

12 . The system of claim 11 , wherein the key performance indicators include a water drop rate, an interface quality, a water clarity, a basic sediments and water metric, an unresolved emulsion metric, or any combination of these.

13 . The system of claim 11 , wherein the first conditions and the additional conditions include a concentration of demulsifier, a bottle test temperature, a bottle test agitation energy, a bottle test mechanical energy, an agitation energy, a mechanical energy, a volume of an emulsion sample used in the bottle test, a volume of free water used in the bottle test, and any combination of these.

14 . The system of claim 11 , wherein the threshold limit corresponds to values of 10% different or less to the one or more reference key performance indicators.

15 . The system of claim 11 , wherein the trained machine-learning model performs an optimization algorithm that emulates bottle test procedures to predict desired key performance indicators based on a set of predictors that describe emulsion characteristics, test conditions, and emulsion breaker formulation and dosage.

16 . The system of claim 11 , wherein the operations further include:

providing a set of constraints to the second trained-machine learning model, wherein the constraints include at least one of reducing a cost of an emulsion breaker formulation, limiting maximum concentrations of one or more chemicals in the emulsion breaker formulation, requiring minimum concentrations of one or more chemicals in the emulsion breaker formulation, and wherein the second trained machine-learning model uses the set of constraints when generating emulsion breaker formulations.

17 . The system of claim 11 , wherein the second trained machine-learning model performs an optimization algorithm that emulates bottle test procedures to predict desired key performance indicators based on a set of predictors that describe emulsion characteristics, test conditions, and emulsion breaker formulation and dosage.

18 . A computer program product tangibly embodied on non-transitory computer-readable storage medium and comprising instructions that, when executed by a processor, cause the processor to perform operations including:

receiving input corresponding to one or more reference key performance indicators;

generating first conditions for a first bottle test using a trained machine-learning;

outputting the first conditions for the first bottle test for use in generating a first result including one or more key performance indicators;

receiving input corresponding to the first result;

generating additional conditions for an additional bottle test using the trained machine-learning model and the first result;

outputting the additional conditions for the additional bottle test for use in generating an additional result including one or more key performance indicators, wherein the additional conditions are different from the first conditions;

receiving input corresponding to the additional result;

determining that the one or more key performance indicators of the additional result are within a threshold limit of one or more reference key performance indicators;

outputting an indicator that the additional conditions used to obtain the additional result are within a threshold limit of one or more reference key performance indicators;

receiving input corresponding to second results including one or more key performance indicators, wherein the second results are for a second bottle test performed using an incumbent emulsion breaker formulation and the additional conditions;

generating a revised emulsion breaker formulation using a second trained machine-learning model, wherein the revised emulsion breaker formulation is different from the incumbent emulsion breaker formulation;

receiving input corresponding to new results including one or more key performance indicators for a new bottle test performed using the revised emulsion breaker formulation and the additional conditions;

determining that the one or more key performance indicators of the new result are within a threshold limit of one or more incumbent key performance indicators; and

outputting the revised emulsion breaker formulation.

Assignments (2)

Continuity (1)

References Cited (19)