3D LITHOLOGY PREDICTION USING ELECTRICAL PROPERTY MODELS AND MACHINE LEARNING

Wei，N.1; Zou，Q.2; Wang，Y.3; Tong，B.2; Liu，K.2; Shi，Y.3; Yang，D.1

2022

83rd EAGE Conference and Exhibition 2022

3

2124-2128

Geophysical identification of lithology for reservoirs in igneous rocks can be complicated because the relationship between physical properties and lithology is often not straightforward. In this paper, we propose an automated data-driven machine learning method to capture such relationship and make lithology prediction in Liaohe Oilfield, China. We first make labelled samples using the 3D electrical resistivity and chargeability models and the lithology from well logging. Then a random forest classifier architecture is constructed and optimized for implicitly learning the relationship between the two electrical properties and three lithological types. After training, the random forest can achieve an average prediction accuracy of 84% on the validation set. Finally, the trained optimal random forest is applied to predict the 3D distribution of igneous rocks for the entire area. The results show that the three lithology units in the area basically form a layered structure, and the igneous rocks predicted by the proposed method is consistent with the previous interpretation.

First

English

National Key Research and Development Program of China[2018YFC0603605];

2-s2.0-85142688549

Scopus

1.Southern University of Science and Technology,China
2.PetroChina Liaohe Oilfield Company,China
3.BGP Inc.,China National Petroleum Corporation,China

Wei，N.,Zou，Q.,Wang，Y.,et al. 3D LITHOLOGY PREDICTION USING ELECTRICAL PROPERTY MODELS AND MACHINE LEARNING[C],2022:2124-2128.