High resolution CO2 monitoring using Deep Learning: Kimberlina reservoir

Nov 13, 2023 | CWP Blog

Posted by Ahmed Ahmed

I am working to get higher resolution of CO2 monitoring. A hybrid approach combined EFWI and physics-based information with a convolutional neural network capable of predicting temporal changes in compressional velocity and saturation of CO2 for acoustic models. I am extending it for elastic media with multi-component data sets. This will make this technique applicable to a wide range of time-lapse projects dealing with hydrocarbon production and CO2 sequestration.

Prof. Ilya Tsvankin and I are building on a joint project between CWP and INPEX (the largest Japanese oil/gas company) to utilize elastic full-waveform inversion (EFWI) of time-lapse seismic data for anisotropic media. In 2023, INPEX and CWP extended this project to monitoring of CO2 sequestration (a potentially important technology that should help reduce carbon emissions) using deep learning. Yanhua Liu successfully applied that workflow to the synthetic Kimberlina data set provided by the Department of Energy and I am expanding the scope of this research by generalizing the developed methodology for elastic and anisotropic media. The developed neural network uses the multicomponent recorded seismic data to predict the subsurface attributes simultaneously. We’re not just doing research; we’re shaping the future of environmental monitoring. Stay tuned for updates, and let’s make a difference together! Your thoughts and suggestions? Always welcome!

Prediction of reservoir parameters from full-waveform inversion combined with convolutional neural network.