The Property Model - Reservoir Model Design - page 95

This tendency to introduce bias when
upscaling from a fine-scale well-log to the reser-
voir model can lead to significant errors. Similar
blocking errors are introduced for the case of
facies modelling (Fig. 3.31 ) - such that modelled
volume fractions of a sandy facies can differ from
the well data (due to blocking) in addition to bias
related to modelling net sand properties. The
errors can be contained by careful tracking of the
correct N/G value in the modelling process.
The common assumption in reservoir flow
simulation is that the N/G ratio is used to factor
down the cell porosity and the horizontal perme-
ability, k h , in order to derive the correct inter-cell
transmissibility. However, no explicit N/G factor
is generally applied to the vertical permeability,
k v , as it is assumed that this is independently
assigned using a k v /k h ratio. This is illustrated
in Fig. 3.32 . A potential error is to double calcu-
late the N/G effect, where, for example the geol-
ogist calculates a total bock permeability of
600 mD and the

Continuous log Discrete log Upscaled log

f f f

N/G sand

0.55

0.85

0.60

0.25

0.40

Sand flag
Cement flag

Fig. 3.30 Upscaling of net-sand logs

reservoir engineer then

is likely to change as a function of scale between
well data and full-field reservoir simulation
model. This is illustrated in Fig. 3.30 for a
simplified workflow. There are several important
biasing factors which tend to occur

multiplies this again by 0.6 to give k x ¼ 360 mD.

When using the N/G approach the main
products from the geological model to the reser-
voir simulation model are as follows:
(i) Model for spatial distribution of N/G;
(ii) Net sand properties, e.g., ø, k h, S w ;
(iii) Multi-phase flow functions for net-sand,
e.g., k ro (S w );
(iv) k v /k h ratios to be applied to each cell;
(v) Information on stratigraphic barriers and
faults.
The N/G ratio approach is widely used and can
be consistently and successfully applied through
the re-scaling process - from well data to geologi-
cal model to reservoir-simulation model. How-
ever, significant errors can be introduced and care
should be taken to ensure that the model correctly
represents the underlying assumptions made.

in this

process:
￿ Blocked sand intervals are likely to contain
non-sand, and the converse (blocking refers to
the process of creating a discrete parameter
from a higher frequency dataset).
￿ Upscaling will bias the volume fractions in
favour of the majority volume fraction. This
is illustrated in Fig. 3.30 , where for example
in the top layer the N/G sand increases from
0.55 to 0.75 to 1.0 in the transition from con-
tinuous log to discrete log to upscaled log.
￿ Cemented sand is not the same as shale, and
will typically be included as part of the shale
fraction (unless care is taken to avoid this).
￿ Since we require net-sand properties we must
filter the data accordingly. That is, only the
fine-scale net-sand values for k and ϕ are
included in the upscaled net-sand values.
￿ We have to assume something about the non-
net sand volume - typically we assume it has
zero porosity and some arbitrary low (or zero)
value for vertical permeability.

3.5.3 Total Property Modelling

Total Property Modelling refers to an approach
where all rock properties are explicitly modelled
and where the cut-offs are only applied after
modelling (if at all). In this way cut-offs, or net to
gross criteria, are not embedded in the process. This