Field development optimization, in which well configuration, well types, and well controls are determined, represents a computationally demanding mixed integer nonlinear programming problem. Such problems may require very large numbers of function evaluations, and if each of these corresponds to a detailed flow simulation, the optimization can become intractable. In this paper, we incorporate a set of surrogate treatments (STs) into the field development optimization problem. The basic ST is a variant of a recently developed surrogate procedure for optimizing well rates. It entails the solution of two optimization problems that both involve simplified physics (unit-mobility ratio displacement) and can be solved very efficiently. In the first problem, we find optimal well-rate ratios (i.e., the fraction of total injection or production allocated to each well), while in the second problem we determine optimal overall field injection and production rates. This ST is incorporated into a particle swarm optimization (PSO) framework. Three treatments are considered for subsequent optimization steps. All of these approaches involve full-physics simulations, and two of the methods entail the use of mesh adaptive direct search (MADS). The ST-based procedures are evaluated for two different 3D problems involving waterflood (with mobility ratios of 2 and 5) and water-alternating-gas (WAG) injection. The surrogate treatments are compared with standard approaches involving PSO, MADS, and a PSO-MADS hybrid. Extensive optimization results demonstrate that the ST-based methods provide consistent improvement in optimizer performance. For example, in the WAG case, the ST-based approach gives an optimal net present value that is 3.2% higher than that achieved using standard PSO-MADS, while also providing a 2.4 × computational speedup.

确定油井配置，油井类型和油井控制的现场开发优化代表了一个计算要求很高的混合整数非线性规划问题。这样的问题可能需要大量的功能评估，并且如果每个评估都对应于详细的流程模拟，则优化可能会变得很棘手。在本文中，我们将一组替代处理（ST）合并到现场开发优化问题中。基本的ST是最近开发的用于优化井速的替代程序的变体。它需要解决两个优化问题，这两个问题都涉及简化的物理原理（单位迁移率比位移）并且可以非常有效地解决。在第一个问题中，我们找到了最佳的井速比（即 分配给每口井的总注入量或产量的百分比），而在第二个问题中，我们确定最佳的整体现场注入量和生产率。该ST被合并到粒子群优化（PSO）框架中。对于后续的优化步骤，考虑了三种处理方法。所有这些方法都涉及全物理模拟，其中两种方法需要使用网格自适应直接搜索（MADS）。针对两个不同的3D问题（包括注水率（流动比为2和5）和水交替气（WAG）注入），评估了基于ST的过程。将替代处理与涉及PSO，MADS和PSO-MADS混合的标准方法进行比较。大量的优化结果表明，基于ST的方法在优化器性能方面提供了持续改进。例如，