Open AccessCompressional-Shear Velocity Model of “Toki” Field using Support Vector Regression, Offshore Niger Delta
Author(s) -
Adelere F Adeniran,
Ahzegbobor P. Aizebeokhai
Publication year - 2019
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1299/1/012085
Subject(s) - niger delta , geology , shear (geology) , support vector machine , shear velocity , anisotropy , vector field , regression analysis , submarine pipeline , mineralogy , statistics , delta , geotechnical engineering , mathematics , machine learning , geometry , mechanics , petrology , physics , computer science , optics , turbulence , astronomy
Shear sonic log is invaluable for fluid and lithological classification. For most fields in the Niger Delta, shear sonic log are rarely acquired along with the compressional sonic log. Where acquired, they are usually very few relative to the number of wells. Hence, there is a need to derive shear velocity from the compressional sonic log. Most of the available models such as Castagna’s mud-rock model are not calibrated to suit the Niger Delta basin. Existing localized models are based on non-robust linear models such as the Ogagarue's localized compressional and shear velocity models for Niger Delta sedimentary region. These models are not reliable in the presence of hydrocarbon and anisotropy. A robust support vector regression (SVR) machine learning algorithm has been used to predict the relationship between compressional velocity and shear velocity. This study shows that in the Niger Delta, shear velocity can be predicted from compressional velocity with relatively high accuracy by using machine learning algorithms such as support vector regression. The mean-square error (MSE) obtained using Castagna’s and Ogagarue's models compared with acquired data are 1.8 and 2.3 times that of the value obtained using support vector regression respectively.