Unsupervised Machine Learning Techniques for Improving Reservoir Interpretation Using Walkaway VSP and Sonic Log Data

In this paper, we present a detailed analysis of the possibility using unsupervised machine learning techniques for reservoir interpretation based on parameters obtained from geophysical measurements that are related to elastic properties rocks. Four different clustering algorithms were compared, including balanced iterative reducing and hierarchies, Gaussian mixture model, k-means, spectral clustering. Measurements with vertical resolutions used. The first set input was walkaway VSP survey. second one acquired in well full-wave sonic tool. Apart study used clustering, two data pre-processing paths analyzed context matching resolution both methods. validation final results carried out lithological identification medium an drill core. performed Silurian rocks (claystone, mudstone, marly claystone) lying under overburdened Zechstein formation (salt anhydrite). This is known high attenuating seismic signal properties. presented shows only multilevel acquisition Poland.