Field Development Optimization with Stochastic Gradient Method: Application to a Multi-Reservoir Carbonate Field in the Middle East

Abstract

Abstract Field development planning activities involve decisions that entail multi-billion-dollar investments. Making right design choices is crucial to the techno-economic success of the project. In this paper we demonstrate how state-of-the-art numerical model-based optimization techniques can support asset teams in this complex decision-making process. Optimization of real-life field development cases can be computationally very demanding due to the cost of running large amounts of large-scale reservoir simulations, the large number of variables to be optimized and the necessity of accounting for uncertainties, among other reasons. In this work we employ TNO’s EVEReST optimization technology leveraging the StoSAG stochastic gradient-based method to achieve optimized solutions in computationally efficient manner. Besides its computational attractiveness, the StoSAG method also renders the optimization framework flexible to be customized to any specific optimization problem, e.g., optimization of any type of field development decisions, coupling to any industry-standard reservoir simulators and handling any type of objective and constraint functions. The optimization framework has been used to optimize the expansion of the development plan of a large field in the Middle East, in particular to support decisions concerning the drilling of (up to) 38 new wells and the upgrading of surface facilities to accommodate incremental production. The challenge is posed by the field being comprised of multiple carbonate reservoirs for which multiple types of decisions need to be optimized, often simultaneously i.e. well target locations, trajectory design features, number and type of wells and their distribution across the different reservoirs as well as the drilling sequence. An economic objective function was used with the operational constraints per reservoir accounted for within the framework. Optimization found non-trivial optimal solutions resulting in significant improvements in the economic objective. The optimal strategy revealed improved distribution of wells (and types) among the reservoirs, more suitable drilling order and superior well locations and trajectories compared to the initial strategy. Optimization found well locations and trajectories taking into account complex local well interaction in already densely populated reservoirs with wells. The optimized solutions were benchmarked against the development plan designed by the asset team, and the comparison of strategies confirmed the added value of numerical optimization as a tool to expedite the search of improved development strategies. The nature of the employed optimization method allowed optimized solutions to be achieved with a reduced number of reservoir simulations, which was crucial for the success of this study due to the time-consuming reservoir simulations involved. This showcases the computational advantage of the optimization method and highlights EVEReST as an enabler technology for optimization under uncertainty, where an ensemble of large-scale models is to be considered and the number of required reservoir simulations tends to be even larger.