Methods and systems for predicting fluid type while geosteering
A method for predicting fluid properties includes measuring one or more measured fluid properties of a mud gas at a surface of a wellbore. The method also includes measuring one or more mud gas properties of the mud gas at the surface of the wellbore. The method also includes predicting one or more first predicted fluid properties using one or more pre-trained machine-learning (ML) models. The one or more first predicted fluid properties are predicted based at least partially upon the one or more mud gas properties. The method also includes comparing the one or more measured fluid properties to the one or more first predicted fluid properties. The method also includes re-training the one or more pre-trained ML models to produce one or more re-trained ML models in response to the comparison.
1. A method for predicting fluid properties, the method comprising:
measuring one or more measured fluid properties of a mud gas at a surface of a wellbore;
measuring one or more mud gas properties of the mud gas at the surface of the wellbore;
predicting one or more first predicted fluid properties using one or more pre-trained machine-learning (ML) models, wherein the one or more first predicted fluid properties are predicted based at least partially upon the one or more mud gas properties;
comparing the one or more measured fluid properties to the one or more first predicted fluid properties;
re-training the one or more pre-trained ML models to produce one or more re-trained ML models in response to the comparison;
predicting one or more second predicted fluid properties using the one or more re-trained ML models, the one or more second predicted fluid properties comprise a fluid type, a fluid composition, a gas-oil ratio (GOR), or a combination thereof;
comparing the one or more measured fluid properties to the one or more second predicted fluid properties; and
determining a geosteering response for a downhole tool that is drilling the wellbore in response to the comparison being less than a predetermined threshold.
2. The method of claim 1 , wherein the one or more measured fluid properties and the one or more first predicted fluid properties comprise a fluid type, a fluid composition, a gas-oil ratio (GOR), or a combination thereof.
3. The method of claim 1 , wherein the one or more measured fluid properties comprise a fluid composition, and wherein the fluid composition comprises a relative amount of organic molecules with six or more carbons in the mud gas.
4. The method of claim 1 , wherein the one or more mud gas properties comprise a composition of molecules in the mud gas, a pressure of the mud gas, a temperature of the mud gas, or a combination thereof.
5. The method of claim 1 , wherein the one or more measured fluid properties and the one or more mud gas properties are measured while the downhole tool that is drilling the wellbore is in a deviated portion or a lateral portion of the wellbore.
6. The method of claim 1 , wherein the one or more re-trained ML models are produced in response to the comparison being greater than the predetermined threshold.
7. The method of claim 1 , further comprising generating a signal with geosteering instructions for the downhole tool that is drilling the wellbore based at least partially upon the one or more re-trained ML models.
8. The method of claim 1 , further comprising geosteering the downhole tool that is drilling the wellbore within a subterranean formation based at least partially upon the one or more re-trained ML models.
9. A non-transitory computer-readable medium storing instructions that, when executed by at least one processor of a computing system, cause the computing system to perform operations, the operations comprising:
measuring one or more measured fluid properties of a mud gas at a surface of a wellbore, wherein the one or more measured fluid properties comprise a fluid type, a fluid composition, a gas-oil ratio (GOR), or a combination thereof, and wherein the wellbore comprises a deviated or lateral wellbore;
measuring one or more mud gas properties of the mud gas at the surface of the wellbore, wherein the one or more mud gas properties comprise a composition of molecules in the mud gas, a pressure of the mud gas, a temperature of the mud gas, or a combination thereof;
predicting one or more first predicted fluid properties using one or more pre-trained machine-learning (ML) models, wherein the one or more first predicted fluid properties are predicted based at least partially upon the one or more mud gas properties, and wherein the one or more first predicted fluid properties comprise the fluid type, the fluid composition, and the GOR;
performing a first comparison of the one or more measured fluid properties to the one or more first predicted fluid properties;
re-training the one or more pre-trained ML models to produce one or more re-trained ML models in response to the first comparison being greater than a first predetermined threshold;
predicting one or more second predicted fluid properties using the one or more re-trained ML models, wherein the one or more second predicted fluid properties comprise the fluid type, the fluid composition, the GOR, or a combination thereof;
performing a second comparison of the one or more measured fluid properties to the one or more second predicted fluid properties; and
determining a geosteering response for a downhole tool that is drilling the wellbore in response to the second comparison being less than the first predetermined threshold.
10. The non-transitory computer-readable medium of claim 9 , further comprising converting the one or more measured fluid properties into a converted data format, wherein the one or more measured fluid properties in the converted data format are compared to the one or more first predicted fluid properties and to the one or more second predicted fluid properties.
11. The non-transitory computer-readable medium of claim 9 , further comprising generating a signal with geosteering instructions for the downhole tool based at least partially upon the geosteering response.
12. The non-transitory computer-readable medium of claim 9 , wherein the re-training is performed using a first portion of a combined dataset, and further comprising testing the one or more re-trained ML models using a second portion of the combined dataset, wherein the combined dataset comprises the one or more measured fluid properties and one or more training fluid properties that were used to train the one or more pre-trained ML models.
13. The non-transitory computer-readable medium of claim 12 , wherein the first portion is greater than the second portion.
14. A computing system, comprising:
one or more processors; and
a memory system comprising one or more non-transitory computer-readable media storing instructions that, when executed by at least one of the one or more processors, cause the computing system to perform operations, the operations comprising:
measuring one or more measured fluid properties of a mud gas at a surface of a wellbore, wherein the one or more measured fluid properties comprise a fluid type, a fluid composition, a gas-oil ratio (GOR), or a combination thereof, and wherein the wellbore comprises a deviated or lateral wellbore;
comparing the one or more measured fluid properties with one or more corresponding measured fluid properties that are measured in a pilot well;
converting the one or more measured fluid properties, the one or more corresponding measured fluid properties, or a combination thereof into a converted data format;
measuring one or more mud gas properties of the mud gas at the surface of the wellbore, wherein the one or more mud gas properties comprise a composition of molecules in the mud gas, a pressure of the mud gas, a temperature of the mud gas, or a combination thereof;
normalizing the one or more mud gas properties to produce one or more normalized mud gas properties;
predicting one or more first predicted fluid properties using one or more pre-trained machine-learning (ML) models, wherein the one or more first predicted fluid properties are predicted based at least partially upon the one or more normalized mud gas properties, wherein the one or more first predicted fluid properties comprise the fluid type, the fluid composition, the GOR, or a combination thereof;
performing a first comparison of the one or more measured fluid properties in the converted data format to the one or more first predicted fluid properties;
re-training the one or more pre-trained ML models to produce one or more re-trained ML models in response to the first comparison being greater than a first predetermined threshold;
predicting one or more second predicted fluid properties using the one or more re-trained ML models, wherein the one or more second predicted fluid properties comprise the fluid type, the fluid composition, the GOR, or a combination thereof;
performing a second comparison of the one or more measured fluid properties in the converted data format to the one or more second predicted fluid properties; and
determining a geosteering response for a downhole tool that is drilling the wellbore in response to the second comparison being less than the first predetermined threshold.
15. The computing system of claim 14 , wherein the one or more pre-trained ML models comprises a first pre-trained ML model for the fluid type, a second pre-trained ML model for the fluid composition, and a third pre-trained ML model for the GOR.
16. The computing system of claim 14 , further comprising transmitting a signal with geosteering instructions to the downhole tool based at least partially upon the geosteering response.
17. The computing system of claim 14 , wherein the re-training is performed using a first portion of a combined dataset, and wherein the combined dataset comprises the one or more measured fluid properties in the converted data format and one or more training fluid properties that were used to train the one or more pre-trained ML models.
18. The computing system of claim 17 , further comprising testing the one or more re-trained ML models comprising:
generating one or more outputs using the one or more re-trained ML models based upon one or more inputs, wherein the one or more inputs comprise a second portion of the combined dataset; and
comparing the one or more outputs with a ground truth of the second portion of the combined dataset, wherein the one or more second predicted fluid properties are predicted in response to the comparison of the one or more outputs with the ground truth being within a second predetermined threshold.