Deep Learning for Seismic Imaging and Interpretation (DLSII)

Carbon Capture, Utilization, and Storage (CCUS) represents a pivotal solution for mitigating greenhouse gas emissions and ensuring a sustainable transition towards a low-carbon future. In seeking cost-effective strategies to implement CCUS, one area that holds significant promise is the utilization of seismic data for cost-effective site characterization. This approach can greatly reduce costs associated with evaluating suitable geological reservoirs for CO2 storage. In this context, we aim to develope deep learning techniques for automatic seismic interpretation, fault detection, seismic image enhancement and alignment to other data sources (e.g., automated well-to-seismic tie). This project is industrially funded by PETRONAS and is a collaboration under PETRONAS Centre of Excellence in Subsurface Engineering and Energy Transition (PACESET).

seismic-unet attention_weights
(left figure) Standard U-Net architecture for segmentation. (right figure) attention weights at layer 1

Project team at Heriot-Watt University (HWU)

  • Principal Investigator: Ahmed H. Elsheikh
  • Postdoctoral Research Associate: Haifa AlSalmi
  • PhD student: Chin Tee Ang
  • PhD student: Sharif Rahman

Project Publications

  • Haifa AlSalmi, Ahmed H. Elsheikh, Automated seismic semantic segmentation using Attention U-Net, Geophysics, (2023), URL
  • A.S, Abd Rahman, Ahmed H. Elsheikh, MS Jaya, Seismic Reflectivity Inversion Using a Semi-Supervised Learning Approach, SPE Annual Technical Conference and Exhibition, San Antonio, Texas, USA, October 2023, URL
  • C.T. Ang, M. Sajid, Ahmed H. Elsheikh, H. Alsalmi, Attention Mechanism Neural Network for Seismic Facies Classification, SPE Annual Technical Conference and Exhibition, San Antonio, Texas, USA, October 2023, URL