Robust identification of the optimal well count and corresponding well locations/trajectories under the influence of subsurface uncertainty is a very challenging yet crucial element of field development plans. The well-count optimization (WCO) component of the problem involves an integer-valued optimization parameter because of the discrete nature of the well count. On the other hand, the well-location, well-trajectory, and well-opening-time optimization components typically feature real-valued continuous optimization parameters. The combined optimization problem encompasses both integer- and real-valued parameters. This class of problems is referred to as the mixed-integer nonlinear optimization problem in the literature. It is one of the most challenging classes of optimization problems and requires specialized optimizers. We have developed a new optimization protocol and accompanying optimizers to enable the joint optimization of well-count, well-locations (including well-trajectory parameters), and well-opening time within the framework of an industrial-grade subsurface field development optimization (sFDO) platform which encompasses state-of-the-art well-location optimization (WLO) capabilities. We have extended effective fault-tolerant global and local optimizers to solve the mixed-integer nonlinear optimization problem associated with the presence of well count in addition to real-valued parameters and optionally including well-opening-time parameters. We have introduced a novel well-priority-order concept to enumerate the wells subject to WCO. The resulting novel capability is simply coined as the simultaneous well-count optimization (SimWCO) method.
The SimWCO method is first validated on a small yet well-known benchmark model. Investigations include assessing the effects of the objective-function type, well-priority order, and robust optimization including the optimization of well-opening-time parameters. SimWCO has been field tested on two real-life field development optimization problems. SimWCO has been compared with computationally intensive concurrent WCO and cumbersome-to-apply reverse- and forward-creaming WCO techniques as part of the field testing work. Reservoir A case involves optimization of well count and locations of up to three vertical production wells. It is a robust optimization application involving multiple subsurface realizations. The optimal well count and locations are sought in a single perforation target zone in this test. Reservoir B case involves the optimization of up to six deviated production wells in addition to eleven active production wells. This case features well-by-well assignment of perforation-target zones, a new development in sFDO. Results of this work demonstrate the viability of simultaneous optimization of well count, well locations, and optionally well-opening times on realistic real-life field development optimization problems using an industry-grade distributed parallel optimization framework. We also quantitatively demonstrate that SimWCO is more than three-fold computationally more efficient compared to alternative techniques in the Reservoir B field test.