Robust data-driven process for improved description and optimal management of oil and gas reservoirs

你有没有想过倒着解决问题？也就是说，从答案上看，试图发现问题？在一般意义上，这是这里描述的研究计划的主要目标。它的最终目标是开发计算效率高的技术，利用地面上所有可用的信息来表征或描述地下“储油层”(即自然储存石油和天然气流体的多孔地下岩石)的结构和性质。从一个特定的储集层中可以回收的石油和天然气的预期数量，直接与其大小和内部多孔介质的性质有关。因此，我们可以想象，了解油藏的形状和大小以及其内部特性对于整个油藏管理过程是多么重要。问题是，没有直接进入储集岩的通道。这意味着，建设水库的这一“形象”不是一个小问题。事实上，如上所述，这是一个逆问题。从数学上讲，我们将其描述为“不适定问题”，因为有几种解决方案都是可以实现的，但只有一种是正确的。要找到反问题的正确解决方案，策略是考虑尽可能多的数据。对于油藏描述问题，可以考虑两类数据：静态数据(如岩心、测井和地震数据)和动态数据(如不稳定压力、饱和度和流量)。目前，大多数可用的技术只允许我们使用有限来源的数据。其结果是，储集层性质的不确定性比可能的高，储集层管理不能优化。拟议的研究计划的目标是开发计算效率高的技术来生成油藏模型，该技术使用静态和动态生产数据。成功开发更有效的油藏描述方法将为石油运营商提供改进的油藏管理技术，从而增加石油和天然气的产量，并提高效率、降低成本和保护环境。

Have you ever thought of solving a problem backwards? That is, from the answer, tried to discover the question? In a general sense this is the main objective of the research program described here. Its final goal is to develop computationally efficient techniques that use all available information at the ground surface, to characterize or describe the structure and properties of an underground 'oil reservoir' (that is, porous subsurface rock in which oil and gas fluids are naturally stored). The expected amount of oil and gas that can be recovered from a particular reservoir, is directly related to its size and internal porous media properties. Thus, we can imagine the tremendous importance of knowing the reservoir shape and size, as well as its internal properties, for the overall reservoir management process. The problem is that there is no direct access to the reservoir rock. This means that the construction of this "image" of the reservoir is not a trivial problem. It is, in fact, an inverse problem, as mentioned above. Mathematically speaking, we describe this as an 'ill posed problem', since several solutions are equally achievable, but only one is correct.To find the correct solution to an inverse problem, the strategy is to consider as much data as possible. For the reservoir description problem two classes of data can then be considered: static data (such as core, log, and seismic data) and the dynamic data (such as transient pressures, saturations and flow rates). At present, the majority of the available techniques only allow us to use data from limited sources. As a result, the uncertainty in reservoir properties is higher than it could be, and reservoir management cannot be optimized. The objective of the proposed research program is to develop computationally efficient techniques for generating reservoir models, which uses both static and dynamic production data. Successful development of more efficient reservoir description methodologies will provide oil operators with improved reservoir management techniques that can lead to increased oil and gas production and more efficient, less costly and environmentally safer operation.