Deep learning methods for fault characterization and upscaling of reservoirs

• Grant number: 24/08939-9
• Support Opportunities: Scholarships in Brazil - Post-Doctoral
• Start date: August 01, 2024
• End date: April 23, 2025
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computer Systems
• Agreement: Equinor (former Statoil)
• Principal Investigator: Hélio Pedrini
• Grantee: Sandro Pereira Vilela
• 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

This post-doctorate research work plan focuses on two primary tasks in the context of the Energy Production Innovation Center (EPIC) involving Equinor/Unicamp/Fapesp.The first task is the use of image processing and machine learning techniques to analyze computed tomography (CT) images, nuclear magnetic resonance (NMR), and borehole images (BHI) in order to detect and characterize structures such as pores, fractures, and vugs in carbonate rocks. To accomplish this task, the project will involve data segmentation to identify the contours of these structures within the images (PURSWANI et al., 2020; XIONG; JIANG; XU et al., 2021; MACDONALD et al., 2022). Novel machine learning algorithms will be developed and applied to evaluate the geometric parameters of these structures, such as their size, shape, and orientation. By combining these data with information about the fluid properties of the rocks, geologists can gain insights into the behavior of fluid flow within the reservoir and potentially develop more effective methods for extracting resources from these formations (LI et al., 2021).The second task is related to reservoir upscaling, where researchers typically use a combination of experimental data, such as core samples and well logs, and numerical simulations, which are used to create models of fluid behavior. These models can be used to estimate properties such as permeability, porosity, and fluid saturation at larger scales, based on data obtained at smaller scales. One of the key challenges in reservoir upscaling is accurately capturing the complex, heterogeneous nature of fluid flow behavior at smaller scales. Deep learning techniques will be used to identify patterns and relationships within the data at small scale (TREHAN and DURLOFSKY, 2018; ZHU et al., 2021), allowing for more accurate predictions of fluid behavior at larger scales. This can lead to more effective resource management, as well as improved production strategies. However, the use of deep learning techniques in reservoir upscaling also presents some challenges, such as the need for large amounts of high-quality data and the potential for overfitting, topics that will be addressed through data augmentation and pre-conditioning techniques. (AU)