Efficient big data assimilation through sparse representation: A 3D benchmark case study in petroleum engineering.

Area of Science:

• Geosciences
• Petroleum Engineering
• Computational Seismology

Background:

• Data assimilation combines models and observations for improved geophysical estimates.
• Ensemble-based methods are state-of-the-art but computationally intensive for big data.
• Existing methods face challenges with large datasets and uncertainty quantification.

Purpose of the Study:

• Extend wavelet-based sparse representation from 2D to 3D seismic data assimilation.
• Address computational challenges posed by big data in geophysical modeling.
• Improve efficiency and accuracy of ensemble-based seismic history matching.

Main Methods:

• Developed a 3D wavelet-based sparse representation procedure.
• Applied the procedure within an ensemble-based seismic history matching framework.
• Utilized the Brugge field 3D benchmark case with extensive seismic data.

Main Results:

• The 3D sparse representation procedure significantly reduces seismic data size.
• Key data features are preserved during the data reduction process.
• The method demonstrates extreme efficiency in handling large-scale seismic data.

Conclusions:

• The extended sparse representation procedure is highly effective for 3D seismic data assimilation.
• This approach enables computationally feasible and accurate ensemble-based history matching.
• The methodology is a crucial step towards real-world field applications in geosciences.