SPE 150863

Proper Selection of Upscaling Techniques for Different Production Processes

Ahmed M. Shehata, Ahmed H. El-Banbi, M. Helmy Sayyouh, Cairo University

Copyright 2012, Society of Petroleum Engineers

This paper was prepared for presentation at the North Africa Technical Conference and Exhibition held in Cairo, Egypt, 20–22 February 2012.

This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents of the paper have not been
reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect any position of the Society of Petroleum Engineers, its
officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written consent of the Society of Petroleum Engineers is prohibited. Permission to
reproduce in print is restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract
One main challenge in reservoir modeling is the computational costs of flow simulation. Although computer power has
significantly increased in the past few decades, effective upscaling to reduce computation time in heterogeneous
reservoirs still remains a challenge.
This research demonstrates the results of a comprehensive study on common upscaling techniques for porosity and
permeability. This work aims at providing guidelines on which upscaling methods are superior for different
depletion/production processes. The processes investigated include straight depletion, water injection, miscible and
immiscible gas injection.
Several fine grid heterogeneous models were developed using actual field data and heterogeneity was evaluated using
Lorenz and Dykstra Parsons Coefficients. Those different models were used to generate benchmark results for
comparison with coarser models developed at different levels of coarsening and with different upscaling algorithms.
Several scenarios were investigated in order to identify which upscaling algorithms result in the closest match with the
fine model results. The results of coarse models were treated statistically to quantify their deviation from the fine models
results and rank the upscaling algorithms accordingly. It was found that the depletion process affects the choice of the
appropriate upscaling algorithm.
The coarse grid models lose some accuracy with increasing the upscaling ratio, as suspected. Also we found that
several upscaling algorithms work equally well in depletion processes. However, only flow-based upscaling techniques
(especially with open boundary condition) worked well for water injection processes. For miscible and immiscible gas
injection processes, the use of flow based upscaling techniques is generally a must, while the flow-based upscaling
technique with closed boundary conditions shows slightly better results than other flow-based upscaling techniques.
For miscible gas injection processes, in particular, the use of which upscaling technique is sensitive to the heterogeneity
index as well specially for models with Lorenze Coefficient higher than 0.5.

Introduction
Advances in geostatistical techniques allow reservoir properties, such as porosity and permeability, to be described by
more and more fine models. However, these detailed reservoir models cannot be used directly for numerical simulation
because they would require high amount of time and computational effort and also an elevate memory storage capacity.
Upscaling is the methodology for transferring the properties from a fine scale to a coarse scale with an adequate
number of cells in order to be used in reservoir simulation.
The properties addressed for Upscaling are both of static and dynamic nature. Static properties (Permeability and
Porosity) can be upscaled through an averaging scheme or through single phase flow calculations while dynamic
properties (Fractional flow, Relative permeability …) are upscaled through multi components flow calculations. In this
study, permeability and porosity variation were studied since heterogeneity of a reservoir is mainly caused by these two
properties.
Porosity was upscaled using volumetric weighted arithmetic mean. Permeability was upscaled using several algorithms
in PetrelTM such as: arithmetic mean weighted, harmonic mean weighted, geometric mean weighted, root mean square
weighted, arithmetic-harmonic mean, harmonic-arithmetic mean, diagonal tensor method with (Open, Closed, and
2 SPE 150863

Closed-K) boundary conditions, and full diagonal tensor method with (Open, Closed, and Closed-K) boundary
conditions.1
Sensitivity analysis was performed to study how influential the two properties were on the output parameters such as
field production oil rate, field production water rate, field production gas rate, field gas oil ratio, field injection water rate,
field injection gas rate, filed oil production total, and field gas production total.

Methodology
Upscaling is the concept of creating a relationship between a fine-scaled geological grid and the coarser simulation grid.
In this study, the PetrelTM software was used to provide the necessary upscaling tools to easily coarsen very large
reservoir models to sizes acceptable to commercial fluid reservoir simulators. Upscaling processs is split into two
stages:
1. Creating the coarser geometry for the simulation grid as shown in Fig. 1.
2. Scaling up the property.

Fig. 1: Upscaling Process

Prior to Scaling Up the property, the coarse grid cells for each model will be generated. In this paper we will generate
two coarse models so will generate two different coarse grids. The upscaling of properties can be divided into two main
categories of scale up methods including averaging and flow based tensor upscaling. The averaging methods are used
for most properties. The flow based tensor upscaling is used especially for permeability. The behavior of permeability is
quite different from that of other properties since the effective permeability is not only different from the static but is
usually also varies in different directions.
In this paper for porosity upscaling1, the volumetric weighted arithmetic mean was applied to preserve the pore volume
locally and globally throughout the reservoirs. For permeability several algorithms were applied such as arithmetic mean
weighted, harmonic mean weighted, geometric mean weighted, root mean square weighted, arithmetic-harmonic mean,
harmonic-arithmetic mean, diagonal tensor method with (Open, Closed, and Closed-K) boundary conditions, and full
diagonal tensor method with (Open, Closed, and Closed-K) boundary conditions. APPENDIX-A is a review for the
different Upscaling techniques used in this paper.
The dynamic reservoir simulation will then be used to predict the reservoir performance at the fine grid and coarse grid
levels under different different depletion/production processes (Natural Depletion, Water Injection, and Gas Injection).
Apart from the grid cells with porosity and permeability parameters, any data required for the simulation (i.e. relative
permeability, initial fluid distribution, and fluid PVT properties) will be treated the same at the fine and coarse scales.
In this paper, the reservoir simulations for natural depletion and water injection drive mechanisms are executed through
a Black-Oil simulator. The reservoir simulations for gas injection drive mechanism are executed through Compositional
simulators.
Through sensitivity analysis the influence of upscaled porosity and permeability were studied on output parameters
such as field oil production oil, field Water production rate, field gas production rate, field gas oil ratio, field water cut,
field injection water rate, and field injection gas rate. The results of the fluid flow performance at the coarse grid model
will then be compared against its fine grid fluid model. The error between the fine model and coarse models is
calculated by using the following equation.

The selection of appropriate Upscaling algorithm for each model will be judged according to how well the fluid flow
prediction made at the coarser level mimics the prediction of field performance at the finer level.
SPE 150863 3

Description of Fine and Coarse Models used

Fine grid heterogeneous model was developed using actual geological, geophysical, petrophysical and reservoir
engineering data. The model is a 3-phase (Oil, Gas, and Water) and described on a regular cartesian grid. The
dimensions of the model are 3,814 meters long by 2,500 meters wide by 52 meters thick. The model was initialized with
a Water-Oil Contact of -2,990 meters subsea and a Gas-Oil Contact of -2,910 meters subsea.
The fine scale cell size is 40 m x 40 m x 2.5 m. The fine scale model has 95 x 63 x 20 (119700 cells). The permability
and porosity distribution are shown in Fig. 2 and Fig. 3. The vertical permeability of the model was altered from the
horizontal. The model used here has a Kv/kh of 0.1.
The fine model will be coarsened from 95 x 63 x 20 (119700 cells) to two different sizes. Fig. 4 through Fig. 6 presents
the 3D view for both fine and coarse grids. Table 1 - presents the difference between the grid properties for the fine
model and the corases models.

Fig. 2: Permeability distribution for fine model Fig. 3: Porosity distribution for fine model

Fig. 4: 3D view for fine model Fig. 5: 3D view for coarse-1 model Fig.6: 3D view for coarse-2 model

Table 1: Grid size properties

Grid Size Properties

Dimension, m
Model Nx Ny Nz No. of Cells
Areal

Fine 40 40 95 63 20 119700

Coarse‐1 80 80 47 31 10 14570

Coarse‐2 160 160 23 16 5 1840

4 SPE 150863

Numerical Simulation Results

In the following section, a simulation model has been used to analyze the effect of upscaling techniques for different
production processes such as Natural Depletion, Water Injection, and Gas Injection. Table 2 presents the used
upscaling algorithms for porosity and permeability for this study.

Table 2: Summary of the studied simulation runs

Upscaled techniques for Natural Depletion Drive Mechanisms

In this case, the exercise was performed using the natural depletion drive mechanism. The model contains five
production wells (D-1, D-2, D-4, D-5 and D-6). All wells are vertical and completed in all the layers of the simulation
model. A 3D view of wells location in the model is presented in Fig. 7. The fine model and the coarse models were run
for ten years and the results were compared. Fig. 8 throug Fig. 11 present comparison between fine and coarse models
in case of natural depletion reservoirs.

Fig. 7: 3D view for wells location

Fig. 8 and Fig. 9 present the variation of field oil production rate and field gas production rate with time respectively for
upscaling ratio 2:2:2. From these figures it is clear that the error between fine model and coarse model is negligible.
Fig. 10 and Fig. 11 present the variation of field oil production rate and field gas production rate with time respectively
for upscaling ratio 4:4:4. On the other hand, as the upscaling ratio increased from 2:2:2 to 4:4:4 a slightly deviation
appears between the fine and coarse models at early time. This deviation is fading after three years. We found that the
results did not differ regardless of the upscaling techniques, but the upscaling ratio is the main parameter in reservoirs
with depletion drive mechanism.
SPE 150863 5

It is worth mentioning that the main driving mechanism is oil expansion, and it seems that the results were not sensitive
to which upscaling technique was used as shown in Table 3.

Fig. 8: Field gas production rate for fine vs. coarse-1 models Fig. 9: Field oil production rate for fine vs. coarse-1 models

Fig. 10: Field gas production rate for fine vs. coarse-2 models Fig. 11: Field oil production rate for fine vs. coarse-2 models

Table 3: Error comparison of the different upscaling techniques for depletion drive mechanism
6 SPE 150863

Upscaling Techniques for Water Injection

This case performed by using the same model that was used in the previous case considering the water injection. The
following features were applied in this model:
• Five wells were used for water injection (WINJ-1, WINJ-2, WINJ-5, WINJ-6, and WINJ-7) and three wells for
production (D-1, D-2, and D-6) as shown in Fig. 12.
• The model was run for 10 years.
• All wells are vertical and completed in all the layers of the simulation model.

Fig. 12: Wells location in simulation grid

Fig. 13 and Fig. 14 illustrate the comparison between fine model and coarse models with upscaling ratio 2:2:2. Fig. 13
shows the variation of field water production rate with time. It indicates that there is a noticeable difference between
both cases in early times. This difference is decreases with time. Fig. 14 presents the field gas oil ratio with time. It is
clear that some upscaling techniques resulted in less error than the other. This difference is small at the early time and
increases with time. Fig. 15 and Fig. 16 illustrate the comparison between fine model and coarse models with upscaling
ratio 4:4:4. Fig. 15 shows the variation of field water production rate with time. It indicates that there is an apparent
difference between both cases of using coarse and fine grid at the beginning that gradually decreases with time. Fig. 16
shows the field gas oil ratio of the fine and coarse cases with time. At the early time there is a slightly difference
between the fine and coarse cases, while at late time the difference increases.
We found that the possible algorithm that could be used to represent the fine scale fluid flow behavior is the flow based
upscaling technique with open boundary condition. Also, as the upscaling ratio increased from 2:2:2 to 4:4:4 we found
that loosing the vertical heterogeneity will affect on the results and increase the error between the fine model and
coarse model as shown in Table 4.

Fig. 13: Field water production rate for fine vs. coarse-1 models Fig. 14: Field gas-oil ratio for fine vs. coarse-1 models

Fig. 15: Field water production rate for fine vs. coarse-2 models Fig. 16: Field gas-oil ratio for fine vs. coarse-2 models
SPE 150863 7

Table 4: Error comparison of the different upscaling techniques for water injection

Upscaling Techniques for Gas Injection

Gas injection has become very important as a means to increase oil recovery in oil field development. In this section we
are focusing on upscaling for immiscible and miscible gas injection. For the immiscible gas injection we used the
produced hydrocarbon gas to reinject it into the reservoir, while for the miscible gas injection we used CO2 as injection
gas.

Upscaling Techniques for Immiscible Gas Injection

Fig. 17 presents the 3D views for immiscible gas injection fine model. In this case one well was used for gas injection
(WINJ-1) and four wells for production (D-1, D-2, D-5, and D-6). The model was run for 10 years. Fig. 18 throug Fig. 21
present comparison between fine and coarse models.

Fig. 17: 3D view for model of immiscible gas injection

Fig. 18 and Fig. 20 show the comparisons between the field oil production rates with time of fine and coarse models for
upscaling ratio 2:2:2 and 4:4:4 respectively. It is apparent that at the early time there is no difference between the fine
and coarse models, but after one year some coarse models results show some error more than other. This error
decreases with time and reaches the minimum at the end of time.
Fig. 19 and Fig. 21 display results for field gas production rates with time of fine and coarse models for upscaling ratio
2:2:2 and 4:4:4 respectively. Here we see a slightly error at the beginning between the fine and coarse models in case
of upscaling ratio 2:2:2. This error disappears at the end of time particularly after 6 years, while in case of upscaling
ratio 4:4:4 we see a significant error between the fine and coarse models. This error decreases with time and disappear
after 8 years.
8 SPE 150863

It is clear from these Figures that in case of upscaling ratio 4:4:4 the error is higher than in case of upscaling ratio2:2:2.
Also, we found that the diagonal tensor with closed boundary conditions worked well for immiscible gas injection drive
mechanism. Table 5 presents the errors for all the upscaling techniques for reservoir with immiscible gas injection drive
mechanism.

Fig. 18: Field oil production rate for fine vs. coarse-1 models Fig. 19: Filed gas production rate for fine vs. coarse-1 models

Fig. 20: Field oil production rate for fine vs. coarse-2 models Fig. 21: Field gas production rate for fine vs. coarse-2 models

Table 5: Error comparison of the different upscaling techniques of immiscible gas injection
SPE 150863 9

Effect of Injection Rate on Immiscible Gas Injection

After testing the effect of immiscible gas injection on the selection of upscaling technique, then we investigated
the effect of the injection rate on the selection of the upscaling technique for immiscible gas injection. We
performed different injection rates scenarios such as 1,000,000, 2,000,000, and 5,000,000 m3/day. Table 6
presents the results for all the upscaling techniques for reservoir with different gas injection rates.
This table indicates that the error between the fine and the coarse models increases when the injection rate
increases. Also, we found that for different injection rates, the flow based upscaling technique with closed
boundary conditions gives the best results.
Table 6: Error comparison of the different upscaling techniques for different injection rates

Upscaling Techniques for CO2 Miscible Injection

Gas injection processes are most effective when the injected gas is nearly or completely miscible with the oil in the
reservoir. Recent activity in miscible flooding has focused on the CO2-miscible drive so in this section we are focusing
on upscaling a continuous CO2 miscible injection process to determine the suitable upscaling techniques. Fig. 22
presents 3D views for CO2 miscible injection. The following features were applied in this model:
• One well was used for gas injection (WINJ-1) and three wells for production (D-2, D-5, and D-6).
• The model was run for 10 years.

Fig. 22: 3D view for model of miscible gas injection

The comparison between the fine and coarse models for continuous miscible gas injection are summarized in Fig. 23
through Fig. 26. Fig. 23 shows the field oil production rate with time for upscaling ratio 2:2:2 and Fig. 24 show the field
oil production rate with time for upscaling ratio 4:4:4. From these Figures it is clear that there is a deviation between the
fine and coarse models at the early time, while all the upscaling techniques give the same results. This deviation
decreases with time and ends after 2 years for upscaling ratio 2:2:2 and extended for 4 years for upscaling ratio 4:4:4.
Fig. 25 and Fig. 26 show comparison between the gas productions with time for upscaling ratio 2:2:2 and 4:4:4
respectively in case of continuous miscible gas injection drive mechanism. From these Figures it is clear that there is a
difference between the fine and coarse models from the early time and some upscaling techniques give less error than
other techniques. This difference between the fine and coarse models in case of upscaling ratio 4:4:4 are higher than
10 SPE 150863

the deviation between the fine and coarse models in case of upscaling ratio 2:2:2. Table 7 presents the errors for all the
upscaling techniques for reservoir with miscible gas injection drive mechanism. It was found that the flowbased with
closed boundary conditions is more suitable for miscible gas injection. This is shown by the smaller error resulted.
We noticed that the error in case of miscible gas injection is higher than in case of immiscible gas injection. This error
may be due to the high injection pressure or miscibility effect. Also, we found that in both cases miscible and immiscible
the suitable upscaling technique is the flowbased with closed boundary condition.

Fig. 23: Field oil production rate for fine vs. coarse-1 models Fig. 24: Filed gas production rate for fine vs. coarse-1 models

Fig. 25: Field oil production rate for fine vs. coarse-2 models Fig. 26: Field gas production rate for fine vs. coarse-2 models

Table 7: Error comparison of the different upscaling techniques for CO2 miscible injection
SPE 150863 11

Quantify the Impact of Reservoir Heterogeneity on CO2 Miscible Injection

A vital factor that influences the oil and gas reservoirs performances is the reservoir heterogeneity. Heterogeneity in a
hydrocarbon reservoir is referred to as nonuniform, non-linear spacial distribution of rock properties. In this section the
effect of reservoir heterogeneity on the upscaling techniques for miscible gas injection driving mechanism is tested.
Initially, several models with different heterogeneities were generated using petrel software. Then the difference
between the fine and coarse models for different heterogeneities in gas injection driving mechanism is examined. To
quantify the heterogeneity levels in the present models, the Dykstra-Parsons coefficient, VDP and Lorenz coefficient, LC
are applied. Table-8 shows the classification of the different heterogenous reservoir models. The ranges of
Heterogeneity Index are:
• HI = 0, Ideal homogeneous reservoir
• 0 < HI < 0.25, slightly heterogeneous
• 0.25 < HI < 0.5, heterogeneous reservoir
• 0.5 < HI< 0.75, very heterogeneous reservoir
• 0.75 < HI < 1, the reservoir is extremely heterogeneous.
Table 8: Classification of the different heterogeneous reservoir models
Dykstra Parsons
Model Lorenz Coefficient Heterogeneity Type
Coefficient
Model-1 0 0 Homogeneous
Model-2 0.14 0.11 Slightly heterogeneous
Model-3 0.31 0.26 Heterogeneous
Model-4 0.67 0.64 Very-Heterogeneous
Model-5 0.79 0.73 Extremly-Heterogeneous

Upscaling Techniques for Different Heterogeneous Models with CO2 Miscible Injection
In this section we focused on upscaling of continuous CO2 miscible injection process in order to determine the influence
of varying reservoir heterogeneity on the selection of upscaling algorithm. In this section the different upscaling
techniques will be applied for coarse model with upscaling ratio 2:2:2.
For homogeneous reservoir Fig. 27 and Fig. 28 present a comparison of the total oil production, and total gas
production with time respectively for fine and coarse models in case of miscible gas injection. We observed that the
curves obtained for the coarse grid is practically coincident to the curve obtained for the fine grid. The results from the
coarse and fine models are not differ due to there are no properties variations in both models.

Fig. 27: Field oil production total for coarse homogeneous models Fig. 28: Field gas production total for coarse homogeneous models
vs. fine vs. fine

Here the results of testing slightly-heterogeneous reservoir using different upscaling algorithms are summarized. The
heterogeneity index of this reservoir is 0.13 using lorenz coefficient. Fig. 29 and Fig. 30 present comparisons between
the total oil production and total gas production for fine model and coarse models in case of continuous miscible gas
injection drive mechanism.
These figures indicate that at early time there is no difference between the fine and coarse cases, while at the late time
there is a deviation between the fine and the coarse cases. It is obvious that the difference is the same by using the
different techniques of upscaling.
12 SPE 150863

Fig. 29: Field oil production total for slightly-heterogeneous coarse Fig. 30: Field gas production total for slightly-heterogeneous coarse
models vs. fine models vs. fine

The effect of heterogenous reservoir was studied using different upscaling algorithm for CO2 miscible injection. The
heterogeneity is 0.31. Fig.31 shows the total oil production with time. This figure indicates a noticeable difference
between the fine and coarse cases from the start date. Fig. 32 Presents comparisons between the total gas production
for fine and coarse models. Also, this figure indicates a noticeable difference between the fine and coarse cases from
the start date. These figures indicate clearly that the error between the coarse model and the fine model is the same by
using different upscaling algorithms.
We found that for models with heterogeneity lower than 0.5 there is a difference between the fine and coarse models,
while all the upscaling technique give the same results.

Fig. 31: Field oil production total for heterogeneous coarse models Fig. 32: Filed gas production total for heterogeneous coarse models
vs. fine model vs. fine model

The results of comparing the fine model versus the coarse models for very-heterogeneous reservoir are shown in Fig.
33 and Fig. 34. The heterogeneity index for this model is 0.67. Fig. 33 and Fig. 34 present comparisons between the
total oil production and total gas production for fine and coarse models in case of continuous miscible gas injection drive
mechanism respectively. The comparison showed that there are differences between the results of the fine and coarse
models.
It was found that the full tensor with closed boundary conditions is more suitable. This is shown by the smaller error
resulted as shown in Table 9.

Fig. 33: Field oil production total for very-heterogeneous coarse Fig. 34: Fied gas productionl total for very-heterogeneous coarse
models vs. fine models vs. fine model
SPE 150863 13

Table 9: Error comparison of the different upscaling techniques for very heterogeneous model

Finally, we examined the effect of using different upscaling techniques in extremely heterogeneous reservoir. The
heterogeneity index of this model is 0.79.
It is clear from Table 10 that the flow analysis of error calculated for the different upscaling techniques showed
that the best method was also the full-tensor upscaling technique using closed boundary conditions.
Fig. 35 shows the total oil production with time. This indicate a significant different between fine coarse and fine
models. This difference started from the early time and increase with time. Fig. 36 shows the total gas production
with time. This figure presents a slightly deviation between the coarse model results and fine model result. We
found that for miscible gas injection processes, the use of which upscaling technique is sensitive for reservoirs
with heterogeneity index higher than 0.5.

Fig. 34: Field oil production total for extremely- heterogeneous Fig. 36: Field gas production total for extremely -heterogeneous
coarse models vs. fine coarse models vs. fine

Table 10: Error comparison of the different upscaling techniques for extremely heterogeneous model
14 SPE 150863

Guidelines For Upscaling

This section provides guidelines for the Upscaling. The following is based on the results of the simulation runs carried
out in this study.
• It is recommended to keep the heterogeneity of the static coarse model to be as close as possible to that of static
fine model. The importance of this recommendation increases as we go from depletion drive to immiscible
displacement. It becomes most important in case of miscible displacement.
• We recommended using the flow based upscaling with open boundary condition in case of water injection, while
using the flow based upscaling with closed boundary condition in case of gas injection.
• It is recommended to generate models with different upgridding level using the best upscaling technique and run
these models and then compare the results with the fine model.
• It is preferred to use the finite difference numerical method more than the harmonic average petrel solver.

Conclusion
Based on the simulation results and analyses shown above, several conclusions are made as follows:
• The use of different existing algorithms resulted in different overall field performance for the various driving
mechanisms.
• For natural depletion drive mechanism, the results did not differ regardless of the Upscaling techniques, but the
Upscaling ratio is the main parameter.
• In case of natural depletion mechanism, when the upscaling ratio increased from 2:2:2 to 4:4:4, the error between
the fine model and its coarse model increased from 5.4% to 62.4%.
• Only flow-based upscaling techniques (especially with open boundary condition) worked well for water injection
processes.
• For immiscible gas injection drive mechanism, it was found that the flow based upscaling technique with closed
boundary conditions worked well.
• After testing the injection rate effect for immiscible gas injection, we found that the diagonal tensor upscaling
technique with closed boundary conditions gives the best results and the error between the coarse model and the
fine model increase when the injection rate increase.
• In case of miscible gas injection drive mechanism, based on the results obtained from the models and the
sensitivity studies, it may be concluded that the full tensor with closed boundary conditions is the most effective
technique.
• The error resulting from different upscaling techniques in case of miscible gas injection is higher than in case of
immiscible gas injection.
• Our results show that in the upscaling miscible and immiscible processes one may need different boundary
conditions from waterflood processes.
• After testing the Heterogeneity effect for miscible gas injection, we found that in case of homogenous reservoirs,
there was no differences in the results of both models (fine and coarse). Also, in case of low heterogeneous
reservoirs, results of using different upscaling techniques showed negligible variations. However, results of high
heterogeneous reservoirs showed significant variations; therefore “full-tensor” was used.
• It is worth to try more than one upscaling technique to check the effect of the upscaling technique on the quality of
the simulation results.
References
1. PetrelTM Software Manual: Version 2008.1” Schlumberger, 2008.
2. Farid A. Allam, Ahmed H. El-Banbi, Sami Bustami, Tamer H. Saada, and Ismail I. Fahmy “Analysis of Upscaling
Techniques in Heterogeneous Reservoirs with Different Driving Mechanisms”, SPE 90292 presented at the SPE
Annual Technical Conference and Exhibition in Texas (2004).

Nomenclature
FOPR = Field Oil Production Rate
FWPR = Field Water Production Rate
FGPR = Field Gas Production Rate
FGIR=Field Gas Injection Rate
FGOR = Field Gas Oil Ratio
FGIT=Total Gas Injected from the field
FOPT = Total Oil Produced from the field
FGPT=Total Gas Produced from the field
3
m /day = Cubic meters per day
SPE 150863 15

APPENDIX-A: Upscaling Methods

In the following section an overview is given on some of the upscaling and flow-based upscaling methods that have been used
2
in reservoir simulation.
Arithmetic mean weighted
∑
∑
Harmonic mean weighted
∑
∑
Geometric mean weighted

∑
Root mean square weighted

Arithmetic-harmonic method
In this method the cells in each plane perpendicular to one of the axis direction (flow direction) are arithmetically averaged to
arrive at an approximate permeability for that plane. The plane permeability’s are then harmonically averaged to give the
permeability for the coarse grid cell in that axis direction. This is repeated for each of the I, J and K directions to produce the
three different output permeability properties.
Harmonic-arithmetic method
In this each line of cells in each axis direction is harmonically averaged to give an approximate permeability for each line.
These lines are then arithmetically averaged to give the upscaled permeability of the coarse grid cell for that axis direction.
Again, this process is repeated for each primary grid direction.
Flow-based tensor upscaling
Flow-based upscaling involves performing a flow simulation on the block of fine cells coinciding with each coarse cell to
determine a representative coarse cell permeability value.
The basic idea of flow based upscaling is that during simulation, the average flow for a given pressure gradient in the coarse
grid should remain the same as in the finer grid. This is accomplished by finding the flow field for three different cases with
pressure drops in the I, J and K direction. First the pressure field is calculated, and then a flow field is derived from that. The
average flow rate (in the three pressure cases) is extracted from the finer grid cells and imposed on the corresponding coarse
cell. There are three different boundary conditions open, closed, and closed K. In open boundary condition a pressure drop is
applied in one direction at a time, while a decreasing pressure is applied to the remaining boundaries. Cross flow between
cells is allowed. Closed boundary condition means that you apply a constant pressure drop in one direction at a time to
compute the permeability in this direction. The remaining boundaries are closed. The closed K boundary condition is identical
to the open boundary except that the top and bottom of the cell are closed when applying a horizontal pressure drop.