Análise multiatributos sísmicos e aprendizado de máquinas aplicados à detecção de rochas ígneas no pré-sal da Bacia de Santos

Abstract

The carbonate reservoirs of the pre-salt section are among the most important recent discoveries in recent decades, transforming the Santos Basin into one of the main exploration centers for hydrocarbons in Brazil. These reservoirs are marked by intercalations with volcanic intrusions of different ages. The presence of these rocks reduces the potential of the reserves and may pose risks to the drilling stage. The high depth of the section of interest, and consequent reduced seismic resolution, associated with the petrophysical characteristics of igneous intrusions, make its identification challenging. Automatic classification techniques, in the form of clustering algorithms, allow the simultaneous analysis of high-dimensional data, grouping them into classes that share characteristics along these dimensions. In the context of reflection seismic data, high dimensionality is achieved from the extraction of several seismic attributes to classify different seismic facies, with the potential to identify features not visible in the amplitude data. The potential of three algorithms, K-Means, Self-Organizing Kohonen Maps and Gaussian Mixture Models, is evaluated in the identification of facies corresponding to the presence of igneous intrusions in seismic data from the Mero Field area, aiming to develop a routine for application in other areas of similar geological context, on the exploratory front, while there is no availability of advanced seismic processes, such as elastic seismic inversions or other high computational cost products, which can be carried out by the exploration geoscientist, in your own workstation.