You are on page 1of 8

DOI: 10.1002/ente.

201500010

Improved Method to Identify Hydraulic Flow Units for


Reservoir Characterization
Abouzar Mirzaei-Paiaman,*[a] Hadi Saboorian-Jooybari,[a] and Peyman Pourafshary[b]

Identification of hydraulic flow units (HFUs) is an important This makes the reservoir characterization very complicated
part of reservoir characterization. Rock samples within and sometimes an erroneous process. An improved method,
a given HFU are expected to have the same mean hydraulic referred to as FZI star method (FZI*), is presented here
radius. We show that the famous reservoir quality index-flow using the base form of the KozenyCarmen (KC) equation,
zone indicator (RQI/FZI) technique and its recent modifica- opposed to RQI/FZI method which relies on the generalized
tions do not use the concept of mean hydraulic radius. Each form of the KC equation, by proper consideration of the
predicted HFU by these methods may contain the samples mean hydraulic radius concept. The presented method is
with different pore structures, and further the rocks with sim- verified using a large set of capillary pressure measurements.
ilar structures may be distributed in more than one HFU.

Introduction Existing Models


Identification of hydraulic flow units (HFUs) is a key step in Kozeny[4] and Carmen[5] considered the porous medium to be
reservoir characterization and is usually performed using composed of a bundle of capillary tubes. They applied Pois-
routine core analysis (RCAL) data. Rock samples within soulles and Darcys laws and invoked the concept of the
a given HFU are expected to have the similar pore structure mean hydraulic unit radius[6] to derive a basic relationship
(i.e., the same mean hydraulic radius). The reservoir quality between porosity and permeability, given by Equation (1). In
index-flow zone indicator (RQI/FZI) method[1] has been this study this equation is referred to as the base form of
widely used for study of HFUs. This method and its recent the KC equation.
modifications[2, 3] are derived from the generalized form of
the Kozeny-Carmen (K-C) equation.[4, 5] As will be shown 2
rmh 1
later, the methods that are derived from the generalized k
2t2
form of the K-C equation do not use the concept of mean
hydraulic radius.[6] The consequence would be inaccurate
classification of HFUs, with different extents depending on in which k is the permeability, f is the effective porosity, rmh
the complexity of the rock pore structure. For example, each is the effective or mean hydraulic unit radius (defined as the
HFU identified by these methods may contain samples with ratio of cross sectional area to wetted perimeter), and t is
different pore structures, and further the samples with similar the tortuosity.
pore geometrical attributes may be distributed in more than The relationship between the mean hydraulic radius, rmh,
one HFU. The main contribution of this work is to present and the surface area per unit grain volume, Sgv, and porosity
an improved method for identification of HFUs by proper can be expressed as:[4, 5]
consideration of the mean hydraulic radius concept. We de-
velop a modified FZI method (here referred to as FZI* or
1
FZI star) with a better resolution which ensures that each rmh 2
Sgv 1
HFU only contains the samples with almost identical pore
geometrical characteristics. FZI* is derived using the base
form of the KC equation with consideration of all available Using Equation (2) and substituting for rmh in Equation (1)
pore geometrical attributes that KC equation provides. In gives the generalized form of the KC equation as:
derivation of the FZI* method, we emphasize that the effec-
[a] Dr. A. Mirzaei-Paiaman, H. Saboorian-Jooybari
tive or mean hydraulic radius should be the same within Department of Petroleum Engineering
a HFU but different among distinct HFUs. Furthermore, the National Iranian South Oil Company (NISOC)
FZI expression should contain the tortuosity term. Generally, Ahvaz (Iran)
E-mail: Mirzaei1986@gmail.com
the FZI should only contain pore geometrical parameters
[b] Dr. P. Pourafshary
and should not contain any parameter controlled by wettabil-
Department of Petroleum and Chemical Engineering
ity of system. Sultan Qaboos University
Muscat (Oman)

Energy Technol. 2015, 3, 726 733 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim 726

3 1 Nooruddin and Hossain[2] developed a modified FZI
k 3 model. The relationship between tortuosity, formation resis-
1 2 FS t2 S2gv
tivity factor FR, and porosity can be written as:[9]

Since both the Equation (1) and Equation (2) have been t FRf 7
written for a circular capillary, therefore the shape factor FS
is introduced to account for non-circular capillary tubes. For The FR can be estimated using Archies equation as:[10]
the specific case of circular tube FS is equal to 2.
Amaefule et al.[1] used Equation (3) to develop the well- a
FR 8
known RQI/FZI model as: m

in which a is the lithology factor and m is the cementation



LogRQI Log LogFZI 4 exponent. Inserting Equation (8) in Equation (7) and using
1
the general KC equation gives:

where RQI and FZI, both presented in mm, are defined as:
2m1 1
k 9
1 2 FS a2 S2gv
r
k
RQI 0:0314 5
Nooruddin and Hossain[2] used Equation (9) to write:

in which permeability is in millidarcy and f is unitless.


m
LogRQI Log LogFZINH 10
1
1
FZI p 6
FS tSgv in which FZINH is the modified FZI expression expressed in
mm defined by Nooruddin and Hossain as:[2]
Graphically speaking, based on Equation (4), on a loglog

plot of RQI versus 1 all samples with similar FZI values 1
FZINH p 11
will lie on a straight line with unit slope and constitutes an aSgv FS
individual HFU. In reservoir characterization studies where
macroscopic properties exist in the form of RCAL data the m
Hence, on loglog plot of RQI versus 1, all samples with
numerical value of FZI for each plug can be determined similar FZINH values will lie on a straight line with slope of
q unity. Samples that lie on the same straight line claimed to
1 k
using FZI 0:0314 . have similar pore throat attributes constituting an individual
Although there should exist one single FZI value for each HFU.
HFU, in real field-case practice, a distribution for each FZI Izadi and Ghalambor[3] coupled the Poiseuilles flow equa-
around its true mean is encountered. When multiple HFUs tion and Darcys law, taking into account the irreducible (or
exist, the overall distribution of FZI is a superposition of the connate) water saturation SWC in the porous medium which
individual distributions around their mean FZIs.[7] Identifica- was considered as a bundle of tortuous capillary tubes. They
tion of each mean FZI value or the corresponding HFU re- derived the following:
quires decomposition of the overall FZI distribution into its

constituting elements which can be performed using cluster
analysis techniques like discrete rock type (DRT) and proba- LogRQIIG Log 1 Swc 2 LogFZI 12
1
bility plots. These clustering methods have been discussed in
details elsewhere.[7, 8] in which FZI is defined in Equation (14) and expressed in
Based on the concept of HFUs, each pore geometrical mm. Furthermore, RQIIG is the modified RQI given in mm by
family is expected to have similar mean hydraulic radius Izadi and Ghalambor as:[3]
(rmh). The approach which is used by Amaefule et al.[1] to de-
velop the RQI/FZI model (i.e., substituting for rmh in Equa- r
tion (1) by a combination of Sgv and f and then forming k p
RQIIG 0:0314 1 Swc 13
groups in Equation (3) by differentiating between micro
(e.g., Sgv) and macroscopic (e.g., f) rock properties) means
that systems with similar Sgv but with different f (i.e., differ- in which permeability is in mD. In this case the FZI was de-
ent rmh values) are identified wrongly as a single HFU. This fined as:
flaw in the RQI/FZI model causes inaccurate classification r
of HFUs, to different extents depending on the complexity 0:0314 1 k
FZI 14
of rock structures. 1 Swc 1:5

Energy Technol. 2015, 3, 726 733 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim 727
Similarly, all samples with similar FZI values will lie on FZI* takes into account all the pore related parameters
a straight line with unit slope on a on loglog plot of RQIIG that the KC equation provides. This FZI* model considers

versus 1 1 Swc 2. the effects of the mean hydraulic radius and tortuosity and
Since, the RQI/FZI models proposed by Nooruddin and does not include any additional parameter controlled by the
Hussain[2] and Izadi and Ghalambor[3] use the same general wetting conditions, which improves the performance of the
methodology as Amaefule et al.[1] in derivation of their modi- proposed model.
fied models, the same problem also exists in using these
models. Besides, the FZINH index proposed by Nooruddin
Model verification
and Hussain[2] does not consider the tortuosity properly in
determination of flow units. In fact, substituting the tortuosi- Our developed model was used to identify the HFUs in the
ty in the general KC equation and removing the factors that Bangestan oil reservoir in Ahvaz field which is located in the
control the texture of the rock from the FZI index, not only south west of Iran. This reservoir is mainly composed of
does not enhance the characterization of HFUs, but causes limestone with occasional dolomitized limestone intervals.
less accurate determination of HFUs than the original RQI/ Due to the complex nature of pores system, the more hetero-
FZI model. It should also be noted that Nooruddin and Hus- geneous and anisotropic a reservoir is, the more complicated
sain model is more complicated than the original FZI ap- the identification of the HFUs.[11] A high degree of heteroge-
proach due to the need for the cementation factor (m), neity in this reservoir is indicated by computing the value of
which is usually measured during SCAL experiments and is 0.9 for Lorenz coefficient. The existing HFU identification
unknown from the results of RCAL tests. In addition, the techniques failed to determine the HFUs in such a heteroge-
connate water saturation which is a function of wettability neous reservoir. Extensive sets of RCAL and SCAL data
has adverse effect on the HFU determination approach de- from the reservoir are available which enhances the valida-
veloped by Izadi and Ghalambor.[3] It means that by using tion of the new model. RCAL data of vertical (501 samples)
the Izadi and Ghalambor approach, the rocks with the same and horizontal (1680 samples) plugs from five wells are used.
pore structure but different wettability lie on two different The oilwater capillary pressure (OWCP) and mercury injec-
HFUs. tion capillary pressure (MICP) data of vertical and horizon-
tal plugs are available for checking different HFU identifica-
tion methods.
Development of an improved method to identify Figure 1 and Figure 2 show the HFUs identified by FZI
HFU and FZI* methods based on DRT and probability plot clus-
tering techniques for, respectively, vertical and horizontal
It is clear that the samples of the same HFU must have the plugs. The summary of the analyses shown in these figures is
similar effective or mean hydraulic radiuses. Hence, we sug- presented in Table 1. For each case, the number of clusters
gest using the base form of the KC equation, which includes and in each cluster the number of samples and the corre-
the pure form of the mean hydraulic unit radius to develop sponding range of FZI, FZI*, permeability, and porosity
the model for HFU. By taking the square root followed by values are also included. As summarized in this table, the
a logarithm of both sides of Equation (1), this equation can numbers and the characteristics of HFUs predicted by FZI
be re-written as: and FZI* differ significantly. Furthermore, the HFU classifi-
cation result is severely affected by the choice of the cluster-
p p ing scheme. A critical question arises here that must be an-
r
Log k log Log pmh 15 swered: which of the classifying methods (FZI or FZI*) and
t FS
clustering techniques (DRT or probability plot) lead to the
best classification of the samples into HFUs? The only way
and if presenting the permeability in mD:
to check the reliability of each model is by analyzing the ca-
p pillary pressure of the plugs as a dynamic parameter affect-
p
Log0:0314 k log LogFZI * 16 ing the fluid flow in the reservoir. The rock samples classified
as a HFU should have similar capillary curves as they have
in which, FZI* is the modified FZI in mm defined as: similar dynamic behavior. In other words, the capillary pres-
sure curves for the samples in each predicted HFU are ex-
r pected to stay together and make a clear separation from
FZI* pmh 17
t FS other HFUs.
p Only limited and selected number of plugs undergo SCAL
p
Hence, on the log-log plot of 0.0314 k versus , all sam- tests because these experiments are time consuming and ex-
ples with similar FZI* values will lie on a straight line with pensive. This is the reason that in the preceding capillary
unit slope. In reservoir characterization studies the numerical pressure plots only some of the identified HFUs may be ob-
value of FZI* for each plug can be determined using served. Figure 3 and Figure 4 show the available measured
q OWCP data for, respectively the vertical and horizontal
k
FZI 0:0314 . plugs based on the FZI and FZI* methods by using different

Energy Technol. 2015, 3, 726 733 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim 728
Figure 1. Use of the different HFU identification methods when undergoing different clustering techniques for the vertical plugs. a) FZI and DRT analysis, b)
FZI* and DRT analysis, c) FZI and probability plot analysis, d) FZI* and probability plot analysis.

Figure 2. Use of the different HFU identification methods when undergoing different clustering techniques for the horizontal plugs. a) FZI and DRT analysis, b)
FZI* and DRT analysis, c) FZI and probability plot analysis, d) FZI* and probability plot analysis.

Energy Technol. 2015, 3, 726 733 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim 729
Table 1. The summary of the analyses using both FZI and FZI* methods based on DRT and probability plot techniques for the vertical and horizontal plugs.

Method Direction No. of clusters (DRT Cluster No. of FZI or FZI* Permeability Porosity range No. of clusters (probability Cluster No. of FZI or FZI* Permeability Porosity range
analysis) data range [mm] range [mD] (fraction) plot analysis) data range [mm] range [mD] (fraction)
1 20 0.10.25 0.0050.577 0.0460.201 1 33 0.130.28 0.0052.159 0.0420.25
2 275 0.250.79 0.00623.185 0.0290.314 2 315 0.281.11 0.00626.422 0.0190.314
3 143 0.82.49 0.006109.397 0.0150.271 3 81 1.112.24 0.016109.397 0.0150.271
Vertical 7 4 45 2.557.34 0.009880.2 0.0070.251 5 4 52 2.286.91 0.009880.2 0.0070.251
5 12 9.1522.01 0.0282.166 0.0050.02 5 20 7.2184.24 0.02886.877 0.0040.025
6 5 29.174.17 0.0686.877 0.0040.025

Energy Technol. 2015, 3, 726 733


7 1 84.24 2.297 0.007

FZI 1 5 0.110.22 0.0460.558 0.0920.281 1 24 0.110.3 0.0462.849 0.0810.281


2 702 0.260.79 0.00920.238 0.0280.288 2 743 0.30.86 0.00935.772 0.0280.292
3 543 0.82.48 0.007173.634 0.0140.331 3 415 0.862.05 0.007164.12 0.0140.331
4 231 2.527.87 0.0182068.959 0.0090.294 4 316 2.078.6 0.0182068.959 0.0090.294
Horizontal 9 5 125 7.9624.93 0.022955.539 0.0040.207 6 5 143 8.7141.38 0.022651.92 0.0040.167
6 60 25.277.92 0.094255.315 0.0030.048 6 39 42.96909.69 0.096255.315 0.0020.040
7 11 81.9205.08 0.34675.418 0.0030.02
8 2 262.41393.78 6.354133.548 0.0040.009
9 1 909.69 3.514 0.002

1 17 0.010.02 0.0050.091 0.0190.187 1 41 0.010.04 0.0050.225 0.0070.187


2 179 0.030.08 0.0091.326 0.0050.216 2 81 0.040.06 0.020.577 0.0090.201
Vertical 5 3 232 0.080.25 0.0614.115 0.0040.272 5 3 273 0.060.18 0.0287.041 0.0040.272
4 69 0.250.69 0.729109.397 0.0070.314 4 80 0.180.4 0.56830.354 0.0070.314
5 4 0.81.87 86.877880.200 0.0250.251 5 26 0.410.84 2.231173.541 0.0070.271

FZI* 1 11 0.010.02 0.0070.051 0.0140.092 1 40 0.010.03 0.0070.118 0.0140.139


2 351 0.020.08 0.011.209 0.0050.281 2 215 0.030.06 0.0180.75 0.0050.281
3 859 0.080.27 0.02620.238 0.0030.288 3 672 0.060.16 0.0266.498 0.0040.284
Horizontal 6 4 394 0.250.79 0.346164.12 0.0030.331 6 4 462 0.160.34 0.09628.278 0.0030.288
5 60 0.802.48 1.234955.539 0.0020.285 5 227 0.350.79 0.491164.12 0.0040.331
6 5 2.493.74 36.053 0.0090.294 6 64 0.793.74 3.5142068.959 0.0020.294
2068.959

2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim


technique.

the DRT technique.

because of existence of microfrac-


tests must be done on unfractured
vertical and horizontal plugs

is possible that the deviations


plifications made by this equation
the FZI* method the clustering

tures in the samples. Fourth, the


the observed deviations might be
little overlapping is observed be-
the case of using the FZI*

samples. In other words, some of


laboratory calculations. Third, the
perimental measurements and
model. The first and the most im-
unexpected deviations from the
might be several reasons for such
better classification of HFUs than
tering Scheme gives a significantly
method the probability plot clus-
method. When applying the FZI*
cation of HFUs than the FZI
cases of clustering, the FZI*
ods undergoing DRT and proba-
MICP data for, respectively the
Figure 5 and Figure 6 show the
cation of HFUs than the DRT
probability plot analysis, the FZI*
clustering schemes. In both cases

tion of HFUs than the FZI


method yields a better classifica-

lar when studying heterogeneous


portant one is that the new
curves of different HFUs. There
tween the capillary pressure
of clustering using either DRT or

samples were saturated and aged


orders. On the other hand, some
method, it is observed that some
analyses shown in Figures 3 to 6,
Concerning the aforementioned
gives a significantly better classifi-
using probability plot analysis

730
have been caused by errors in ex-
carbonated rocks.[1215] Second, it
curacy in the analysis, in particu-
may be the main source of inac-
method presented here is based
on the KC equation. Some sim-
curves do not obey the expected
it should be noted that even in
method provides a better classifi-
bility plot clustering schemes.
Likewise the OWCP data, in both
based on the FZI and FZI* meth-
method. In addition, when using
Figure 3. The OWCP data for vertical plugs based on different methods. a) FZI and DRT analysis, b) FZI* and DRT analysis, c) FZI and probability plot analysis,
d) FZI* and probability plot analysis.

Figure 4. The OWCP data for horizontal plugs based on different methods. a) FZI and DRT analysis, b) FZI* and DRT analysis, c) FZI and probability plot analy-
sis, d) FZI* and probability plot analysis.

Energy Technol. 2015, 3, 726 733 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim 731
Figure 5. The MICP data for vertical plugs based on different methods. a) FZI and DRT analysis, b) FZI* and DRT analysis, c) FZI and probability plot analysis,
d) FZI* and probability plot analysis.

Figure 6. The MICP data for horizontal plugs based on different methods. a) FZI and DRT analysis, b) FZI* and DRT analysis, c) FZI and probability plot analy-
sis, d) FZI* and probability plot analysis.

Energy Technol. 2015, 3, 726 733 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim 732
with the formation water for about two weeks before the In other hand, the classical equations presented in this
measurement of the OWCP curves in order to simulate ini- work may also be modified by introducing the fractal theory
tial wettability state of the reservoir before migration of the to obtain novel results. For example, Cai and Yu[15] modified
oil. As described above, not giving enough time for the aging the classical LucasWashburn (LW) equation by introducing
can be a source of error for validation of the new model with the fractal dimension for tortuous capillaries representing
the OWCP data. Fifth, since the MICP tests were performed the heterogeneity of flow in porous media. The modified LW
under very high injection pressures to fill almost all of pores equation may be called LW-CAI equation.
of the samples, loss of the original texture is a possibility.
Thus, seeing some deviations from the expected trend, partic-
ularly at high mercury saturations, can be attributed to the Acknowledgements
destruction of the original pore structures at very high injec-
tion pressures. The first and second authors thank National Iranian South
Oil Company (NISOC) and National Iranian Oil Company
(NIOC) for permission to publish this work.
Conclusions
In this work, a modified FZI method was presented using
the base form of the KC equation. The new model was veri- Keywords: capillary pressure FZI star hydraulic flow
fied using abundant capillary pressure data from a carbonated unit kozenycarmen equation mean hydraulic radius
reservoir. The following conclusions can be drawn from this
work: [1] J. O. Amaefule, M. Altunbay, D. Tiab, D. G. Kersey, D. K. Keelan,
* The base form of the KC equation, which includes the Enhanced reservoir description using core and log data to identify hy-
mean hydraulic radius term can be used to develop a more draulic flow units and predict permeability in uncored intervals/wells
reliable method for identification of HFUs. SPE Annual Technical Conference and Exhibition, 3 6 October,
1993, Houston, Texas.
* The literature methods of identifying HFUs developed
[2] H. Nooruddinn, M. Hossain, J. Pet. Sci. Eng. 2011, 80, 107 115.
using the general form of the KC equation do not use the [3] M. Izadi, A. Ghalambor, SPE Reservoir Eval. Eng. 2013, 16, 257
concept of mean hydraulic radius in a consistent way. Each 264.
HFU identified by the existing methods may contain the [4] J. Kozeny, ber kapillare Leitung des Wassers im Boden, Sitzungsber-
ichte. Royal Academy of Science Vienna, Proc. Class I, 1927, 136,
samples with different pore structures, and further the sam- 271 306.
ples with similar pore geometrical attributes may be distrib- [5] P. C. Carmen, Trans. AIChE 1937, 15, 150 166.
uted in more than one HFU. [6] R. B. Bird, W. E. Steware, E. N. Lightfoot, Transport Phenomena,
* Identification of HFUs is significantly impacted by the Wiley, New York, 1960.
[7] M. Abbaszadeh, H. Fujii, F. Fujimoto, SPE Formation Evaluation
choice of different clustering methods. The probability plot 1996, 11, 263 271.
clustering technique gives better results than the famous [8] M. S. El Sharawy, J. Appl. Sci. Res. 2013, 9, 4271 4287.
DRT method. [9] M. R. J. Wyllie, W. D. Rose, J. Pet. Technol. 1950, 2, 105 118.
* Employing the FZI* method along with the probability [10] G. E. Archie, Trans. AIME 1942, 146, 54 62.
[11] H. Saboorian-Jooybari, G. H. Mowazi, S. R. Jaberi, A new approach
plot clustering technique leads to much more satisfactory for rock typing used in one of the Iranian carbonate reservoir (a case
identification of HFUs. study). Paper SPE 131915 presented at the International Oil and Gas
Conference and Exhibition, Beijing, China, 8 10 June 2010. DOI:
10.2118/131915-MS.
Future works [12] J. Cai, B. Yu, M. Zou, L. Luo, Energy Fuels 2010, 24, 1860 1867.
[13] J. Cai, E. Perfect, C. L. Cheng, X. Hu, Langmuir 2014, 30, 5142
In this work, we used only two clustering techniques of DRT 5151.
and probability plot analysis. Other available clustering tech- [14] R. Masoodi, E. Languri, A. Ostadhossein, J. Colloid Interface Sci.
niques may also be used. In particular, artificial intelligence 2013, 389, 268 272.
[15] J. Cai, B. Yu, Transp. Porous Media 2011, 89, 251 263.
techniques can be considered as efficient clustering methods.
The new rock typing method presented in this study can fur-
Received: January 11, 2015
ther be verified using more SCAL data from other reser- Revised: February 13, 2015
voirs. Published online on April 15, 2015

Energy Technol. 2015, 3, 726 733 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim 733

You might also like