水气交替 (WAG) 注入是一种提高采收率 (EOR) 技术，用于克服与气体注入相关的问题，包括重力超控、粘性指进和通道。然而，WAG 注入的成功受到储层特性、注入器/生产器连接以及注入气水比 (GWR) 的影响。本研究提出了一种混合模型，该模型结合了电容/电阻模型 (CRM)、机器学习 (ML) 模型和石油模型，以评估和优化碳酸盐油田中的 WAG 注入器。

通过结合基于物理的降阶模型 (CRM) 和 ML 模型，可以实现注入井和生产井之间的井间连通性。CRM 和 ML 的混合组合提高了结果的可信度，消除了单独使用 CRM 模型的缺点。

接下来，将注入连通性结果与降阶幂律油模型相结合，评估注入量增加对石油产量的影响。最后，根据注入井对石油产量的贡献潜力对其进行优化和排序。

建议的工作流程应用于一个大型复杂碳酸盐油田，该油田有 93 多个生产井和 47 个注入井。用于获取井间连通性的 CRM 和 ML 的混合组合可将各井相互比较，从而提高结果的可信度。

使用 ML，汇总从多个信号获得的结果，以进一步识别和验证注入器/生产器对连通性。 CRM 与石油模型的进一步结合有助于根据变化的气水比 (GWR) 评估额外的采油量。 CRM 优化和 ML 连通性结果的结合有助于快速准确地对 47 个注入器进行排名和优先排序。选择了五个注入器进行现场测试，结果显示实施建议的注入计划后，石油产量显着提高。

结合 CRM 和 ML 实现连通性的混合模型解决了单个模型的缺点。通过新颖地使用混合模型来识别 WAG 注入的有效注入器可加速决策。所提出的方法可以扩展到具有许多注入器和生产器的类似 WAG 注入油田，以帮助优化注入策略。这种新方法有助于石油公司当前的数字化战略加速决策，尤其是在无法进行历史匹配的成熟油藏中。

在此处查找有关 OnePetro 的完整论文（SPE 218282）。

Water-alternating-gas (WAG) injection is an enhanced oil recovery (EOR) technique used to overcome problems related with gas injection, including gravity override, viscous fingering, and channeling. The success of WAG injection is influenced, however, by reservoir characteristics, injector/producer connections, and the injected gas water ratio (GWR). This study proposes a hybrid model that combines the capacitance/resistance model (CRM), a machine-learning (ML) model, and an oil model to assess and optimize WAG injectors in a carbonate field.

Interwell connectivity between injection and production wells is obtained by combining a physics-based reduced-order model (CRM) and an ML model. The hybrid combination of CRM and ML increases confidence in results, eliminating shortcomings associated with CRM models used alone.

Next, the results obtained from injection connectivity are combined with reduced-order power-law oil model to evaluate the impact of injection rate increase on oil production. Finally, injectors are optimized and ranked based on their potential to contribute to oil production.

The proposed workflow is applied to a large, complex carbonate field with more than 93 production wells and 47 injection wells. The hybrid combination of CRM and ML used to obtain interwell connectivity compares wells with one another, resulting in higher confidence in the results.

Using ML, results obtained from multiple signals are aggregated to further identify and verify the injector/producer pair connectivity. The further combination of CRM with an oil model helps evaluate additional oil recovery based on the changing gas/water ratio (GWR). A combination of CRM optimization and the ML connectivity results help rank and prioritize 47 injectors quickly and accurately. Five injectors are selected for field testing, and results show significant improvement in oil production after implementation of suggested injection schedules.

Hybrid models that combine CRM and ML to obtain connectivity address the shortcomings of the individual models. The identification of efficient injectors for WAG injection by the novel use of hybrid models accelerates decision-making. The approach presented can be extended to similar WAG injection fields with many injectors and producers to help optimize the injection strategy. This new approach helps with current digitization strategies in oil companies to accelerate decision making, especially in mature reservoirs where history matching is not available.

Find the complete paper (SPE 218282) on OnePetro here.