DATA-DRIVEN REAL-TIME GEOSTEERING USING SURFACE DRILLING DATA

Abstract

In this thesis I present a method for estimating lithology or deriving formation properties from real-time surface drilling data. This information can then be used to enhance real-time geosteering capabilities. Current approaches for geosteering often rely on data from an MWD sub. Because of the position of the MWD sub relative to the bit, the MWD sub is relating information that is depth- and time-delayed relative to the bit. In this work, I use a data-driven approach that relies on the use of Hidden Markov Models (HMM) and Change Point Detection (CPD) algorithms to relate surface drilling signals to formation properties/lithology. My approach views the surface drilling data as a multichannel time-series signal which is an advantage over prior approaches that have attempted to derive variations in lithology from surface drilling data. I finally test my approach with two field datasets: the first is the Volve Oilfield in offshore Norway, and the second is the Meramec formation in the Oklahoma STACK. In both case studies, I demonstrate that the use of HMM and/or CPD can substantially enhance resolution of lithology changes from surface drilling data alone and can therefore be a promising approach for use in real-time geosteering decisions.