Determining optimal well placement and control is essential to maximizing production from an oil field. Until recently, these have usually been treated as separate problems; namely one of determining optimal control schemes for wells already in place, or one of placing wells assuming some fixed control strategy. Determining a truly optimal configuration of wells, however, requires that the control parameters be allowed to vary as well, since the objective (e.g. net present value of produced oil, computed from a reservoir simulation) has a complicated dependence on both types of parameter. This presents a challenging optimization problem, since the behaviour of the objective with respect to these two types of parameter is quite different. Furthermore, evaluating the objective function is expensive, and one is usually limited to using black-box optimization approaches due to the unavailability of gradient information from the simulator.
In this talk I will discuss how to address the placement and control optimization problem jointly using approaches that combine a global search strategy (particle swarm optimization, or PSO) with a local generalized pattern search (GPS) strategy. These two searches can be hybridized within a single algorithm, or applied to the problem in a sequential way. I will present the results of several numerical experiments and discuss which approaches are most successful for different production scenarios.