Geological Realism and Uncertainty in History Matched Reservoir Predictions

Vasily Demyanov and Junko Hutohaean and Goncalo Simoes and Eduardo Barrella and Leonardo Azevedo and Mike Chrsitie. ( 2017 )
in: 2017 RING Meeting, pages 2, ASGA

Abstract

Reservoir predictions encounter several key challenges: adequate representation of uncertainty, agreement with the observed reservoir response and retaining geological realism consistent with prior conceptual knowledge of the reservoir structure, which is subject to uncertainty. The work will highlight recent research advances to tackle these challenges within a Bayesian history matching formalism. Reliability of reservoir predictions depends how well uncertainty in the model representation is aligned with the model calibration to dynamic data, namely history matching. We will demonstrate the value of multi-objective history matching in increasing reliability of reservoir predictions under uncertainty in reservoir model description. The issue of adequate representation of reservoir geological uncertainty at different levels of the modelling workflow was addressed through history matching across multiple combination of reservoir scenarios. History matching was performed with the range of different geological scenarios including reservoir top surface and fault network interpretations, facies definitions and spatial distributions. The spread of history matched predictions across the considered scenarios was obtained with Bayesian averaging based on multiple history matched models. The essential factor in history matching is that the update preserves geological realism to ensure reliable predictions. Therefore, the model update should be consistent with the geological model features. Geostatistical model perturbation is one of the possible ways to make geologically consistent updates and obtain the dynamic data match of the desirable quality. We will illustrate how geostatistical history matching is achieved across multiple stochastic realisations with respect to local variation and uncertainty in spatial continuity.

Download / Links

BibTeX Reference

@INPROCEEDINGS{Demyanov2017,
    author = { Demyanov, Vasily and Hutohaean, Junko and Simoes, Goncalo and Barrella, Eduardo and Azevedo, Leonardo and Chrsitie, Mike },
     title = { Geological Realism and Uncertainty in History Matched Reservoir Predictions },
 booktitle = { 2017 RING Meeting },
      year = { 2017 },
     pages = { 2 },
 publisher = { ASGA },
  abstract = { Reservoir predictions encounter several key challenges: adequate representation of uncertainty, agreement with the observed reservoir response and retaining geological realism consistent with prior conceptual knowledge of the reservoir structure, which is subject to uncertainty. The work will highlight recent research advances to tackle these challenges within a Bayesian history matching formalism. Reliability of reservoir predictions depends how well uncertainty in the model representation is aligned with the model calibration to dynamic data, namely history matching. We will demonstrate the value of multi-objective history matching in increasing reliability of reservoir predictions under uncertainty in reservoir model description. The issue of adequate representation of reservoir geological uncertainty at different levels of the modelling workflow was addressed through history matching across multiple combination of reservoir scenarios. History matching was performed with the range of different geological scenarios including reservoir top surface and fault network interpretations, facies definitions and spatial distributions. The spread of history matched predictions across the considered scenarios was obtained with Bayesian averaging based on multiple history matched models. The essential factor in history matching is that the update preserves geological realism to ensure reliable predictions. Therefore, the model update should be consistent with the geological model features. Geostatistical model perturbation is one of the possible ways to make geologically consistent updates and obtain the dynamic data match of the desirable quality. We will illustrate how geostatistical history matching is achieved across multiple stochastic realisations with respect to local variation and uncertainty in spatial continuity. }
}