Sie sind auf Seite 1von 8

SPE 98828

Permeability Prediction Using Genetic Unit Averages of Flow Zone Indicators (FZIs) and
Neural Networks.
C.I. Uguru, U.O. Onyeagoro, J. Lin, J. Okkerman and I.O. Sikiru,
Shell Petroleum Development Company of Nigeria Limited

the very good quality sands and to overestimate the


Copyright 2005, Society of Petroleum Engineers Inc.
permeability in the low quality sands. This observation was
This paper was prepared for presentation at the 29th Annual SPE International Technical
Conference and Exhibition in Abuja, Nigeria, August 1-3, 2005. recorded lately for the cored intervals of several wells one of
This paper was selected for presentation by an SPE Program Committee following review of
which is highlighted in Figure 1.
information contained in an abstract submitted by the author(s). Contents of the paper, as
presented, have not been reviewed by the Society of Petroleum Engineers and are subject to
correction by the author(s). The material, as presented, does not necessarily reflect any Figure 1
position of the Society of Petroleum Engineers, its officers, or members.
Plot of calculated permeability (modified Kozeny-Carman)
versus Core Permeability

Abstract 100000

A significant improvement in the prediction of permeability


10000
from wireline logs has been achieved for the Niger Delta Overestimated
reservoirs using the concept of Flow Zone Indicators 1000 Zone
(developed by Amaefule et al) based on genetic unit
(Modified kozeny Carman)

classification as well as the application of neural networks. 100


Calculated Permeability

Core
Permeability
This concept is building on earlier reservoir description work 10
whereby reservoirs in the Niger Delta have been classified Underestimated
(mD)

according to the environment of deposition and by the 1 Zone


lithofacies associations (i.e. genetic units) in these
depositional environments. Using core data, averages of the 0.1
0.1 1 10 100 1000 10000 100000
FZIs have been computed for each genetic unit. FZI values Core Permeability (mD)
have then been assigned to reservoir intervals without core
data whose respective genetic units have been identified from
log data using for example neural networks. The permeability
values in the uncored reservoir intervals have thus been In full field dynamic simulations reservoirs whose
estimated using the permeability-FZI-porosity relationship for permeability were estimated using the said model could not
identified genetic unit in those wells or through multiple non- achieve acceptable history matches without significant
linear regression using neural networks. Clear benefits of this permeability multipliers being applied to the predicted. The
improved estimation of permeability have been achieved in the multiplying factor has in some cases been as high as 10. This
process of history matching well behaviour in full field observation, with its attendant serious business impact, has
reservoir simulation models. lead to the initiation of this study.

Introduction This paper seeks to highlight the very significant


A successful reservoir model requires that the reservoir be improvements to permeability modelling through the
properly characterised in terms of the lithology, porosity, application of the concept of Flow Zone Indicators (FZIs).
permeability and hydrocarbon saturation. Knowledge of the Two complementary techniques that have been used to
permeability distribution is critical to field development facilitate the application of FZIs in permeability modelling are
planning. Recent experience in SPDC has shown that there is a also discussed. These are:
disparity between the calculated permeability and that (i) the reservoir classification technique and
measured on cores. Up to 1996 the most widely applied (ii) multiple regression using neural networks.
technique for permeability modelling in SPDC was a modified
form of the Kozeny-Carman equation (Ref.: 1). This model The improvements to permeability prediction have been
has however been observed to underestimate permeability in recorded in the following areas:
2 Permeability Prediction Using Genetic Unit Averages of Flow Zone Indicators (FZIs) and Neural Networks.] SPE 98828

(i) matching of core measurements, 4.4


(ii) matching of welltest results and k = 0.136 2
(03)
(iii) history matching in full field simulations. Swi

Literature Review For this equation the porosity and the irreducible water
Most permeability models are based on the perception that saturation are expressed as percentages (%).
there is indeed a relationship between the porosity and the
permeability. The generalised expression is always such that Limitations of the Wyllie and Rose Permeability models
permeability is made proportional to space within the reservoir The major problem experienced in the application of the
matrix that is available for fluid occupation (usually a power Wyllie and Rose-type expressions is the lack of a proper
of the porosity) and inversely proportional to space not technique for the estimation of the irreducible water saturation
available to fluids (normally a power of the surface to volume across a reservoir zone of interest. In the past the irreducible
ratio, irreducible water saturation or mineral volume fractions water saturation has been approximated as the water saturation
present in the rock of interest). Some of these techniques of at the topmost part of the reservoir. This value of irreducible
estimating permeability are highlighted in the paragraphs water was then used as a constant in either the Morris & Biggs
following. equation or the Timur equation for the entire reservoir. One of
the drawbacks of this is that a constant value for the
Wyllie and Rose Permeability Models irreducible water saturation tends to indicate that the reservoir
The Wyllie and Rose concept is that permeability can be is homogeneous which is not always the case. The water
expressed by the empirical relationship between the saturation at the topmost part of the reservoir may also be
permeability on the one hand and the porosity and irreducible much higher than the irreducible water saturation that actually
water on the other hand given by the expression: characterises the reservoir. This will be the case if the top of
the reservoir is not sufficiently separated from the free water
C x level. In such a case the permeability will always be
k= (01) underestimated. Furthermore, with a constant value for the
Swiy irreducible water saturation, the permeability becomes a
function of porosity only. Thus the effects of variations in pore
where, size distribution, grain size, textures and variations in capillary
k is the permeability behaviour as a result of diverse lithofacies or genetic units
C is a constant of proportionality encountered in a reservoir are not accounted for. This problem
is the porosity can be taken care of if capillary curves for the reservoir are
available.
Swi is the irreducible water saturation
x and y are just powers of porosity and irreducible water On the whole, the Wyllie and Rose type of permeability
saturation respectively. expression, which is a function of the porosity and irreducible
water saturation, should not be applied in reservoir studies
There are two major forms of the above expression namely the unless a reliable means of estimating irreducible water is
Morris and Biggs equation and the Timur equation. these are developed. The results from applying this technique will be
specified below. greatly improved if irreducible water saturation is obtained for
each facie/genetic unit in the reservoirs of interest.
Modified Kozeny-Carman (minerals fraction model)
Morris and Biggs model By considering the reservoir to be composed of a bundle of
For the Morris and Biggs equation, the permeability is given capillary tubes Kozeny and Carman (Ref.:2) derived the
by the expression following: following expression for permeability
C 3
k 1/ 2 = (02)
Swi 1 3
k = (04)
where, F S (1 )2
C = 250 for reservoir zones with gases, while s gv
C = 80 for reservoir zones with oil.
The porosity and the irreducible water saturation are measured where,
in decimal fractions. = porosity
Fs = pore throat shape factor
Timur model = tortuosity
The expression for the Timur equation is:
S gv = surface area to grain volume ratio.
3 Permeability Prediction Using Genetic Unit Averages of Flow Zone Indicators (FZIs) and Neural Networks.] SPE 98828

Determining all the various constituents of this expression is Figure 2


near impossible from logs which makes the application of the Plot of Permeability-Porosity relationship for
Kozeny-Carman equation as it is given above to be limited in a 100% Quartz rock using the modified Kozeny-Carman model
scope for reservoir studies. A modified form of the Kozeny-
Carman relation was thus developed within SPDC. In this
model (which was developed for the Niger Delta) the natural 100,000.00

logarithm of the permeability was considered to be linearly 10,000.00

perturbed by the volume of minerals present in the reservoir


rock. The modified Kozeny-Carman equation was thus given 1,000.00

Permeability (mD)
as:
100.00

Log e ( k ) = A + 3 * Log e 2 * Log e ( 1 ) + BiVi Calculated Perm


(Modified Kozeny
Carmann for a 100%
10.00

(05) Quartz system)

Core Data (perm. Vs


porosity)
1.00

where,
k = permeability in mD 0.10
0.00 0.10 0.20 0.30 0.40 0.50

A is a constant Porosity (Fractions)

is the total porosity in fractions


Vi = volume fraction of the ith mineral in fractions.
Bi = constant for the various minerals. For the various
minerals it has the following values: The Flow Zone Indicator Concept
B(quartz) = 3.324395 Amaefule et al (Ref.: 3 and 4) recognised that the pore
B(silt) = -3.32003590 geometry and pore throat size distribution and not only the
B(dispersed and structural shales) = -13.44 porosity control the hydraulic quality of a rock. These
attributes, pore geometry and pore throat size distribution are
The mineral volume fractions for quartz and silt as well as the functions of mineralogy (i.e. type, abundance, morphology
volume fractions for the volume fractions for the dispersed and location relative to pore throats) and texture (i.e. grain
and structural and laminated shales are usually determined size, grain shape, sorting and packing). It was also observed
from statistcal evaluation of log responses. that various permutations of the geological attributes indicate
that there are distinct rock units with similar pore throat
Limitations of the Modified Kozeny-Carman attributes. Amaefule et al then postulated that the ratio of
It has been observed that using the modified Kozeny-Carman permeability to effective porosity was a unique index of the
relation underestimates the actual permeability in the high hydraulic quality of a reservoir rock with respect to the pore
permeability reservoir zones. This observation has been made throat size distribution (which is the key determinant of
from the comparison of computed permeability values with hydraulic behaviour). Thus, by modifying the Kozeny-Carman
those measured on cores taken from Niger Delta reservoirs. equation the reservoir quality index (RQI) was defined by the
Figure 2 shows the computed values of permeability (modified given expression:
Kozeny-Carman) for the porosity range 0 to 0.5 (in fractions)
and for very clean sands (i.e. only sandstone). From the given
k
figure it can be seen that the permeability in a clean sand at 35 RQI = 0.0314 (06)
percent porosity will have a maximum permeability value of
9000mD according to the modified Kozeny-Carman equation.
The core measurements of permeability for this value of or equivalently,
porosity could be as high as (and in some cases be above)
20,000mD. With the uncertainty in the permeability estimated
1
with the modified Kozeny-Carman at 40%, an allowance for RQI = 0.0314 * (07)
maximum random error would mean that the permeability at FsS gv 1
35 percent porosity could vary from 5000mD to 13000mD
which is quite below that estimated from cores at even lower
From the foregoing, RQI can be estimated through
porosity values.
measurements on reservoir rock samples. Amaefule et al then
defined a term now known as the flow zone indicator (FZI)
which is given by

0.0314
FZI = (08)
FS S gv
4 Permeability Prediction Using Genetic Unit Averages of Flow Zone Indicators (FZIs) and Neural Networks.] SPE 98828

The FZI can be rearranged in terms of the measurable RQI as Based on the above criteria, the permeability models that
given below involved the determination of irreducible water saturation,
dominant grain size and sorting factors have been classified as
RQI being less practical in application than the FZI based method.
FZI = (09) The FZI based approach requires just the determination of the
Z core measurements of permeability and porosity as primary
data for a particular reservoir. This method (i.e. the FZI based
where model) was therefore chosen as the basis for developing a
suitable permeability modelling technique for SPDC. The
main challenge was to device a means for the lateral
Z = is the pore volume to grain volume ratio. application of FZI values derived from core data to reservoir
1
intervals that have no core data. Two approaches were
considered, namely:
This FZI is intrinsic to a reservoir and is such that rocks or
(i) Facies' or Genetic unit averages of FZI values.
associations of rocks with identical FZI values are expected to
(ii) Multiple non-linear regression of core derived FZI values.
have identical hydraulic behaviours. According to Amaefule et
Both Techniques have been investigated and the results are
al, the FZI is a unique parameter that incorporates the
presented in the chapter following.
attributes of texture and mineralogy in the discrimination of
distinct pore geometries and facies. In terms of the FZI the
Data Analysis
permeability can be written as
The objective for the data analysis was to obtain a basis for the
lateral application of the results of core analysis to other parts
3 of the reservoir where no core data have been taken. Two
k = 1014( FZI ) 2 (10)
items of information were identified as having the potential for
(1 ) 2
providing the link required for the lateral application of core
derived measurements. These are:
Thus a knowledge of the FZI will serve as a basis for the (i) The Facies or Genetic units identified on cores.
proper estimation of the permeability in any reservoir interval. (ii) The petrophysical log responses across the cored interval
There are available techniques for estimating the FZI in any of a reservoir.
logged point in a reservoir. The permeability-porosity
relationship for different FZI values is given in Figure 3. The genetic units present in a cored reservoir interval are
usually specified in the routine geological description of cores.
Figure 3 Genetic units can also be described from petrophysical log
Plot of Permeability versus porosity for responses. Thus the first analysis conducted was to check if
different FZI values. genetic units had predictably distinct characterising FZI
values. The second analysis was to examine the possibility of
100000 predicting FZI values directly from log responses. Both sets of
10000
investigation yielded positive results, which are highlighted
below.
Permeability (mD)

FZI = 1
1000 FZI = 2
FZI=3
100 FZI=5
FZI=7 Genetic Unit Averages of FZIs
FZI=10
10
ZFI=15
Genetic reservoir units (simply lithofacie associations) are the
1 FZI=20 result of a practical subdivision of a reservoir into components
FZI=25
which have a consistent range of reservoir properties, a
FZI=30
0.1
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 consistent external geometry, and a set of log responses
Porosity (fractions) (electrofacies) by which they can be consistently recognised.
Dividing reservoirs into genetic units provides the link that
ensures that the reservoir property data measured from core
Summary of Permeability Models' Review are properly incorporated into the volume cells (voxels) used
The criteria for the choice of a permeability modelling in reservoir modelling.
technique have been based on the following:
(i) The availability of a method for accurately determining the Core data available to the authors have been identified as
constituent parameters of the model. being sourced from three distinct geological environments of
(ii) The ease/precision of determining the constituent deposition that have been identified as being present in the
parameters of the model. Niger Delta. These are the Continental/Fluvial,
(iii) The ease of applying the model to reservoirs of interest. Tidal/Estuarine and the Coastal/Distributary depositional
5 Permeability Prediction Using Genetic Unit Averages of Flow Zone Indicators (FZIs) and Neural Networks.] SPE 98828

environments. The deltaic deposits of the Tertiary Niger delta


Permeability versus Porosity for
are a heterogeneous succession comprising major sandstone Coastal/Distributary Depositional Environment
bodies alternating with claystones and thinly interbedded
sandstones. 28 Genetic Reservoir Units (including shales) genetic
units/FZIs
have been identified within these depositional environments 100000.000

Foreshore
based on available core data, log responses, and depositional 10000.000
Distributary Channel

environments for the Tertiary Niger Delta reservoirs. Upper Shoreface

Core Permeability (mD)


Mouth Bar
1000.000 Transgressive Lag
Transgressive Upper Shoreface
Cross plots of permeability versus porosity were made for the 100.000
Middle Shoreface
Proximal Lower Shore Face
genetic units in each environment of deposition. These cross Distal Lower Shoreface

plots showed distinctly that the respective genetic units had 10.000
FZI = 1
FZI = 2

characteristic FZI ranges. The cross plots are presented in 1.000


FZI=3
FZI=5
Figure 4, 5 and 6 for the Continental/Fluvial, Tidal/Estuarine FZI=7
FZI=10
and Coastal/Distributary environments of deposition .100
0.000 0.050 0.100 0.150 0.200 0.250 0.300 0.350 0.400 FZI=20

respectively. Core Porosity (Fractions)


FZI=30

Figure 4 A summary of the FZIs for the 28 genetic units are given in
the tables following.
Permeability versus Porosity for
Continental/Fluvial Depositional Environment Table 1:
genetic
FZIs for Continental/Fluvial Depositional Environment
100000
units/FZIs
Genetic Reservoir Unit FZI FZI FZI(ma
10000
(min) (Mean) x)
F lu v ia l L a g
Fluvial Lag 19.52 24.22 28.92
C o r e P e r m e a b il it y (

F lu v ia l C h a n n e l
P o in t B a r
1000
P o in t B a r H e te r o lith ic s
F ZI = 1
Fluvial Channel Sandstone* 13.84 18.04 22.24
F ZI = 2
100
F Z I= 3 Fluvial Channel Top ID
F Z I= 5
10
F Z I= 7 Heterolithics**
F Z I= 1 0
F Z I= 2 0
F Z I= 3 0
Crevasse Splay ID
1

Point Bar Sandstone 8.51 11.59 14.67


0 .1
0 0 .0 5 0 .1 0 .1 5 0 .2 0 .2 5 0 .3 0 .3 5 0 .4 Point Bar Heterolithics*** 1.44 2.72 4.00
C o r e P o r o s ity (fr a c tio n s )
Coastal Plain Sandstone*** ID
Coastal Plain Heterolithics ID
Mud Shale 0.10
ID implies insufficient data to
Figure 5 compute a representative
P e rm e ab ility ve rs u s P o ro s ity fo r average
T id a l/E s tu arin e D ep o s itio n a l E n viro n m e n t
g en etic

1000 00
u n its/F ZIs

Estu a rin e Ch a n n e l L a g
Table 2:
Estu a rin e Ch a n n e l S a n d sto n e FZIs for Tidal/Estuarine Depositional Environment
S a n d F la t

Genetic Reservoir Unit FZI FZI FZI(ma


1000 0
C o re P e rm e a b ility (m

T id a l Ba r

(min) (Mean) x)
Cre va sse Ch a n n e l S a n d sto n e
1000
T id a l Ch a n n e l

Estuarine Channel Lag 19.18 23.75 28.32


T ra n sg re ssive S a n d sto n e
100 T id a l F la t

Estu a rin e Ch a n n e l He te ro lith ics

10 F ZI = 1 Estuarine Channel Sandstone 11.79 14.27 16.75


F ZI = 2

1
F ZI= 3

F ZI= 5
Estuarine Channel Top 2.52 4.55 6.58
0.1
F ZI= 10 Heterolithics
F ZI= 20
0 0.05 0.1 0.15 0.2 0.25

C o re P o ro s ity (fra c tio n s )


0.3 0.35 0.4
F ZI= 30 Tidal Bar 9.29 11.06 12.83
Tidal Channel Sandstone 6.99 9.68 12.37
Tidal Flat 1.52 2.20 2.88
Figure 6 Crevasse Channel Sandstone 4.70 6.73 8.76
Sand Flat 10.05 13.46 16.87
Transgressive Sand 2.72 3.76 4.80
Marine Shale 0.10
6 Permeability Prediction Using Genetic Unit Averages of Flow Zone Indicators (FZIs) and Neural Networks.] SPE 98828

dependence on log responses that indicate permeable zones


(e.g. gamma ray and SP logs), porosity (e.g. density and
Table 2: neutron logs), permeability (e.g. shallow and deep resistivity
FZIs for Coastal/Distributary Depositional Environment logs). The proposed equation has the following form:
Genetic Reservoir Unit FZI FZI FZI(ma
(min) (Mean) x) FZI = A0 GR A1 SP A 2 FDC A3 CNLA4 RES A5 ....
Foreshore 14.57 19.35 24.13
Distributary Channel Sandstone 6.04 8.88 11.72 where GR, SP, FDC, CNL, RES are the respective log
Upper Shoreface 6.02 7.54 9.06 responses for gamma ray, spontaneous potential, density,
Mouth Bar 3.40 4.69 5.98 neutron and resistivity logs while A0 , A1, A2 etc are
Transgressive Lag 2.53 4.24 5.95 constants that are to be determined through the data obtained
Transgressive Shoreface 3.68 4.28 4.88 in the cored interval. When these constants highlighted above
Middle Shoreface 3.01 4.23 5.45 are determined, the FZI may then be estimated in other parts
of the reservoir using the respective log responses. With the
Proximal Lower Shoreface 1.39 2.10 2.81
FZI determined as a continuously varying parameter along
Distal Lower Shoreface 0.68 1.24 1.80 hole in any well, the permeability can then be estimated using
Marine Shale 0.10 the expression developed by Amaefule et al. That is
NDGN = Niger Delta Genetic Number
FZI = Flow Zone Indicator The tedium associated with determining the constants has
been completely by-passed by the training and application of a
The maximum and minimum values for the respevtive FZIs neural network in regression mode. Log responses have been
have been taken as one standard deviation about the mean. trained successfully to identify FZIs derived from cores from
With these tables, the identification of the genetic units in the reservoirs in the Niger Delta. These networks have been used
reservoir enables the assigning of FZI ranges to the respective to predict permeability in reservoir intervals other than those
genetic units. This in turn enables the prediction of minimum, from which the cores were taken. In Figure 8 the logs and core
mean and maximum values for the permeability at each point permeability data (from the Niger Delta) used to train a neural
in the reservoir. A plot of permeability computed using network are presented.
genetic unit averages of FZIs against permeability on cores
taken from resrvoirs in the Niger Delta is presented in Figure Figure 8
7. Display of logs and core permeability
data that were used to to train a
Figure 7 regression neural network
Density Log
* permeability
Calculated Permeability (using Genetic Unit Gam ma Ray Log measured on core
Neutron Log
averages of FZIs) vs Core Permeability
100000

10000
Permeability
Calculated

1000

100

10

1
1 10 100 1000 10000 100000

Core Permeability

The plot indicates that there is good agreement between


computed and measured values.

Estimation of FZIs via Multiple Linear Regression (directly


from log responses)
Ambastha and Moynihan (Ref.: 5) proposed a technique for
using the results of estimated FZI in cored intervals to predict
the FZI in uncored intervals using multiple regression. By this
technique the FZI was expressed as having a functional
7 Permeability Prediction Using Genetic Unit Averages of Flow Zone Indicators (FZIs) and Neural Networks.] SPE 98828

The FZIs derived from the said neural network were used to reservoir unit with respect to its genetic unit. Figure 10 which
predict permeability values in other reservoir intervals with is a plot of FZIs against stress highlights this phenomenon that
core data for comparison with measured data. A comparison FZIs are stable under stress unlike permeability and porosity.
of the permeability derived from the application of FZIs from
the neural network with measured core permeability is shown Figure 10
in Figure 9.
A plot of FZIs versus stress for different core plugs

Figure 9 30

A comparison of horizontal permeability measured on core with Each line represents


25 the measurements on
that from nuclear magnetic tools and that predicted using the
a unique core plug
FZI technique in conjunction with neural networks
20

FZI Values
Permeability Permeability Permeability
predicted from predicted from predicted the
nuclear nuclear FZI technique
magnetic magnetic and neural 15
tools: using the tools: using the networks
Timur formula inter echo times

10

0
0 1000 2000 3000 4000 5000 6000 7000

Stress (psi)

Permeability
measured on core

Application of Flow Zone Indicators in Permeability


modelling.
Any of the two methods described above namely, reservoir
classification into genetic units or multiple non linear
In all cases as also shown in Figure 9 there was excellent regression may be employed in the lateral application of core
agreement between computed and measured values of derived FZIs.
permeability. For the particular example highlighted in Figure
9, permeability predicted by the application of FZIs derived The impact of FZI based permeability models on dynamic
from the neural network were better than those predicted by reservoir simulation.
nuclear magnetic resonance logging tools. The greatest benefit of the techinque being presented for
permeability modelling has been in the numerical simulation
The good agreement between computed permeability values of reservoirs behaviour and performance. Before now the
and core measurements no doubt increases the value of all problems encountered in the history matching process of
available core data as they could be used as analogues for dynamic reservoir models for reservoirs whose permeability
reservoirs without core data. This method is easier and quicker had been modelled using other methods were as follows:
as it does not need a detailed geological description of the
reservoir as a prerequisite for permeability modelling. To (1) Well inflow: There was usually an underestimation of the
minimise the uncertainty due neural network error any trained productivity of reservoirs in historically good producing
network should be tested in a reservoir interval with core data. sands due to low values of estimated permeability.
Better results are generally obtained if the core data used to (2) Production history: There were mismatches in the
train a neural network is taken from a reservoir with a similar watercut and GOR profiles due to permeability
geological environment of deposition as the target reservoir. distributions that bore no resemblance to measured values
or any geological depositional pattern.
Uncertainty in permeability derived from FZIs (3) Reservoir Performance: there were often incidents of
The uncertainty in the determination of permeability on cores wrong flow directions during the history match process.
is 25%. The total relative error in estimating permeability (4) It was virtually impossible to predict or match well test
using FZIs is 27% which makes it more attractive than the results.
modified Kozeny-Carman which has an estimated random (5) Low Permeability: In order to obtain a reasonable history
error of at least 50%. The core permeability and porosity data match it was usually necessary to apply arbitrary
used in this study were measured at 400psia. This is because permeability multipliers which literarily reduced reservoir
only routine core analysis was performed on most of the data classification to guess work.
sets. Nevertheless, the impact of stress on FZIs is minimal and
certainly not sufficient to change the classification of a
8 Permeability Prediction Using Genetic Unit Averages of Flow Zone Indicators (FZIs) and Neural Networks.] SPE 98828

The problems highlighted above have been significantly


reduced by the application of FZIs in permeability modelling. A very useful dimension of the scope of applicability of the
The time for the history match process has been reduced a FZI technique is that core data can be used as analogues for
great deal even for large and complicated reservoirs. The reservoirs other than the parent reservoir (i.e. the reservoirs
application of neural networks has also reduced the time from which the cores were taken). This utility is facilitated by
required to characterise a reservoir adequately to hours instead the availability of neural network software. The ability to
of weeks no matter the size of the reservoir. reproduce core measurements of permeability indicates that
the application of FZIs is a viable alternative to coring in low
Conclusion budget operation scenarios.
Of the permeability modelling techniques reviewed, the
technique based on FZIs is recommended for application. It is This inter reservoir application of analogues is independent of
easy to apply and the data required is easily available. Since depth up to 12000ft. No data have been tested below this
the FZI is a number that accounts for the cumulative effects of depth. Nevertheless the effects of diagenesis might require that
pore throat sizes and distribution, texture (i.e. grain size, grain a different set of FZIs should be developed for the deep
shape, sorting and packing), tortuosity, and surface to grain reservoirs.
volume ratio on permeability, it serves as a very good basis to
describe and predict the hydraulic behaviour of a reservoir. References
The FZI is also a reservoir parameter that is stable under 1. Evaluation package for radioactive sands in the Niger
stress. The good results obtained while using FZI based Delta, Shell Group Research Report: Ogagarue E.E.
permeability models in dynamic reservoir simulation attest to 2. Kozeny and Carman*: Reference not sighted but is
this. extensively quoted by other authors.
3. Enhanced Reservoir Description: Using Core and Log
There is scope for further development in the area of Data to Identify Hydraulic (Flow) Units and Predict
modelling anisotropies. Progress in this area has been limited Permeability in Uncored Intervals, SPE-26436. J.O.
due to insufficient data. In practice vertical permeability Amaefule & Altunbay M.
values are measured at wider intervals along a core than the 4. . Mature Niger Delta Reservoirs: Integration of Core and
horizontal permeability values, thus, neurological networks Log Data for flow unit zonation and Permeability
based on vertical permeability data will require further testing Prediction , SPENC 97. Amaefule J.O.,.Njoku RF.A,
on field data before being recommended for wide application. Egbogah E.O., Keelan D.K., Rasolovoahangy R.K. and
Udegbulam E.O.:
5. A Simple and Accurate Method for an Integrated
Analysis of Core and Log Data to Describe Reservoir
Heterogeneity, The Journal of Canadian Petroleum
technology (January 1996). Ambastha A.K. & Moynihan
T.J.

Das könnte Ihnen auch gefallen