The Application of Machine Learning On Heterolithics Identification in Brunei Darussalam
Nur Aishah Helena Mohd Sharif, Chunming Xu, Yee Yung Liew, Amir Sabli, Idrus Puasa, Lisa Thieme, Tom Savels, Leytzher Muro · 2018
Machine learning algorithms have been used to predict the presence of heterolithics from conventional well logs such as gamma-ray, resistivity, density and neutron, by exploiting the Boomerang workflow. Thirty wells have been selected from a number of producing fields that represent both onshore and offshore Brunei. These wells are typically traverses across various depositional sub-environments laterally and also vertically. The wells represent the training data to generate a model, which can then be used to quantify the precision of the model together with the hundreds of wells in the study area. Results show that Support Vector Mechanism typically yield higher average precision (>60%) for all identified Boomerang facies compared to Decision Tree and Random Forest (each with average less than 30% precision). The relatively higher precision from Support Vector Mechanism could be due to: (i) the non-linear nature of reservoir properties, (ii) the uncertainties in the estimation of various reservoir properties, and (iii) heterogeneity of the datasets.