# US 12,223,420 B2 · Artificial learning fracture system and method for predicting permeability of hydrocarbon reservoirs

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 12,223,420 B2
- Document ID: US12223420B2
- Status: Granted
- Kind: B2
- Application number: 17/039,403
- Filed: 2020-09-30
- Published: 2021-04-01
- Granted: 2025-02-11
- Earliest claimed: 2019-10-01
- Assignees: Chevron U.S.A. Inc.
- Applicants: Chevron U.S.A. Inc.
- Inventors: Baosheng Liang; Chaoshun Hu; Min Li
- Web page: https://patentgrep.com/patent/US12223420B2

## CPC

- G06N 3/08: Learning methods
- G01V 20/00: Geomodelling in general
- G06N 3/04: Architecture, e.g. interconnection topology
- G06N 3/045: Combinations of networks
- G06N 3/0455: Auto-encoder networks; Encoder-decoder networks
- G06N 3/0464: Convolutional networks [CNN, ConvNet]
- G06N 3/09: Supervised learning
- E21B 2200/22: Fuzzy logic, artificial intelligence, neural networks or the like
- G01V 2210/6246: Permeability
- G06Q 10/04: Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem" (market predictions or forecasting for commercial activities G06Q30/0202)
- G06Q 50/02: Agriculture; Fishing; Forestry; Mining
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G01V: GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS (means for indicating the location of accidentally buried, e.g. snow-buried, persons A63B29/02)
- E21B: EARTH OR ROCK DRILLING; OBTAINING OIL, GAS, WATER, SOLUBLE OR MELTABLE MATERIALS OR A SLURRY OF MINERALS FROM WELLS
- G06Q: INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR

## Abstract

A method is described for predicting permeability including receiving a 3-D earth model including a volume of interest; generating 2-D property images; receiving 2-D fracture images; training a physics-guided neural network using the 2-D fracture images; and predicting permeability using the physics-guided neural network applied to the 2-D property images. The method is executed by a computer system.

## Claims

### Claim 1

A computer-implemented method of permeability prediction, comprising:
a. receiving, at a computer processor, a 3-D earth model including a volume of interest;
b. generating, via the computer processor, 2-D property images;
c. receiving, at the computer processor, 2-D fracture images;
d. training a physics-guided neural network (PGNN) using the 2-D fracture images;
e. providing the 2-D property images to the PGNN as input; and
f. predicting, by the PGNN, permeability corresponding to the 2-D property images;
and wherein the PGNN is a multi-channel whole-image deep neural network including convolutional and deconvolutional layers with considerations of skip connectors to avoid vanishing gradient issues.

### Claim 2 (depends on 1)

The method of claim 1 wherein the 2-D property images include one or more of porosity, water saturation, Poisson's Ratio, reservoir pressure, and clay content.

### Claim 3 (depends on 1)

The method of claim 1 wherein the PGNN includes a convolutional layer for multi-scale feature.

### Claim 4 (depends on 1)

The method of claim 1 wherein the PGNN is trained with consideration of structure constraints.

### Claim 5 (depends on 1)

The method of claim 1 further comprising displaying the permeability on a graphical user interface.

### Claim 6

A computer system comprising:
one or more processors;
memory; and
one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs including instructions that when executed by the one or more processors cause the system to execute:
a. receiving, at the one or more processors, a 3-D earth model including a volume of interest;
b. generating, via the one or more processors, 2-D property images;
c. receiving, at the one or more processors, 2-D fracture images;
d. training a physics-guided neural network (PGNN) using the 2-D fracture images;
e. providing the 2-D property images to the PGNN as input; and
f. predicting, by the PGNN, permeability corresponding to the 2-D images;
and wherein the PGNN is a multi-channel whole-images deep neural network including convolutional and deconvolutional layers with considerations of skip connections to avoid vanishing gradient issues.

### Claim 7

A non-transitory computer readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by an electronic device with one or more processors and memory, cause the device to execute
a. receiving, at the one or more processors, a 3-D earth model including a volume of interest;
b. generating, via the one or more processors, 2-D property images;
c. receiving, at the one or more processors, 2-D fracture images;
d. training a physics-guided neural network (PGNN) using the 2-D fracture images;
e. providing the 2-D property images to the PGNN as input; and
f. predicting, by the PGNN, permeability corresponding to the 2-D property images;
and wherein the PGNN is a multi-channel whole-image deep neural network including convolutional and deconvolutional layers with considerations of skip connections to avoid vanishing gradient issues.