Predicting well site failure modes and times using machine learning analytics and dynacard classifications
Systems and methods for real-time monitoring and control of well operations at a well site use machine learning (ML) based analytics at the well site. The systems and methods perform ML-based analytics on data from the well site via an edge device directly at the well site to detect operations that fall outside expected norms and automatically respond to such abnormal operations. The edge device can issue alerts regarding the abnormal operations and take predefined steps to reduce potential damage resulting from such abnormal operations. The edge device can also anticipate failures and a time to failure by performing ML-based analytics on operations data from the well site using normal operations data. This can help decrease downtime and minimize lost productivity and cost as well as reduce health and safety risks for field personnel.
1. A method of anticipating failure modes at a well site, comprising:
receiving dynacards from a remote terminal unit (RTU) at the well site;
performing machine learning (ML) based analytics on an edge device at the well site using the dynacards from the RTU;
identifying dynacard classifications on the edge device for the well site operations from the ML-based analytics performed on the dynacards;
determining on the edge device at the well site that a predefined number of dynacards having a specified dynacard classification has been identified;
performing ML-based analytics on the edge device at the well site using the predefined number of dynacards having the specified dynacard classification;
predicting a failure mode and a time to failure on the edge device at the well site from the ML-based analytics performed on the predefined number of dynacards having the specified dynacard classification; and
initiating a responsive action on the edge device based on the failure mode and the time to failure, the responsive action including at least one of logging a date and time, sending an alert message to a control system, adjusting a motor speed of a rod pump, or shutting off power to the rod pump, depending on a severity of the failure mode and the time to failure.
2. The method of claim 1 , wherein predicting a failure mode and a time to failure comprises predicting a dynacard shape associated with the failure mode on the edge device at the well site from the ML-based analytics performed on the predefined number of dynacards and the specified dynacard classification.
3. The method of claim 1 , wherein the predefined number of dynacards comprises a predefined number of consecutive dynacards.
4. The method of claim 3 , wherein the specified dynacard classification is a normal classification.
5. The method of claim 1 , wherein predicting a failure mode comprises predicting one of: a fluid pound failure or a gas interference failure.
6. The method of claim 1 , wherein performing ML-based analytics using the dynacard classifications includes inputting the dynacard classifications into one or more ML models.
7. The method of claim 6 , wherein the one or more ML models are trained using historical data, further comprising generating augmented training data using the historical data.
8. An edge device installed at a well site and operable to anticipate failure modes at the well site, comprising:
a processor, and
a storage device coupled to the processor and storing computer-readable instructions for performing a failure anticipation function thereon;
wherein the failure anticipation function, when executed by the processor, causes the edge device to:
obtain dynacards and dynacard classifications for the well site;
identify a predefined number of dynacards having a specified dynacard classification;
perform ML-based analytics using the predefined number of dynacards having the specified dynacard classifications classification;
predict a failure mode and a time to failure from the ML-based analytics performed on the predefined number of dynacards having the specified dynacard classification; and
initiate a responsive action on the edge device based on the failure mode and the time to failure, the responsive action including at least one of logging a date and time, sending an alert message to a control system, adjusting a motor speed of a rod pump, or shutting off power to the rod pump, depending on a severity of the failure mode and the time to failure.
9. The edge device of claim 8 , wherein the failure anticipation function further causes the edge device to predict a dynacard shape associated with the failure mode from the ML-based analytics performed on the predefined number of dynacards having the specified dynacard classification.
10. The edge device of claim 8 , wherein the predefined number of dynacards comprises a predefined number of consecutive dynacards.
11. The edge device of claim 10 , wherein the specified dynacard classification is a normal classification.
12. The edge device of claim 8 , wherein the failure anticipation function causes the edge device to predict a failure mode by predicting one of: a fluid pound failure or a gas interference failure.
13. The edge device of claim 8 , wherein the failure anticipation function causes the edge device to perform ML-based analytics by inputting the dynacard classifications into one or more ML models.
14. The edge device of claim 13 , wherein the one or more ML models are trained using historical data and augmented training data generated using the historical data.
15. A method of anticipating failure modes at a well site, comprising:
obtaining dynacards and dynacard classifications for the well site on the edge device at the well site;
identifying on the edge device at the well site a predefined number of dynacards having a specified dynacard classification;
performing ML-based analytics on the edge device at the well site using the predefined number of dynacards having the specified dynacard classification;
predicting a failure mode and a time to failure on the edge device at the well site from the ML-based analytics performed on the predefined number of dynacards having the specified dynacard classification; and
initiating a responsive action on the edge device based on the failure mode and the time to failure, the responsive action including at least one of logging a date and time, sending an alert message to a control system, adjusting a motor speed of a rod pump, or shutting off power to the rod pump, depending on a severity of the failure mode and the time to failure.
16. The method of claim 15 , wherein predicting a failure mode and a time to failure comprises predicting a dynacard shape associated with the failure mode on the edge device at the well site from the ML-based analytics performed on the predefined number of dynacards having the specified dynacard classification.
17. The method of claim 15 , wherein the predefined number of dynacards is a predefined number of consecutive dynacards.
18. The method of claim 17 , wherein the specified dynacard classification is a normal classification.
19. The method of claim 15 , wherein predicting a failure mode comprises predicting one of: a fluid pound failure or a gas interference failure.
20. The method of claim 15 , wherein performing ML-based analytics using the dynacard classifications includes inputting the dynacard classifications into one or more ML models.