Advances on the hierarchical and stochastic multiple well stratigraphic correlation in WeCo

in: 2020 RING Meeting, ASGA

Abstract

We present a global update on the WeCo code for the stochastic correlation of multiple wells. This code is an interesting algorithmic framework to assess and reduce the uncertainties about strata in sedimentary basins and reservoirs. The input is a set of subvertical wells carrying measurements and interpretations such as well logs and facies, and a set of mathematically formulated correlation rules representing the stratigraphic concepts deemed relevant for the correlation problem. The output is a set of possible correlations, which are stratigraphic lines (or labels) mapping each well to a global stratigraphic column. These lines define the global connectivity of the stratigraphic intervals crossed by the wells, and in particular the possible stratigaphic hiatuses. To address this huge computational challenge, the base algorithm starts by independently correlating nearby wells using an n-best version of dynamic time warping. The produced sets of correlations for these correlated wells are then aggregated by correlating groups of wells until a global set of correlations is produced. To improve the ability of the approach to efficiently explore the search space and to better reproduce the hierarchical reasoning used in expert-based correlation, we extend this method to account for nested stratigraphic orders. For this the base method is applied on lower-order stratigraphic data first, and then the higher-order solutions are constrained by the stochastic results of the lower-order correlations. \\\\

Download / Links

BibTeX Reference

@inproceedings{ANTOINE_RM2020,
 abstract = { We present a global update on the WeCo code for the stochastic correlation of multiple wells. This code is an interesting algorithmic framework to assess and reduce the uncertainties about strata in sedimentary basins and reservoirs. The input is a set of subvertical wells carrying measurements and interpretations such as well logs and facies, and a set of mathematically formulated correlation rules representing the stratigraphic concepts deemed relevant for the correlation problem. The output is a set of possible correlations, which are stratigraphic lines (or labels) mapping each well to a global stratigraphic column. These lines define the global connectivity of the stratigraphic intervals crossed by the wells, and in particular the possible stratigaphic hiatuses. To address this huge computational challenge, the base algorithm starts by independently correlating nearby wells using an n-best version of dynamic time warping. The produced sets of correlations for these correlated wells are then aggregated by correlating groups of wells until a global set of correlations is produced. To improve the ability of the approach to efficiently explore the search space and to better reproduce the hierarchical reasoning used in expert-based correlation, we extend this method to account for nested stratigraphic orders. For this the base method is applied on lower-order stratigraphic data first, and then the higher-order solutions are constrained by the stochastic results of the lower-order correlations. \\\\ },
 author = { Antoine, Christophe AND Caumon, Guillaume },
 booktitle = { 2020 RING Meeting },
 publisher = { ASGA },
 title = { Advances on the hierarchical and stochastic multiple well stratigraphic correlation in WeCo },
 year = { 2020 }
}