Leveraging Machine Learning for Exception Diagnostics: Insights from Quest CCS Datasets

This study summarizes the application of two automated machine learning (ML) detection workflow applied on fiber optics datasets acquired at the Quest CCS site.

Distributed Temperature Sensing (DTS) data are acquired as part of the Measurement, Monitoring and Verification (MMV) plan since 2017, offering an extensive dataset for further evaluations.

Our analysis includes the training and evaluation of ML models to predict temperature responses and detect anomalous trends in the data that may indicate out-of-zone CO2. Additionally, we discuss the implementation of the workflow on real-time data and the insights gained from this process.

Induced seismicity is also continuously monitored at Quest using a downhole geophone system, which is part of the base MMV plan. For a period of nine months, as part of a technology trial, microseismic activity was simultaneously measured using the fiber optic technology available at the injection wells. A deep learning model was developed to automatically detect microseismic events. To achieve optimal results, maximizing event detection and reducing false positives, the model was trained independently on the DAS dataset.

The results of our study are promising and highlight the importance of integrating advanced technologies with traditional monitoring methods to achieve more effective and reliable monitoring solutions.