HM-WELLSTOC: A hierarchical graph-based method for multiple well stochastic stratigraphic correlation

in: 2018 Ring Meeting, ASGA

Abstract

Stratigraphic well correlation is an essential and difficult step of subsurface studies, which de- termines the stratigraphic architecture and therefore impacts sediment budget calculations and subsurface heterogeneities. Stochastic correlation can potentially generate a set of likely scenarios, but exploring this space is computationally very challenging, especially in the presence of multi- ple wells. We introduce the HM-WELLSTOC: A hierarchical method for multiple well stochastic stratigraphic correlation. This method automatically produces likely scenarios from a large set of vertical or subvertical wells. It generalizes the n-best version of Dynamic Time Warping (DTW). It starts by generating the n-best correlations from sets of two wells, and stores the result as a directed acyclic transition graph whose nodes represent stratigraphic units and edges represent possible the vertical stacking of these units. Groups of wells are then correlated by larger and larger clusters untill all units have been associated. This is achieved by a variant of n-best correlations method directly acting on the transition graphs instead of the wells themselves. As the graph size increases exponentially, the resulting graphs are built to ensure minimal distances between the possible cor- relations, and are pruned to keep the number of solutions manageable. This method is easy to parallelize and can handle hundreds of wells quite efficiently.

Download / Links

BibTeX Reference

@INPROCEEDINGS{,
    author = { Antoine, Christophe and Caumon, Guillaume },
     title = { HM-WELLSTOC: A hierarchical graph-based method for multiple well stochastic stratigraphic correlation },
 booktitle = { 2018 Ring Meeting },
      year = { 2018 },
 publisher = { ASGA },
  abstract = { Stratigraphic well correlation is an essential and difficult step of subsurface studies, which de-
termines the stratigraphic architecture and therefore impacts sediment budget calculations and
subsurface heterogeneities. Stochastic correlation can potentially generate a set of likely scenarios,
but exploring this space is computationally very challenging, especially in the presence of multi-
ple wells. We introduce the HM-WELLSTOC: A hierarchical method for multiple well stochastic
stratigraphic correlation. This method automatically produces likely scenarios from a large set of
vertical or subvertical wells. It generalizes the n-best version of Dynamic Time Warping (DTW). It
starts by generating the n-best correlations from sets of two wells, and stores the result as a directed
acyclic transition graph whose nodes represent stratigraphic units and edges represent possible the
vertical stacking of these units. Groups of wells are then correlated by larger and larger clusters
untill all units have been associated. This is achieved by a variant of n-best correlations method
directly acting on the transition graphs instead of the wells themselves. As the graph size increases
exponentially, the resulting graphs are built to ensure minimal distances between the possible cor-
relations, and are pruned to keep the number of solutions manageable. This method is easy to
parallelize and can handle hundreds of wells quite efficiently. }
}