Geoscience Reference
In-Depth Information
Fig. 2.17
Different rock body types as an illustration of the deterministic/probabilistic spectrum
The balance of deterministic and probabilistic
influences on a reservoir model is not as black
and white as it may at first seem. Consider the
simple range of cases shown in Fig.
2.17
,
showing three generic types of rock body:
1.
Correlatable bodies
(Fig.
2.17
, left). These
are largely determined by correlation choices
between wells, e.g. sand observations are made
in two wells and interpreted as occurrences of
the same extensive sand unit and are
correlated. This is a deterministic choice, not
an outcome of a probabilistic algorithm. The
resulting body is not a 100 % determined
'fact', however, as the interpretation of conti-
nuity between the wells is just that - an inter-
pretation. At a distance from the wells, the sand
body has a probabilistic component.
2.
Non-correlated bodies
(Fig.
2.17
, centre).
These are bodies encountered in one well
only. At the well, their presence is determined.
At increasing distances from the well, the loca-
tion of the sand body is progressively less well
determined, and is eventually controlled
almost solely by the outcome from a probabi-
listic algorithm. These bodies are each partly
deterministic and partly probabilistic.
3.
Probabilistic bodies
(Fig.
2.17
, right). These
are the bodies not encountered by wells, the
position of which will be chosen by a proba-
bilistic algorithm. Even these, however, are
not 100 % probabilistic as their appearance
in the model is not a complete surprise.
Deterministic constraints will have been
placed on the probabilistic algorithm to
make sure bodies are not unrealistically large
or small, and are appropriately numerous.
So, if everything is a mixture of determinism
and probability, what's the problem? The issue is
that although any reservoir model is rightfully a
blend of deterministic and probabilistic pro-
cesses, the richness of the blend is a choice of
the user so this is an issue of model design. Some
models are highly deterministic, some are highly
probabilistic and which end of the spectrum a
model sits at influences the uses to which it can
be put. A single, highly probabilistic model is not
suitable for well planning (rock bodies will prob-
ably not be encountered as prognosed). A highly
deterministic model may be inappropriate, how-
ever, for simulations of reservoirs with small
rock bodies and little well data. Furthermore,
different modellers might approach the same res-
ervoir with more deterministic or more probabi-
listic mindsets.
The balance of probability and determinism in
a model is therefore a subtle issue, and needs to
be understood and controlled as part of the model
design. We will also suggest here that greater
happiness is generally to be found in models
which are more strongly deterministic, as the
deterministic inputs are the direct carrier of the
reservoir concept.
