Full text loading...
-
Machine Learning For Improved Characterization of the Yamama Reservoir In A ME Gas Field
- Publisher: European Association of Geoscientists & Engineers
- Source: Conference Proceedings, 84th EAGE Annual Conference & Exhibition, Jun 2023, Volume 2023, p.1 - 5
Abstract
Summary
Accurate lithofacies classification and petrophysical property modeling represent essential steps to preserve the reservoir heterogeneity and capture the uncertainty space for reservoir modeling and field development optimization. It is crucial to model the discrete lithofacies distribution as a function of well logging attributes for the prediction at the unsampled intervals. In this research, the Decision tree (DT) and Random Forest (RF) were sequentially adopted for lithofacies classification and permeability regression applied to data from the Yamama reservoir in a middle-eastern gas field.
The measured discrete lithofacies distribution was comparatively modeled as a function of well-log data: water saturation, shale volume, total porosity, and effective porosity. The classification accuracy was 92% for both DT and RF. Therefore; the RF-predicted lithofacies were then combined with the well-log data to model the permeability. For accurate modeling and to avoid overfitting, cross-validation of random sub-sampling was performed on the data set to split it into two sub-sets; 70% training subset for modeling and 30% testing subset for prediction. The adjusted R-squared was 0.971 and 0.915 for the RF and DT regression models, respectively. The resulting model can be later used to predict the facies and permeability at other wells of completely unavailable core measurements.
© EAGE Publications BV