Deep learning methods for shear log predictions in the Volve field Norwegian North Sea

Shear logs are required to calculate reservoir characterization or geomechanics parameters. Shear logs are also used for various seismic analysis applications, such as Amplitude Versus Offset (AVO) inversion and multicomponent seismic interpretation. Furthermore, shear logs or their inverse shear velocity logs are an essential component for rock physics analysis in order to constrain seismic inversion results for potential reservoir pay intervals, in reservoir characterization practices. Often, shear logs are missing to reduce well logging costs, or the data are of poor quality due to poor borehole conditions, or cycle-skipping. This thesis discusses artificial neural networks (ANNs) for shear log predictions using data from the Volve field, in the Norwegian North Sea. In this thesis I use deep neural networks or feedforward neural networks, and I propose convolutional neural networks and recurrent neural networks, to predict shear logs from gamma ray, bulk density, neutron porosity, true formation resistivity, compressional sonic and depth logs. Multiple efforts have been made to predict shear logs using machine learning methods. These studies mostly used one well each for training data and for blind testing. To the best of my knowledge, a comprehensive study that includes well logs, including shear data, from multiple wells in a single field to synthesize shear logs using machine learning methods is lacking. In this thesis I explore the robustness of deep learning methods for shear log (DTS) prediction, using a quantitative measure of accuracy scoring for prediction, such as the coefficient of determination R-squared, and the Root Mean Squared Error (RMSE), using six wells that contained DTS data. I use deep learning methods to synthesize shear logs using all wells with shear log data in the Volve field, Norwegian North Sea. This thesis will provide geoscientists with a tested deep learning approach to synthesize shear logs on a full field scale data set, and to assist in shear prediction for reservoir characterization applications such as AVO inversion, or geomechanical parameters calculation such as shear modulus.

Name thesis:degree_name

Master of Science (M.S.)

Level thesis:degree_level

Masters

Discipline thesis:degree_discipline

Geophysics

Grantor dc:publisher

Colorado School of Mines. Arthur Lakes Library

Year dc:date.issued

2020

Author dc:creator

• Al Ghaithi, Aun

Advisor dc:contributor.advisor

• Prasad, Manika

Committee members dc:contributor.committeemember

• Zerpa, Luis E.
• Behura, Jyoti
• Wang, Hua

Statement dc:rights

• Copyright of the original work is retained by the author.

Language dc:language.iso

eng, English

Identifier

T 9023

OAI identifier oai:identifier

oai:repository.mines.edu:11124/176291