整篇论文的目的是提出一种混合方法，将物理学与数据驱动方法相结合，以高效、准确地预测共同开发下的非常规油井的性能。作者提出的方法是油藏图网络（RGNet）模型。通过降低系统复杂性，同时保持基本物理原理，该模型提供了一种高效、准确的方法来建模、历史匹配和预测非常规井。与需要几小时到几天才能运行的完整物理模型相比，所描述的模型只需要几秒钟到几分钟。
非常规油藏开发是一个复杂的过程，涉及水力压裂水平井的选井、时机、井距和完井设计。
The aim of the complete paper is to propose a hybrid approach that combines physics with data-driven approaches for efficient and accurate forecasting of the performance of unconventional wells under codevelopment. The method the authors propose is the reservoir graph network (RGNet) model. By reducing system complexity while maintaining fundamental physics, the model provides an efficient and accurate way to model, history-match, and predict unconventional wells. Compared with a full-physics model that takes from hours to days to run, the described model only takes from seconds to minutes.
Developing unconventional reservoirs is a complex process involving well targeting, timing, spacing, and completion design for horizontal wells with hydraulic fractures.
