Uncertainty assessment in stratigraphic well correlation: a new stochastic method for carbonate reservoir.

in: Proc. 31st Gocad Meeting, Nancy

Abstract

An application of well correlation is to subdivide a reservoir into stationary intervals to support geostatistical modelling of static reservoir properties. In this scope, sequence stratigraphic well correlation appears to be an efficient technique, especially in carbonate sedimentary systems resulting from numerous and interdependent genetic processes. However, due to the complexity of sedimentary layers, well correlation is a hazardous process which is subject to many uncertainties. In this work, we propose to account for these uncertainties by generating several possible realizations of well correlations. The method is based on the Dynamic Time Warping (DTW) algorithm, whose efficiency has already been demonstrated in speech recognition and bio-informatics. Three derivative versions of the DTW algorithm are introduced: a stochastic one for correlating two wells (2D DTW); a multi-well one, offering the possibility to take into account for the 3D organization of stratigraphic architectures by analysing all input wells at once. Due to computation time and memory requirements, this version is limited to the correlation of up to ten wells. To address this issue, an iterative multi-2D DTW based on well correlation propagation is proposed. In addition, the proposed algorithms integrate stratigraphic order of well markers to perform the well correlation hierarchically (low frequency stratigraphic events are correlated first, and then are used to constrain the correlation of higher-order events). To perform stratigraphic correlation with DTW algorithms, one has to compute the likelihood of each possible horizon. Because of the high significance of chronostratigraphy in siliciclastic and carbonate reservoir modeling, the presented likelihood computation is based on paleo-angles consistency and facies and sedimentary profile coherency. Our stochastic correlation algorithm and the influence of stratigraphic correlation uncertainties on static reservoir modelling are demonstrated on outcrop data of the Cretaceous southern Provence Basin (Provence, S-E France).

Download / Links

BibTeX Reference

@INPROCEEDINGS{,
    author = { Lallier, Florent and Caumon, Guillaume and Borgomano, Jean and Viseur, Sophie and Royer, Jean-Jacques and Antoine, Christophe },
     title = { Uncertainty assessment in stratigraphic well correlation: a new stochastic method for carbonate reservoir. },
 booktitle = { Proc. 31st Gocad Meeting, Nancy },
      year = { 2011 },
  abstract = { An application of well correlation is to subdivide a reservoir into stationary intervals to support geostatistical modelling of static reservoir properties. In this scope, sequence stratigraphic well correlation appears to be an efficient technique, especially in carbonate sedimentary systems resulting from numerous and interdependent genetic processes. However, due to the complexity of sedimentary layers, well correlation is a hazardous process which is subject to many uncertainties. In this work, we propose to account for these uncertainties by generating several possible realizations of well correlations.
The method is based on the Dynamic Time Warping (DTW) algorithm, whose efficiency has already been demonstrated in speech recognition and bio-informatics. Three derivative versions of the DTW algorithm are introduced: a stochastic one for correlating two wells (2D DTW); a multi-well one, offering the possibility to take into account for the 3D organization of stratigraphic architectures by analysing all input wells at once. Due to computation time and memory requirements, this version is limited to the correlation of up to ten wells. To address this issue, an iterative multi-2D DTW based on well correlation propagation is proposed.
In addition, the proposed algorithms integrate stratigraphic order of well markers to perform the well correlation hierarchically (low frequency stratigraphic events are correlated first, and then are used to constrain the correlation of higher-order events). To perform stratigraphic correlation with DTW algorithms, one has to compute the likelihood of each possible horizon. Because of the high significance of chronostratigraphy in siliciclastic and carbonate reservoir modeling, the presented likelihood computation is based on paleo-angles consistency and facies and sedimentary profile coherency. Our stochastic correlation algorithm and the influence of stratigraphic correlation uncertainties on static reservoir modelling are demonstrated on outcrop data of the Cretaceous southern Provence Basin (Provence, S-E France). }
}