The application of the k-means clustering method in the prediction of the primary and secondary mineral composition of sandstones: a case study from Lower and Middle Jurassic rocks in the Polish Basin

Abstract

Comprehensive knowledge of both the primary and secondary mineral composition of rocks, as well as the diagenetic processes occurring within them, stands as a fundamental requirement for accurately estimating reservoir rock parameters across various geological sectors, such as petroleum geology, geothermal energy, and carbon capture and storage. In the era of widespread automatization, machine learning methods are increasingly being used for interpretation, which, with the support of appropriate datasets and the experience of interpreters, make it possible to draw a variety of conclusions about geological processes occurring within selected geological formations, as well as entire sedimentary basins. We describe the application of the k-means clustering method for the rapid prediction of primary and diagenetic mineral composition using the the example of Lower and Middle Jurassic sandstones in the Polish Basin - one of the main aquifers and a potential reservoir formation. The model was based on the correlation between neutron porosity (NPHI), bulk density (RHOB), interval transit time (DT), deep resistivity (LLD), total natural gamma-ray (GR) and spectral gamma-ray values (K, Th, U), correlated with the results of petrographic, petrophysical and qualitative geochemical analysis. This correlation was the basis for distinguishing 5 different sandy petrofacies with variable primary and diagenetic characteristics typical of Jurassic sandstones in the Polish Basin.