| dc.description.abstract | Assisted History Matching (AHM) has become a common practice in reservoir simulations. The algorithms used try to obtain through a function the models that best approximates the production data recorded during the reservoir life. There are many different AHM tools that try to optimize the process by improving the exploration of parameter uncertainties and speeding up the convergence to one or more solutions.
ResX is an AHM software that works with ensemble-based methods and integrates static and dynamic modelling in the history matching process. This thesis's motivation is to understand and study ResX’s geological reliability, as well as the concepts and workflows involved in the methodology. The objective is to identify the limitations and propose future improvements that help integrate the static model with the dynamic model in a better way.
For this purpose, a synthetic model (“Real Model”) was created from scratch, with no real field data, from which all its characteristics and parameters are known. Three production wells and one injector were included, and a simulation case was run. After this, a new model (“Field Model”) was built from well logs and production data of the synthetic model. With it, a prior collection of models (Initial Ensemble) was created to be history matched using ResX. The HM process in this study consisted of experiments that challenged ResX and tested how the software reacts to inputs that are intentionally wrong, like unrealistic contacts and porosity ranges. The results of the ensemble were analyzed and compared with the Real Model to check ResX geological reliability.
The results indicate that ResX methodology and workflows are still ignoring the concept model of geologists. There is a limitation to test certain parameters, like the presence of faults or geological objects (channels, fan lobes, etc.). The inclusion of this could be a great improvement to the methodology and would allow a better geological understanding of the fields. Also, although ResX being a useful HM tool, the software is not intelligent enough to identify by its own that some parameter uncertainty ranges are incorrect and tends to use extreme porosity and permeability values if allowed, which can lead to unreliable geological models if not use consciously. Future HM experiments could help identifying more limitations to propose further improvements. | |