Deep learning a poroelastic rock-physics model for pressure and saturation discrimination

Determining saturation and pore pressure is relevant for hydrocarbon production as well natural gas [Formula: see text] storage. In this context, seismic methods provide spatially distributed data used to determine fluid migration. A method developed that allows the determination of reservoir from data, more accurately rock-physics attributes velocity, attenuation, density. Two models based on Hertz-Mindlin-Gassmann Biot-Gassmann are developed. Both generate poroelastic pressure, saturation, other parameters. The inverted with deep neural networks derive porosity attributes. demonstrated a 65 m unconsolidated high-porosity at Svelvik ridge, Norway. Tests most suitable structure network carried out. Saturation can be meaningfully determined under condition gas-free baseline known an accurate campaign, preferably cross-well seismic. Including attenuation increases accuracy. Although training requires hours, predictions made in only few seconds, allowing rapid interpretation results.