Prediction of hydrocarbon depth for seabed logging (SBL) application using Gaussian process

Abstract

Seabed Logging (SBL) is an application of marine Controlled-Source Electromagnetic (CSEM) technique to characterize hydrocarbon-filled layers underneath the seabed remotely in deep water regions. This technique maps structure of subsurface electrical resistivity in the offshore environment. Basically, exploration of offshore hydrocarbon is based on the contrast of electrical resistivity between hydrocarbon reservoir and its surrounding sea sediments. Modelling offshore hydrocarbon is a core analysis and time consuming task. Current numerical modelling techniques used in SBL application involve meshes and complicated mathematical equations. Thus, a simple supervised learning method which is Gaussian Process (GP) is proposed to process synthetic SBL data which are generated through Computer Simulation Technology (CST) software to predict the depth of hydrocarbon. This statistical model is able to provide additional hydrocarbon information by utilizing the prior information. 1-dimensional (1-D) forward GP models have successfully been developed to predict the presence of hydrocarbon and as the continuation work, 2-dimensional (2-D) forward GP model is then developed to be used as the standard profile in order to predict the depth of hydrocarbon. This shall give indication that GP can be used as the methodology to predict the depth of hydrocarbon reservoir underneath the seabed.