Optimizing Well Trajectories in Steam-Assisted-Gravity-Drainage Reservoir Development

Abstract

Summary An important task before starting a steam-assisted-gravity-drainage operation is optimizing the location of the well trajectories. There are shale barriers and other heterogeneous features at different positions in these reservoirs. Steam cannot pass a thick shale barrier. Also, if there is a shale barrier between the injector and producer, oil cannot drain to the producer. For these reasons, optimizing the injector- and producer-well trajectories is important. Running the reservoir simulator is time-consuming, and running trial-and-error cases to optimize the producer and injector trajectories is not practical. On the other hand, robust well-trajectory optimization must consider uncertainty in the reservoir parameters. This optimization problem cannot be solved by calling the simulator for many possible trajectories and geostatistical realizations. To tackle this problem, a new semianalytical approximate thermal simulator modeled after Butler's theory (a proxy) has been developed and tested on different synthetic and realistic history-matched 2D and 3D models (Dehdari and Deutsch 2013; Dehdari 2014). Many modifications, extensions, and improvements have been made to Butler's original model. This proxy can be used as a substitute to the reservoir simulator. In this paper, two methods are tested for well-trajectory optimization. The first method is based on random sampling from a 3D box that has been selected for drilling wells. Then, the differential-evolution-optimization algorithm has been used for automatically improving the trajectory location. The second method is based on parameterizing the trajectory by use of a Hermite spline, then optimizing the parameters of the spline. The producer and injector trajectories of a realistic model with a single realization and a synthetic model with 100 realizations have been optimized by use of these methods.