DeepLight

1.

Research Plan:
Machine Learning in
Oil and Gas Industry
deeplightventures.com

2.

World hydrocarbon resources
A map of world oil reserves, 2013.
https://www.reddirtreport.com

3.

602 [SSTVD]
SSTVDSATuse
2217.5
534 [SSTVD]
SSTVDSATuse
well correlation
838 [SSTVD]
588 [SSTVD]
558 [SSTVD]
637 [SSTVD]
SSTVDSATuse
SSTVDSATuse
SSTVDSATuse
2182.5
2200
2202.5
732 [SSTVD]
SSTVDSATuse
SSTVDSATuse
781 [SSTVD]
SATuse
SSTVD
2205
852 [SSTVD]
2185
2195
836 [SSTVD]
SSTVDSATuse
827 [SSTVD]
807 [SSTVD]
2185 SATuse
SSTVD
SSTVDSATuse
SSTVDSATuse
2192.5
2220
2197.5
2220
2210
J1-0_top
J1-0_top
2207.5
2187.5
2195
GZ
2222.5
GZ
2197.5
J1-0_top
SSTVDSATuse
2200
2187.5
NZ
GZ
2205
2200
J1-1_top
J1-1_base
2222.5
2212.5
2190
2190
2207.5
GZ
2185
808 [SSTVD]
GZ
2202.5
2205
2187.5
GZ
J1-1_top
2210
2190
2197.5
2225
2202.5
2225
GZ
2215
2207.5
2190
2212.5
2192.5
2202.5
2205
2227.5
GZ
2210
J1-1_base
2192.5
2192.5
2227.5
2217.5
2195
2212.5
2215
NZ
2217.5
3d physical
model
hydrodynamic
simulation
[Thenin, Larson, 2014]
NZ
2230
GZ
GZ
GZ
2230
J1-0_base
J1-0_base
GZ
GZ
2195
2205
GZ
3d model
frame
2215
2210
2212.5
GZ
2195

4.

Partnership
Laboratory on Machine Learning
in Oil & Gas Industry
Research and Innovation Projects:
• applied projects
• partnership with oil/gas companies
Student Training:
• student thesis projects
• publications
• student professional activities

5.

Interpretation

6.

Fault Detection 0
Image Recognition
(CNN)
Generative Models
GAN, VAE, Bayesian

7.

Fault Detection 1
Image Recognition
(CNN)
Fault detection in slices
Generative Models
GAN, VAE, Bayesian
Fault surface construction

8.

Well logging
www.saltworkconsultants.com
sanuja.com
infolupki.pgi.gov.pl

9.

Petrophysical model
Image Recognition
(CNN)
Generative Models
GAN, VAE, Bayesian
Convert well logs to petrophysical models
http://wallace-international.com/

10.

Lithological model
Image Recognition
(CNN)
Generative Models
GAN, VAE, Bayesian
Convert well logs to rock types
http://www.reddoggeo.com/
[Hall, 2016]

11.

Well Correlation
Image Recognition
(CNN)
Generative Models
GAN, VAE, Bayesian
602 [SSTVD]
SSTVDSATuse
2217.5
534 [SSTVD]
SSTVDSATuse
838 [SSTVD]
588 [SSTVD]
558 [SSTVD]
637 [SSTVD]
SSTVD
SATuse
SSTVDSATuse
SSTVDSATuse
SSTVD
SATuse
2182.5
2200
2202.5
732 [SSTVD]
SSTVDSATuse
781 [SSTVD]
SATuse
SSTVD
2205
852 [SSTVD]
827 [SSTVD]
807 [SSTVD]
2185 SATuse
SSTVD
SSTVDSATuse
SSTVDSATuse
2185
2195
836 [SSTVD]
SSTVDSATuse
2192.5
2220
2197.5
2220
2210
J1-0_top
J1-0_top
2195
GZ
GZ
2222.5
SSTVDSATuse
2200
2187.5
NZ
GZ
2205
2200
J1-1_top
J1-1_base
2222.5
2212.5
2190
2190
2207.5
GZ
2185
2187.5
2197.5
GZ
2202.5
2205
J1-0_top
2187.5
808 [SSTVD]
2207.5
GZ
J1-1_top
2210
2190
2197.5
2225
2202.5
2225
GZ
2215
2207.5
2190
2192.5
2202.5
2205
2227.5
GZ
2210
J1-1_base
2192.5
2192.5
2212.5
2227.5
2217.5
2195
2212.5
GZ
2215
2217.5
GZ
2230
NZ
J1-0_base
GZ
GZ
2230
NZ
GZ
2195
2205
GZ
2215
2210
2212.5
J1-0_base
GZ
2195

12.

Production

13.

Production 0
Physical Systems
(explicit PDE)
Production Modelling using Proxy Models. History matching
of physical Proxy Models to production data.
Simple equations: