Production-Increase Potential Evaluations after Refracturing Low-Shale-Oil-Producing Wells via...

Area of Science:

• Petroleum Engineering
• Machine Learning Applications
• Reservoir Development

Background:

• Shale reservoir development increasingly relies on horizontal wells, with a growing proportion exhibiting low production.
• Effective refracturing is crucial for maximizing recovery from these wells, necessitating accurate production potential evaluation.

Purpose of the Study:

• To develop a novel method for assessing the repeated-fracturing potential of low-producing horizontal wells in shale reservoirs.
• To identify critical geological and engineering parameters influencing refracturing success.

Main Methods:

• A dataset of 149 wells with 27 parameters was compiled.
• An optimized XGBoost algorithm was developed and compared against seven other machine learning models.
• Feature importance and variance inflation factor analysis were used to screen key parameters.

Main Results:

• The optimized XGBoost model achieved an R-squared value of 0.904 on test data, outperforming other algorithms.
• Six critical parameters were identified: injection volume/intensity, geological reserves, fluid volume, formation length, and fracturing stages.
• The model successfully assessed 29 candidate wells, identifying two for prioritized refracturing.

Conclusions:

• The proposed XGBoost-based method provides a reliable approach for evaluating refracturing potential in low-producing shale wells.
• Prioritizing wells based on identified parameters can optimize refracturing decisions and enhance shale oil development.
• This research offers a theoretical and technical basis for improving the efficiency of shale oil recovery through targeted refracturing.