Fault MLReal: a fault delineation study for the Decatur CO₂ field data using neural network predicted passive seismic locations

Carbon Capture, Utilization, and Storage (CCUS) projects are crucial for mitigating greenhouse gas emissions and combating climate change. These projects involve the injection of captured CO₂ into subsurface geological formations for long-term storage, necessitating detailed knowledge of subsurface geology and geomechanics. Critical tools for ensuring the stability and safety of CCUS operations are passive source relocation and fault detection, which provide key insights into subsurface structures and fluid migration pathways. However, the implementation of these techniques is challenging as they require high-quality seismic data and advanced computational methods. To address these challenges, we present Fault MLReal, a novel deep learning method tailored for passive source relocation and fault detection in CCUS projects. Applying data domain-adaptation techniques, Fault MLReal enables training a neural network with synthetic data and testing it on field data. We demonstrate the efficacy of this approach in a field case study involving CO₂ injection microseismic data from the Decatur area. The successful relocation of passive seismic events and identification of faults not only improved the understanding of the subsurface structures but also facilitated prevention of potential geological hazards. In summary, our work highlights the significant value that machine learning, specifically Fault MLReal, can add to CCUS operations, potentially transforming the understanding of subsurface geology and geomechanics in the field and paving the way for more successful CCUS deployments in the fight against climate change.