SPE Annual Technical Conference and Exhibition, (4), 2024
DOI: 10.2118/221062-ms
Abstract In water-drive reservoirs, predicting variable water influx to individual wells poses significant challenges due to geological heterogeneity and diverse production strategies. This study introduces a novel capacitance-resistance model (CRM) that robustly incorporates natural water influx into oil reservoirs from aquifer systems, eliminating the need for average reservoir and aquifer pressure versus time data. The proposed aquifer-integrated CRM addresses the pressure change equation between the aquifer and reservoir systems using fundamental CRM parameters including aquifer pore volume and productivity index. The development process focuses on the CRM with producer representation (CRMP) and reformulates it to calculate instantaneous water influx based on reservoir and aquifer pressure differences, naming this model CRMPAQ. The history matching algorithm utilizes Sequential Least-Squares Programming (SLSQP) to manage both inequality and equality constraints, employing linear inequality constraints to ensure that the summation of aquifer-producer connectivity and well-to-well connectivity does not exceed unity. Additionally, a linear equality constraint ensures that the summation of producer pore volumes equals the total reservoir pore volume. The CRMPAQ model is designed to match total production rates and bottomhole pressures (BHPs), if available. Numerical experiments conducted using CRMPAQ on both 3D synthetic and realistic field datasets demonstrate its effectiveness. Comparisons with results from the commercial simulator ECLIPSE, which accounts for saturation dependency on well productivity, further validate the model. The realistic field dataset comprises three wells from a heterogeneous reservoir with natural water influx, frequent shut-ins, and variable skin effects, with daily records of production rates and BHPs. The results indicate that CRMPAQ yields comparable outcomes and effectively emulates the behavior of high-fidelity reservoir simulators in terms of total liquid production rates and BHP data. CRMPAQ is significantly more computationally efficient; two to three orders of magnitude faster than high-fidelity simulator history matching. Furthermore, when matching BHP data using flow rate history, CRMPAQ accurately accounts for the effects of frequent shut-ins as well as the variable productivity indices on parameter estimation. The model was tested on synthetic datasets and validated with a realistic field dataset (used in SPE-210102-PA). The realistic dataset considers three wells produced from a heterogeneous reservoir with natural water influx, frequent shut-ins, and variable skin effects. The wells' production rates and BHP data are recorded at daily time intervals.