• High-quality fracture network map is essential to investigate the integrity of the reservoir for Carbon Capture Utilization and Storage (CCUS).
  • We have developed a workflow to map and visualize the fracture network using Integration of data collected from actual field measurements including:
    • Acoustic image logs: obtained from each well to identify the fracture intensity, orientation and depth.
    • Fiber optics (DAS/DTS and DSS) obtained from two monitoring wells.
    • Microseismic event during well stimulation.

This workflow highly relies on acoustic image log analysis which is:

  • Highly influenced by human bias (different experts reported significantly different number of fractures, orientations and depths – see figure to right)
  • It is extremely time consuming and expensive and  prevents deployment in real time (e.g., takes months to interpret)

Automated Machine Learning Workflow developed to:

  • Eliminate the human bias in fracture identification using image logs by using high frequency drilling vibration data
  • Eliminate the extremely long time required for interpretation of image logs  (Using only 10% of  interpreted log that enables a near real time application)
  • Optimize the machine learning model selection and hyperparameters optimization (Eliminate human bias in model and hyperparameter selection)

Benefits:

  1. Image log interpretation time and cost dropped  by ~10 fold (only 10%-15% of the interpreted image log is required) that enabled the near real time fracture mapping.
  2. Automated machine learning workflow eliminated the human bias in model and hyperparameters selection.
  3. High quality fracture network mapping and visualization is achieved using high frequency drilling vibration data.