Using a Forward Model as Training Model for 3D Stochastic Multi-well Correlation

in: Second Conference on Forward Modelling of Sedimentary Systems, EAGE

Abstract

Conditioning a forward stratigraphic model to seismic or well data is still a challenge. So, such model cannot usually be used to run static or dynamic reservoir studies. In common methodologies, training images are built from FSM and used with Multiple Point Statistics methods to integrate the information in static geocellular models. Similarly, we present a method that uses a FSM as training model to generate 3D stratigraphic correlations of a set of units identified along wells. The wells are correlated iteratively, each new well being correlated to the result of the previous correlation. This ensures to take the 3D disposition of the wells into account. When the probability of association of the units is computed, we use the Dynamic Time Warping algorithm to build a consistent stratigraphic correlation. First results on synthetic data are presented, using a forward model built with the Sedsim algorithm and three wells.

Download / Links

BibTeX Reference

@inproceedings{edwards:hal-01402553,
 abstract = {Conditioning a forward stratigraphic model to seismic or well data is still a challenge. So, such model cannot usually be used to run static or dynamic reservoir studies. In common methodologies, training images are built from FSM and used with Multiple Point Statistics methods to integrate the information in static geocellular models. Similarly, we present a method that uses a FSM as training model to generate 3D stratigraphic correlations of a set of units identified along wells. The wells are correlated iteratively, each new well being correlated to the result of the previous correlation. This ensures to take the 3D disposition of the wells into account. When the probability of association of the units is computed, we use the Dynamic Time Warping algorithm to build a consistent stratigraphic correlation. First results on synthetic data are presented, using a forward model built with the Sedsim algorithm and three wells.},
 address = {Trondheim, Norway},
 author = {Edwards, Jonathan and Lallier, Florent and Caumon, Guillaume},
 booktitle = {{Second Conference on Forward Modelling of Sedimentary Systems}},
 doi = {10.3997/2214-4609.201600384},
 hal_id = {hal-01402553},
 hal_version = {v1},
 month = {April},
 publisher = {{EAGE}},
 title = {{Using a Forward Model as Training Model for 3D Stochastic Multi-well Correlation}},
 url = {https://hal.univ-lorraine.fr/hal-01402553},
 year = {2016}
}