Field Production Optimization Using Closed-Loop Direct Feedback Control of Intelligent Wells: Application to the Brugge Model

摘要

Optimized production strategies using intelligent wells have been shown in numerous studies to improve economic performance. However, most optimization methods are model-based, effective only if the reservoir model captures the range of all possible reservoir behaviors at the individual well and completion level. This is seldom the case. Furthermore, reservoir models are rarely predictive at the spatial and temporal scales required to identify control actions. Motivated by this, recent studies have shown that direct feedback control, triggered by monitoring at the surface or downhole, can increase net-present-value (NPV) and mitigate reservoir uncertainty. This approach does not neglect model predictions entirely; rather, a model-based approach is used to optimize adjustable parameters in a generic feedback control algorithm. We evaluate the benefits of using direct feedback control for multi-well production optimization using the synthetic Brugge field case study. We test three inflow control strategies. Two are based on direct feedback control, but differ in the level of monitoring and control. In the first feedback control strategy, all monitoring and control is taken at surface, using surface multiphase flow meters and on/off well-head control valves. In the second, monitoring and control can take place either at surface or downhole, using on/off well-head and variable completion inflow control valves, in response to measurements from surface and downhole multiphase flow meters. These control strategies are optimized on a subset of the published model realizations; the other realizations are then used to simulate unexpected reservoir behavior. For benchmarking purposes, we implement a third, reactive rule based approach, heuristically developed with prior reservoir knowledge of the truth model. We also compare our results to previously published, model-based inflow control strategies developed by optimizing NPV with perfect knowledge of the Brugge truth case. Our results suggest that closed-loop direct feedback control, implemented at surface and downhole, can yield significantly higher NPV compared to surface feedback control alone. Moreover, despite the simplicity of the direct feedback control approach, the NPV returned is higher than a heuristic reactive approach, particularly when reservoir behavior is unexpected. In contrast to model-based optimization techniques, direct feedback control is straightforward to implement and can be easily applied in real field cases.