Proceedings of the 3rd (2011) CUTSE International Conference

Miri, Sarawak, Malaysia, 8-9 Nov, 2011 (224 -- 917)

Application of Hydraulic Flow Units and Intelligent

Systems for Permeability Prediction in a Carbonate
Reservoir
Tohid Nejad Ghaffar Borhani Seyed Hossein Emadi
Chemical engineering department, Faculty of chemical and Chemical engineering department, Faculty of chemical
natural resources engineering engineering
University Technology Malaysia University of Tehran
Johor Bahru, Malaysia Tehran, Iran
[email protected] [email protected]
Abstract—An accurate prediction of absolute permeability is A. Traditional approach
5B

necessary for developing effective reservoir characterization

Despite the fact that this approach is usefull in some cases,
programs.Permeability predictions in complex carbonate
there is no valid unique empirical permeability estimation for
reservoirs are generally complicated by the local variations in the
reservoir properties. An improvement in the prediction of
all porous media. One of these correlations is Carman-
permeability values from standard well logs and core data Kozeny’s that shows the dependency of permeability on
through the utilization of Hydraulic Flow Units Approach (HFU) average grain size, tortuosity and flow zone index [1, 2]. Tixier
and an intelligent network is presented in this work. Besides, developed a simple model to calculate permeability as a
combined geological-petrophysical background of HFU approach function of porosity and residual water saturation [3]. Morris &
was explained. Flow zone indicator (FZI), a unique parameter for Biggs had also spotted the permeability as a function of
each hydraulic unit, was used to characterize each rock type. The porosity and connate water saturation [4]. Timur and also Coats
number of hydraulic flow units and mean values of FZI for each and Dumanoir developed another models [5, 6].
HFU were calculated from the measured porosity and
permeability. The Cored well’s permeability data is drawn based Generally, in the traditional approach the permeability
on FZI and Hydraulic Flow Units theory. Using this approach, estimation may be achieved via two different methods:
the optimal number of HFUs was found to be equal to five and 1-Simple linear regression and modifications
the average relative error between the core permeability data
and the calculated values was 3.4%. In addition, a correlation In this method a linear relationship between core porosity
coefficient, R2, greater than 0.9 in each HFU was observed. Also, and the logarithm of core permeability is developed as:
values of FZI for un-cored wells calculated using Adaptive
Network Fuzzy inference system (ANFIS) and well log data. (1)
Some core data points were not used to train the (ANFIS) and
Where a, b are the constant coefficients that different values
were used for validation. The set of logs NPHI, RHOB, PHIE and
GR were used as inputs for the ANFIS and the log-scale FZI was
have been reported for them in different models. In these
the output. The calculated FZI values obtained from ANFIS were models permeability estimation for un-cored well will be
fairly consistent with that of core data. Also, the average relative estimated by making a simple regression equation between
error between ANFIS's calculated and the measured well log-porosity and core-porosity or establishment of some
permeabilites was found to be 5.9%. The presented methodology novel techniques such as neural network, fuzzy logic, genetic
was successfully applied to a large data set of laboratory and well algorithm and etc.
logging measurements from the largest Iranian carbonate gas
2-Emperical models
field.
In these methods, the correlations are set among various
Keywords-component; Permeability; porosity; Hydraulic Flow well log results and/or core analysis data such as core porosity,
Unit (HFU); Reservoir Rock Typing (RRT) clay content, grain size, tortuosity, water residua saturation and
etc. Because of depositional characteristics variety in different
I. INTRODUCTION
0B fields, the empirical models are applied locally [7].
Correct estimation of permeability is important in different B. New approaches
6B

sections of reservoir engineering and reservoir simulation.

The important similarity of new developed models is their
Permeability of reservoir rock is affected by different
geological background. In these methods the depth of earth is
parameters such as porosity, kind and distribution of rock
divided into the layers and zones which are lithologically and
materials, mineralogy of rock matrix, residual saturation, and
minerlogically homogeneous. Then petrophysical properties
size of pore throat. The common practice to estimate
may be characterized in each layer. Recently geologists and
permeability is to define a relationship between porosity and
reservoir engineers attempt to correlate the permeability via
permeability. Mainly there are two major approaches to
other rock and fluid properties in these layers [15]. Major
estimate permeability as follows:
 Proceedings of the 3rd (2011) CUTSE International Conference
Miri, Sarawak, Malaysia, 8-9 Nov, 2011 (225 -- 917)

concept of these models is their geological or petrophysical continuous laterally and vertically and has similar flow and
background. These methods are known as reservoir rock typing bedding characteristics. The main mathematical relations of
[RRT]. Hydraulic flow units (HFU) method is one of the most HFU method can be achieved by following equations.
important approaches in this category. In the HFU method, by
defining the FZI (flow zone indicator) parameter, certain Porous media can be represented by a bundle of capillary
numbers of hydraulic units in a reservoir are identified. In each tubes. Combination of Darcy law and Poiseuille law for
unit, petrophysical properties are homogenous. straight cylindrical tubes yields to:

The HFU method was first defined by Bear [8] and then (2)
developed by other investigators [9, 15]. In this paper hydraulic
flow units (HFU) was used as a petrophysical rock typing tool For a realistic porous media Kozeny and Carmen have
in a carbonate reservoir from south pars gas field in Iran. Also, modified Eq 2. by adding a tortuousity factor � as:
by employing the Adaptive Network Fuzzy inference system
(ANFIS) permeability was estimated for un-cored interval. (3)
II. FIELD DESCRIPTION
About 50 percent of gas reserves in Iran have been stored in Where �� is the shape factor, �� is effective porosity and
the South Pars gas field. This huge gas field, which is the ��� is surface area per unit grain volume. Eq. 3 can be written
largest offshore gas field in the world, is located about 100 in field unit as:
kilometers from south shore of Iran in Persian Gulf. This gas
field covers an area of 9700 square kilometers in Iran and Qatar (4)
territorial waters. Gas accumulation of this field is mostly
limited to the Permian–Triassic stratigraphic units that became
prospective during the 1970s following delineation of By using Eq. 4, three parameters are defined as follows:
enormous gas reserves. These units known as the “Kangan– (5)
Dalan Formations” constitute very extensive natural gas
reservoirs in this field and Persian Gulf area, which composed
of carbonate – evaporate series also known as the Khuff Where RQI is reservoir quality index.
Formation [16, 17].
(6)
III. STATEMENT OF THE PROBLEM
A. Concept of HFU. Where ϕz is the pore volume to grain volume ratio or
Several authors have various definitions of flow units, normalized porosity.
which are resultant of the depositional environment and
diagenitic process [9]. Bear defined the hydraulic (pore (7)
geometrical) unit as the representative elementary volume of
the total reservoir rock within which the geological and Where FZI is flow zone indicator. Eq. 4 can be rewritten as:
petrophysical properties of the rock volume are the same [8]. (8)
Ebanks defined hydraulic flow units as a mappable portion of
the reservoir within which the geological and petrophysical Or
properties that affect the flow of fluid are consistent and
predictably different from the properties of other reservoir rock (9)
volume [10]. Hear et al. defined flow unit as a reservoir zone
that is laterally and vertically continuous, and has similar Finally, the relationship between permeability and FZI is:
permeability, porosity, and bedding characteristic [11]. Gunter
et al. defined flow unit as a stratigraphically continuous (10)
interval of similar reservoir process that honors the geologic
framework and maintains the characteristic of the rock type RQI, FZI, and ϕz are used for HFU classification which will
[12]. According to Tiab [13], a hydraulic flow unit is a
be discussed in more details in the next sections.
continuous body over a specific reservoir volume that
practically possesses consistent petrophysical and fluid B. Hydraulic flow units zonation process
properties, which uniquely characterize its static and dynamic Different methods may be applied to use FZI for zonation
communication with the wellbore. of the reservoir and defining flow units. Some of these
Amaefule et al. proposed the hydraulic flow unit concept to techniques are as follow: cluster analysis, probability plots
be used as a principle for subdividing reservoir in different [neural networks, multivariable regression, fuzzy logic, and
rock types reflecting different pore-throat attributes [14]. A multi-linear graphical clustering. Fuzzy logic, neural network,
hydraulic flow unit (HFU) is a representative volume or section nonlinear regression, Ward’s algorithm and etc. These methods
of a reservoir rock. In each HFU, geological and petrophysical are subdivided into three main categories [7]:
properties are different from properties of other sections of the
reservoir. Thus, a flow unit is a reservoir zone that is
 Proceedings of the 3rd (2011) CUTSE International Conference
Miri, Sarawak, Malaysia, 8-9 Nov, 2011 (226 -- 917)

1) Log-Log plot of RQI versus �� . This method is very

simple analysis, but it is clearly not sufficient to distinguish
different rocks’ HFU and estimate their boundaries. All
samples with similar FZI values will lie on a straight line with
slope equal to one. Samples that lie on the same straight line
have similar pore throat attributes and thereby constitute a
hydraulic unit. The mean value of FZI can be determined from
the intercept of the unit slope straight line with �� = � .
2) Histogram analysis. As FZI distribution is a superposition
of multiple log-normal distributions, a histogram of FZI can be
used to define “n” number of normal distributions for “n”
number of HFUs.
3) Probability plot. The probability plot or cumulative Figure 1. Log of Permeability versus Porosity
distribution function is the integral of histogram (probability
TABLE I. DEPTH AND THICKNESS OF OBSERVED ZONES IN WELL SP13
density function). As this plot is smoother than the histogram it
is more useful to define HFUs because identification of clusters Formation Zone Well Name
becomes easier. SP13
Top(m) Thick.(m)
In this paper to determine the number of hydraulic flow Kangan K1 2830.5 105.5
units, three different mentioned ways were compared. Finally K2 2936 47
we used the sum of square errors method in company with a Dalan K3 2983 120
visual method to obtain accurate HFU number determination. K4 3103 175.5

IV. METHODOLOGY AND RESULTS

In this paper, data obtained from conventional core analysis
and Log data in wells SP13 and SP9 were applied. Data of well
SP13 was used for producing the model and validation. Then
the model final test was evaluated by SP9 data banks.
Before any data analysis, all the fractured samples were
removed. To find the fractured samples, all of the samples were
sorted versus depth by plotting their permeability and porosity
in parallel. Each sample with very high permeability and low
porosity was considered as a fractured sample. But according
to exploration contractor team report if the fractured were
induced by the core recovery and handling process the core Figure 2. All layers, Permeability versus Porosity
samples were rejected, and if these were natural open fractures,
there was the risk of getting rid of real data which might have
an important influence on the simulation model or may be a
different rock type that needs to be catered for. So they were
not omitted.
At first, we used all core porosity and core permeability of
each core sample and established the core porosity-core
permeability semi logarithmic cross plot as is shown in Fig. 1.
This figure shows a lot of scatter, which clearly indicates that
porosity alone, is not enough to explain the permeability
variation. A correlation between Log K and ϕ indicates that the
correlation coefficient is equal to 0.31. This poor fitting and
scatter of the data points could be attributed to the change of
lithology (i.e., more than one rock type existed within reservoir Figure 3. Dolomite and Limestone for All Layers
with different fluid flow properties). The combined results of
both macro and microscopic studies indicate that the Upper The cored intervals in these four layers and thickness of
Dalan and Kangan successions can be subdivided into four each zone as observed in the core analysis are summarized in
distinct reservoir units including the K1, K2, K3 and K4 units. Table I. The porosity and permeability of well SP13 were
related to each other in each layer as is shown in Fig. 2. There
are three main types of lithology which forming carbonate
rocks: Limestone, Dolomite, and Anhydrite. Considering the
lithological background of the pars gas field reservoir, the data
was categorized based on the lithology of each sample as is
shown in Fig. 3. Whereas the permeability of the anhydrate is
 Proceedings of the 3rd (2011) CUTSE International Conference
Miri, Sarawak, Malaysia, 8-9 Nov, 2011 (227 -- 917)

supposed to be zero, all anhydrate samples were ignored. After By plotting Log RQI against Log ϕz , in Fig. 4, the optional
very precise study on dolomite and Limestone samples number of HFUs that exists in the reservoir may be determined
permeability-porosity relationships, it has been completely by applying the Iterative Multi-linear Regression Clustering
cleared that lithology can be pondered as an influential Technique. The proposed method is briefly summarized as
parameter in determination of different rock types. In following steps:
summary, depositional system has important role on creating
different types of lithologies. 1. Calculate the RQI values and φz from Equations (5) and (6),
respectively, using the core data given in Fig. 1.
In order to develop a relation between porosity and
permeability and classify the rock types, eight rock types were 2. Plot RQI vs. ϕz in logarithmic space as Fig. 4.
selected based on the lithology (dolomite or limestone) and the 3. Use a reasonable initial guess of the intercept of each
layer of the samples (K1 to K4). These rock types are straight line equation: the mean value of each HFU (or each
summarized in Table II. RTij in this table is the indicator of FZI).
rock types where i shows the layer number (K1=1, K2=2,
K3=3 and K4=4) and j shows the lithology (Dolomite=1, 4. Assign core sample data to the nearest straight line.
Limestone=2). The anhydrate lithology in each section of the 5. Recalculate the intercept of each HFU using least squares
reservoir has been pondered as a rock type with zero regression equations.
permeability.
6. Compare the new and old values of the intercept for every
TABLE II. RESERVOIR ROCK TYPES straight line. If the difference is not within the acceptable
tolerance, updates the intercept values and go to step 4.
Rock Type Minimum Square Relationship between k and �
RT11 Layer K1, Dolomite � = 0.24� 0.72 7. Repeat steps 3 through 6 until the optimal location of each
RT12 Layer K1, Limestone � = 0.148� 0.71 straight line is found in which the error sum of squares is a
RT21 Layer K2, Dolomite � = 0.44�1.6 minimum for the desired number of HFUs.
RT22 Layer K2, Limestone � = 0.017�1.56
RT31 Layer K3, Dolomite � = 0.1� 2 By applying the above procedure to core data in Fig. 1, the
RT32 Layer K3, Limestone � = 0.025�1.8 resulting error of sum squares is plotted against HFUs in Fig. 7.
RT41 Layer K4, Dolomite � = 0.2�1.3
This figure shows that the optimal number of HFUs is equal to
RT42 Layer K4, Limestone � = 0.09�1.1
five. This means that five rock types exist in the studied
The correlation coefficients of these semi-logarithmic
reservoir. The five unit slope lines are plotted with the data
relationships between the permeability and porosity were
from Fig. 1, and are shown in Fig. 4. The values of intercept
clearly improved related to correlation coefficient of scattered
(FZI mea n ) are used to calculated permeability from the
cross plot of log K and ϕ. It means the best correlation between
R RR R

following equation:
porosity and permeability can be obtained if rocks with similar
lithology and fluid flow properties are identified and grouped (11)
together.
A. The relationship between Porosity and Permeability based
9B

The calculated permeability is plotted against the measured

on Hydraulic Flow Unit (HFU) Method ones in Fig. 8. As the average relative error (ARE) between
At the first, three parameters ϕz , RQI and FZI, were core permeability and the calculated ones is equal to 3.4%, this
calculated based on the previous mentioned formulas by cross-plot indicates the accuracy of calculated permeability.
applying well SP13 core data. Then, all data were sorted based (12)
on their FZI parameter value in increasing order. To determine
the number of hydraulic flow units, three different ways were
applied and compared the results. As it is clearly obvious in Fig. 9, this classification method
Graphical clustering using classical RQI− �� plot is the can conduct a very competent numerical classification among
simplest analysis, but it is clearly not sufficient to distinguish the data. The depth of each group was very scattered. The
between different rocks HFU and estimate their boundaries relation between permeability and porosity for each HFU was
(Fig. 4). Another graphical approach is analysis of histogram summarized in Table III. As you see the R2 of each correlation
P P

and probability plots, which are shown in Fig. 5 and 6. clearly indicates the accuracy of HFU approach in permeability
Theoretically, these plots should be able to distinguish correlating with porosity.
individual distributions for each HFU, which are commonly
TABLE III. RESERVOIR ROCK CLASSIFICATION BY HFU METHOD
normal or log-normal type. However, due to overlapping of
data, the individual distributions are not distinguishable on the Layers Minimum Square Relationship between k and �
histogram (Fig. 5). The cumulative probability plot allows HFU 1 � 2 = 0.9168 � = 244768� 2.69
picking out at least 5 HFU and estimate FZI boundaries for HFU 2 � 2 = 0.9761 � = 81529� 2.9879
these rock types (Fig. 6). Finally we used the sum of square HFU 3 � 2 = 0.9252 � = 4488.3� 2.8759
HFU 4 � 2 = 0.9372 � = 905.63� 3.2457
errors method in company with one of three above visual
HFU 5 � 2 = 0.9519 � = 15.52� 2.459
method to obtain accurate HFU number determination as
follow:
 Proceedings of the 3rd (2011) CUTSE International Conference
Miri, Sarawak, Malaysia, 8-9 Nov, 2011 (228 -- 917)

As a first step, the data set was divided into two sets, one
for the network training and the second for training validation.
The validation set was taken from several data points, one from
each of the flow zones. The set of logs NPHI, RHOB, PHIE
and GR were used as inputs for the ANFIS. Each of these
records was normalized, subtracting the mean and dividing by
the standard deviation. The log-scale FZI was output of the
system. FZI determined from ANFIS was matched to FZI
obtained from effective porosity and core permeability with
correlation coefficient (R2) of 0.93. The values of FZI
calculated from core data were correlated with FZI determined
from well log data (at the corresponding depth of core data),
and the evaluation is in agreement with Eq.13:
(13)
Figure 4. Plot of reservoir quality index vs. normalized porosity, Well SP13

By substituting Equation (11) in Equation (13), the

B. Permeability estimation in un-cored intervals/wells with
permeability can be predicted from well log data.
Adaptive Network Fuzzy inference system (ANFIS)
In this new method, a direct mapping between well log data (14)
and core data for a given set of experimental points is
attempted. Adaptive Network Fuzzy inference system (ANFIS)
is an ideal tool for this type of problem. First an appropriate In Fig. 10, the correlation between the core data of
model for well/log data and FZI in a cored well is determined. permeability and the values predicted by the ANFIS for the
Then the method is generalized to un-cored well to obtain FZI selected model was shown. The average relative error between
calculated permeability with ANFIS and measured
from its log data.
permeability was found to be 5.9%. As a final test, the
predicted and measured permeabilities are shown for the cored
interval in Well SP 9. Fig. 11 shows the core permeability as a
function of measured depth, adding the predicted permeability
using Log data by the method of ANFIS.

Figure 5. Histogram of Flow zone indicator, Well SP13.

Figure 7. Determination of the optimal number of HFUs, Well SP13.

Figure 8. Core permeability vs. calculated permeability after final HFU

definition , Well SP13.
FZI
Figure 6. Probability plot of FZI for well SP13.
Proceedings of the 3rd (2011) CUTSE International Conference
Miri, Sarawak, Malaysia, 8-9 Nov, 2011 (229 -- 917)

V. CONCLUSION
The HFU methodology was applied to the South Pars
carbonate depositional reservoir to estimate the absolute
permeability from standard well logs uses data analysis
techniques that have been successful in other areas of science
and engineering, such as neural network and principal-
component analysis. Through the various bases of rock typing
utilizing special core analysis data, such as depth variation
classification in scattered plot, layer and sub-layer
classification, lithology and combination of lithology by layer
classification and Hydraulic Flow Units (HFU) classification,
the HFU technique certainly classified more homogenous areas
in the reservoir than other ones. Flow Zone Indicator (FZI) is
Figure 9. Permeability vs. Porosity, Well SP13. an effective parameter in correlating rock and fluid properties.
It cleared that lithology can be pondered as an influential
parameter in determination of different rock types. Using
Adaptive Network Fuzzy inference system (ANFIS), flow zone
indicator can be determined in un-cored wells.
REFERENCES
1. Kozeny, J.: “Uber Kapillare Leitung des Wassers im Boden,
Stizurgsberichte”, Royal Academy of Science, Vienna, Proc. Class I.
V. 136, 271. 1927.
2. Carmen, P.C.: “Fluid Flow through Granular Beds”, Trans, AIChE.
V. 15, 150. 1937.
3. Tixier, M.P.: “Evaluation of permeability From Electric Log
Resistivity Gradient”, Earth Sci. J., 2, 113. 1949.
4. Morris, R.L. and Biggs, W.P.: “Using Log Derived Values of Water
Saturations and Porosity”, Proceeding of SPWLA, Eighth Annual
Logging Symposium, Denver, Paper X, 1. 1967.
Figure 10. Permeability determined from ANFIS versus Permeability 5. Timur, A.: “An Investigation of Permeability, Porosity, and Residual
measured from core. Water Saturation Relationships”, Proceeding of SPWLA, Ninth
Annual Logging Symposium, New Orleans, Paper K, 1. 1968.
6. Coates, G.R. and J.L. Dumanoir.: “A New Approch to Improved Log
Derived Permeability”, The Log Analyst, 17. 1981.
7. Abbaszadeh., M., Fujii., H., Fujimoto., F.,: “Permeability Prediction
by Hydraulic Flow Units-Theory and Applications”, SPE Formation
Evaluation, December 1996.
8. Bear, J.: “Dynamics of Fluids in Porous Medi”, Elsevier, New York,
1972.
9. Tiab, D. and Donaldson, Petrophysics: “Theory and Practice of
Measuring Reservoir Rock and Fluid Transport Properties”, Gulf
Publishing Co., Houston, Texas, E.C. (1996)
10 Ebanks, W. J. “The Flow Unit Concept-An Integrated Approach to
Reservoir Description for Engineering Projects. Am. Assoc. Geol.
Annual Convention, 1987.
11 Hear, C. L., Ebanks, W. J., Tye, R. S. and Ranganatha, V.
Geological Factors Influencing Reservoir Performance of the Hartzog
Draw Field, Wyoming. J. of Petrol. Tech, Aug. 1984, pp. 1335-1344.
12 Gunter, G. W., Finneran, J. M., Hartman, D. J. and Miller, J. D.
“Early Determination of Reservoir Flow Units Using an Integrated
Petrophysical Method SPE 38679. SSPE Annual Technical
Conference and Exhibition, San Antonio, TX, 5-8 October 1997.
13 Tiab, D. Advances in Petrophysics, Vol. 1-Flow Units. Lecture Notes
Manual, University of Oklahoma, 2000.
14. Amaefule, J.O., Altunbay, M., Tiab, D., Kersey, D.G., and Keelan,
D.K.: “Enhanced Reservoir Description: Using Core and Log Data to
Identify Hydraulic (Flow) Units and Predict Permeability in Uncored
Intervals/Wells”, SPE 26436, presented at 68th Annual. Tech. Conf.
And Exhibit. Houston, Tx, 1993.
15. Svirsky, D., Ryazanov, A. , and Pankov, M.: “Hydraulic Flow Units
Resolve Reservoir Description Challenges in a Siberian Oil Field”,
Figure 11. Permeability determined from ANFIS versus Permeability SPE 87056, presented at the Asia Pacific Conference, Malaysia. 2004.
measured from core, Well SP 9. 16. International Energy Agency (IEA), World Energy Outlook 2008 -
Chapter 12 - Natural gas resources and production prospects, Page 298
17. Aali, J., Rahimpour-Bonab, H., Kamali, M.R., : “Geochemistry and
origin of the world's largest gas field from Persian Gulf, Iran”, Journal