Characterization of West-Africa Submarine Channel Reservoirs: A Neural Network Based Approach to Integration of Seismic Data

Abstract

Abstract Advances in deep water drilling technology have cleared the path for new domains of hydrocarbon exploration along the Atlantic margins. Such deep marine settings are generally characterized utilizing seismic data while in development phase. Yet, interpreting facies information from seismic amplitude data requires an extensive and tedious work of an expert geologist/geophysicist. This process is not only costly in terms of work hours but also invites human inconsistencies into interpretations. This paper attempts to solve the above problem by proposing a neural network based automation method that learns how to relate the seismic data to facies objects like channels. The neural net training is based on the manual/CAD interpretation of a small portion of the available seismic amplitude data by an expert geologist/geophysicist. The trained neural net is shown to be able to automatically detect channel body features in the remaining portion of the data. As the case study, the paper includes a West-Africa submarine channel reservoir where the expert interprets a seismic amplitude data set of a turbidite sequence and where the neural network manages to estimate channel facies morphologies from seismic data.