XGSleeve: Detecting Sleeve Incidents in Well Completion by Using XGBoost Classifier

Sahand Somi¹ David Cooper² Peng Wang² Sheikh Jubair^1*

• ¹Alberta Machine Intelligence Institute, University of Alberta, Canada
• ²Kobold Completions Inc., Canada

The sliding sleeve is an essential component for regulating fluid flow during hydraulic fracturing in shale oil extraction. However, concerns still need to be addressed regarding its reliability resulting in attempts to open the sleeve multiple times, resulting in an inefficient process. Although using downhole cameras can be used to verify sleeve opening or closing, these cameras are extremely expensive. A few research is suggested to analyze downhole data for detecting sleeve incidents instead of cameras. This project introduces a novel machine-learning approach called XGSleeve, which integrates hidden Markov model-based clustering and the XGBoost model to identify sleeve incidents and act as the operator's visual representative. This model offers valuable insights into downhole activities, detecting sleeve incidents with 86\% precision, resulting in reducing the number of attempts to open or close the sleeve. By providing a more reliable and accurate method for malfunction detection, our model enhances efficiency and safety and holds the potential to transform sleeve incident detection in the oil and gas industry. Our XGSleeve model not only improves the detection of sleeve incidents but also paves the way for further advancements in data-driven decision-making within the oil and gas industry. As technology continues to evolve, integrating machine learning and artificial intelligence into daily operations can lead to broader optimization of processes, reduced environmental impact, and increased sustainability. Ultimately, the successful implementation of XGSleeve and similar approaches will contribute to the long-term growth and resilience of the oil and gas sector while promoting responsible resource management on a global scale.