APPLICATION OF FAST RESERVOIR SIMULATION CAPACITANCE-RESISTANCE METHOD TO PREDICT THE HOT WATER FLOODING PERFORMANCE

Abstract

A range of methods is available to assess a reservoir performance. Development and application of fast methods to evaluate the performance of a recovery method and provide a general picture of injectors/producers connectivity is critical to manage a reservoir. Capacitance Resistance Model (CRM) is a useful tool for improving real-time flood management, as it allows rapid modeling and simulation of gas and water flood recovery processes. The CRM approach is based on signal processing methods in which injection rates are accepted as input signals and production flow rates are considered as reservoir response or output signals. The model offers key advantages, including simplicity, immediate results, and optimal performance even with minimal initial data. Over recent years, enhancements in CRM have established it as a reservoir management tool, enabling essential tasks like history matching of production data, forecasting production rates, scheduling injection rates, detecting injection leakage, and estimating fracture distribution (Sayarpour, 2008). In this study, we expanded the application of CRM to predict the behavior of hot water injection processes. Systems identification is applied for history matching using only injection/production data from commercial simulator to characterize the reservoir models where injection of hot water was applied, evaluating interwell connectivities and time constants. Four case studies were developed with two different injection fluid types. These included a homogeneous model with a five-spot well pattern (Case 1), models featuring high-permeability streaks (Case 2 and 3), and a heterogeneous reservoir model (Case 4). In these cases, bottomhole pressures and production rates remained constant, while injection rates fluctuated over the simulation period. The first three cases were analyzed to predict reservoir performance analytically under specific conditions for homogeneous scenarios. The highest calculated average error was observed during Case 2 for both total liquid production and oil production rates (10.84% and 11.79%, respectively), while the minimum average error values were found in Case 4, with values of 6.50% for liquid rates and 5.76% for oil production rates. In all cases, the results of the developed models exhibited satisfactory agreement with those of a grid-based commercial simulator. We considered these hypothetical cases where modifications were applied to generate a more reliable evaluation of interwell connectivity and time constants, and used the R-squared value of the model as a fitting parameter for history matching processes. This approach, applied across multiple cases, yielded excellent evaluations of both reservoir performance and well connectivity.