Substituição de simuladores de fluidos em metodologias de ajuste de histórico automático com o uso de regressão simbólica

12/05/2017

A geologically consistent reservoir model that is able to properly reproduce the behavior of the simulated reservoir may be attained by geostatistical history matching. In this domain, the existing methodologies for geostatistical models have become important due to their capacity of reproducing production data while at the same time allowing to evaluate the uncertainty on the adjusted models. However, due to the long time used in this process, especially because of the countless number of fluid flow simulations necessary for a good history matching, there is the need of finding a more efficient method to obtain the production curves of these simulations. This work studies the use of an algorithm of symbolic regression as an alternative for the use of fluid flow simulators, in order to reduce the time spent with such simulations. This work starts with the study of the symbolic regression tool used in it, especially its capacity for the desired application, and then the partial replacement the fluid flow simulator on the simulation of petrophysical models of reservoirs on a geostatistical history matching loop. Watt-field is the reservoir in which the methodology was tested. The results achieved for the studied reservoir show an equivalence on the quality between the original and the proposed methodology with symbolic regression. However, the possibility of, in less time, increasing the number of simulations per iteration on the proposed methodology allows an even more efficient convergence due to the ampliation of the exploration space of the model’s parameters.

Authors in the community:

• 

  E

  Erik Eidy Shimabukuro

  ist424461

Supervisors of this institution:

• 

  L

  Leonardo Azevedo Guerra Raposo Pereira

  ist168809

civil-engineering - Civil engineering

por - Portuguese

Instituto Superior Técnico