GESGB London Evening Lecture – June 2023 (In person)

GESGB members can register free of charge via the membership platform

Please note: to guarantee access, you must register prior to the event.

If you are interested in becoming a member of GESGB please visit: https://www.ges-gb.org.uk/membership/

Kindly sponsored by:

Speaker: Shervin Rasoulzadeh, Technical Team Lead – GeoSoftware UK

Topic: Integration of rock physics and deep machine learning for reservoir characterisation of a complex geology oil field

Date/Timings: Tuesday 27 June 2023 – 18:00-19:00 (doors open at 17:30)

Venue: The Finery – 23 Great Castle Street, London, W1G 0JA

Catering: Self-funded bar

Kindly sponsored by:

Integration of rock physics and deep machine learning for reservoir characterisation of a complex geology oil field

Abstract

Supervised machine learning (ML) techniques are effective in predicting reservoir properties from seismic and well log data. However, a significant amount of training data is required. To overcome the shortage of training samples, this study synthesized training data based on available seismic data, well logs, and reservoir conditions. A hybrid Theory-Guided Data Science (TGDS) model was used to generate pseudo wells and synthetic seismic gathers reflecting different reservoir conditions. A Convolutional deep Neural Network (CNN) was trained using the synthetic seismic data to estimate 5 different reservoir properties simultaneously. The trained CNN produced comparable results to deterministic seismic inversion and showed a higher correlation with well data in some cases. The study focused on an oil field with complex lithology, used lithology-specific rock physics models and systematic changes to simulate expected reservoir conditions. The method involved preparing training data, generating synthetic reservoir property well logs, calculating synthetic gathers, and optimizing the CNN parameters. The study demonstrates the efficacy of incorporating rock physics models for augmenting training data and using CNNs for predicting elastic and reservoir properties, even in cases of complex lithologies with limited well data.

Shervin Rasoulzadeh, Technical Team Lead – GeoSoftware UK

Shervin Rasoulzadeh is Technical Team Lead for Europe and West Africa team within GeoSoftware C.V. Company and based in the UK. By education he has got MSC in mine engineering (exploration). Shervin has more than 20 years of experience working in the oil and gas industry and has completed several QI projects either deterministic or geostatistical seismic reservoir characterization projects around the world. In his current role as technical team lead, Shervin provides technical consultancy for geoscience project services, knowledge transfer, training, and support to users of GeoSoftware applications.