Abstract
The objective is to couple tens of full-field giant reservoirs and apply optimized-based well management system to the all the flowing wells to manage a high level production plateau and injection system.
In this work each individual reservoir simulation model is a giant field, with a large historical (70y) and prediction (200y) time periods, 10s of millions of cells and thousands of wells, many injection and production constraints and drilling targets. Each of these giant fields requires a parallel supercomputer to run each individual model in a reasonable time. To couple many of these models they are all executed as one combined parallel job, which is subdivided in groups of processes dedicated for individual reservoirs. A layered communication hierarchy is used, which facilitates coordination and synchronization of the individual reservoirs and the optimization and allocation of the production and injection targets of every well and group of wells in the coupled system. Great effort was placed on making the user workflow very simple; having established a history and prediction field-scale model, the coupling can be achieved by a straightforward manipulation of the individual field-scale models, and high-level constraints and targets can be added to the coupled system. To alleviate some run time/resource issues, the level of detail required for each model was chosen by activating a coarsening facility within the reservoir simulator. This reduced the simulation resources needed for individual fields, but exacerbated the potential bottleneck of the optimization engine, as all well and group constraints were preserved in the coupled model.
To stress test this system, up to 15 giant models were coupled, with over 15 thousand wells managed by an optimized based well management system. The single node responsible for allocation of all the wells was not found to be a bottleneck in the system.
The journey from the detailed full-field giant model to a coupled super-giant coupled model has many challenges. Here, these are alleviated by tight and efficient integration of the coupling optimized-based well management system and the massively parallel reservoir simulator. This paper demonstrates a new industry high water mark for the coupling of reservoir models.
