24. 利用简化代理模型进行页岩资源开发优化: 以Barnett页岩为例

OPTIMIZATION OF SHALE RESOURCE DEVELOPMENT USING REDUCED-PHYSICS SURROGATE MODELS

本文提出了一种将计算优化应用于页岩气藏开发的一般工作流程。首先从包含非均质地质、高分辨率裂缝网络、双孔隙、双渗透区域和气体解吸的详细全物理模拟模型开始，方法首先涉及生成一个更简单、计算效率更高的缩减物理代理模型。通过类似于历史匹配的程序调整代理模型，使其结果与全物理模型紧密一致。然后使用代理模型进行油气田开发优化。在优化过程中，定期对代理模型进行重新训练，以匹配当前最佳解决方案的全物理表示。在这里考虑的优化中，我们应用了直接搜索技术（广义模式搜索），并寻求确定一组水平井的最佳位置、长度和裂缝阶段数量。在两个涉及具有巴尼特页岩代表性非均质属性的三维模型的例子中，优化被证明可以提供净现值是基础案例设计的两倍以上的场景。

CMG软件应用情况

在这项工作中，所有的全物理模拟都是使用CMG的IMEX黑油模拟器（CMG，2010）进行的。该软件用于执行包括双孔隙、双渗透模拟、显式裂缝描述、局部细化网格、非达西流动效应和气体解吸在内的复杂模拟。CMG的IMEX模拟器作为一个高效的工具，帮助研究者在优化过程中评估不同的油气田开发方案。

Abstract

The economics of oil and gas field development can be improved significantly by using computational optimization to guide operations. In this work, we present a general workflow for applying optimization to the development of shale gas reservoirs.Starting with a detailed full-physics simulation model, which includes heterogeneous geology, highly-resolved fracture networks, dual-porosity, dual-permeability regions, and gas desorption, the approach first entails the generation of a much simpler, and much more computationally efficient, reduced-physics surrogate model. The reduced physics model is tuned using a procedure akin to history matching to provide results in close agreement with the full-physics model. The surrogate model is then used for field development optimization. During the course of the optimization, the surrogate model is periodically retrained to match the full-physics representation of the current best solution.

In the optimizations considered here, we apply a direct search technique (generalized pattern search) and seek to determine the optimal locations, lengths, and number of fracture stages for a set of horizontal wells. In two examples, involving three-dimensional models with heterogeneous properties representative of the Barnett Shale, optimization is shown to provide field development scenarios with net present values that are more than double those of base case design.