Well and Facility Placement Optimization in Multiple Flow Units Reservoirs

Abstract

After the discovery of an oil or gas field, an optimal field development plan must be developed. Selecting the optimal plan leads not only to cost savings, but also to maximized reservoir performance, and hence leads to a maximized projects’ net present value. The development plan mainly consists of two crucial tasks that affect the profitability of the project: well placement optimization and facility placement optimization. Well placement optimization allows placing wells in zones containing highest quantities of hydrocarbons initially in place. This requires the adoption of a flexible algorithm that accounts for non-communicating flow units in the reservoir and considers various well settings for each flow unit. Following well placement optimization, optimizing facility placement and size is crucial to ensure optimal drilling, production, gathering, and processing of hydrocarbons. This requires highly efficient and flexible facility placement algorithms that account for topological complexities and various constraints, leading to optimal hydrocarbon recovery while decreasing capital cost. In this work, well placement optimization is extended to multiple flow units reservoirs, followed by well completion to selectively produce hydrocarbons from intended zones. This allows for production from the entire reservoir while controlling fluid and sand production. Black hole optimizer is employed to optimally placing wells of different types (vertical/horizontal), spacing, and depths in multiple flow units. Open hole and perforation completion types are adopted. Additionally, to optimally place facilities, this work develops a clustering-based algorithm that optimizes facility placement layout in oil and gas field development projects. The developed algorithm honors facility nodes’ capacities and considers realistic well trajectories and pipeline paths. Four clustering approaches are implemented and compared: k-means, Hierarchical, Gaussian Mixture Models, and DBSCAN. The clustering-based algorithm is compared with a non-gradient algorithm, Particle Swarm Optimization, in terms of total facility cost and computational requirements. The comparison is performed on different scenarios and different levels of complexities that demonstrate the features of the developed algorithm. Results show that the developed algorithm significantly outperforms PSO in terms of cost minimization but presents higher computational load challenge.