Chapter 12: Limitations of Data-Driven Reservoir Modeling

Like any other technology that is developed for specific purposes, top-down modeling (TDM) has many limitations, and it is most powerful when it is used under the right circumstances. One limitation of TDM is that it is only as valid as the degree to which the available data are representative of the fluid flow through the reservoir that is the subject of the TDM. This is an important issue because in data-driven modeling, we are limited by the physics that is historically present in the operations and is captured by the data.

That said, it is important that we not take this statement superficially. In other words, if operators only refer to their traditional experience and education with reservoir and operational characteristics as they apply to reservoir modeling and fluid flow through porous media, then operators may conclude that since they do not have permeability measurements from many locations in the field, and since knowledge about formation permeability is critical in estimating flow (production), TDM cannot be applied. This is an incorrect interpretation of the limitation of the TDM that is being addressed here. The reason is that because TDM uses artificial intelligence to make its estimations and to learn the fluid flow behavior in a reservoir, it can use other data, such as well logs, gamma ray, density porosity, neutron porosity, sonic, spontaneous potential, and resistivity, to create an internal representation of formation permeability and then correlate these data to production from the reservoir.