Unisim-iv: benchmark proposal for light oil carbonate reservoir with high co2 content

The Brazilian pre-salt fields are carbonate reservoirs with good quality oils, but they can present high amount of CO2 in dissolution, which leads to a high amount of produced gas and can limit oil production. Therefore, the development and management of fields with those characteristics are complex tasks that involve many decisions, with a large number of variables to be considered. Thus, numerical simulation plays an important role in overcoming the challenges that arise from the management of these fields, integrating different subjects such as geosciences and reservoir characterization, data assimilation, production facilities, production optimization processes, economic evaluation, and decision analyses under uncertainty. Open source benchmarks are often used in numerical simulation studies to evaluate and compare techniques and methods, using the same comparison basis. The objective of this paper is to present UNISIM-IV, a set of carbonate benchmarks analogous to a pre-salt field, adding new possibilities to the scientific community and organizations that can improve workflows in the context of reservoirs with the characteristics mentioned above. The benchmark is divided into four different cases: UNISIM-IV-2019, UNISIM-IV-2022, UNISIM-IV-2024, and UNISIM-IV-2026, where the date refers to the date of the analysis. The main differences among these cases involve the stage of field’s life cycle, ranging from early development phase (2019) to a developed reservoir with eight years of production (2026). Thus, the available history data and the mapped uncertainties differ between the cases. The users can choose the case that best suits their needs, depending on specific research objectives. Each of these cases comprise: (1) an ensemble of prior uncertainties, (2) production, injection, and pressure history data, and (3) a history-matched simulation model suggested as a base case. There is also a reference case, named UNISIM-IV-R, which consists of a model with a very refined grid and known information used as the “true response” to generate all data that could be measured in a real field, such as production history and well logs.