Comparing stochastic fault networks before flow simulation using geometric criteria.

in: Proc. 33rd Gocad Meeting, Nancy

Abstract

Stochastic structural modelling is now able generate a large ensemble of possible geological models. However, efficient comparison between these possible models is needed, for instance to select representative ensemble members for history matching. In this paper, we propose to compute distances between stochastically generated fault networks. We define six distance metrics to group fault networks which have similar dynamic behaviour: (1) the number of faults, (2) the number of fault blocks, (3) the minimum distance between a well and the fault network, (4) the number of faults between an injection and a producer well, (5) the angle of the fault network projection from a well, and (6) the projected fault network area. These criteria are tested on 563 synthetic fault networks and allow to define groups using a K-means clustering.

Download / Links

BibTeX Reference

@INPROCEEDINGS{ChauvinGM2013,
    author = { Chauvin, Benjamin P. and Julio, Charline and Caumon, Guillaume },
     title = { Comparing stochastic fault networks before flow simulation using geometric criteria. },
 booktitle = { Proc. 33rd Gocad Meeting, Nancy },
      year = { 2013 },
  abstract = { Stochastic structural modelling is now able generate a large ensemble of possible geological models. However, efficient comparison between these possible models is needed, for instance to select representative ensemble members for history matching.
In this paper, we propose to compute distances between stochastically generated fault networks. We define six distance metrics to group fault networks which have similar dynamic behaviour: (1) the number of faults, (2) the number of fault blocks, (3) the minimum distance between a well and the fault network, (4) the number of faults between an injection and a producer well, (5) the angle of the fault network projection from a well, and (6) the projected fault network area. These criteria are tested on 563 synthetic fault networks and allow to define groups using a K-means clustering. }
}