Geomodeling Network Newsletter Discusses Fault Modeling Techniques

The Geomodeling Network Newsletter May 2009

H as it really been 2 months since the last Geomodeling Network

newsletter? Based on the number of (kindly) reminder emails I received
in my inbox this morning asking where the May edition is, I think it must
be!

This month’s newsletter is one of the best yet and has contributions from
E&P companies, software vendors as well as a great discussion taken
from our online forum.

Talking of emails, you may have spotted that this newsletter does not
have the Blueback Reservoir watermark running through it. The reason
for this is that a couple (2) of you recently contacted me requesting that
this should be removed and thus making it easier to read. Never being
one to shirk from my responsibilities, (especially when the elderly are
concerned), I have removed the offending watermark – the jumbo-print
version should be available for the next release :o)

Anyway, for the next 20 or so minutes sit back, relax, grab a cup of coffee
and enjoy the latest offering of the Geomodeling Network newsletter.
Many thanks to those members who took the time to contribute the
interesting articles contained in this version, it’s very much appreciated.

And finally, as our network quickly approaches the 800 members mark, I
hope some of you will take inspiration from the articles and discussions of
this (and previous newsletters). If you do get the urge to make a
contribution for future versions, drop me an email with your thoughts –
the next one is not due out until the end of July 2009, so you have plenty
of time!

Mitch Sutherland
mitch.sutherland@blueback-reservoir.com

Page 1 The Geomodeling Network – Sponsored by Blueback Reservoir www.blueback-reservoir.com


Table of Contents

1. How “Good Looking” are your Faults?
The first of a series of short articles that will look at faults and fault geometry,
using straightforward structural geological principles.
Titus Murray & Merrick Mainster - FaultSeal Pty Ltd Page 3

2. Rock types and flow zones
Practical methods for defining rock types, their use in property models and flow
zone characterisation.
Steve Cannon – Senior Staff Geologist at DONG E&P UK Ltd Page 8

3. The Petrosys Plug-in for Petrel
The Petrosys Plug-in for Petrel allows geoscientists and engineers utilizing Petrel
to present their insight, integrated with information from many other data
sources, through the Petrosys map interface.
Scott Tidemann, Global Sales & Marketing Manager at Petrosys Page 18

4. What problems have you had using horizontal well data
within your models?
This was a question placed on the Geomodeling Network discussion forum
which generated a fair bit of response from our members
Brian Casey – Geological Consultant at Oxy Page 19

5. EAGE 2009
This year’s event is in Amsterdam. Page 26

6. The Blueback Toolbox (a Petrel plug-in) – update Page 27



Member Articles, Reviews & Questions

The most exciting phrase
to hear in science, the one
that heralds new
discoveries, is not 1. How “Good Looking” are your Faults?
'Eureka!' but 'That's Titus Murray & Merrick Mainster, FaultSeal Pty Ltd
funny...'
This is the first of a series of short articles that will look at faults and fault
Isaac Asimov geometry, using straightforward structural geological principles. We aim
to help your understanding by showing examples from our software
application FaultRisk that we use on a daily basis when assessing fault
seal capacity in our consulting business.

Within a faulted 3D model it is important to understand the uncertainty
related to the position and throw of the faults in the model. In many of
the models we come across in consulting projects we see that the
structure is “watertight” it is also in some cases geologically improbable.
This is generally due to faults not being imaged in seismic but they are
actually inferred from the absence of a reflector.

Due to the inherent problems of seismic fault imaging, the following
uncertainties arise in:

Position of the footwall;
Throw on the fault;
Shape of the fault;
Stratigraphic thicknesses;
Growth across the fault;
Tectonic inversion.
The best way to define fault displacement is based on detailed well
correlation but wells are generally only drilled on one side of the fault!

Throw Profiles

The displacement on a fault should vary systematically across and down
the fault plane. Faults should have a point of maximum displacement with
a zero throw at the tips of the fault, the throw diminishes radial from the



maximum displacement. See the diagram below which shows the footwall
and hanging wall of a faulted stratigraphic layer.

In most faulted reservoir cases the lateral variation of displacement is the
key factor and when looking at the displacement it is common to review
these profiles.

If the profile is not consistent is it likely that the fault is segmented or
there is another problem with its interpretation in some way as shown in
the diagram below.

Gulfax Field Examples

As an example of this type of analysis we will look at the Gulfax model
that ships as a demonstration example set in Petrel.
“The use of solar energy
has not been opened up Fault polygons have been made from the Petrel grid and imported into
because the oil industry FaultRisk™. The picture below shows the FaultRisk™ mapping interface,
with a structure contour map loaded.
does not own the sun.”
Ralph Nader



As the fault polygons are made from the Petrel grid it includes a set of XYZ
coordinates that can be split into hanging wall (down-thrown) and foot
wall (up-thrown) lines. This diagram shows either side of the fault as a set
of points that we can edit or modify.

“The past history of our
globe must be explained by
what can be seen to be
happening now. No powers When reviewing the displacement profile in FaultRisk™ any potential
anomalies in the profiles can easily be identified. In the case below a
are to be employed that are
“Bow Tie” displacement can be seen.
not natural to the globe, no
action to be admitted except
those of which we know the
principle.”
James Hutton



Looking at another fault from the Petrel model, segmentation along the
fault’s length can be observed.

“The oil can is mightier
than the sword.”
Everett Dirksen

In this fault there are some anomalous cross cutting faults in the hanging
wall



One has to look out for
engineers -- they begin with
sewing machines and end
up with the atomic bomb.
Marcel Pagnol

that generate a complex displacement profile.

In an ideal world one would review the seismic data to look for fault
linkages and branch lines. Pragmatically in a fault seal analysis the fault
can be split into two or three segments to look for leak points from the
compartment.

This style of analysis is quick and easy to do and will greatly improve the
quality, and accuracy of your 3D models and help you amend the Petrel
model with good looking faults. When investigating the probability of
fault leakage and/or across fault flow this analysis is a vital step in the
workflow.

The next article will look at throw length ratios and fault segmentation.

If you have any queries on this article or fault seal issues please contact us
at titus@faultseal.com and/or visit our website www.faultrisk.com.



2. Rock types and flow zones
Steve Cannon, DONG E&P UK Ltd

This article attempts to present some practical methods for defining rock
types, their use in property models and flow zone characterisation. Rock
typing is common practice in Middle Eastern carbonate reservoirs
because they tend to be extensively cored with vast conventional and
special core analysis datasets used for petrophysical interpretation. Each
petrophysical data point will usually be associated with a petrographic
description based on thin section analysis, and often SEM data as well:
such comprehensive and consistent datasets are less common in clastic
reservoirs. Such detailed studies can have their downside however,
especially when an over-enthusiastic sedimentologist defines 85
lithotypes in a field where 40% are in non-reservoir sections and the rest
just subsets of the about eight major petrophysical rocktypes; some
simplification is required before they can be used for reservoir modelling.
The list below attempts to define some of the nomenclature commonly in
use to define different levels of description:

Lithofacies/lithotype: the character of a rock described in terms of its
visible components; structure, colour, mineral composition, grainsize,
sorting etc: the smallest scale purely geological description of a rock.

Facies: a mappable unit of rock that forms under certain conditions of
sedimentation, reflecting a particular process or environment. A facies
may be defined by the types of component lithofacies and is an
interpretation rather than a description of any unit.

Facies association: a group of facies that together define a sedimentary
unit with a common depositional setting; again this is an interpretation
based on an understanding of the different components and their process
of deposition. The recognition of a specific facies and facies associations
defines the depositional environment of a group of rocks.

Rock type: a rock with a well defined porosity network leading to a
unique porosity-permeability relationship and saturation profile: the



porosity network is the result of a predictable depositional and diagenetic
history.

Petrofacies/petrotype: terms used to integrate petrophysical
relationships with lithofacies/lithotype descriptions: petrophysics in a
geological context. The term petrofacies is also used by many researchers
to define only mineralogical/petrographical classes.

Flow unit/flow zone: a mappable unit of the total reservoir volume
within which geological and petrophysical properties that effect fluid flow
are internally consistent and predictably different from other reservoir
rock volumes. Other terms used are hydraulic flow unit and genetic
hydraulic unit that attempt to standardise or map different petrotypes
within discrete depositional environments. A flow unit, while ideally
being related to geologically defined depositional package, may not
correspond with discrete facies boundaries and may not be laterally and
vertically contiguous.

Essentially all these terms fall into two categories, either purely
descriptive or largely interpretative: both are important to understand
when characterising a reservoir, especially for dynamic modelling. But
what does all this mean to the different subsurface disciplines? To a
geologist, a flow unit is a discrete facies object such as a channel or a
carbonate shoal; to a petrophysicist it is correlatable zone with similar
petrophysical properties; to a reservoir engineer it is a layer in a model
that has a consistent and predictable dynamic response to flow in the
simulator. To a reservoir modeller it is all of these things! Petrophysicists
and engineers still often think in terms of simplified zone average
property models as the way forward in determining in-place volumes and
reserve estimates, hence the concept of discrete zones rather than the
more stochastic idea of reservoir objects with common petrophysical
properties.

Amaefule et al (1993) developed a method of reservoir description using
core and log data to identify hydraulic flow units and predict permeability
in un-cored intervals. This method has been used in many ways to define
both rock types and flow units, and is based on well founded
experimental methods developed over many years. The method is
depends on understanding the pore geometry of a rock and relating this
to the mean hydraulic unit radius of the pore throats: mean hydraulic


“It puzzles me how they radius realtes porosity, permeability and capillary pressure
know what corners are measurements. A similar approach was published by Kolodzie (1980)
good for filling stations. based on work down by Winland of Amoco. Pore geometry is a function
of the mineralogy and texture of a rock, which means that different
Just how did they know
lithofacies may have similar pore throat attributes: in this way different
gas and oil was under
facies may belong to the same rock type.
there?”
Dizzy Dean

Theoretical background
The basis for all rock typing is Darcy's Law and Pouseille's theory for
capillary bundles under laminar flow: these are used to derive a
relationship between porosity and permeability for a capillary bundle.

2 2
e r2 e r r
e mh
k 2 2 2
8 2 2 2 Equation 1

The mean hydraulic radius is function of grain surface area (Sgv) and
effective porosity.

2 e 1 e
S gv
r 1 e rmh 1 e Equation 2

Carmen and Kozeny obtained the following relationship by substituting
for mean hydraulic radius.

3
e 1
k 2 2
2
1 e Fs S gv Equation 3

where F is a shape factor, which for a circular cylinder is 2. In real rocks
the Kozeny constant (Fsτ2) can vary between 5 to100. Because it is a
quot;variable constantquot;, varying between hydraulic units in a reservoir a

Page The Geomodeling Network – Sponsored by Blueback Reservoir www.blueback-reservoir.com
10


further mathematical transformation is required: dividing both sides by
effective porosity and taking the square root of both sides the following
relationship is derived:

k e 1
e 1 e F s S gv Equation 4

where permeability, k is in μm2.

Presenting permeability in millidarcies, then a Reservoir Quality Index
(RQI) can be defined:

k
RQI 0.0314
e Equation 5

e
The Flow Zone Indicator (in μm) is related to RQI by the term z
1 e
where φz is the ration between pore volume and grain volume.

Thus Flow Zone Indicator,

1 RQI
“Sometimes, I guess there FZI
Fs S gv22
Equation 6
z
just aren't enough rocks.”
Forrest Gump

On a log-log plot of RQI against φz (PhiZ) all samples with a similar FZI will
lie on a straight line with unit slope; samples with other FZI values will lie
on parallel lies. Samples that lie on the same straight line have the same
pore throat attributes and therefore constitute an hydraulic unit, even
though they may represent different facies. Permeability can also be
calculated from this relationship using the appropriate hydraulic unit or
FZI relationship:

3
2 e
k 1041( Fzi ) 2
1 e Equation 7

11


Corbett et al (2005) present an excellent case study of braided river
sandstones which goes through the workflow and offers some interesting
interpretations of the effect of grain-size and sorting on capillary pressure
measurements.

Workflow
There are two main phases to the process; firstly analysis of core data to
determine the various petrophysical relationships and secondly
application within a reservoir modelling workflow where the relationships
are integrated with log data.

Data analysis

1. Routine core porosity and permeability data should be characterised in
terms of an appropriate lithofacies or facies scheme. It is essential to
have a consistent dataset, and any outlying data removed. To ensure
data integrity later in the process it is important that depth
correspondence between core and log data is accurate. The data should
be plotted in the normal way and general porosity permeability
relationship established for one or two dominant facies groups, if
sufficiently different: these can be used to produce a permeability curve
from the wireline log derived porosity.
2. In a spreadsheet, the flow terms PhiZ, RQI and FZI should be calculated
for each core analysis point and sorted in order of decreasing values of
FZI. As a first approximation the results can be plotted as two groups,
greater than and less than a value FZI equal to one on a log-log plot of
PhiZ against RQI: the better rock types will be greater than one. Any
recognizable rock type variations will be apparent as parallel groups of
data. Generally this will reveal the better rock types, and correspond with
those that have previously been identified from the facies breakdown as
better quality.
3. Plotting porosity against permeability classified by FZI will allow the
generation of a predictive permeability relationship for each rock type.
The quality of the predicted permeability can be compared with the
original core permeability using a Q-Q plot: an organised plot of
comparing the same points Alternatively the relation shown in Equation
12


quot;The box said 'Required 7 can be used to predict permeability and again be compared with the
Windows 95 or better'. So, original data.
I installed LINUX.quot; Application

4. Within the reservoir modelling package calculate the FZI terms using the
log derived porosity and permeability. Using a property calculator
develop a series of logical statements to classify each of the rock types
according to the core defined scheme. Visually check the results against
cored intervals to ensure reasonable correspondence and consistency:
anything over an 80% correspondence is acceptable; 60 to 80% is a
common result. The key is to ensure that the extremes are captured,
especially low permeability layers that could form barriers/baffles, and
high permeability streaks that might dominate flow in the reservoir.
5. Block the data to grid scale and check that the detailed description of rock
types is retained in the upscaled well data.
6. Either use the rock types directly to populate a model zone or build a
detailed facies model and populate each facies/object with the
appropriate rock type. Reservoir properties, porosity, permeability and
water saturation can then be distributed according to the rock type
relationship defined in the analysis stage. When modelling S w, a direct
link to core-based capillary pressure data can be established with respect
to height above a local or regional free water level. The value of SCAL
data cannot be over-emphasised; capillary pressure measurements
should be representative of the different rock types recognised. A data
base of as little as ten samples can be sufficient to characterise a series of
reservoir rocktypes; fifty is even better!

Other applications
In an attempt to standardise all possible lithofacies in terms of hydraulic
flow zones, Corbett & Potter (2004) used the Amalaefule methodology to
create 10 Global Hydraulic Units (Figure 1) that utilised fixed FZI lower
boundaries against which they plotted core derived porosity and
permeability from different depositional environments. This is an
alternative to the clustering method described above and relies on

13


mapping these predetermined classes on the porosity-permeability
crossplot; they coined the term petrotyping.

Because the rock types in the petrotyping
FZI Global Hydraulic Unit approach are quot;globalquot; in the sense they are
predetermined, the base map can be used
48 10 to determine whether a reservoir
24 9 comprises one or more rock types.
12 8 Different reservoirs can be compared
6 7 quickly using this method as well as a rapid
3 6 technique for screening and selecting
1.5 5 samples for further analysis. This method
0.75 4 can also be used to decide the appropriate
0.375 3 scale of cells needed to capture the
0.1875 2 porosity and permeability distribution in a
0.0938 1 reservoir model; ideally at the smallest
scale,

each grid block should contain an individual global hydraulic unit.

Using an example data set from a series of braided river deposits, Corbet
et al (2005) demonstrated the workflow output. Figure 2 shows the
results of the PhiZ:RQI log-log plot and the four hydraulic units indentified
with the corresponding FZI values. Figure 3 recasts the data in terms of a
familiar Phi:K plot with the predictive relationships calculated for each
hydraulic unit or rocktype. Figure 4 shows the result of grainsize and
sorting analysis for each hydraulic unit/rock-type: grain size shows a clear
contrast whereas sorting has little impact.

Conclusions
Rock-typing can be a challenging process, but ultimately very satisfying:
engineers would much rather talk about a rock-type than a lithofacies
because it infers some sort of numerical consistency, even when it is
directly related to depositional unit! But these methods must be used

14


quot;If at first you don't with care; a detailed understanding of what rock-types might be expected
succeed; call it version and how they can be grouped is required. Rock-typing works when the
1.0quot; geologist is in control of the input and the output, especially when that
output is going to be used to populate a static model.

References
Enhanced Reservoir Description: Using Core and Log Data to Identify
Hydraulic (Flow) Units and Predict Permeability in Uncored
Intervals/Wells: Amaefule et al, SPE 26436 (1993)

Use of Flow Units as a Tool for Reservoir Description: A Case Study:
Guangming et al, SPE 26919 (1995)

Permeability Prediction by Hydraulic Flow Units – Theory and Application:
Abbaszadeh et al, SPE 30158 (1996)

Early Interpretation of Reservoir Flow Units Using and Integrated
Petrophysical Method: Gunter et al, SPE 38679 (1997)

Petrotyping: A basemap and atlas for navigating through permeability and
porosity data for reservoir comparison and permeability prediction:
Corbett & Potter, SCA2009-30 (2004) (Society of Core Analysts)

The geochoke test response test response in a catalogue of systematic
geotype well test responses: Corbett et al, SPE93992 (2005)

15


Figures

Figure 1: Global hydraulic units as applied to a shallow marine sandstone

Plot of RQI vs. Phi(z) for well X2

10
HU-1 FZI = 2.509

HU-2 FZI = 1.233

HU-3 FZI = 0.685

HU-4 FZI = 0.323
1
RQI

0.1

0.01
0.01 0.1 1
Phi(z)

Figure 2: Log-log plot of PhiZ against RQI used to define different hydraulic units

16


Crossplot of (k vs. Phi) for different
Hydraulic Units, Well X2

1000

100

10

1
k, mD

0.1

0.01

0.001

0.0001
0 0.05 0.1 0.15 0.2 0.25

Phi, frac.

Figure 3: Porosity-permeability cross-plot broken down by hydraulic units

Grain Size and Sorting for each HU
4

3.5
Grain Size and Sorting

3

2.5
, Phi Units

2

1.5

1

0.5

0
G7HU1 G7HU1 G7HU2 G7HU3 G7HU4 G7HU5

Hydraulic Units

Figure 4: Impact of grainsize (violet) on definition of hydraulic units

17


When NASA first
started sending up
astronauts, they
3. The Petrosys Plug-in for Petrel
Accelerate exploration, improve productivity.
discovered that pens Get collaborative mapping results more easily.
would not work in zero Scott Tidemann, Petrosys

gravity. To combat this
problem, NASA The Petrosys Plug-in for Petrel allows geoscientists and engineers utilizing
Petrel to present their insight, integrated with information from many
scientists spent a decade
other data sources, through the Petrosys map interface. This enables asset
and $12 million teams to accelerate decision making through consistent use of Petrosys
developing the ball point mapping and surface modelling as their focus moves from the regional
pen that writes in zero 8 8227 2799 > Americas: 1888 PETROSYSthe reservoir scale. 6555 > Calgary: +1 403 537
Australasia: +61
overview through the field to
> Europe: +44 141 420
gravity, upside>down,
5600 Web: www.petrosys.com.au
Harness the power of the Petrosys plug-in for Petrel to:
underwater, on almost
any surface including Start Petrosys mapping, surface modelling or 3D viz from icons in the
glass and at Petrel application.
temperatures ranging Effectively map and present
from below freezing to opportunities by directly
over 300C. incorporating Petrel 3d seismic
horizons and 3d model grids using
When confronted with Petrosys map colorfill and 3D viz
displays. Compute and map
the same problem, the contours for the structures.
Russians used a pencil.
Integrate decision making, using a
range of other Petrosys display
options to overlay geoscience and Your vital information comes together, with
cultural data from OpenWorks, both applications working side by side to
support collaborative workflows and
GeoFrame, ArcSDE, SMT, PPDM and understanding.
many other data sources directly
accessible through Petrosys.

Map in many coordinate reference systems (CRS); the underlying CRS of
maps can be switched to effectively map surfaces in regional
interpretation situations.
18


Use Petrel seismic data as a direct input data source in Petrosys gridding
workflows.

Effectively combine 2d & 3d
interpretive workflows, using
Petrosys surface modelling
functions such as volumetrics and
well tie. Use direct data inputs and
efficient import/export facilities,
while creating repeatable workflow
processes.

Import faults, model grids/horizons
and seismic data to Petrosys. Export
Petrosys grids directly into Petrel
projects. Petrosys effectively and efficiently handles the
* Petrel is a mark of Schlumberger. mapping of Petrel models, including faults, colorfill
display and posting of surface values.

4. What problems have you had using
horizontal well data within your models?
Taken from a discussion posted on the Geomodeling Network discussion
forum.
Brian Casey, Oxy

Many modern fields are dominated by horizontal wells, but modelers are
reluctant to use this data due to:
- Zonal bias
- Imprecise tops
- Imprecise well path locations
- Other?

19


We handle zonal bias through debiasing workflows. Imprecise well paths
and imprecise tops can be managed through careful data selection and
the use of Zone Log. Do group members have other modeling issues and
solutions for the use of horizontal well data in their models?

Brian

Li bin – geologist, Tiandi Energy
The well path locations are relatively reliable, the tops could be corelated
using chosed logs. the reservior quality and facies also could be analysed
and evaluated. all these information can be used in modeling

Samir Benmahiddi – production geologist, Sonatrach
the way I see it, regardless the zonal bias, continuous lateral data
sampling from a decent number of horizontal wells, well distributed
throughout the reservoir, should help to check / adjust the main reservoir
attributes anisotropy and variograms, should provide hints on lateral
heterogeneity and deposits architecture, particularly if you come to run
imaging tools (which is quite difficult I agree but still less than coring, and
always worth to try) with intent to collect some imagefacies and dipmeter
data allowing much better facies mapping at least.
But still it should be framed with a robust sedmentary and stuctural
conceptual model and avoid mixing fractures/faults with simply a
lamination of high contrast steeply dipping on image log !
Otherwise, tops issue is only important when target is a tiny window and
you add up uncertainties on depth because the cable length stretch and
such ...which is less likely to happen.

Anders Ørskov Madsen –Senior Consultant, Blueback Reservoir
I use horizontal well data regularly for building reservoir models. Intead of
using the whole horizontal well I have in some cases simply cut out a
section where I had high confidence to which zone/stratigraphic unit it
was drilled through. In other case I have made a pseudo trajectory for the
well due to the depth uncertainty for long horizontal wells, in order to
place it correctly in the model.

Petter Abrahamsen –Research Director, NCC
We are curently making a software for getting the surface consistent with
the zonation in wells. This assumes the zonation is correct. A (vertical)
correlated uncertainty on the well path location is possible to include but
hasn't been prioritized so far. The easy part is to ensure that surfaces
cross the well trajectories at the correct locations. The hard part is to
ensure that they do NOT cross the trajectories at the wrong locations.

20


Here is a link with some more details:
http://www.nr.no/pages/sand/area_Cohiba
“It is only when they go The introduction in the manual (pdf) gives a good overview.
wrong that machines
remind you how powerful We essentially use kriging to interpolate the well picks. The challenge is to
use the additional constraints from the horizontal wells in a consistent
they are” manner so that we can provide realistic uncertainty description in terms
of simulations (Monte Carlo) and prediction errors.
Clive James
Thorbjorn Pedersen –Chief Geoscientist, Oxy
Interesting to see Madsen's approach using a pseudo trajectory to place
the horizontal well quot;correctlyquot; in the model. What is the depth accuracy
of the model in the first place? Do you have vertical well tops near the toe
end of the Horizontal trajectory? Or is the depth model for horizons
constrained by depth converted seismic horizons? In that case what is
bigger, the depth uncertainty to the depth conversion or the depth to the
toe end of the horizontal well?

Holger Rieke –Principal Geologist, StatoilHydro
Peter,
Could you please elaborate why you choose kriging for the interpolation
between well picks? Thickness of reservoir zones or the depth surface to
well tie are not accurately computed using kriging unless you select a
large variogram i.e. at least half the distance of well spacing. The kriging
result then resembles convergent or global b-spline (depending on which
software you prefer). Those algorithms are in my opinion more suitable to
extrapolate these types of data.

The reason we using kriging is to be able to quantify uncertainty. The
uncertainty is (indirectly) described by the shape of the variograms and
the standard deviations (sill). The depth uncertainty is quantified by
prediction error maps (kriging error) or by a set of simulated realizations
depending on usage.

The shape of the variograms can be chosen so that the result is similar to
spline interpolation. This is visually appealing but rarely realistic in natural
phenomena. Moreover, the variogram shapes and the sill can be
estimated from well picks so that we can confirm consistency between
interpolation method and data.

We never use zone thickness data directly since these are only available
in vertical wells. We always use surface depth data (well picks) and

21


“Geologists don't wrinkle, constraints on the surfaces in the horizontal sections. This is to avoid the
use of pseudo-data that are hard to make and even harder to justify.
they show lineation”
Finally note that we consider many surfaces, and the intervals (zones)
between them, simultaneously. Variograms for all interval thicknesses are
specified. This can amount to 20 or more different variograms. The big
advantage of considereing all surfaces simultaneously is that well data
influence surfaces below and above. In particular horizontal sections in
thin zones lock surfaces above and below very accurately.

I hope this didn't obscure things rather than clarified them :-).

Thorbjorn Pedersen –Chief Geoscientist, Oxy
I like Petter's approach assessing all surfaces and their associated
variograms at once. This allows for a holistic look into the relationship
between surfaces, their controlling data and subsequently the total
structural/stratigraphic architecture.
However, dependent upon the well density and geologic setting, I would
still maintain that from time to time usage of zone thickness data may be
required to maintain a morphology that is in line with the respective
sedimentological setting defined by core data or infered from regional
context. These cases generally tend to be used in models of fields in their
early stages of development.

Tim Wynn –Senior Reservoir Geologist, AGR-TRACS
We have built several models with a large number of horizontal wells and
found problems with the use zone logs option in Make Zones (zone logs
not honoured, random spikes etc).

In a large 'layer cake' stratigraphy reservoir we used pseudo tops shifted
by the requisite isochore thickness, this was quite successful, particularly
a there was no seismic data constraint. However, care had to be taken
around the faults so it was quite time consuming.

We have considered using psuedo well paths (polygons) clipped to the
required zones and shifted up by an arbitrary amount. These polygons
could then used as part of the input data for the surface. These would
only be required where the surface cuts a well where it shouldn't

Both these options are quite time consuming and result in pseudo data so
they are not perfect but they do ensure the surfaces honour the zone
logs.

22


To the optimist, the glass Anders Ørskov Madsen –Senior Consultant, Blueback Reservoir
Answer to Thorbjorn question about 'pseudo trajectory'
is half full.
To the pessimist, the It depends on the well density. In cases where I have shifted the
horizontal well trajectories it has usually been on fields with a bunch of
glass is half empty. vertical or deviated wells near the hz wells highlighting the depth
To the engineer, the glass uncertainty of the horizontal wells. In many cases the hz wells have been
more than 25000 ft MDRT with an depth uncertainty of +/- 50 ft TVD at TD
is twice as big as it needs (e.g. chalk wells in the Danish North Sea), so it has been neceassry to
to be shift these wells in order to use them as input in a 'base case' structural
depth model.

Note that we can use trends for the zone thicknesses so we can impose
interpreted sedimetological trends. The trends can be globally adapted to
data (well picks and trajectories) or kept untouched. The simplest trend is
of course a constant (e.g. 20m). The kriging essentially interpolates the
difference between the trends and the observed data. Trends can also
include velocity fields, travel times, anomalies, pinch outs and all kinds of
weird geological features but thats another story.

As Tim Wynn comments, the biggest challenges are really areas close to
faults where simple layer cake models can fail. Normal faults will squeze
zone thicknesses to zero and this requires special care to avoid opening
up the faults. Reverse faults are even worse since this requires multi-z
values at surface locations near the fault. This is currently not handled but
we are discussing how to integrate surface and fault models in a proper
and efficient way.

Brian Casey –Geological Consultant, Oxy
Further to Anders comments, it is not just extremely long reach horizontal
wells with + 50 ft TVD error that should concern us. Horizontal well
placement is relative to our grid dimension, so both the geologist and
simulation engineer should be concerned. Even if no static properties are
attributed to a horizontal well, dynamic performance must still be
matched. If the well is mis-located in the grid the engineer will make the
adjustments. Better to be pro-active…

Thank you to Petter and everyone else for contributing to this discussion.
We may wish to test some of these processes for placing horizontal wells
“correctly” in the model. Also, I had not previously considered Petter’s
approach toward horizontal well placement uncertainty. I can see a lot of
applicability, in both the static and dynamic modeling.

There is considerable support for using the static properties of horizontal
23


wells, and as Samir observed, these wells provide information on lateral
heterogeneity and depositional architecture. [that we might not observe
in the vertical wells.] How we handle that lateral reservoir data may need
further discussion.

Petter, perhaps you would like to further address handling multi-z values
in the case of reverse faults (and recumbent folding), and the proper
integration of fault and surface data. I would be very interested in your
thoughts and approach, but do not wish to bury that discussion within
one about horizontal well placement. It deserves its own topic.

Juan Cottier –Subsurface Manager, Blueback Reservoir
2 ideas to add in here of a purely pragmatic nature:

Firstly, I have built all sorts of models (in PETREL) using horizontal wells
(West African, UK, Danish and Norwegian producing fields) and have
found that from quot;get the job donequot; approach that if one takes the MAKE
HORIZONS and MAKE ZONES steps slowly and try to achieve
incrementally improving results then a decent job can be done in most
cases. In PETREL I tend to turn off the quot;use in geomodelingquot; option for
most horizontal wells and repeat and repeat layer by layer slowly turning
on each well or even each individual well top. This allows simple QC of the
results and the quot;problem childrenquot; can be more easily identified.

Secondly, those of us who have been around more than 10 years have
seen how deviation surveying has changed and how so much more
confidence can be put upon surveys. However, MWD surveys still do not
get close to a wire/slickline survey in terms of accuracy so it is worth
checking where you survey comes from. Also if drilling we'll often get a
MWD survey and then some days later there maybe a wireline survey ...
so has the project trajectory been updated? Anoother little wrinkle I came
across was the elipses of uncertainty on MWD surveys in particular. As
they are based upon Hall's Effect the uncertainty varies in geographical
regions and azi/inclination. In the Ivory Coast we were drilling horizontal
wells to the south, running along the earth's magfield and the lateral
uncertainty was huge. Similar thing for UK north sea t 60 degrees (ish)

Right ... I'm dragging on a bit, I'll stop now.

Keith Milne –Petroleum Geologist
This subject has generated a lot of discussion because we regularly come
across fields with a mixture of vertical and horizontal wells.
Regarding the point about producing a sensible zonation, lets not forget

24


that the well picks and position of the horizontal well takes more
quot;If GM had kept up
interpretation that a vertical well and there may be more than one
with technology like the alterative. I do not expect any software to be able to solve the problem
computer industry has, without some additonal data points to control the zonation, especially if
seismic control is poor or if the borehole passed through a fault. If logs
we would all be driving
have been run that enable dip to be estimated, then this will assist in
$25 cars that got 1000 determining the zones along the well. One approach is to make a cross
MPG.quot; section before trying to model - this is what would be typically done
during actual drilling of the well, using all available data sources.
Bill Gates
Knut Midtveit –Sales Manager, Roxar
I have seen many attempts to solve the problem of handling horizontal
wells and getting the model to honour the well data including zonelog in
horizontal wells. This range from manual thus very tedious approach to
clever scripting methods, to what I feel is a more holistic approach that
Peter Abrahamsen talk about, so I look forward to that.

Roxar has been including adjust model to zonelog functionality for a
couple of years, and we are now seeing it being successfully use on some
pretty large fields with many horizontal wells. I find it interesting to
observe that some companies focus a lot on this and spend lot of time
getting the model right in order to plan wells optimally, and others accept
that no one has a good solution.
Furthermore some companies regularly shift well positions manually
though, and others object strongly to the concept of shifting the well
position.

Personally I hope that we will get a solution where you consider all your
data with a certainty and allow both seismic envelopes and wells to be
changed according to the uncertainty of each data type.

Until we have a such a solution try the adjust to zone log in RMS, and yes
it can handle faults to.

Ahmad Nazhri Mohd Zain –Geological Modeler, Saudi Aramco
How do we handle the large upscaling issue with regards to the horizontal
wells?. I assume that having a 25m x 25m grid lateral grid dimensions is
acceptable to take into account the horizontal section of the wells, thus
for a 1km horizontal section, you will have around 40 grid cells. I work
with Ghawar and I cannot have anything smaller than 250m x 250m grid
spacing otherwise my static model will be in the hundreds of millions
cells. How do I handle 6000 data points (6 inch sampling rate) in a 1km
horizontal section? That 1km will be blocked into 4 grid cells only. You will

25


not be able to compare the blocked wells to the raw dataset as the
horizontals will introduce a severe bias due to the amount of samples in
the horizontal sections.

This is a major issue at the moment in our modeling group.

Petter Abrahamsen – Research Director, NCC
For surface modelling this is not a big problem in practice since the well
geometry is very regular. Sampling the well at approximately the grid
spacing is sufficient in our experience. We can go back and check all the
data points although this is hardly necessary (in our experience). We have
tested our approach on Troll (North Sea) which is a giant field. But
working with Ghawar is of course an even greater challenge due to the
volume of data. It would be nice to do a practical test on such a huge
field...

Upscaling for petrophysics is of course a different issue since comparing
e.g. permeability on plug scale and on modelling scale is non-trivial.

5. EAGE 2009 – Amsterdam 8th to 11th June
http://www.eage.org/events/index.php?eventid=103

I am sure that a lot of you will have attended, exhibited and indeed
presented at previous EAGE’s. In the past, these conferences have been
held in fantastic cities such as Rome, Vienna and Madrid, as well as
Leipzig. The event organizers have chosen another great city to host this
year’s event, with Amsterdam being the chosen one for 2009.

There are many things that pop into my mind as I think about Amsterdam
(notably canals and tulips and a certain brewery). However, distractions
aside, the EAGE is shaping up to be quite an event, not least for Blueback
Reservoir.

Throughout the entire conference, Blueback will be exhibiting at stand
#2538 where we will have a number of staff available to discuss
geomodeling consulting opportunities, Bridge and the Blueback Toolbox,
software development on the Ocean framework, as well as the
Geomodeling Network itself.

26


If any of you are attending this event then please feel free to swing by our
booth for an informal chat and some Blueback hospitality.

6. The Blueback Toolbox (a Petrel plug-in)
Our Toolbox has been widely available for a month or so now and already
we are seeing a great take-up with this FREE software. Indeed the
reception we have received for the Toolbox is such that users are already
seeing the benefits to using plug-in technology to supplement their
existing Petrel workflows. A few users have already requested additional
plug-in suggestions which we are planning to have ready for the next free
release of the Blueback Toolbox – these suggestions include:

- Facies Maps
- Shift Well Log
- Merge Seismic Cubes
- Cube flattening seismic volume attribute
- Make empty seismic cube

If you would like access to the Blueback Toolbox, then please refer to the
March 2009 edition of the Geomodeling Network newsletter on how to
download the software and request a license. Or drop an email to
paul.hovdenak@blueback-reservoir.com and he will get back to you with
information on what you need to do.

Requests for the newsletter No6

The next newsletter is planned for a July 2009 release, so please send articles to
me at the following email address for inclusion (mitch.sutherland@blueback-
reservoir.com).

Finally, please take advantage of the Geomodeling Network discussion board on
LinkedIn to initiate comments on any Geomodeling subject of interest to you, or
to respond to any of the articles in this newsletter – all I ask is that you respect
other people’s opinions.

Fin
27

Geomodeling Network Newsletter Discusses Fault Modeling Techniques

Recomendados

Recomendados

Mais conteúdo relacionado

Semelhante a Geomodeling Network Newsletter Discusses Fault Modeling Techniques

Semelhante a Geomodeling Network Newsletter Discusses Fault Modeling Techniques (12)

Último

Último (20)

Geomodeling Network Newsletter Discusses Fault Modeling Techniques