Geoscience Reference
In-Depth Information
This allows the stress field to operate as
a control on the fracture flow properties,
generating a tensor permeability matrix for the
network properties still need to be calibrated to
static and dynamic data (e.g. image logs, seismic
data and well tests), but now the effects of stress
are also included. The stress field in itself may be
hard to determine accurately, and alternative
stress field scenarios may need to be evaluated,
ideally within a framework of multiple determin-
istic scenarios.
may need to be significantly adjusted. It is not
uncommon for orders of magnitude permeability
adjustments to be made in order to reconcile
models with production data; in matrix-only
reservoirs the permeability adjustments are
more normally no more than a factor 2 or 3, or
none at all.
This leads to some general implications for
the use of fracture models:
1. In inversion-style workflows the production
data is often treated as representative for the
field; if based on a single well test, this is
unlikely to be the case.
2. The inversion process is itself non-unique.
3. Because of the above, base-case fracture
models are of even less value than base-case
matrix models - multi-model uncertainty-
handling
6.7.2.6 Forward-Modelling or Inversion?
In discussing carbonate reservoirs (Sect.
6.6
), the
point was made that once a significant number of
assumptions have been made in a complex
modelling workflow it can be questioned whether
attempts to forward-model reality from limited
data are still valid. This is particularly the case
for fractured reservoirs, where necessary static
field data is limited and the sensitivity to the
missing data is high. For example, fracture per-
meability is particularly sensitive to fracture
aperture, yet
in situ
aperture data is extremely
difficult to determine. Average aperture can be
back-calculated if fracture density and gross frac-
ture porosity are known, but fracture porosity is
also difficult to measure and even estimates of
fracture density are usually built on very limited
data sets. Many assumptions must be made, in
addition to those routinely made for modelling
the matrix properties. By definition, fractured
reservoir models involve a greater degree of
approximation than models
based
on
alternative
fracture
concepts and scenarios is essential.
6.8
Fit-for-Purpose Recapitulation
The preceding sections have discussed different
reservoir types in terms of their geology,
identifying the key issues for reservoir modelling
with an underlying assumption that we are gener-
ally talking about oil fields. However, as pointed
is as important as the type of rock system. The
effects of fluid physics have to be considered
alongside the effects of rock architecture.
To reiterate the underlying principle, for any
given reservoir architecture (e.g. fluvial, shallow
marine, carbonate, or structurally complex fields)
the impact of the reservoir heterogeneities on
flow depends on the fluid system. Using the
A gas reservoir is only sensitive to 3 orders
of magnitude of permeability variation;
An oil reservoir is sensitive to 2 orders
of magnitude of permeability variation;
Heavy oil reservoirs or lighter crudes under
secondary recovery (waterflood) are sensitive
to 1 order of magnitude of permeability
variation.
for un-fractured
reservoirs.
Because of this, there is a greater reliance on
using production data, and in this sense fracture
models are typically 'inverted' from production
data rather than forward-modelled from logs and
core. Indeed, one of the most useful roles of DFN
software is to reconcile well test datawith potential
fracture network properties - the missing fracture
data is effectively inverted from the well test data.
This is captured in the workflow shown in
Fig.
6.62
in which the forward modelling steps
culminate in the need to 'calibrate to well tests',
after which several of the fracture parameters
