Geological reservoir modelling/

Geologisk reservoarmodellering TGB4275,

Spring 2015 Lecture 9
The Property Model
Well data
Modelling property distributions
Upscaling & REV
Multiscale modelling

Arve Nss

Practical information
Exercise #6 (REV) to be published March 17th Deadline March 24th at 16:00
NTG Exercise March to be published published March 24th
Revised lecture plan Cut Offs NTG Thursday 19th March

Lecture on Thursday 26th March - TBA

Review of Lecture 8
Determinism and Probability
Property modelling workflow

Determinism and probability

Property modelling workflow

Object or pixel based workflow:
Given that the blocked wireline estimates are good, the properties is distributed in the 3D grid by
conventional interpolation or simulation methods.
Kriging methods
Sequential Gaussian Simulation (SGS)
Truncated Gaussian Simulation (TGS)

The distribution can be done directly or via a rock model (e.g. object based model).
Caution with the statistical validity if the experimental spatial correlation structure.
Use the conceptual model to also steer property modelling

Rock model

Property model

Property model

Porosity
Total porosity is the sum of all
connected and isolated pore volumes.
Interconnected porosity (sometimes
called effective porosity) is the sum
of all pore volumes that are connected
and potentially can contribute to flow.
Presence of clay minerals are
particularly challenging when defining
porosity.
The clay minerals can contain micro
porosity and parts of their associated
fluids can be movable.

Intergranular
porosity
Heavy
mineral

Clay (with
micro porosity)

Quartz

Cement
Mica

Permeability
Navier-Stoke equation can be used to
describe the fluid flow behaviour in
the pore network.
However, the description and solution
is very complicated (if possible).
A macroscopic description of fluid
flow is given by the Darcy law:

Intergranular
porosity
Heavy
mineral

Clay (with
micro porosity)

Cement
Mica

Darcy law is valid for incompressible fluids, flowing

through a homogeneous medium at low speed.
7

Quartz

Lectures 9
1. 3D reservoir gridding
2. Model elements What to model
3. Model build How to model
4. Determinism and probability
Find the balance
Essential statistics

1. Understanding porosity and

permeability

Some definitions
Homogeneous

Homogeneous
The material is uniform in composition
and consists of one phase.
Heterogeneous

Heterogeneous
The material is non-uniform in
composition and consists of two or more
phases.

Isotropic
The measurement is direction
independent.

Heterogeneous and isotropic

Homogenous and isotropic

kx=ky

Anisotropic
The measurement is direction dependent.

Heterogeneous and anisotropic

kx>ky

Homogenous and anisotropic

Some definitions

Homogeneity and heterogeneity is

also a function of measurement
volume:
Measurement 1: Homogeneous
(and isotropic)
Measurement 2: Heterogeneous
(and anisotropic).

10

Measurement 2

Measurement 1

Property modelling workflow

Porosity
log

Well

Calibration

Permeability
log

Well in
model

Correlation

Un-cored

+ spatial correlation structure

= Blocking/Averaging/Upscaling
11

Property distribution commonly simulated

using SGS, TGS in a pixel or object a based
workflow.
Can also use seismic directly in the property
modelling workflow (see chapter 3).

Property modelling workflow

Porosity
log

Well

Calibration

Un-cored

Permeability
log

Correlation

Core plugs are

representative samples.
Wireline log is measuring
a parameter related to
porosity and the same
structures as the core
plugs..

The correlation between

porosity and permeability
from core plugs is valid
at wireline log and
reservoir grid scale.

The average porosity and

permeability is representative
for the grid cell scale
property.

= Blocking/Averaging/Upscaling
12

Well in
model

+ spatial correlation structure

Property distribution commonly simulated

using SGS, TGS in a pixel or object a based
workflow.
Can also use seismic directly in the property
modelling workflow (see chapter 3).

Communication between disciplines

Geologist

Reservoir Engineer

Petrophysicist

Data

Model

13

Wavy bedded,
fine grained
heterolithic
sand interpreted as
tidal channel infill
on a lower delta
plain

Free Fluid

Cap. bound water

Clay bound water
VSH
Dry clay
Rock matrix

VMA

k, Sw, , Pc,
Kro, Krw

Data Model Truth

We must not confuse the observations (data) with the model (a hypothesis) and
both of these with the truth (an unknown).
Often the data are forced to fit the models.
The data are however a limited subset of the subsurface: Data Model Truth
Our models should be consistent with data, but should aim to capture a wider range
(approaching the truth)
14

Wireline log measurements

True value

Wireline logs are recordings against depth of a rock

characteristics.
Wireline tools responding to a natural or
spontaneous phenomena with a detector.

E.g. Gamma-ray and Neutron tool.

Wireline tools responding to an induced phenomena

with a source and a receiver.

Precision and accuracy (quality of

measurement).
True value

Measured
value X1

Measured
value X2

Average
Property
value
Precision
Accuracy

15

Detector

E.g. Resistivity, density and velocity.

Response
volume

Wireline
log

Core
plugs

Wireline log measurements

True value

Vertical resolution
Dependent on tool design (e.g. distance between receiver
and emitter).
Features thinner than the vertical resolution can still be
recorded, but only a fraction of the true value.

Shoulder
bed
effect

With many thin features within the response volume, an

average is recorded.

Response volume or volume of investigation is a

function of vertical resolution and depth of
investigation.
Shoulder bed effects.
At the transition between beds with different properties, an
average of the two beds are recorded.

16

Depth of
investigation

Vertical
resolution

Dependent on tool design and well and formation properties.

Detector

Depth of investigation

Response
volume

Wireline
log

Core
plugs

Well data and depositional units

The volume of investigation (or vertical resolution) of the well data must
be taken into account when interpreting the response in relation to
sedimentary features.
17

Permeability variations
Permeability varies with several order of magnitudes and is directly linked to the
productivity of a reservoir.
In the conventional property modelling workflow, an average or representative
permeability (and all other properties) must be specified.
This process is generally known as upscaling.

Conventional property modelling workflow

18

Permeability variations
Standard deviation of permeability
increases when the mean increases.
Useful variability measures for
permeability:
Coefficient of Variation (CV); a normalized
measure of dispersion.

Can be estimated with:

Where (p) is the standard deviation and

is the mean of the samples.

The N0 statistic gives the number of samples

needed to estimate the mean value with a 20%
tolerance (at a 95% confidence level)

19

Defining an average or upscaled permeability will be

easier for homogenous reservoirs (with little
variations)

Upscaling
Discrete variable
Facies A

N=10

Facies B

N=15

Homogenous medium
Upscaling

Facies B

Heterogeneous
medium

Upscaling method:
Dominant element

Homogenous medium
Upscaling
=20%

N=10

=30%

N=15

Continuous variable

High resolution
Fine scale
Microscopic

=26%

Upscaling method:
Arithmetic average

20 10 + 30 15
= 26%
10 + 15
Low resolution
Coarse scale
Macroscopic

Upscaling
Homogenization
Effective medium theory
Continuum approach

Upscaling is the method for replacing a high resolution, fine scale

heterogeneous medium with a homogeneous low resolution,
coarse scale medium.

20

Upscaling
Discrete variable
Facies A

N=10

Facies B

N=15

Homogenous medium
Upscaling

Facies B

Heterogeneous
medium

Upscaling method:
Dominant element

Homogenous medium
Upscaling
=20%

N=10

=30%

N=15

Continuous variable

High resolution
Fine scale
Microscopic

=26%

Upscaling method:
Arithmetic average

20 10 + 30 15
= 26%
10 + 15
Low resolution
Coarse scale
Macroscopic

Upscaling
Homogenization
Effective medium theory
Continuum approach

Upscaling

Real rock medium

21

Upscaled rock medium

21

Upscaling
Discrete variable
Facies A

N=10

Facies B

N=15

Homogenous medium
Upscaling

Facies B

Heterogeneous
medium

Homogenous medium
Upscaling
=20%

N=10

=30%

N=15
=

Continuous variable

=26%

Upscaling method:
Arithmetic average

20 10 + 30 15
= 26%
10 + 15

=20%

N=10

=30%

N=15

=30%

=26%

Heterogeneous
medium

=25%

Low resolution
Coarse scale
Macroscopic

=24%

High resolution
Fine scale
Microscopic

22

Upscaling method:
Dominant element

=26%

Upscaling
Discrete variable
Facies A

N=10

Facies B

N=15

High resolution
Fine scale

Low resolution
Coarse scale

Homogenous medium
Upscaling

Facies B

Heterogeneous
medium

Homogenous medium
Upscaling

All fine scale

values are
known.

=20%

N=10

=30%

N=15

Continuous variable

Only a sub set

of all possible
fine scale
values are
known.

Core plugs

=26%

Upscaling method:
Arithmetic average

20 10 + 30 15
= 26%
10 + 15

Grid cell
5m

x 15
20 cm2
100 m

300 cm2
High resolution
Fine scale

23

Upscaling method:
Dominant element

50000 m2
Low resolution
Coarse scale

Additive and non-additive properties

1=0.1
2=0.2

V1=V2

Upscaling

=0.15

Additive properties are:

Independent on the spatial distribution
of the different components
Can be upscaled analytically by a

Upscaling

=0.15

volume average equation (arithmetic

average).
Commonly called bulk properties

Upscaling

24

1 1 + 2 2
=
1 + 2

=0.15

=
1

Other additive properties: density, mass,

sand fraction, saturation, in general
scalar values (a physical quantity that
can be represented by a number at a
particular point).

Additive and non-additive properties

k=1 mD
k=100 mD

V1=V2

Arithmetic
average, kA
Upscaling

kx=50.5

Non-additive properties are:

Dependent on the spatial distribution
of the different components
Cannot in general be upscaled

Harmonic
average, kH
Upscaling

kx=1.98

analytically but need a fine scale

model where the components are
explicit defined.

Upscaling

25

k=<kA,kH>

Other non-additive properties: resistivity,

velocity, in general a vector property
(a vector is a property with a magnitude
and a direction, a tensor is a method to
express a vector).

25

Permeability averages
100% sand
0% mud

Sandstone
100 mD
Mudstone
0,1 mD

0% sand
100% mud

26

26

Permeability averages
100% sand
0% mud

Sandstone
100 mD
Mudstone
0,1 mD

0% sand
100% mud

Layered system

27

Permeability averages
100% sand
0% mud

Sandstone
100 mD
Mudstone
0,1 mD

0% sand
100% mud

Layered system

28

Permeability averages
Random distribution
100% sand
0% mud

Sandstone
100 mD
Mudstone
0,1 mD

0% sand
100% mud

Layered system

29

Upscaled permeability
The upscaled permeability is
defined as the permeability of a
hypothetical homogenous block
as the real rock medium.
The upscaled permeability can either
be measured or calculated given the
fine scale distribution of permeability
values.

The upscaled permeability need

in general to be expressed as a
tensor.
kxy: permeability in x-direction given a
pressure gradient in y-direction.

30

Upscaled permeability
For some simple systems, the
upscaled permeability can be
calculated analytically.
For a plane parallel system:
Vertical permeability is the harmonic
average.
Horizontal permeability is the
arithmetic average.

For a 2D random permeability

field with a log-normal distribution
(and low variance), the upscaled
permeability is the geometric
average:

31

Upscaled permeability

For some simple systems, the

upscaled permeability can be
calculated analytically.

Core plugs are representative

samples.
Wireline log is measuring a
parameter related to porosity
and the same structures as
the core plugs.
The correlation
between porosity
and permeability
from core plugs is
valid at wireline log
and reservoir grid
scale.

For a plane parallel system:

Vertical permeability is the harmonic
average.
Horizontal permeability is the
arithmetic average.

For a 2D random permeability

field with a log-normal distribution
(and low variance), the upscaled
permeability is the geometric
average:

32

The average porosity and

permeability is
representative for the grid
cell scale property.

Upscaled permeability

Lerkendal Stadium (100 x 70 m)

Core plugs are representative

samples.
Wireline log is measuring a
parameter related to porosity
and the same structures as
the core plugs.
The correlation
between porosity
and permeability
from core plugs is
valid at wireline
log and reservoir
grid scale.

, k..
33

The average porosity and

permeability is
representative for the grid
cell scale property.

Boundary conditions
The boundary conditions (BC) specify
how the flow is behaving on the
boundaries.

Fixed boundary
conditions (FBC)

P1

Linear boundary conditions (LBC)

P2

No flow

The BC influence how P is applied in

step 2 of the numerical upscaling.
Fixed boundary conditions (FBC)

No flow

Linear boundary
conditions (LBC)

P1

P2

No flow across
boundaries (i.e. similar
to laboratory core
flooding)
Obtain the diagonal
elements of the
permeability tensor (kxx,
kyy and kzz)

Linear pressure drop

allowing flow across
boundaries.

Output is a non-symmetric
9-elements permeability
tensor

Periodic boundary conditions (PBC)

Periodic boundary
conditions (PBC)

34

Periodic or repeatable
structure

Output is a symmetric 9elements permeability

tensor.

Permeability tensor
The orange layers have lower
permeability than the yellow layers.
=

An applied flow in the +x direction gives

a flux in the z direction (i.e. upwards).
A PBC is used here.
This results in a vertical flow and
requires a kzx permeability term in the
Darcy equation.
Two things to note:
The off-diagonal terms are ~10% of kxx.
The kxx is higher than kzz.

In 3D the tensor contains 9 elements.

If the area or volume of the
permeability grid is large enough, the
upscaled permeability is independent
of the boundary conditions effective
permeability (keff).

35

Upscaling

kx=kz: Isotropic
kxkz: Anisotropic

Numerical upscaling of permeability

Px

For a general permeability distribution, the upscaled

permeability cannot be calculated analytically.

k1
k2

For more in-depth overview of this topic, see Renard and Marsily
(1997) Not part of the course material.

Upscaling steps:
1.

Starting point: A fine-scale grid of permeability values

(arbitrary distribution).

2.

Apply a pressure gradient P in the x-direction.

3.

Calculate the fine-scale pressures for each grid cell.

4.

Calculate the inter-bloc flows in the x-direction using Darcys

law:

5.

Calculate the total flow in x-direction by summing individual

flows between any two planes.

6.

Calculate the permeability for all the grid cells (kblock) using
Darcys Law for the upscaled block.

7.

Repeat from step 2 for z-direction.

Upscaling is the method for replacing a high resolution, fine scale heterogeneous
medium with a homogeneous low resolution, coarse scale medium.
The upscaled (or block) permeability is defined as the permeability of an
homogeneous block which under the same pressure boundary conditions will give
the same average flows as the heterogeneous block.
36

Upscaling

k=<kA,kH>

Measurements and averaging

A measurement f represents a finite volume of the porous media and an average of the
properties in this volume.

Fine grained sandstone

P1
P2

P1: Grain
P2: Pore

37

= 1 , 2

The measurement is an average of P1

and P2 (i.e. comparable to upscalig).

Measurements and averaging

A measurement f represents a finite volume of the porous media and an average of the
properties in this volume.

P1

P2

P1: Grains (grey white)

P2: Pore (black)

38

Definition: The representative elementary

volume (REV) is the smallest volume over
which a measurement can be made that will
yield a value representative of the whole.
The correlation lengths () must be much
smaller than the sample volume.

Measurements and averaging

A measurement f represents a finite volume of the porous media and an average of the
properties in this volume.

Flaser to wavy bedded lithofacies

P1
P2

P1: Mudstone (dark brown)

P2: Sandstone (light brown)

39

Definition: The representative elementary

volume (REV) is the smallest volume over
which a measurement can be made that will
yield a value representative of the whole.
The correlation lengths () must be much
smaller than the sample volume.

The Representative Elementary Volume (REV)

Flaser to wavy bedded sandstone

Non-additive properties

Dependent on
spatial
distributions of
components

Additive properties

40

Dependent on
volume fraction of
components
Bulk property

The Representative Elementary Volume (REV)

Definition: The representative elementary volume
(REV) is the smallest volume over which a
measurement can be made that will yield a value
representative of the whole.

41

All model equations contain some geometrical

parameter that is difficult to measure on the real rock.
When using the continuum approach, the permeability
parameter allows us to pass from the microscopic
(detailed) level to a macroscopic level. The
macroscopic model or rock should be on a REV to
ensure effective parameters.

The Representative Elementary Volume (REV)

The REV is important for several issues:
1.

Upscaling at a REV gives a physical

and statistical meaningful value.
Permeability upscaled at an arbitrary
volume is called kblock, while a
permeability value upscaled at REV
(or larger) is called an effective
2)
permeability, keffective.

2.

Understanding, interpretation and

calibration of (well data)
measurements.

3.

Geological length scales.

4.

The flow equations (e.g. Darcy Law)

is defined in a REV.

42

1)

REV
keffective
kblock

Property modelling workflow

Porosity
log

Well

Calibration

Un-cored

Permeability
log

Correlation

Core plugs are

representative samples.
Wireline log is measuring
a parameter related to
porosity and the same
structures as the core
plugs..

The correlation between

porosity and permeability
from core plugs is valid
at wireline log and
reservoir grid scale.

The average porosity and

permeability is representative
for the grid cell scale
property.

= Blocking/Averaging/Upscaling
43

Well in
model

+ spatial correlation structure

Core-log calibration and correlation

Should be easy

Need attention
Heterogeneous system from
core plug to wireline log
scale

Near homogeneous system

from core plug to wireline
log scale

44

1m

1m

Core
plug

Core
plug

Wireline log

Core plugs are measuring above

REV and are representative (the
measurements are relatively equal).
Wireline logs are measuring the
same structures as core plugs and
these can be used for calibration.
Porosity-permeability correlation
probably not different at core plug
and wireline log scale.

Wireline log

Core plugs are measuring below

REV and are not representative
(need to be averaged).
Wireline logs are not measuring the
same structures as core plugs.
Porosity-permeability correlation
probably different at core plug and
wireline log scale.

Sedimentary hierarchy

Clinoform

Clinoform surfaces

Sedimentary heterogeneities at multiple

length scale.
Potentially, all these can influence fluid
flow.

45

Sedimentary hierarchy

Lobe

Lobe
element
Bed

46

Sedimentary heterogeneities at multiple

length scale.
Potentially, all these can influence fluid
flow.

Test

Sedimentary hierarchy

Full field simulation model

47

Object based models

Test
Pixel based model
(Processed based
models)

2 mm

Sedimentary hierarchy

Full field simulation model

Simulation cell
50 000 m3

5m

48

Test

2 mm
100 m

Sedimentary hierarchy

How many scales to model and upscale?

Which scales to focus on?
How to best construct model grids?
Which heterogeneities matter most?

Full field simulation model

Simulation cell
50 000 m3

5m

49

2 mm
100 m

How many scales to model?

Three main scale transitions
(Para)sequence scale

See also Fig. 4-18 in Reservoir Modelling Design

Facies association scale

Lithofacies Lamina scale

scale
(pore and grains)

Core
plug

Wirline
log

30 m

90m

30 m

Simulation model
(~ 100 x 100 x 10 m)

Geological model
(~ 50 x 50 x 1 m)

3) Geomodel to reservoir simulator

An upscaling step often required due
to computational limitations.

50

Lithofaices model
(~ 1 x 1 x 1 m)
2) Lithofacies to geomodel
Lithofacies is distributed in a
geological model according to
conceptual model.
Trends are modelled.

Lamina or core plug properties

(~ 2 x 2 x 2 cm)
1) Pore to lithofacies
Lithofacies is the basic concept in
sedimentary description
Pore scale curves are applied to the
lithofacies model and upscaled.

How many scales to model?

Conventional upscaling approaches

See also Fig. 4-18 in Reservoir Modelling Design

Averaging of well data directly into the flow simulation grid: Fast and simple, but

Upscaling approaches

ignores effect of geological structures not modelled.

Core
plug

Wirline
log

Single phase upscaling only in z. The sim-grid is designed with same x, y as

geo-grid. Uses averaging, but tries to take into account thin layers and barriers. Useful
when good seismic is available that constrains lateral variation

Single phase upscaling in x, y and z. Try to take into account multiscale

effects. Upscaling tools are available (RMS, Petrel) giving diagonal tensors. Neglects
multiphase flow effects.

Multiphase upscaling in x, y and z. Try to take into account both multiscale and
multiphase effects. Time and resource demanding. Less seldom applied. Ongoing
research in this field.

30 m

90m

30 m

Simulation model
(~ 100 x 100 x 10 m)

51
See also Fig. 4-18 in Reservoir Modelling Design

Geological model
(~ 50 x 50 x 1 m)

Lithofaices model
(~ 1 x 1 x 1 m)

Lamina or core plug properties

(~ 2 x 2 x 2 cm)

How many scales to model?

Conventional upscaling approaches

See also Fig. 4-18 in Reservoir Modelling Design

Averaging of well data directly into the flow simulation grid: Fast and simple, but

Core
plug

Upscaling approaches

ignores effect of geological structures not modelled.

Wirline
log

Single phase upscaling only in z. The sim-grid is designed with same x, y as

geo-grid. Uses averaging, but tries to take into account thin layers and barriers. Useful
when good seismic is available that constrains lateral variation

Single phase upscaling in x, y and z. Try to take into account multiscale

effects. Upscaling tools are available (RMS, Petrel) giving diagonal tensors. Neglects
multiphase flow effects.

Multiphase upscaling in x, y and z. Try to take into account both multiscale and
multiphase effects. Time and resource demanding. Less seldom applied. Ongoing
research in this field.

30 m

These four degrees of upscaling complexity help define the number and dimensions
of models required.
90m
1
3
2
The number of scales modelled is typically related to the complexity and precision of
answer sought.
30 m
Improved oil recovery (IOR) strategies and reservoir drainage optimisation studies
areSimulation
often model
the reason for starting
multi-scale approach.
Geologicala
model
Lithofaices model
Lamina or core plug properties
(~ 100 x 100 x 10 m)

52
See also Fig. 4-18 in Reservoir Modelling Design

(~ 50 x 50 x 1 m)

(~ 1 x 1 x 1 m)

(~ 2 x 2 x 2 cm)

Which scale to model?

The models we make for upscaling should be at a
REV because:
1.

Upscaling at a REV gives a physical and statistical meaningful

value (keffective).

2.

The flow equations (e.g. Darcy Law) is defined in a REV.

1)

REV
keffective
kblock

53

Which scale to model?

The physical size of the REV is dependent on:
The property (additive or non-additive)
Geology (i.e. geological length scales)

54

54

Multiscale REV
Note that this is a conceptual sketch and the variation in
measured property and scale on the x-axis will be
property and case dependent.

55

Multiscale REV - Pore scale

Moderately
sorted fine
sand

Poorly
sorted
medium
sand

56

Multiscale REV -Lithofacies scale

57

Multiscale REV - Facies association/Flow unit scale

58

Measured property

Multiscale REV
Relation to conventional measurements

59

59

Nordahl & Ringrose, 2008, Identifying the Representative Elementary Volume for permeability in
heterolithic deposits using numerical rock models, Mathematical Geoscience, vol. 40, p. 753-771.

Multiscale REV

60

Which heterogeneities matter?

Overview of important heterogeneity types for different depositional settings.

Both small- and large scale heterogeneities are important.
When faults are present and studied, they are typically important.
61

Which heterogeneities matter?

The balance of forces concept gives the framework for udnerstanding which scale
most affect flow, e.g. capillary trapping is likely to be important for rocks with
permeability variations at the small scale, while fluid segregation is more important
over larger (vertical) distances and for larger permeabilities.
Relate also to Floras Rule lecture 3.
62

Lecture 10 - Property model

1. Cut-offs, NTG & Kv/Kh
2. Multiphase flow and upscaling

Multiphase flow and upscaling

Lectured by Philip Ringrose

63

Acknowledgements to:
Statoil for data, models and
experience
Kjetil Nordahl, Philip
Ringrose and Mark Bentley
for inspiration and examples
Kjetil Nordahl for providing
the majority of the lecture
notes
Geological Reservoir Modelling:
Arve Nss
[email protected]
www.statoil.com

64