24Nov2025
10:00 Doctoral defense room 353 of IC3,5
Topic on
Model Uncertainty Quantification to Improve Oil Reservoir Simulations
Student
Jeanfranco David Farfan Escobedo
Advisor / Teacher
Anderson de Rezende Rocha - Co-advisor: Gabriel Cirac Mendes Souza
Brief summary
Proxy models are simplified representations of complex numerical models used to increase the computational efficiency of a process. Proxy development is challenging due to the inherent high dimensionality of the spaces they model. This research takes a broad and holistic approach to address these challenges. As the study of proxy models evolves toward a deeper understanding of complex reservoir characteristics, our work follows this trend. We begin by proposing a supervised method to approximate the simulator's response using a single Artificial Neural Network (ANN). This network is designed to predict the behavior of multiple cumulative production curves simultaneously, diverging from the conventional approach in the literature, which typically relies on training multiple ANNs for each cumulative production curve. The next step involves reconstructing the original geological uncertainties. In reservoir engineering, returning to the original space after the Data Assimilation for Uncertainty Reduction (DAUR) process is essential to validate the proxy's ability to retrieve information while effectively reducing the input dimensionality. Finally, we address the challenge of quantifying proxy model uncertainty, a less explored problem in the literature, by using a set of Meta Activation Functions (MAFs) to assess forecast confidence through uncertainty analysis of proxy outputs. We evaluate the three proposed methods using a dataset from an offshore Brazilian field, demonstrating the effectiveness and practicality of our approach.
Examination Board
Headlines:
| Anderson de Rezende Rocha | IC / UNICAMP |
| Karla Tereza Figueiredo Leite | IME/UERJ |
| George Luiz Medeiros Teodoro | DCC / UFMG |
| Marcelo da Silva Reis | IC / UNICAMP |
| Celmar Guimaraes da Silva | FT / UNICAMP |
Substitutes:
| Marcos Medeiros Raimundo | IC / UNICAMP |
| Alessandra Davólio Gomes | CEPETRO / UNICAMP |
| Adriano Alonso Veloso | DCC / UFMG |
| João Paulo Papa | FC / UNESP |