Log data-driven predictive models and feature ranking in reservoir characterization

Miah, Mohammad Islam (2020) Log data-driven predictive models and feature ranking in reservoir characterization. Doctoral (PhD) thesis, Memorial University of Newfoundland.

[img] [English] PDF - Accepted Version
Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission.

Download (3MB)


Log-based reservoir characterization is one of the widely used techniques to estimate the reservoir properties and make decisions for hydrocarbon production. Use of the machine learning tools is becoming a more accessible approach for data-driven model development. The objective of this research is to identify and rank the most contributing log variables considering their relative performance for prediction of water saturation and rock strength using the machine learning tools. The single layer and multi-layer perception (MLP) artificial neural network (ANN) and the kernel function-based least-squares support vector machine (LS-SVM) techniques are employed for model development. The models can capture the non-linear behavior and high-dimensional complex relationships among real field log data variables. The mutual information (MI) is used to investigate the dependency of predictor subset variables in a model and to rank log variables according to their importance. The connectionist models are also examined to find reliable data-driven predictive models to estimate reservoir properties and rock strength. A new correlation is developed to obtain the in-situ rock strength of the siliciclastic rocks using the most important log parameters. The model predictions are compared/validated against the measured values as well as results obtained from existing log-based correlations. The approaches suggested in this study (connectionist and MI strategies) can assist engineers/operators to run a few numbers of logging tools for prediction of reservoir and rock properties to save the exploration costs. Also, it is expected that the introduced robust data-driven predictive models will enable engineers to better manage the wellbore stability and formation analysis in terms of technical, economic, and environmental aspects.

Item Type: Thesis (Doctoral (PhD))
URI: http://research.library.mun.ca/id/eprint/14483
Item ID: 14483
Additional Information: Includes bibliographical references.
Keywords: Mutual information, Machine learning, Reservoir properties, Water saturation, Uniaxial compressive strength, Rock mechanics, Artificial intelligence
Department(s): Engineering and Applied Science, Faculty of
Date: May 2020
Date Type: Submission
Digital Object Identifier (DOI): https://doi.org/10.48336/vhdd-jm60
Library of Congress Subject Heading: Hydrocarbon reservoirs--Mathematical models.

Actions (login required)

View Item View Item


Downloads per month over the past year

View more statistics