Case Study: Assessment of Predictive Capability of Reservoir Simulators for Waterfloods in Carbonates: How Realistic is My Simulation Model?

Abstract Conventional reservoir modeling practices have proved not to be sufficient to capture the true complexity of reservoir architecture and multiscale heterogeneity which represent serious challenges to production performance. Therefore, the objective of this study is to evaluate various techniques to build realistic reservoir models to improve the accuracy of the waterflood recovery forecast in carbonate reservoirs. A systematic simulation study was performed to investigate the subject problem leading to a proposed solution. Petrophysical properties utilized in the model are based on a geological description from the Permian basin. To mimic the fluid flow in the reservoir during production history, the way of representing the dynamic performance of the geological model, rather than the static model itself was considered as the key factor for this study. Therefore, capturing geological heterogeneity in discretized flow models becomes crucially important. Hence, different geological constraints were applied to the models, amongst them, no flow thin layers at different intervals, depositional sequence, and permeability cutoffs. At the equivalent pore volumes of injection, the effect of well spacing on the recovery factor was also investigated and compared to the literature data and leading correlations. The results of the case study indicate that the sweep is significantly high in reservoir models because they are not able to capture inherent reservoir heterogeneities. However, we show that even when heterogeneity is added, the waterflood recovery factor forecast is not drastically affected. The results were validated by comparing them with field data and infill drilling correlations, suggesting that reservoir models tend to provide more optimistic forecasts. In addition, conventional methods to perform the history match, such as permeability cutoffs and relative permeability endpoints, that can influence the mean free path between the wells, do not have the desired effect leading to unrealistic results in predictions. At present, conventional practices in reservoir modeling are not completely honoring the field data considering the complexities and the sub-grid physics, leading to the overestimation of waterflood recovery factors. This case study suggests a methodological approach and provides more insights in terms of the effect of heterogeneities, thus estimating more realistic water flood recovery factors from a simulation study. Also, emphasizes the importance of closely representing the reservoir architecture for realistic reservoir modeling. Multiple techniques are proposed here to identify the problem and mitigate the challenges. We also propose a measure of the mismatch between the predictions and the observations for various approaches. Overall, the key takeaway from this study emphasizes the need for greater caution when placing trust in simulation studies predicting waterflood recovery factors (RF) for subsequent field development management and planning.