An oil or gas field may comprise a number of isolated reservoir units, each of which may have been studied separately with their own simulation models. If these reservoirs are to be linked to common surface facilities, they are effectively coupled by global production (and perhaps injection) constraints. Their production and injection streams must be combined to obey these constraints. Two approaches are commonly applied to this problem: either amalgamating the simulation models into a single model of the whole field, or using production planning methods on pre-constructed production profiles of the individual reservoirs. The former requires a great deal of effort from the engineers, and may result in a large unwieldy model, while the latter may not be accurate because of rate-dependency effects which are ignored.
This paper introduces a third approach to the problem, which we call Reservoir Coupling. The individual simulation models are still run as separate processes, with minimal change to their data. But they are coupled to a master process which handles the global production and injection constraints and determines the flow rate targets to be applied to the individual reservoirs at each time step. Inter-process communication is handled by the PVM message passing system. This can spawn processes over a network of heterogeneous Unix computers, enabling the individual simulation models to be run in parallel on separate workstations.
The paper describes the Reservoir Coupling algorithm that has been implemented in a commercial black oil simulator, and discusses the advantages and disadvantages of the technique in comparison with the other two approaches. A simulation study is described which illustrates the application of Reservoir Coupling to three isolated reservoirs subject to global production and injection constraints.