| CPC G01V 9/007 (2013.01) [G01V 1/30 (2013.01); G01V 3/18 (2013.01); G01V 11/00 (2013.01); G01V 20/00 (2024.01); G06F 30/23 (2020.01); G06F 30/27 (2020.01); G06F 30/28 (2020.01); G06N 3/04 (2013.01); G06N 3/045 (2023.01); G06N 3/08 (2013.01); G01V 99/00 (2013.01); G01V 2210/61 (2013.01); G01V 2210/6169 (2013.01)] | 20 Claims |
|
1. A method for a physics-driven deep learning-based inversion coupled to fluid flow simulators, comprising:
obtaining, with a Magnetic Resonance (MR) scanner or a pulse Nuclear Magnetic Resonance (NMR) logging tool configured to generate a magnetic field and measure a presence of protons in the magnetic field, measured data for a subsurface region, wherein the measured data includes MR data and NMR data;
obtaining, by a computer processor, prior subsurface data for the subsurface region;
obtaining, by the computer processor, a physics-driven standard regularized joint inversion for at least two model parameters;
obtaining, by the computer processor, a case-based deep learning inversion characterized by a contracting path and an expansive path, wherein the case-based deep learning inversion includes a training phase where network hyperparameters are learned from the MR data, the NMR data, and the prior subsurface data and the case-based deep learning inversion includes a testing phase where an optimized pseudoinverse operator is used to predict resistivity measurements;
forming, by the computer processor, the physics-driven deep learning inversion with the physics-driven standard regularized joint inversion, the case-based deep learning inversion, and a rock-physics coupling operator based on a penalty function;
forming, by the computer processor, a feedback loop between the physics-driven standard regularized joint inversion and the case-based deep learning inversion for re-training the case-based deep learning inversion;
generating an inversion solution for reservoir monitoring, by the computer processor, using a hybrid coupled approach of physics-based and deep learning-based inversions with the feedback loop to converge to a true model distribution through an iterative approach;
guiding, with a geosteering system, a drill bit based upon the resistivity measurements or acoustic measurements;
adjusting the geosteering system using the inversion solution based on the contracting path and the expansive path, and
steering, by the computer processor, the drill bit based on monitoring, in real time, a detected sensor signature proximate the drill bit,
wherein when the resistivity measurements are employed, an upper boundary and a lower boundary of the subsurface region are computed from geological models using inversion techniques,
wherein when the acoustic measurements are employed, the upper boundary and the lower boundary of the subsurface region are calculated based on a travelling time of reflected sonic waves and a corresponding sonic velocity of formation rocks, and
wherein the monitoring comprises comparing the detected sensor signature to a known cap rock sensor signature, a known pay zone sensor signature, and a known bed rock sensor signature.
|