Geoscience Reference
In-Depth Information
Fig. 2.45
The Franken
Field:
Top image
: the seven
zone subdivision;
bottom
image
: trended SIS model
conceptual reservoir model is to make a
sketch section of the reservoir, or request a
sketch, showing the desired architecture.
2. The model concept needs to be expressed in
terms of the chosen
modelling elements
, the
selection of which is based on not only a con-
sideration of the heterogeneity, but with a view
to the fluid type and production mechanism.
Some fluid types are more sensitive to hetero-
geneity than others; if the fluid molecules do
not sense the heterogeneity, there is no need to
model it - this is 'Flora's Rule.'
3. Rock models are mixtures of
deterministic
and
probabilistic
inputs. Well data tends to
be statistically insufficient, so attempts to
extract statistical models from the well data
are often not successful. The road to happi-
ness therefore generally lies with strong deter-
ministic control, as determinism is the most
direct method of carrying the underlying res-
ervoir concept into the model.
4. To achieve the desired reservoir architecture,
the
variogram model
has a leading influence if
pixel-based methods are employed. Arguably
the
variogram range
is the most important
geostatistical input.
(f) Make full use of the visualisation tools, espe-
cially the ability to scroll through the model
vertically, layer by layer, to look for anoma-
lous geometries, e.g. spikes and pinch-outs.
2.8.2 Synopsis - Rock Modelling
Guidelines
The first decision to be made is whether or not a
rock model is truly required. If rock modelling
can add important controls to the desired distri-
bution of reservoir properties, then it is clearly
needed. If, however, the desired property
distributions can be achieved directly by property
modelling, then rock modelling is probably not
necessary at all.
If it is decided that a rock model is required, it
then needs some thought and design. The use of
system default values is unlikely to be successful.
This chapter has attempted to stress the fol-
lowing things:
1. The
model concept
needs to be formed before
the modelling begins, otherwise the modeller
is 'flying blind'. A simple way of checking
your
(or
someone else's) grasp of
the
