Intégration de modèles approchés pour mieux transmettre l’impact des incertitudes statiques sur les courbes de réponse des simulateurs d’écoulements

Gaetan Bardy. ( 2015 )
Université de Lorraine

Abstract

Although it is common to use many different numerical models for the static description of underground reservoirs and their associated uncertainties, fluid flow uncertainties are only based on a few dynamic simulations for performance reasons. The objective of this thesis is to better transmit the impact of static uncertainties on flow simulator responses without increasing computation time, using approximated models (proxies). Research has been undertaken in 2 directions: - Implementation of new proxies based on Fast Marching in order to better approach fluid propagation behavior in a reservoir using only a few parameters. This allows to obtain response curves close to those provided by the flow simulator in a very short period of time. - Set up of a mathematical minimization’s procedure in order to predict flow simulator’s response curves using an analytical model and distances between proxy responses computed on each model. The methods developed during this PhD were applied on two different real cases in order to validate them with industry data. Results have shown that our new proxy improves the quality of the information about fluid behavior compared to the available proxy even though ours can still be improved. We also highlight that our minimization procedure better assesses dynamic uncertainties if the proxy used is reliable enough.

Download / Links

BibTeX Reference

@PHDTHESIS{BardyPhD2015,
    author = { Bardy, Gaetan },
     title = { Intégration de modèles approchés pour mieux transmettre l’impact des incertitudes statiques sur les courbes de réponse des simulateurs d’écoulements },
     month = { "oct" },
      year = { 2015 },
    school = { Université de Lorraine },
  abstract = { Although it is common to use many different numerical models for the static description
of underground reservoirs and their associated uncertainties, fluid flow uncertainties are only based on a few dynamic simulations for performance reasons. The
objective of this thesis is to better transmit the impact of static uncertainties on flow
simulator responses without increasing computation time, using approximated models (proxies).
Research has been undertaken in 2 directions:
- Implementation of new proxies based on Fast Marching in order to better approach fluid
propagation behavior in a reservoir using only a few parameters. This allows to obtain
response curves close to those provided by the flow simulator in a very short period of time.
- Set up of a mathematical minimization’s procedure in order to predict flow simulator’s
response curves using an analytical model and distances between proxy responses computed
on each model.
The methods developed during this PhD were applied on two different real cases in order to
validate them with industry data. Results have shown that our new proxy improves the quality of the
information about fluid behavior compared to the available proxy even though ours can still be
improved. We also highlight that our minimization procedure better assesses dynamic uncertainties
if the proxy used is reliable enough. }
}