Physics-Based Model Accelerates Deepwater Greenfield Development

_ This article, written by JPT Technology Editor Chris Carpenter, contains highlights of paper SPE 228190, “Accelerating Field-Development Planning for Deepwater Greenfield With Multiple Realizations,” by Davud Davudov, SPE, Resermine, and Alireza Iranshahr, SPE, and Namit Jaiswal, SPE, Shell, et al. The paper has not been peer-reviewed. _ Field-development planning (FDP) is a time-consuming process that can take huge numbers of simulation runs to determine optimal development. Computation time and complexity increase even further with multiple realizations. In this research, a reduced physics-based model is presented that accelerates the ability to screen infill-location optimization. In particular, the fast marching method (FMM) is used to rank and optimize well locations and trajectories to maximize oil production for a given reservoir with multiple realizations. Overcoming FDP Challenges A strong FDP involves a development strategy that works best for all likely scenarios, not just the most likely. The conventional method for analyzing development scenarios combines optimization algorithms with full-physics numerical reservoir simulators to identify the optimal well number, type, and location. However, a significant computational bottleneck exists with this method. For a high-resolution model, it can take several hours to run a single simulation. Individuals must severely restrict the number of scenarios they evaluate when the computational time required to repeat this process for thousands of possible well configurations in dozens of geological realizations becomes expensive. To overcome the computational barrier, the industry increasingly has turned to reduced-order or reduced-physics models. These models, often termed proxies or surrogates, aim to approximate the dynamic behavior of the reservoir with sufficient accuracy for screening and optimization but at a fraction of the computational cost. A certain complexity exists for these proxies. At the simpler end are static indices that combine static properties but fail to capture dynamic fluid flow. Positioned between these proxies and full-physics simulators are semianalytical, reduced-physics models. FMM, which achieves notable computational speedup while maintaining important physical principles, is a particularly effective example. FMM efficiently solves the Eikonal equation, which can be derived as a high-frequency asymptotic solution to the pressure diffusivity equation. The solution yields the diffusive time of flight (DTOF), a scalar field representing the propagation time of a pressure front from a source. The DTOF implicitly accounts for reservoir heterogeneity and provides a dynamic measure of drainage-volume expansion, making it a much more robust proxy of reservoir performance than static indices. This study describes and applies a novel, combined, accelerated workflow for FDP under uncertainty. The primary contribution is the combination of the FMM as a high-fidelity, reduced-physics proxy with a genetic algorithm for robust global optimization. This workflow is applied at scale to a complex deepwater greenfield asset, using a set of 30 distinct geological realizations to optimize well placement and conduct a comprehensive analysis. Although earlier research has shown that FMM is useful for ranking realizations, this paper extends that concept by using FMM-derived maps as the direct objective function for well-placement optimization within and across the complete set of realizations, in addition to ranking. This provides a complete, end-to-end solution for accelerated FDP.