be multiplied with a factor ten in selected areas to
improve match.
2. The outer loop. In this case the reservoir engineer
works together with the geomodeller to modify the
static model that provides input to the dynamic
model.
There is no doubt that the “inner-loop” is most
frequently applied as it is simpler and results will be
quicker. However, it breaks the link with the static
model, and the models can diverge significantly. In
many cases, the reservoir engineer can achieve a sufficient history match with values that cannot be
defended from the static model and concept model.
Because the goal of history matching is to develop a
model or models that can be used to make reliable
predictions of future performance, it is important to
ensure that history match changes are consistent with
the geologic concept that has been inferred from available geologic data. If the history match changes are
not consistent with geologic data, it is less likely that
the predictions of the resulting model will be reliable.
That is, if history match changes are not consistent
with a geologic concept, it will be difficult to extrapolate to other parts of the reservoir, or in some cases,
even to later times (Lun et al. 2012).
Hence, working along the outer loop will more or
less ensure that the model is updated according to the
concept model and in doing so will promote crossdisciplinary consistency. However, as the outer loop is
more demanding and time consuming, a hybrid solution is also attractive. This implies that the reservoir
engineer tests alterations to model parameters in the
inner-loop, but at frequent intervals includes the outerloop in order to calibrate the static model and ensure
that changes made are within acceptable ranges.
22.5 Uncertainty Handling
A key aspect of reservoir modelling is the handling of
uncertainty. With limitations in seismic resolution,
number of available wells etc., there will always be a
significant uncertainty associated with the description
and understanding of the subsurface. The impact of
this uncertainty can to a certain degree be estimated
through constructing alternative scenarios and statistical realisations of the model. The aim of this is to
quantify the impact on static and dynamic volumes,
impact on depletion planning and well placement.
Our insight of the reservoir evolves through the life
of a field. As the field moves from the initial exploration phase, via the field development phase, through
the early production phase and eventually ends up in
the late stage IOR (Improved Oil Recovery) phase, the
amount of available static and dynamic reservoir
information increases. Hence the development of reservoir understanding and modelling is a dynamic and
continuous process. Uncertainty is a complex term,
and can be classified as follows:
Uncertainty Due to Lack of Data Although we
assume we know the geostatistical model, we lack
data in the reservoir. For instance, at some co-ordinate
between wells, we cannot predict if we have facies A
or facies B, as both can be possible. This is illustrated
in Fig. 22.14, where the stochastic method predicts a
different facies and porosity at a single position in the
reservoir.
Uncertainty Due to Lack of Knowledge This relates
more to the “human factor” of uncertainty. As the data
analysis is not definite, we may end up with many
alternatives (scenarios) regarding how to set up the
model. For instance, in a fluvial reservoir, we may
have one case with narrow channels with a NorthSouth direction, while another scenario could be to
have broader channels trending East-West. We usually
treat this by scenario modelling, i.e. building alternative models.
Uncertainty Due to Imprecise Measurements All
the measurements we perform have some uncertainty,
and it is important to take this into account. For
instance, the density and resistivity logs are uncertain
due to uncertainties in well logging, and those
uncertainties propagate in the estimation of the
derived petrophysical properties such as porosity and
saturations.
Uncertainty Due to Unknown Unknowns These
are the uncertainties we do not test for – we simply
do not know their existence. One example could
be that none of the concept models we test are the
right one. Another example is that the software
performs erroneous calculations we are not aware of,
or (usually more frequent) that the reservoir modeller
makes mistakes that are not captured by the quality
control.
22 Introduction to Reservoir Modelling
579
improve match.
2. The outer loop. In this case the reservoir engineer
works together with the geomodeller to modify the
static model that provides input to the dynamic
model.
There is no doubt that the “inner-loop” is most
frequently applied as it is simpler and results will be
quicker. However, it breaks the link with the static
model, and the models can diverge significantly. In
many cases, the reservoir engineer can achieve a sufficient history match with values that cannot be
defended from the static model and concept model.
Because the goal of history matching is to develop a
model or models that can be used to make reliable
predictions of future performance, it is important to
ensure that history match changes are consistent with
the geologic concept that has been inferred from available geologic data. If the history match changes are
not consistent with geologic data, it is less likely that
the predictions of the resulting model will be reliable.
That is, if history match changes are not consistent
with a geologic concept, it will be difficult to extrapolate to other parts of the reservoir, or in some cases,
even to later times (Lun et al. 2012).
Hence, working along the outer loop will more or
less ensure that the model is updated according to the
concept model and in doing so will promote crossdisciplinary consistency. However, as the outer loop is
more demanding and time consuming, a hybrid solution is also attractive. This implies that the reservoir
engineer tests alterations to model parameters in the
inner-loop, but at frequent intervals includes the outerloop in order to calibrate the static model and ensure
that changes made are within acceptable ranges.
22.5 Uncertainty Handling
A key aspect of reservoir modelling is the handling of
uncertainty. With limitations in seismic resolution,
number of available wells etc., there will always be a
significant uncertainty associated with the description
and understanding of the subsurface. The impact of
this uncertainty can to a certain degree be estimated
through constructing alternative scenarios and statistical realisations of the model. The aim of this is to
quantify the impact on static and dynamic volumes,
impact on depletion planning and well placement.
Our insight of the reservoir evolves through the life
of a field. As the field moves from the initial exploration phase, via the field development phase, through
the early production phase and eventually ends up in
the late stage IOR (Improved Oil Recovery) phase, the
amount of available static and dynamic reservoir
information increases. Hence the development of reservoir understanding and modelling is a dynamic and
continuous process. Uncertainty is a complex term,
and can be classified as follows:
Uncertainty Due to Lack of Data Although we
assume we know the geostatistical model, we lack
data in the reservoir. For instance, at some co-ordinate
between wells, we cannot predict if we have facies A
or facies B, as both can be possible. This is illustrated
in Fig. 22.14, where the stochastic method predicts a
different facies and porosity at a single position in the
reservoir.
Uncertainty Due to Lack of Knowledge This relates
more to the “human factor” of uncertainty. As the data
analysis is not definite, we may end up with many
alternatives (scenarios) regarding how to set up the
model. For instance, in a fluvial reservoir, we may
have one case with narrow channels with a NorthSouth direction, while another scenario could be to
have broader channels trending East-West. We usually
treat this by scenario modelling, i.e. building alternative models.
Uncertainty Due to Imprecise Measurements All
the measurements we perform have some uncertainty,
and it is important to take this into account. For
instance, the density and resistivity logs are uncertain
due to uncertainties in well logging, and those
uncertainties propagate in the estimation of the
derived petrophysical properties such as porosity and
saturations.
Uncertainty Due to Unknown Unknowns These
are the uncertainties we do not test for – we simply
do not know their existence. One example could
be that none of the concept models we test are the
right one. Another example is that the software
performs erroneous calculations we are not aware of,
or (usually more frequent) that the reservoir modeller
makes mistakes that are not captured by the quality
control.
22 Introduction to Reservoir Modelling
579