Deep Learning Application for 4D Pressure Saturation Inversion Compared to Bayesian Inversion on North Sea Data

Jesper Sören Dramsch; Gustavo Corte; Hamed Amini; Mikael Lüthje; Colin Macbeth

doi:10.31223/osf.io/zytp2

Deep Learning Application for 4D Pressure Saturation Inversion Compared to Bayesian Inversion on North Sea Data

Jesper Sören Dramsch^*, Gustavo Corte, Hamed Amini, Mikael Lüthje, Colin Macbeth

^*Corresponding author for this work

Department of Applied Mathematics and Computer Science

Heriot-Watt University

Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review

Abstract

In this work we present a deep neural network inversion on map-based 4D seismic data for pressureand saturation. We present a novel neural network architecture that trains on synthetic data and pro-vides insights into observed ﬁeld seismic. The network explicitly includes AVO gradient calculationwithin the network as physical knowledge to stabilize pressure and saturation changes separation. Weapply the method to Schiehallion ﬁeld data and go on to compare the results to Bayesian inversionresults. Despite not using convolutional neural networks for spatial information, we produce mapswith good signal to noise ratio and coherency

Original language	English
Title of host publication	Proceedings of the Second EAGE Workshop Practical Reservoir Monitoring
Number of pages	4
Publication date	2019
DOIs	https://doi.org/10.31223/osf.io/zytp2
Publication status	Published - 2019
Event	Second EAGE Workshop Practical Reservoir Monitoring - American Hotel, Amsterdam, Netherlands Duration: 1 Apr 2019 → 4 Apr 2019

Conference

Conference	Second EAGE Workshop Practical Reservoir Monitoring
Location	American Hotel
Country/Territory	Netherlands
City	Amsterdam
Period	01/04/2019 → 04/04/2019

Keywords

4D seismic
Deep learning
Geophysic
North Sea
Seismic inversion
Time-lapse
Pressure
Saturation

Access to Document

10.31223/osf.io/zytp2

OpenUrl availability

Full text

Cite this

@inproceedings{4e0a209379e841b1a0100b417e830d9f,

title = "Deep Learning Application for 4D Pressure Saturation Inversion Compared to Bayesian Inversion on North Sea Data",

abstract = "In this work we present a deep neural network inversion on map-based 4D seismic data for pressureand saturation. We present a novel neural network architecture that trains on synthetic data and pro-vides insights into observed ﬁeld seismic. The network explicitly includes AVO gradient calculationwithin the network as physical knowledge to stabilize pressure and saturation changes separation. Weapply the method to Schiehallion ﬁeld data and go on to compare the results to Bayesian inversionresults. Despite not using convolutional neural networks for spatial information, we produce mapswith good signal to noise ratio and coherency",

keywords = "4D seismic, Deep learning, Geophysic, North Sea, Seismic inversion, Time-lapse, Pressure, Saturation",

author = "{S{\"o}ren Dramsch}, Jesper and Gustavo Corte and Hamed Amini and Mikael L{\"u}thje and Colin Macbeth",

year = "2019",

doi = "10.31223/osf.io/zytp2",

language = "English",

booktitle = "Proceedings of the Second EAGE Workshop Practical Reservoir Monitoring",

note = "Second EAGE Workshop Practical Reservoir Monitoring ; Conference date: 01-04-2019 Through 04-04-2019",

}

Sören Dramsch, J, Corte, G, Amini, H, Lüthje, M & Macbeth, C 2019, Deep Learning Application for 4D Pressure Saturation Inversion Compared to Bayesian Inversion on North Sea Data. in Proceedings of the Second EAGE Workshop Practical Reservoir Monitoring. Second EAGE Workshop Practical Reservoir Monitoring, Amsterdam, Netherlands, 01/04/2019. https://doi.org/10.31223/osf.io/zytp2

TY - GEN

T1 - Deep Learning Application for 4D Pressure Saturation Inversion Compared to Bayesian Inversion on North Sea Data

AU - Sören Dramsch, Jesper

AU - Corte, Gustavo

AU - Amini, Hamed

AU - Lüthje, Mikael

AU - Macbeth, Colin

PY - 2019

Y1 - 2019

N2 - In this work we present a deep neural network inversion on map-based 4D seismic data for pressureand saturation. We present a novel neural network architecture that trains on synthetic data and pro-vides insights into observed ﬁeld seismic. The network explicitly includes AVO gradient calculationwithin the network as physical knowledge to stabilize pressure and saturation changes separation. Weapply the method to Schiehallion ﬁeld data and go on to compare the results to Bayesian inversionresults. Despite not using convolutional neural networks for spatial information, we produce mapswith good signal to noise ratio and coherency

AB - In this work we present a deep neural network inversion on map-based 4D seismic data for pressureand saturation. We present a novel neural network architecture that trains on synthetic data and pro-vides insights into observed ﬁeld seismic. The network explicitly includes AVO gradient calculationwithin the network as physical knowledge to stabilize pressure and saturation changes separation. Weapply the method to Schiehallion ﬁeld data and go on to compare the results to Bayesian inversionresults. Despite not using convolutional neural networks for spatial information, we produce mapswith good signal to noise ratio and coherency

KW - 4D seismic

KW - Deep learning

KW - Geophysic

KW - North Sea

KW - Seismic inversion

KW - Time-lapse

KW - Pressure

KW - Saturation

U2 - 10.31223/osf.io/zytp2

DO - 10.31223/osf.io/zytp2

M3 - Article in proceedings

BT - Proceedings of the Second EAGE Workshop Practical Reservoir Monitoring

T2 - Second EAGE Workshop Practical Reservoir Monitoring

Y2 - 1 April 2019 through 4 April 2019

ER -

Deep Learning Application for 4D Pressure Saturation Inversion Compared to Bayesian Inversion on North Sea Data

Abstract

Conference

Keywords

Access to Document

OpenUrl availability

Fingerprint

Cite this