Reliable flow modeling of highly heterogeneous/fractured reservoirs
necessarily goes through the calibration of poorly determined geological and/or
petrophysical parameters to field flow measurements. To that end, optimization
procedures based on gradient methods or on gradual-deformation techniques have
been developed in recent years.
This paper proposes a sequential method combining those two approaches. The
case under consideration is a water-bearing reservoir constituted of
heterogeneous, karstic and fractured limestones located near Poitiers, France.
In a preliminary step, drilling, core, and log data acquired in approximately
30 wells were integrated into a geostatistical facies model used as the support
for flow simulation.
First the facies petrophysical properties of this model were calibrated to
well pumping and interference responses within a gradient-based inversion loop.
Flow responses could be reproduced, with the exception of a few
"problematic" observation wells.
Second the gradual-deformation method was applied, globally then locally, to
improve the distribution of facies while keeping the previously optimized
petrophysical properties. The problematic wells' responses could be reproduced
better without altering the other wells’ match. Furthermore, that good match of
calibration wells was obtained on a simplified geostatistical model involving
fewer facies than in the initial model. The gradual-deformation method then
appears as a robust and effective approach to find a model best matching a set
of flow data among equiprobable geostatistical models.
To conclude, the sequential-modeling method demonstrated herein is an
effective way to actually integrate geological and flow data and to link
geosciences and reservoir-engineering skills, for setting up consistent models
of hardly tractable highly heterogeneous reservoirs.
During the past 20 years, the technique of mathematical modeling has been
used extensively in the study of groundwater-resources management and aquifer
remediation (Sun 1994). Concern was especially focused on fluid transfer in
heterogeneous and/or fractured reservoirs. That resulted in conclusive advances
in the characterization and modeling of fractured reservoirs (Cacas et al.
2001). Actually, wellbore information on underground reservoir heterogeneities
and fractures (e.g., core descriptions, image logs, and production profiles) is
now used to condition the geostatistical pixel-based models or the object-based
stochastic models of these reservoirs and to calibrate the hydraulic properties
of major flow heterogeneities such as fractures (Sarda et al. 2002). However,
the problem of reservoir-model construction remains highly undetermined.
The purpose of this paper is to design and validate an inversion method for
calibrating the poorly defined flow models of highly heterogeneous reservoirs
to wellbore dynamic data. The design and the application of that method are
performed on an experimental hydrogeological site (EHS) settled on a karstic
and fractured limestone aquifer located near Poitiers, France. The flow model
is based on a geostatistical distribution of facies. The method involves two
successive steps: (1) the inversion of facies petrophysical properties and (2)
the gradual deformation of the facies distribution. The resulting model is
shown to predict well responses effectively. Finally, the possibility of
further calibration improvement is investigated by means of alternative
flow-modeling approaches, such as the use of a dual-porosity model or a more
accurate modeling of conductive bodies.
© 2009. Society of Petroleum Engineers
View full textPDF
- Original manuscript received:
21 February 2008
- Meeting paper published:
9 June 2008
- Revised manuscript received:
15 September 2008
- Manuscript approved:
22 September 2008
- Published online:
1 June 2009
- Version of record:
1 June 2009