Unsupervised Learning Approach for Robust Monitoring of Carbon Dioxide Plume in the Subsurface

An accurate knowledge of the location, content and movement of the injected CO2 is crucial for risk management at a geological CO2-storage complex. This work presents an unsupervised-learning-based methods that facilitate robust monitoring of the subsurface CO2 plume. Such an approach adapts and scales based on the data without requiring an assumption of the geophysical model. The data-processing workflow was applied to the cross-well tomography data from the SECARB Cranfield carbon geo-sequestration project. The workflow for CO2 visualization incorporates feature extraction, feature selection, and two-level clustering. The multi-level clustering approach was developed to account for data imbalance due to the absence of CO2 in the large portion of the imaged reservoir. The multi-level clustering differentiates the CO2-bearing regions into regions containing negligible, low, medium, and high CO2 content. Overall, the CO2 monitoring achieved a silhouette score, Calinski-Harabasz index, and Davies-Bouldin index of 0.74, 59656, and 0.32, which confirms the high quality and reliability of the unsupervised spatial monitoring of CO2. The robustness of the qualitative levels of CO2 content assigned by the unsupervised approach is confirmed by high adjusted-Rand and homogeneity scores among various distinct approaches. Four statistical tests, namely F-test, mutual information, Tukeys HSD, and boxplot analysis, were performed to determine new geophysical signatures are discovered that are suitable for detecting CO2 presence and content. Further, we find certain geophysical signatures, such as Fourier transform and wavelet transform, to be highly relevant and informative indicators of the spatial distribution of CO2 content. The use of unsupervised learning provides a fast, data-driven, qualitative approximation of CO2 content, distribution, and presence, which serves as a substitute to rock-physics models that have inherent parametric and geophysical assumptions.