Fluid type identification from downhole fluid analysis using machine learning techniques
Embodiments present a method for fluid type identification from a downhole fluid analysis that uses machine learning techniques that are trained and derived from a computer model using pressure, temperature and downhole optical characteristics of sampled fluid.
1. A method, comprising:
collecting optical spectral data for a downhole fluid of a wellbore;
providing the collected optical spectral data to a trained classification module;
processing the collected optical spectral data with the trained classification module configured to determine a fluid type classification based upon an identified maximum separation boundary between fluid classes and comparing at least one of pressure or temperature data for the downhole fluid to the identified maximum separation boundary;
determining a fluid type based upon the fluid type classification based upon the trained classification module; and
controlling components associated with a tool-string in the wellbore based on the determined fluid type.
2. The method according to claim 1 , wherein the optical spectral data is collected from a downhole system.
3. The method according to claim 2 , wherein the downhole system is a downhole fluid analysis module.
4. The method according to claim 1 , wherein the optical spectral data is collected from a downhole probe.
5. The method according to claim 1 , further comprising:
collecting additional data from a downhole fluid analysis module prior to collecting at least one of pressure or temperature data.
6. The method according to claim 1 , wherein the fluid type classification is determined through a support vector machine.
7. The method according to claim 1 , wherein the collected optical spectral data includes at least one of a pressure or a temperature of the downhole fluid.
8. The method according to claim 1 , further comprising:
collecting optical data related to the downhole fluid;
extracting data from the collected optical data to form a qualified dataset;
partitioning the qualified dataset to determine a portion of the qualified dataset related to testing data and a second portion of the qualified dataset related to training data; and
performing a classification model training on the second portion of the qualified dataset related to the training data, resulting in the trained classification model.
9. The method according to claim 8 , further comprising performing an output validation after performing the classification model training.
10. The method according to claim 8 , further comprising:
obtaining external data from an external data source related to fluid properties.
11. The method according to claim 10 , wherein the obtaining the external data from the external data source is from a downhole fluid analysis module.
12. The method according to claim 11 , wherein the downhole fluid analysis module includes a filter array spectrometer, a grating spectrometer and a resistivity sensor.
13. The method according to claim 11 , wherein the downhole fluid analysis module includes one of a density sensor, a viscosity sensor, a pressure gauge, or a temperature gauge.
14. The method according to claim 10 , wherein the classification model is categorically variable.