Towards Managing Uncertain Geo-Information for Drilling Disasters Using Event Tracking Sensitivity Analysis.

Area of Science:

• Petroleum Engineering
• Data Science
• Geophysics

Background:

• Sub-surface drilling faces critical risks from unwanted borehole influx, especially in high-pressure gas formations.
• Existing risk assessments have limitations due to unpredictable formation behaviors, necessitating advanced prediction methods.

Purpose of the Study:

• To develop a computationally efficient method for real-time prediction and prevention of drilling crises.
• To reduce the computational overhead of input feature analysis for time-critical predictions.

Main Methods:

• Applied EventTracker sensitivity analysis, an incremental dimensionality reduction method, to real-world drilling data.
• Utilized 1600 input features and 6 time series outputs across 4 wells.
• Introduced significant input series to Random Forest Classifier and Neural Networks.

Main Results:

• EventTracker significantly reduced input features while maintaining predictive capability.
• Neural Network Classifier performance was enhanced by using EventTracker-selected features.
• EventTracker generated results in milliseconds, allowing ample time for intervention.

Conclusions:

• EventTracker is a highly efficient method for feature selection in real-time drilling safety applications.
• The method offers a viable solution to overcome computational challenges in time-critical prediction scenarios.
• EventTracker facilitates improved accuracy and speed for predictive models in sub-surface drilling.