System and method for failure detection for artificial lift systems

• US 9,280,517 B2
• Filed: 06/22/2012
• Issued: 03/08/2016
• Est. Priority Date: 06/23/2011
• Status: Active Grant

• Current Assignee: University of Southern California
• Original Assignee: University of Southern California
• Inventors: Liu, Shuping, Raghavendra, Cauligi Srinivasa, Liu, Yintao, Yao, Ke-Thia, Balogun, Oluwafemi Opeyemi, Olabinjo, Olanrewaju, Chinnapparaja Gunasekaran, Dinesh Babu
• Primary Examiner(s): Gaffin, Jeffrey A
• Assistant Examiner(s): Smith, Paulinho E
• Application Number:
  US13/531,127
• Publication Number:
  US 20130080117A1
• Time in Patent Office: 1,355 Days
• Field of Search: None
• US Class Current: 1/1
• CPC Class Codes: E21B 44/00 Automatic control systems specially adapted for drilling operations, i.e. self-operating systems which function to carry out or modify a drilling operation without intervention of a human operator, e.g. computer-controlled drilling systems; Systems specially adapted for monitoring a plurality of drilling variables or conditions
• CPC Class Codes: E21B 47/009 Monitoring of walking-beam pump systems
• CPC Class Codes: G06F 17/10 Complex mathematical operations function generation by table look-up G06F1/03; evaluation of elementary functions by calculation G06F7/544

Abstract

A computer-implemented artificial lift detection system, method, and software are provided for failure detection for artificial lift systems, such as sucker rod pump systems. The method includes providing artificial lift system data from an artificial lift system. Attributes are extracted from the artificial lift system data. Data mining techniques are applied to the attributes to determine whether the artificial lift system is detected to fail within a given time period. An alert is output indicative of impending artificial lift system failures.

Claims

1. A method of monitoring a plurality of artificial lift systems in a hydrocarbon reservoir, the method being implemented on a computer system, the method comprising:

monitoring a plurality of attributes of the plurality of artificial lift systems to collect a data set characterizing performance of the plurality of artificial lift systems, each one of the plurality of artificial lift systems having a plurality of attributes;

applying dimensionality reduction techniques to the collected data set to project the collected data set into a reduced dimensional space;

clustering the collected data set in the reduced dimensional space to generate a labeled training data set;

wherein a portion of the collected data having the reduced dimension form clusters that contain known failures collected during a period of time are used to label another portion of collected data during another period of time from the same clusters as containing failures to generate the labeled training data set;

applying a machine learning algorithm using the labeled training data set to boost learning and build a failure detection model for the plurality of artificial lift systems;

applying the failure detection model to another collected data set from the plurality of artificial lift systems; and

automatically generating an alert in response to identification of a data value in said another collected data set from the plurality of artificial lift systems indicating a fault in an artificial lift system in the plurality of artificial lift systems that generated the data value indicating the fault.

2. The method of claim 1, further comprising pre-processing the collected data prior to the applying the machine learning algorithm.

3. The method of claim 2, wherein the pre-processing comprises applying a noise-reducing algorithm to the collected data.

4. The method of claim 1, wherein the applying the machine learning algorithm comprises applying a Bayesian learning algorithm or a decision tree learning algorithm, or both.

5. The method of claim 1, wherein the monitoring comprises measuring attributes collected from pump off controllers of an array of sucker rod pumps located in the hydrocarbon reservoir.

6. The method of claim 1, further comprising partitioning into groups, the groups being ranked in accordance with a metric based on relevancy to failure prediction.

7. The method of claim 6, wherein the metric is further based on data quality.

8. The method of claim 1, wherein the applying the machine learning algorithm comprises:

constructing a classifier as a linear combination of weak classifiers, and

increasing weight of misclassified training examples and decreasing weight of correctly classified training examples.

9. The method of claim 1, wherein, when an alert is generated, a look forward-backward process is executed as a verification of the alert condition.

10. A system of monitoring a plurality of artificial lift systems in a hydrocarbon reservoir, comprising:

a processor configured and arranged to receive a plurality of attributes characterizing performance of the plurality of artificial lift systems, each one of the plurality of artificial lift systems having a plurality of attributes;

the processor being programmed with machine executable instructions for;

applying dimensionality reduction techniques to the collected data set to project the collected data set into a reduced dimensional space;

clustering the collected data set in the reduced dimensional space to generate a labeled training data set;

wherein a portion of the collected data in the reduced dimensional space form clusters that contain known failures collected during a period of time are used to label another portion of collected data during another period of time from the same clusters as containing failures to generate the labeled training data set;

applying a machine learning algorithm using the labeled training data set to boost learning and build a failure detection model for the plurality of artificial lift systems; and

applying the failure detection model to another collected data set from the plurality of artificial lift systems; and

the processor being further programmed with machine executable instructions for automatically generating an alert in response to identification of a data value in said another collected data set from the plurality of artificial lift systems indicating a fault in an artificial lift system in the plurality of artificial lift systems that generated the data value indicating the fault.

11. The system of claim 10, further comprising a plurality of detectors, configured and arranged to monitor the attributes of the artificial lift system, and in communication with a database for collecting the monitored attributes, the database being further in communication with the processor.

13. The system of claim 10, wherein the processor is further being programmed with machine executable instructions for pre-processing the collected data prior to the applying the machine learning algorithm.

14. The system of claim 13, wherein the pre-processing comprises applying a noise-reducing algorithm to the collected data.

15. The system of claim 10, wherein the applying the machine learning algorithm comprises applying a Bayesian learning algorithm or a decision tree learning algorithm, or both.

16. The system of claim 10, wherein the monitoring comprises measuring attributes collected from pump off controllers of an array of sucker rod pumps located in the hydrocarbon reservoir.

17. The system of claim 10, wherein the processor is further being programmed with machine executable instructions for partitioning into groups, the groups being ranked in accordance with a metric based on relevancy to failure prediction.

18. The system of claim 17, wherein the metric is further based on data quality.

19. The system of claim 10, wherein the applying the machine learning algorithm comprises:

constructing a classifier as a linear combination of weak classifiers, and

increasing weight of misclassified training examples and decreasing weight of correctly classified training examples.

20. The system of claim 10, wherein, when an alert is generated, a look forward-backward process is executed as a verification of the alert condition.

12. A non-transitory tangible machine readable medium comprising instructions for executing machine executable instructions for performing a method comprising:

monitoring a plurality of attributes of a plurality of artificial lift systems to collect a data set characterizing performance of the artificial lift system, each one of the plurality of artificial lift systems having a plurality of attributes;

applying dimensionality reduction techniques to the collected data set to project the collected data set into a reduced dimensional space;

clustering the collected data set in the reduced dimensional space to generate a labeled training data set;

wherein a portion of the collected data in the reduced dimensional space form clusters that contain known failures collected during a period of time are used to label another portion of collected data during another period of time from the same clusters as containing failures to generate the labeled training data set;

applying a machine learning algorithm using the labeled training data set to boost learning and build a failure detection model for the plurality of artificial lift systems;

applying the failure detection model to another collected data set from the plurality of artificial lift systems; and

automatically generating an alert in response to identification of a data value in said another collected data set from the plurality of artificial lift systems indicating a fault in an artificial lift system in the plurality of artificial lift systems that generated the data value indicating the fault.