You are on page 1of 12

Setting cut-offs in numerical modeling

Author: Agha Hassan Akram


Co-author: Ammar Ahmad

Aug 2014

Introduction
Oilfield convention requires that hydrocarbon that can never be produced or even
moved towards the producer should not be counted in place. This is done by
applying cut-offs to remove these volumes from the HCIP.
Some references:
Worthington, SPE 84387:

Their purpose is to eliminate those rock volumes that do not contribute


significantly to the reservoir evaluation product.

We should do the core por-perm relationship twice: once initially, before cut
offs are set. After setting cut offs, then establish the relationship again for all
points above the cut off. It may be different.

Upscaled properties and predictive algorithms should relate only to those


intervals that will flow. Applying cut offs before moving to dynamic simulation
helps computational efficiency.

Vsh, Por and Sw cut offs that all tie back to perm and are interrelated are
preferred.

Cobb, SPE 48952:

At a minimum cut offs should represent that portion of the reservoir which
has commercially recoverable hydrocarbon.

In its simplest form, producible net pay can be defined as those portions of
the reservoir that are porous, reasonably permeable, and have producible
hydrocarbons.

Application of cut-offs

We know that below a certain permeability value, fluids will become


practically immobile, and will not contribute measurably to production. Such
rock should be excluded from volumetric calculations of OOIP.

How to set and apply the cut offs? Traditionally we have used por, Sw and
Vclay cut offs. And these cut offs are applied to logs as well as to models.

Usually, a rule of thumb is used, such as in my light oil reservoir, I will use 1
mD as a cut-off.

This cut-off permeability is correlated with porosity using a core por-perm


plot, as well as with Vclay through facies if possible. And this allows the
porosity and Vclay cut-offs to be set.

With the power of 3D modelling, we should use a simpler system of cut-offs:


directly apply a perm cut-off on cells instead of on Por and Vclay. Net flag is
1 if K > 1 mD, for example. The perm in the cell is the best estimate we
have based on Facies, Porosity, Vclay, and Core analysis.

Sw cut-offs are primarily for transition zones. When So is so low that it is not
expected to move, then dont count it as in-place.

Cut-offs are usually only for static models, not for dynamic models. If used in
dynamic models to reduce the number of active cells, do not use the Sw cutoff. The only use we can see for Sw cut offs is in transition zones. Vsh and Por
cut off the tight rock. Sw cuts off rock near the OWC where So drops below
the Sorc critical level. It will kill cells that may conduct aquifer support to the
reservoir. It is only to be applied to the static model for volumetrics.

Use the cut-offs on the fine static model, not on any upscaled model.

Cut-offs should be set after the perm multiplier is applied. Pre-estimate an


approximate perm multiplier as shown in the next section.

Cut-offs will depend on reservoir dynamics: a fractured reservoir can have a


lower cut-off. Similarly, when high perm layers alternate with tight ones, cutoffs will be lower as the high perm pathways allow tight rock to contribute
commercially from a large surface area.

Pre-estimation of the permeability multiplier

Static models are usually populated with permeability values based on core.

For many good reasons, a permeability multiplier has to applied during


calibration or history matching. This can be an order of magnitude or more.
0.1 to 10 are not uncommon as perm multipliers.

The cut-offs should be applied after multiplication. This creates a small issue
of timing: we need volumes-in-place upon completion of the static model, and
before history matching. And the multiplier is only known after history
matching.

One way to pre-estimate the multiplier in time to be used for the static model
match is as follows:

Select representative wells

Obtain measured K*h from pressure build up analysis

Add model K*h (at the well) vertically of all rock connected to the
tested interval. That means not only the perforated interval, but
additional permeable rock that is part of net pay vertically in case of
limited entry or partial perforation.

Divide the first by the second to get the multiplier.

If pressure build ups are not available, get well PI from production rates
and pressures.

Estimate model PI using the model K*h of the connected rock and a
skin guesstimate.

Divide the first by the second to get the multiplier.

In case of a horizontal well, adding up the K*h will not work as described
above. In this case, use the following steps for the multiplier. This can also be
used for vertical wells:

Add the K*h of all cells within a radius of investigation based on the
PBU. Say 1000m.

Divide this total by the number of cells within this radius as seen from
above. i.e. the number of cells laterally in the i and j direction within
the given Rinv.

This is the estimate of model K*h. the rest of the steps are as above.

An idealized way to approach setting cut-offs

Set up a 2D fine cell model, with very small cells. The cell size should be
based upon the small scale heterogeneity of the rock, and in the absence of
other compulsions, core plug size cells should be used. This is because we are
using measurements at core scale to define rock properties. Smaller than
core scale variations are lumped into this measurement.

The model should be as large as a few full sized grid blocks, so it could be,
say, 1000 ft long (x-direction) and 20 ft high (z-direction). It needs only one
cell in the y-direction. It should have a producer at one end, and if
appropriate to the expected depletion mechanism, an injector at the other.

The permeability should be distributed based on the best understanding of


vertical and lateral variations in permeability at core scale.

The model should then be produced using the recovery mechanism expected
in the field. For example, 1 PV of water injected thru it. Similar pressure
gradients should be used: in real reservoirs away from wellbores, it does not
exceed 1 2 psi/ft. And durations: i.e. 30 years.

Different perm multipliers should be applied to the model, and the point at
which recovery becomes negligible is determined. (say URF < 1%)

The average perm of the whole model (after applying the perm multiplier) is
now the cut-off for full size grid cells.

This procedure may be ideal, but is not followed in practice, and is included
for completeness.

Example of proving cut-offs


One way of testing a cut-off permeability number, say 1 mD, is to set different perm
cut-off numbers in the calibrated dynamic model, and to test which is the minimum
number below which there is no significant change in the production forecast. This
needs a few runs. Suppose a run is made at 10mD cut off. Then another run at 5
mD cut off, and we note, for example, that there is a significant increase in
cumulative production. Then at 1 mD, again a significant increase in production.
Then 0.1 mD and we see no increase in production. This means that 0.1 < cut-off <
1 mD.
See Figure below for an example.

Note how setting the cut-off too high reduces the recovery. As we reduce the cut-off,
we reach a point where it no longer increases recovery: 5% in this case. This proves
that below 5% porosity, oil does not contribute to production.

Setting cut-offs in numerical modeling without good core data


In many cases, core tests of the required number and accuracy are not available.
There may be too few tests, there may be a lack of accuracy between the lab test
and in-situ rock and fluid behavior, etc.
To cope with this, an alternative method is proposed, which relies on the presence of
filtrate invasion to indicate that the rock is conductive and should be included in the
net. One reason for doing this is that the fact that a zone tested gas, for example,
only tells you that at least a part of the tested interval has produced. The interval is
usually composed of different quality rock with varying porosity and permeability,
and it is difficult to tell which part of the tested interval produced and which part did
not.
The method proposed is as follows:

Display the moved hydrocarbon or Sw = Sxo Swc on the well section

Make sure there is a reliability flag that tells you when not to believe it. This
may be based on the caliper. Only do this in good quality data zones.

Use the permeability indication of Sw > 0 to decide whether or not rock is


permeable or not.

Decide on the cut-offs using this method. i.e. if por < 4% shows no invasion,
and Por > 4% starts to show invasion of mud filtrate, set the Por cut-off at
4%.

One approach is to set a cut-off flag manually in the Petrel well section after
eyeballing the invasion flag in good data zones. Wherever you see invasion,
flag it as net pay. Then set a flag that is a function of the three cut-offs, i.e.
Netflag if Por >4%, Sw < 75%, Vcl < 25%. Then try different values of the
three cut-offs to get the two flags as similar to each other as possible where
data quality is reliable.

An example is described and shown below


(1) First step is to create a manual PAY / NON PAY Flag based on the following
criteria;
(a) Look for a clean borehole interval (without washout)
(b) Secondly see if you can find an invasion profile against that clean
borehole interval
(c) If the first two conditions are met then manually mark it as a Pay Zone
against the invaded profile ONLY.
(d) If there is no invasion, it is telling us that the rock is too tight to allow
invasion, and hence marked as non-pay.

(2) The second step is to generate PAY / NON PAY Flag from calculator by
applying cutoffs for variables like Sw, Phie, and VCl. The aim then is to match
this flag generated by calculator by applying different values for cut off
variables with the earlier PAY / NON PAY flag generated manually. The values
for the cut off variables where the PAY of these two flags match are your cut
offs. It is important to mention that these PAY Zones are only matched
against a clean borehole interval showing an invasion profile.
Note also that there will be many pay zones marked by the Petrel generated
flag at depths which were not considered of good quality data, and hence not
marked manually by you with the manual net pay flag. This mismatch is
simply ignored. We are only interested in the match in zones of good data.

Manually
marked PAY

Invasion
Profile Sxo Sw

PAY marked
by applying Cut
Offs

Upscaling cut-offs: Calibrating model volumes with the OH log


ground truth.
The fine scale model will require different cut-offs to the OH log for various reasons
including scale of measurement/cell size. Adjusting fine scale model cut-offs needs
a reference point or calibration. The ground truth is the OH log. That is as close as
we can get to the truth in estimating static volumes. We need to ensure that our
model matches the OH log in terms of HCIP. The method to do this is as follows:
Compare model volumes and OH log volumes. To do this, sum up as follows for each
wells OH log, by zone or reservoir:
*h*(1-Sw)*(NTG flag)
Where:
-

is log porosity
H is height of measurement, 6 inches is the common number for logs
(1-Sw) represents the HC saturation
NTG flag is based on cut-offs and is 1 if the measured point is considered
included in the net, and 0 if it is considered to be excluded from the net.
The units of this calculation are length units, i.e. ft

This represents the HCPV in the well as seen by the OH log. This is then summed up
for all the wells in the field, by zone or reservoir.
The same is then done at each well for the fine scale static model after it has been
populated with properties, and cut-offs have been applied. The number is then
compared to that from the OH logs. If it is too large, then the cut-offs should be
modified to reduce it and vice versa. In the case of the fine scale model, h now
becomes the height of a cell, rather than the measurement interval of 6 that we
used for summing up for OH logs.
Note that to get a match, we can tweak the Sw vs Phi transform to increase or
decrease water, or we can change cut-offs. There is a limiting factor on how much
we can tweak the Sw vs Por transform: a good way to assess how well the Sw
transform is working is to eyeball the well section in Petrel with the model Sw
displayed on the log along with the OH log Sw. They should track each other and the
model Sw should not show a systematic bias. Having satisfied ourselves that the Sw
transform is optimal, we are left with cut-offs as the preferred method of ensuring
that volumes are preserved between the OH log and the model.
An example is given below of using this method to adjust cut-offs after satisfying
ourselves that the Sw modeling is adequate:

In the Figure below it can be observed that the filled Sw curve shows the Sw derived
from OH logs while the red colored curve tracking this OH log Sw represents the
modeled Sw. It can be observed that there is a very close match between OH logs
and Sw Modeled; therefore validating an important Sw QC check. We can now
conclude that the modeled Sw in the Fine Grid Model represents the actual Sw
observed on OH Logs.

OH Logs Sw
Modeled
Sw

0.95 % Difference

Table displaying summation of LOG, UPSCALE and MODEL *h*(1-Sw)*(NTG flag).


This QC is performed to confirm if the model has the same amount of GIIP as on Log
Scale.

Deriving properties from porosity


The volumes are directly impacted by how we populate properties in the model, and
how we cut-off the non-net cells.
Consider the Vcl vs Por crossplot shown below based on OH logs, upscaled at the
well to fine grid size. In this case, the OH log has measurements every 6 and the
fine grid has a height of 2m. The Vcl cut-off of 6% in this case, and the por cut-off of
4 pu in this case are marked.

The figure displays Por Vs Vcl plot with a cut off of 4 % Por and 6 % Vcl marked.
It is clear that there is a relationship between Vcl and Porosity. And it is also clear
that if we apply the Vcl cut-off, a few high-por cells will be cut-off, more medium por
cells will be cut-off, and a larger number of low por cells will be cut off.
What are our modeling options?

1. We can establish a best fit line through the points and distribute Vcl
according to this, as a function of porosity. The result will be that all high
porosity points will have low Vcl, and will be included, and all low porosity
points will have high Vclay and will be excluded. So no high or medium por
points will be cut-off. Our resulting volumes will be too high.
2. We can populate Vcl purely at random regardless of porosity, based upon the
overall Vcl histogram. As a result when we apply the Vcl cut-off, roughly equal
numbers of high and low pr cells will be cut off. This will result in volumes
being too low.
3. We can create porosity bins or facies, and in each bin populate Vcl as per the
histogram of Vcl corresponding to its por bin, based on the OH logs upscaled
to fine scale model. This is our best shot at getting correct volumes.
This is illustrated below in an example.

Facies 3 Phie 6-12%


Facies 2 Phie 3-6%

Facies 4 Phie >12%

Facies 1 Phie 0-3%

The figure displays 4 Porosity Bins created to populate VCl. It is important to


mention that all 4 of these porosity bins have their respective histograms according
to which VCl has been populated in the model.
Other properties such as permeability may be modeled the same way. Sw may be
modeled this way, however it could create difficulties in matching modeled Sw to
OH log Sw, and a single line fit may be better. It is common practice to model Sw

and Perm as deterministic functions of porosity, but modeling permeability as a


random function based on the process shown above is more realistic.

You might also like