realisations may be challenging. One possible
approach may be to choose the realisations that are
volumetrically closest to the average off all the model
realisations. However, it will also be important to
make a visual inspection to ensure that the most representative model realisations are chosen, and it may
also be necessary to investigate those realisations
that represent end-members in some sense, such as
volumetrics or connectivity. In a complex situation
with a large number of model realisations, so-called
streamline simulations (which are fast and simplified
flow simulations) may be used as a guide for picking
an appropriate set of model realisations.
The fact that many reservoir models today are
based on stochastic modelling techniques must be
carefully considered when using models for well
planning or other decisions. For example, a horizontal
well optimised for penetrating highly permeable layers
in one model realisation, might be positioned in low
permeability layers in a different stochastic realisation
of the same model, due to the stochastic distribution of
facies and petrophysical properties. Hence, it is of
critical importance to consider multiple realisations
when doing such analysis.
22.4 Constructing the Dynamic
Reservoir Model
While geological (static) models are the domain of
geologists, geophysicists and petrophysicists, the
dynamic models are in the hands of the reservoir
engineers. Flow simulation is a major activity in oil
companies, as such simulations can be a very valuable
tool to understand the reservoir dynamics, optimise the
production and compute recovery factors, Chap. 20
(Fig. 22.15). Flow simulations are quite computationally demanding, and require powerful “numbercrunchers” that are occupied continuously, as each
flow simulation run may take from several hours to
many days.
Most dynamic models inherit both the structural
and property model from the static model, and tight
integration between the disciplines is required for an
optimal result. In addition, the dynamic model needs
to incorporate special data types only relevant for flow
modelling, such as relative permeabilities and a number of temperature and pressure dependent fluid
properties.
Dynamic models can also deal with cell-face
properties. This means that the numerical properties
are not limited to the cell-centre, but also add
properties controlling the flow between two adjacent
cells. This is quite often used to model the effect of
thin barriers and baffles, both of stratigraphical origin
and those created by faults.
22.4.1 3D Gridding for Flow Modelling
The grid used in flow simulation is usually coarser
than those that are used for static models. To save
computational time, flow simulation grids usually
have coarser cells in parts of the reservoir which are
of less interest, such as attached water bodies
(aquifers) which may be important for reservoir
energy, but do not require the detailed cells required
for two- or three-phase flow near oil and gas
producers. This is possible to achieve as the flow
simulator can handle varying cell sizes, in contrast to
property modelling methods which commonly require
quite uniform grids. A comparison of such grids is
shown in Fig. 22.16.
There are cases where the static and dynamic
modelling is performed on the same grid, and this
simplifies the modelling. In addition to the requirement for a relative uniform grid, this also demands an
in-depth analysis of proper scales, so-called REV
(Representative Elementary Volume) analysis, which
is discussed in Sect. 22.4.2.
22.4.2 Upscaling from Static to Dynamic
Models
As previously stated, there are several ways of defining the scale relationship between the geomodel and
the dynamic simulation model. In most cases, the
geomodel is constructed at a finer scale (for accuracy),
and a property upscaling step is necessary in order to
have a simulation model with a number of grid cells
giving reasonable computational time for the dynamic
simulations. We will not go into the details of various
approaches and techniques for upscaling the
parameters of the geological model from geogrid
scale to simulation grid scale, however a few main
principles can be outlined:
22 Introduction to Reservoir Modelling
575
approach may be to choose the realisations that are
volumetrically closest to the average off all the model
realisations. However, it will also be important to
make a visual inspection to ensure that the most representative model realisations are chosen, and it may
also be necessary to investigate those realisations
that represent end-members in some sense, such as
volumetrics or connectivity. In a complex situation
with a large number of model realisations, so-called
streamline simulations (which are fast and simplified
flow simulations) may be used as a guide for picking
an appropriate set of model realisations.
The fact that many reservoir models today are
based on stochastic modelling techniques must be
carefully considered when using models for well
planning or other decisions. For example, a horizontal
well optimised for penetrating highly permeable layers
in one model realisation, might be positioned in low
permeability layers in a different stochastic realisation
of the same model, due to the stochastic distribution of
facies and petrophysical properties. Hence, it is of
critical importance to consider multiple realisations
when doing such analysis.
22.4 Constructing the Dynamic
Reservoir Model
While geological (static) models are the domain of
geologists, geophysicists and petrophysicists, the
dynamic models are in the hands of the reservoir
engineers. Flow simulation is a major activity in oil
companies, as such simulations can be a very valuable
tool to understand the reservoir dynamics, optimise the
production and compute recovery factors, Chap. 20
(Fig. 22.15). Flow simulations are quite computationally demanding, and require powerful “numbercrunchers” that are occupied continuously, as each
flow simulation run may take from several hours to
many days.
Most dynamic models inherit both the structural
and property model from the static model, and tight
integration between the disciplines is required for an
optimal result. In addition, the dynamic model needs
to incorporate special data types only relevant for flow
modelling, such as relative permeabilities and a number of temperature and pressure dependent fluid
properties.
Dynamic models can also deal with cell-face
properties. This means that the numerical properties
are not limited to the cell-centre, but also add
properties controlling the flow between two adjacent
cells. This is quite often used to model the effect of
thin barriers and baffles, both of stratigraphical origin
and those created by faults.
22.4.1 3D Gridding for Flow Modelling
The grid used in flow simulation is usually coarser
than those that are used for static models. To save
computational time, flow simulation grids usually
have coarser cells in parts of the reservoir which are
of less interest, such as attached water bodies
(aquifers) which may be important for reservoir
energy, but do not require the detailed cells required
for two- or three-phase flow near oil and gas
producers. This is possible to achieve as the flow
simulator can handle varying cell sizes, in contrast to
property modelling methods which commonly require
quite uniform grids. A comparison of such grids is
shown in Fig. 22.16.
There are cases where the static and dynamic
modelling is performed on the same grid, and this
simplifies the modelling. In addition to the requirement for a relative uniform grid, this also demands an
in-depth analysis of proper scales, so-called REV
(Representative Elementary Volume) analysis, which
is discussed in Sect. 22.4.2.
22.4.2 Upscaling from Static to Dynamic
Models
As previously stated, there are several ways of defining the scale relationship between the geomodel and
the dynamic simulation model. In most cases, the
geomodel is constructed at a finer scale (for accuracy),
and a property upscaling step is necessary in order to
have a simulation model with a number of grid cells
giving reasonable computational time for the dynamic
simulations. We will not go into the details of various
approaches and techniques for upscaling the
parameters of the geological model from geogrid
scale to simulation grid scale, however a few main
principles can be outlined:
22 Introduction to Reservoir Modelling
575