A Semi-Implicit Approach for Integrated Reservoir and Surface-Network Simulation

Abstract

SummaryConventionally, methods of coupling reservoirs and surface networks are categorized into implicit and explicit approaches. The term "implicit coupling" indicates that the two simulators solve unknowns together, simultaneously, or iteratively, whereas "explicit coupling" indicates that the two simulators solve unknowns sequentially and exchange their boundary conditions at the last coupled time tn.The explicit approach is straightforward to implement in existing reservoir and surface-network models and is widely used. Explicit coupling does have drawbacks, however, because well rate and pressure oscillations are often observed.In this paper, a new semi-implicit method for coupled simulation is presented. This technique stabilizes and improves the accuracy of the coupled model. The "semi-implicit coupling" overcomes the problems found in explicit-coupling methods without requiring the complexity of a fully implicit coupled model.The new approach predicts inflow-performance-relationship (IPR) curves at the next coupled time tn+1 by simultaneously conducting well tests for all wells in the reservoir before actually taking the required timestep. All wells first flow simultaneously to the next coupled time tn+1 with the well rates unchanged from the last coupled timestep. The timestep is rewound, and all well rates are reduced by a uniform fraction and then simultaneously flow again to tn+1. By extrapolating the resulting well pressures, the well's shut-in pressures at time tn+1 are determined, and thus, straight-line IPRs are produced. The new IPR curves approximate better each well's drainage region at tn+1 and each well's shut-in pressure at tn+1 which helps to stabilize the explicitly coupled model.The new coupling technique normally does not require iteration between the reservoir and surface network and normally has the stability and accuracy characteristics of an implicitly coupled approach. Because the well tests already account for individual well-drainage regions, an explicit knowledge of the well-drainage region is not required. Because of the stabilized IPR, the approach also was found to reduce the overall computational time compared with explicit coupling.Applications of the new approach are presented that show significant improvements surpassing explicit coupling in both stability and accuracy.