Assessment of the uncertainties' impact of relative permeability curves on mature field oil production: a case study in unisim-i-h

Resumo

Relative permeability is a crucial parameter in reservoir characterization as it directly influences the behavior of the rock-fluid system and, consequently, oil production. However, estimating this parameter is difficult due to uncertainties that can originate from various sources, such as geological, parametrical modelling, computational, and so on. If not properly examined, these uncertainties might undermine prediction accuracy and impair final production. Over time, several correlations were created to estimate relative permeability better. Although no parametrical correlation can adequately predict the exact shape of the curve, most of them reasonably approximate it within specific uncertainty limits. For mature oil fields, a novel kind of industry for Brazil, the research of relative permeability and associated uncertainties is still in its early stages. The main objective of this study is to numerically investigate the effects of uncertainties on relative permeability curves via software CMG® and the geological model UNISIM-I-H developed by UNICAMP. The LET correlation was used to estimate the relative permeability curves. Initially, a base case known as UNISIM-I-H was simulated, upon which the remaining analyses were built. This case consisted of a 15-year production phase followed by a 5-year production pause, a 10-year production restart, the abandonment of the three less productive wells and the drilling of the three new wells to represent a mature oil field better. Four cases were created with relative permeability parameter uncertainties increased by 5%, 10%, 15%, and 20%. The effects of these uncertainties on the relative permeability curve, production curve, net product value, and recovery factor were then examined. The data analysis revealed that these uncertainties mostly affect the forecasts relating to net product value, reservoir recovery factor, and cumulative production, with the latter being the most impacted. Every uncertainty case increased the maximum and minimum values of the curves. Compared to the base case, an intriguing pattern also showed up, where the maximum increased much more quickly than the minimum, suggesting that the risk of estimating the values for each curve above the actual is significantly higher with increasing uncertainty than the opposite.