Santos Basin has been highlighted throughout recent years, especially because of the discovery of new offshore fields and the huge exploratory development in the pre-salt layer. The optimization of the exploration and production procedures of new oil fields requires studies of the lithological characteristics of the reservoir and adjacent lithological units. The geophysical well logging assists in stratigraphic understanding of such characteristics from indirect records that reflect the physical and chemical composition of the materials. Traditionally, these log interpretations are made from two methods: (I) visual analysis, based on observing different responses of records in well logs; (II) artificial neural networks, using training networks from a first information. In both ways, the lithological interpretations may be open to subjectivity, since these rely on supervised methods. The Self-Organizing Maps (SOM) technique is an evolution of artificial neural networks, and consists of an unsupervised tool for analysis visualization and classification of n-dimensional data, based on the principles of vector quantization and measures of the similarity vector. This project seeks to develop a logical and systematic routine of well log classification using the SOM technique, in view of the interpretation of lithological classes. Therefore, there will be developed visual, supervised and unsupervised analysis which will be compared quanti- and qualitatively. The analysis will be developed in collaboration with Petrobras geologists, Operational Unit of E & P in the Santos Basin (UO-BS), and will rely on well logs provided by the ANP. The generated routine must assist the interpretation of well logs data, which is a fundamental process to oil exploration and production.
News published in Agência FAPESP Newsletter about the scholarship: