GEOMODEL SEISMIC CONDITIONING USING MACHINE LEARNING AND PROBABILISTIC METHODS

• Grant number: 25/04182-3
• Support Opportunities: Scholarships in Brazil - Doctorate
• Start date: September 01, 2025
• End date: August 31, 2029
• Field of knowledge: Physical Sciences and Mathematics - Geosciences - Geophysics
• Principal Investigator: Emilson Pereira Leite
• Grantee: Muhammad Irfan Haider
• Host Institution: Centro de Estudos de Energia e Petróleo (CEPETRO). Universidade Estadual de Campinas (UNICAMP). Campinas , SP, Brazil
• Company: Universidade Estadual de Campinas (UNICAMP). Faculdade de Engenharia Mecânica (FEM)
• Associated research grant: 17/15736-3 - Engineering Research Centre in Reservoir and Production Management, AP.PCPE

Abstract

The main objective of this project is to develop a consistent new workflow for the conditioning of geomodels of carbonate reservoirs to seismic data. Geomodels provide a reference for numerical modeling that helps to better understanding key factors underlying challenges in exploration and development phases of a field, particularly in the case of complex carbonate reservoirs such as those of the pre-salt offshore Brazil (AMARU; LI; TYSHKANBAYEVA, 2022). Seismic 3D data is used to provide 3D structural interpretation as well as to populate the geomodels with reservoir properties, which can be conducted by many strategies and algorithms, depending mainly on the complexity of the geological scenario, data availability and data quality. A workflow that is able to produce a realistic 3D geomodel in a complex geological scenario capturing the various uncertainties and heterogeneities involved needs to be developed. This project will address this challenge by investigating probabilistic and machine learning algorithms in a multistep conditioning of geomodels to seismic data (SONG et al., 2022). It has been discussed in the literature that traditional strategies do not provide a fully realistic model that is able to produce a seismic model comparable to the observed seismic (SHRAGGE et al., 2019). Deep learning algorithms will be used for automatic interpretation of faults and stratigraphic units to construct a structural framework for the geomodel. The developed algorithms will be able to evaluate a set of seismic attributes to provide the best possible interpretation. Prediction of petrophysical properties and petroelastic facies will be conducted using seismic-petrophysics inversion methods based on probabilistic and machine learning techniques (GRANA et al., 2022). Focus will be on the combination of both approaches and each intermediate product will be evaluated in terms of uncertainties and resolutions. Predictions of reservoir properties will require the establishment of adequate rock-physics models that should consider the geological facies described at the well log scale. Special attention will be given to the validation of the seismic modeled traces calculated from the conditioned geomodel. (AU)