You are on page 1of 10

SPE 86935

A New Methodology To Estimate Fracture Intensity Index For Naturally


Fractured Reservoirs
Sandro Arango, Ecopetrol – DYA, Eduardo A. Idrobo*, Hector H. Perez*, Ecopetrol – ICP
*
SPE Members

Copyright 2004, Society of Petroleum Engineers Inc.


Introduction
This paper was prepared for presentation at the SPE International Thermal Operations and The fracture intensity is a key factor to predict petrophysical
Heavy Oil Symposium and Western Regional Meeting held in Bakersfield, California, U.S.A.,
16–18 March 2004. properties in a Naturally Fractured Reservoir. These variables
This paper was selected for presentation by an SPE Program Committee following review of
define optimal well productivity and reservoir
information contained in a proposal submitted by the author(s). Contents of the paper, as development strategies.
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 In this project, by a suitable combination of information such
position of the Society of Petroleum Engineers, its officers, or members. Papers presented at
SPE meetings are subject to publication review by Editorial Committees of the Society of
as geologic and engineering concepts, core interpretation, well
Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper logs, pressure tests, and production data, we have found a way
for commercial purposes without the written consent of the Society of Petroleum Engineers is
prohibited. Permission to reproduce in print is restricted to a proposal of not more than 300 of modeling fracture permeability from fracturing
words; illustrations may not be copied. The proposal must contain conspicuous susceptibility. Our approach was applied to a field producing
acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O.
Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435. from a Cretaceous Formation in a Colombian Field.
This paper is divided in three parts. First, a discussion about
Abstract the core analysis and the petrophysical interpretation is shown.
The characterization of Naturally Fractured Reservoirs In the second part, the methodology itself is discussed.
(NFR) represents additional challenges to that of conventional Finally, we have successfully applied the proposed technique
reservoirs because it is not only required the description of on a reservoir producing from a Cretaceous Formation in a
two separate means (matrix and fracture) but also it is Colombian Field.
necessary to understand their interaction. Among the list of
potential parameters that can contribute to understand the
complexity of NFR is the fracture intensity. This is defined as Methodology
the density of fractures per foot of formation. This attribute is Core Analysis
a key factor for a quantitative prediction of the porosity and Core analyses allowed describing features of fractures such as
permeability of a NFR. Furthermore, it is directly related to intensity, dip, length, aperture, mineralization type, and
the reservoir productivity and can be used to optimize morphology. According to their description, analysis, and
reservoir management decisions. environmental interpretation the Formation Facies distribution
In this paper, we present a new methodology to identify and their respective Fracture Intensity was determined. Seven
fractured intervals in a NFR by combining the answer of a set hundred and sixty six (766.0) ft of core were analyzed.
of conventional electric logs and the information coming from
the physical description of the cores of a fractured formation. Facies Determination. Five Facies were defined. For practical
With this information, a continuous variable fracture intensity purposes, they were grouped into three Facies, based on the
track is generated and related with the fracture permeability answers obtained from electric logs. Limestones was
through two different approaches. In the first approach, the denominated Facies 1 (considering those of the type grain
aperture and fracture intensity were used to estimate the supported: Grainstone and Packstone and the mud supported:
fracture permeability. The second approach is indirect. Using Weakstone and Mudstone). Sandstones was named Facies 2
a non-parametric regression technique, the measured fracture (including, original Facies 2 of conglomeratic sands and
permeability is modeled from variables such as GR, fracture Facies 3 of sandstones) and Clays (represented by Facies 4 of
porosity, Facies, and fracture intensity. In both cases, silstones and Facies 5 of mudstones)
quantitative models of fracture permeability were obtained
under static conditions. Types of Fractures. Natural fractures observed in cores were
We have successfully applied the proposed technique on a classified in four types:
reservoir producing from a Cretaceous Formation in a Sub-verticals. With variable spacing (10-100 millimeters),
Colombian Field. The models obtained were satisfactory length of 10-300 millimeters, and aperture minor to 1
validated with information coming from core analysis, millimeter. They are characterized to present a right profile
stratigraphic and structural models, pressure tests, and and dip related to a stratification plane, seen in the nearest
production data.
2 SPE 86935

clays, between 45 and 90°. They form an interconnected of minimum clay volume (VSHmin) which was employed in
net of fractures through the Formation producer. the models of effective porosity and as cutoff for the selection
Stilolytes. These fractures are related to high concentration of fractured intervals.
of planar efforts inside a rock that can originate surfaces of
weakness and breakup of minerals1. The observed GRlog − GR Sand
stilolytes present 10 to 20 mm of length, amplitude minor VSH min = (1)
GR Shale − GR Sand
to 1 mm and variable spacing. Almost all possess
horizontal orientation and irregular profile. According to Porosity models. From log interpretation the apparent,
the presence of organic matter and mud, these breakup effective, and fracture porosity was obtained. The Density and
structures can contribute as a network for the Neutron logs crossplot was used for the calculation of the total
hydrocarbon flow. porosity. The matrix porosity was obtained starting from the
Sub-horizontal. They present characteristic common to sonic log. The difference between the two previous
the sub-vertical fractures. They dip ranges between 0 and calculations is considered to be the secondary porosity.
45°, fracture spacing between from 10 to 30 millimeters, For the determination of the matrix density, The grain density
longitude between 10 and 100 millimeters, and amplitude data of laboratory tests was used. In non-cored wells, a grain
between 0.5 and 1 millimeter. Although they are partially density of 2.71 gr/cc was assumed for calcareous matrix. The
mineralized with calcite, important apertures are observed fluid density was assumed close to 1.0 gr/cc, considering fresh
for the storage and flow of fluids in the porous media. mud and low API gravity.
Fissures. Fissures have been defined2 as discreet The total and matrix porosity were calculated departing from
heterogeneities with permeability that can be considerably the combination of the equations:
bigger that the matrix permeability that surrounds them.
They are usually associated to calcareous Facies. At RHOB matrix − RHOBlog
ultraviolet light, they show a potential of hydrocarbon and DPHI = (2)
high connectivity. The aperture of this type of fractures RHOBmatrix − RHOB fluid
does not surpass the millimeter. They are practically
horizontal and their longitude varies between 10 and 100 NPHI = NPHI _ LS _ C (3)
mm. The Figure 1 shows the different types of fractures
visualized in the cores of the study field. DPHI + NPHI
NDPHI = (4)
2
Fracture Mineralization. It is considered that the four fracture
types described previously improve the reservoir permeability. DTlog − DTmatrix
SPHI = (5)
However in most of the cases, the fractures seen in the cores DT fluid − DTmatrix
are completed or partially affected by diagenetic processes
(secondary mineralization). The commonly opposing minerals For the total and matrix porosity, the effective porosity curve
are calcita, quartz, glauconite and pyrite. Theoretically, it is was obtained to correct the apparent porosity by the clay
considered that when the fracture is completely occupied for a volume (VSHmin) by the following equation:
secondary mineralization process, the permeability reduction
is significant. Sometimes the partial mineralización can end up PHIE = PHIA*( 1 − VSH min ) (6)
being beneficial for the processes of production of fluids,
because it impedes the closing of the fractures allowing the When considering the sonic log response (fundamentally the
draining of the reservoir.3 compresional wave), the matrix porosity is provided. The
combination of Neutron-Density provides the total porosity
Relationship between Facies and fracture intensity. The and the difference between them is interpreted as
method developed to determine the relationship between fracture porosity4:
fracture intensity and Facies departs from the description and
count of the fractures per foot in the cores to associate it to the PHIS = NDPHI − SPHI (7)
previously described Facies (Limestone, Sandstone and Clay).
The number of fractures per foot seen in the cores was plotted From the porosity analysis, it is important to point out that the
in function of the number of described ft of core (Figure 2). effective total porosity values (PHIE) favorably correlate with
those of secondary porosity (PHIS), both obtained from the
Petrophysical Interpretation interpretation of electric logs. This fact corroborates what is
This interpretation included the determination of the minimum observed from thin section images (Figure 3) in which a poor
clay volume and of the apparent and effective porosity of the contribution from the matrix to the total porosity of the system
system matrix - fractures. is observed.
Clay Volume. The model that best represents the behavior of Fracture Intensity Model through of well logs
the established clay volume for the core analyses and the In general, all formations contain fractures in different
answer of the logs was the one obtained taking of parameters proportions, therefore it is necessary to detect them. There are
for sand and clay in each well. A linear relationship, from the two basic techniques of identification of fractured intervals1:
answer of the log Gamma Ray was used to generate the curve Direct (data from outcrops, cores, sidewall images, and
SPE 86935 3

bottom cameras) or Indirect (well logs, pressure tests, and With this curve and the shaliness cutoff (VSHmin < 0.5) is
seismic interpretation). In the last decades, big efforts have possible to establish the best cutoffs of this variable for each
been done to detect fractures by means of indirect techniques unit and well. The criterion of shaliness was involved by
derived of the interpretation of the electric logs. keeping in mind that with this tool similar answers can be
Because different effects (shaliness, ruggedness, hole obtained in highly shaly areas. Using a simple algorithm, the
geometry, fluids, etc.) can affect the response of electric logs, well depths at which there is probability of fractured zones can
a similar answer identifying fractured zones can not be be obtained:
expected. A methodology that combines the answers of
several tools could be helpful to generate a reliable model for IF DCAL >= DCCUTOFF AND VSHMIN <= 0.5
the identification of fractured intervals. THEN DCFLAG = 1;
ELSE DCFLAG = 0;
In our case, the fractured intervals were identified comparing END IF;
the tool responses with cutoffs obtained of the fracture
intensity core information. The approach was validated with The previous algorithm generates a Boolean parameter that
information coming from FMI logs.5 These cutoffs were would take a value of one (1) where high fracture probability
determined for every lithounit. In order to achieve a better exists and zero (0) where the presence of fracturing is not
determination of each one of these zones, the curves Gamma likely.
Ray and VSHmin were used like lithological discriminators
(Figure 4). Density log correction condition. An indicative of the
It is important to point out that every particular case defines presence of fractures can be obtained from the curve of
the cutoffs to be used in the identification of fractured correction of the density log (DRHO) This curve corrects the
intervals. Many practical applications6,7,8,9,10,11,12,13,14 prove density log for effects of hole ruggedness and mud cake.
this approach. When mud fractures exist, anomalies in the curve (positive or
negative deflections) are displayed, spite of good hole
Decision Tree Analysis conditions.4
We applied a statistical tool called "Decision Tree Analysis", In a constant lithology, the ruggedness of the wellbore obeys
for identifying the more important logs in prediction of almost exclusively to the presence of fractures. In this case,
fracture intensity. The tree-building methodology used in the the correction curve acts as a very sensitive caliper that detects
current study is based on a probability model approach.15,16,17 ruggedness of the hole and presence of fractures.
Classifiers at each node are selected based on an overall
maximum reduction in deviance, for all possible binary splits Deep and shallow resistivity relationship condition. In a full
over all the input variables. The deviance at a given node i is: hydrocarbon fracture system, the mud displaces the fluids near
to the wellbore invading the fractures. The resistivity tool
Di = −2∑ nik log( pik ) (8) response can be used to identify fractured intervals. In general,
k the fractured areas will be represented by a positive
where pik is the probability distribution at i over the classes k differential in readings between the Rt and Rxo curves. With
and nik is the number of cases (y values) assigned to k at i. The this criterion is possible to determine qualitatively the fracture
probabilities are unknown, but can be estimated from the intensity and the fluid type present in the fractures (Figure 5).
proportions at each node, i.e.: For our case, the following equation was used to indicate the
n presence of a high fracturing level.
pik = ik (9)
ni RT _ C
RR = (11)
RXO
Criteria of Definition of Fracture Intensity from Electric
Logs Differential of micro-resistivity log condition. The separation
The criteria used for the determination of fractured intervals of the micro-inverse and micro-normal curves is another
were: caliper differential (DCAL), density log correction parameter to identify fractured intervals. This hypothesis is
(DRHO), deep and shallow resistivity log relationship (RT- also derived from the mud invasion concept, exposed
RXO), differential of micro-resistivity logs (MINV- previously. The micro-inverse curve has a shallower radius of
MNORM), compresional and shear waves relationship of the investigation than the micro-normal and for this reason its
sonic log, secondary porosity, and photoelectric potential log reading is lower in presence of filled fractures with conductive
response (PEF). fluids. The following equation is used to quantify the effect of
fracture intensity from these tools:
Caliper differential condition. The caliper is a measure of the
DML = HMNO − HMIN (12)
diameter of the hole. The fractured areas can be reflected in
the tool by washouts generated by losses of mud causing
Compressional (P) and shear (S) waves of the sonic log
collapses in the wellbore. This fact creates an increment in the
relationship condition. The sonic log is employed to detect
tool reading.18 The caliper is used to calculate the DCAL
fractures given the variation of acoustic attenuation
curve by the following equation:
experienced by a wave when a drastic change in permeability
DCAL = CAL − BS (10) takes place. The relationship between the speed of the waves S
and P is constant for a defined lithology. The fractures can be
4 SPE 86935

identified in those intervals when the time of transit of the Validation of the Fracture Probability Model
shear wave is much bigger (smaller speed) than that of the Once generated the fracture probability model for each well,
compressional wave, for a given lithology. The following the next step is to validate it by comparing the information
equation will be used. coming from cores and images logs.

DTshear Core Information. Seven (7) cores were analyzed and


RDT = (13) described. The fractures were quantified foot by foot. This
DTcompressional
information was scaled between zero (0) and one (1)
representing areas highly fractured with a value of fracture
Secondary porosity condition. The combination of porosity
intensity of one (1) or near to the unit and areas without
logs (Neutron, Sonic, and Density) can be an effective method
fractures with a value of zero (0) or near to zero. The
to indicate the presence of fractures. The difference between
generated curve was denominated COREFRACT and it was
the porosity obtained from the Neutron-Density combination
compared with the logs interpretation derived
and the porosity derived from the sonic log can be interpreted
curve, LOGFRAC.
as fracture porosity (Figure 6). The condition used is:
PHIS = NDPHI − SPHI (14)
Images logs (FMI). Four wells of the study field have
information of images logs and quantitative analysis of the
Photoelectric potential (PEF) log response condition. This fracture intensity. This information was scaled between zero
log is an indicator of the effective photoelectric absorption of (0) and one (1) to assure a convenient base of comparison with
the formation. Fractured intervals, associated to calcareous the core analysis and the log interpretation curves. The curve
formations, with strong mud invasion can be interpreted by FMIFRAC represents a high fracture degree for the values
high values of PEF.11 Additionally, this tool can contribute to near to one (1) while values close to zero (0) mean poor
clarify ambiguous situations where the answer of the logs fracture intensity. Figure 8 compares the curves of fracture
could mean fractured intervals or gas zones. intensity coming from logs (LOGFRAC) core description
(COREFRACT) and images logs (FMIFRACT).

Field Application Incidence of the Fracture Intensity on the Fracture


Permeability for a NFR
We have applied the proposed technique on a reservoir As mentioned, the fracture intensity is an indispensable factor
producing from a Cretaceous Formation in a Colombian Field. for the quantitative prediction of the porosity and permeability
The field has six wells with most of the required information. for a NFR. These variables are intimately related to the
Four wells have image logs. reservoir exploitation strategy and the well productivity. An
empiric relationship relating fracture permeability (measured
Application of the Criteria to Establish the Fracture in a limited number of cored wells) and petrophysical
Probability properties indirectly obtained from log interpretation is
As it was discussed, the degree or fracture intensity of a primordial to NFR characterization.19
formation can be evaluated under different conditions. For our
particular case and keeping in mind that not all the six wells Direct modeling of the fracture permeability
have the seven conditions described previously, a pondered Direct modeling of the fracture permeability is based on the
arithmetic average was established depending on the number analysis of the causes of natural fracturing and its associated
of available conditions for each well: properties (spacing, intensity, aperture, porosity, etc.). The
proposed relationship (analytic y/o empiric) must be able of
LOGFRAC = P1*DCFLG+P2*HDFLAG+P3*RRFLAG+P4*DMLFLAG+
incorporating the mentioned parameters and control factors
P5*RDTFLAG+P6*PEFFLAG+P7*FRACFLG such as lithology, structural position, thickness, etc. To
(15) accomplish this task is necessary to have enough measured
data. This is usually a great limitation.5 Before establishing a
In the wells where there is a missing condition, the weight model to quantify the fracture permeability, it is necessary to
factor (Pi) is made zero (0) and the weighting factors outline a series of concepts.
redistributed (Table 1). The same weighting factors were used
for each condition, because it was considered that each one of Fracture Porosity. If we consider that the porosity depends
them contributes potentially in same proportion in the exclusively of the porous space than the fractures, then the
identification of intervals fractured. This approach leaves an matrix contributes with zero percent (0%) to the total porosity.
open possibility of implementing a methodology to establish, Assuming a simple cylindrical geometry to represent a core
in a hierarchical way, a set of weighting factors for a particular rock, then:
case.
Figure 7 summarizes the use of the conditions described IF * π * r 2 * H f IF * H f
φ= = 0.3281 (16)
previously. It shows the fracture probability curve π *r *H
2 H
(LOGFRAC) generated starting from electric logs.
Based on the fact that the values for the fracture aperture (Hf)
ranged between 1 and 2 mm, a total longitude (H) of 1 ft was
assumed to obtain a set of values to explain the effect of the
SPE 86935 5

fracture intensity on the fracture porosity (Table 2). From this With w in centimeters, IF (# fracture/ft)
table, it can be appreciated that more than 15 fractures/ft are
needed to reach a porosity of 1.0% with an average fracture K f = 2.77 *10 5 w 3 * IF (23)
aperture of 0.2 millimeters and close to 8 fractures/ft to obtain
that same porosity if an average fracture aperture of 0.4 Equations (22) and (23) show that the fracture permeability is
millimeters is used. This points out the fact that the average directly proportional to the cube of the fracture aperture and
fracture porosity hardly exceeds 1% in a NFR. the fracture intensity. This demonstrates the importance of
developing methodologies to obtain good estimates of these
Matrix permeability. To estimate the matrix permeability the parameters under the difficulty of directly determining them.
Darcy´s Law can be used. This applies under steady state, For our field study, average values of fracture aperture ranged
linear, horizontal, laminar and isothermic flow conditions, between 0.1 and 0.5 mm and fracture intensity between 0.5
with constant viscosity and porous space fully saturated with a and 4 fractures/ft.
non-wetting phase. The high values of fracture permeability that are obtained by
For the case of the flow of an incompressible, linear fluid in means of the equations (22) or (23) also indicate the
one dimension, the permeability can be estimated with the importance of fracture intensity and aperture on the production
following equation: of tight reservoirs, which could be considered non-commercial
spite of having important hydrocarbons reserves (Table 3).
QµL
K= (17)
0.001127 A ∆P Inverse modeling of the fracture permeability
The inverse approach focuses in the effects created by the
Fracture permeability. It is defined as the permeability of a
variables involved in the modeling on the attribute to be
rock in which the fracture is only responsible for the flow of
predicted. They have become several intents to understand the
fluids. The matrix is considered practically impermeable for
nature of the fractured systems using the concepts of storage
its low permeability.
coefficient and interporosity derived of the pressure tests
Assuming a flat and planar fracture with aperture (w),
analysis. Unfortunately the analyses to obtain these data are
longitude (L), lateral extension (h) and a constant rectangular
based in very simplified models (as sugar cubes, matches, etc.)
transversal section (hw), viscosity of fluid (µ) that flows to a which do not consider the reservoir heterogeneities. For this
rate (Q) under a differential pressure ∆P. The flow of fluids particular case, it is required to predict the behavior of the
occurs only through the fracture (Figure 9). For this system, it fracture permeability using information obtained from the core
can be demonstrated5 that: basic analyses, the variables derived of the log interpretation,
and some geologic observations.
Wo 2
Kf = (18)
12 Preliminary analysis of the information. The fracture
permeability data was obtained for two wells of the study area.
This last equation is known as Muskat’s expression. Now The fracture permeability values were those obtained from the
remembering that the matrix permeability is practically probe permeameter run on full cores and corrected by
negligible, it can demonstrate that: stress effects.
An initial correlation analysis showed which variables had
w3 bigger impact on the prediction of the fracture permeability. In
Kf = (19)
12 s this modeling process, the permeability is operated in
logarithmic scale (the typical behavior of this variable is log-
The equation (19) is known as Lambs’s expression for the normal). Table 4 shows the variable that has more impact on
fracture permeability. If a total thickness (h) of 1 ft is assumed the prediction of the fracture permeability is LOGFRAC,
and knowing that: followed by the lithological variable GR, the fracture porosity
(also obtained of logs interpretation), and the geologic variable
Total thickness h
IF = = (20) FACIE (related to the lithology present in the Formation).
Spacing S
Modeling inverse technique To establish a relationship
then:
between the dependent variable (log of fracture permeability)
w3 and the independent ones (LOGFRAC, GR, FACIE and
Kf = * IF (21)
12 PHIE), it was necessary to use a multivariable no-parametric
regression. The selected technique was the algorithm GRACE
The equation (21) shows the direct relationship that exists (Graphical Alternating Conditional Expectation), originally
between the fracture intensity with the fracture permeability proposed by Breiman and Friedman20 and successfully applied
for a NFR in the cases where the matrix has little contribution by Lee and Datta-Gupta21.
to the storage and flow. Introducing units in this equation, The optimum transformations generated by GRACE and the
then: final equation to predict the logarithm of the fracture
permeability are:
With w in inches, IF (# fractures/ft)
FACIEtr = −0.02844 * FACIE 2 + 0.12436 * FACIE − 0.023514 (24)
K f = 4.54 *10 6 w 3 * IF (22)
6 SPE 86935

GR _ Ctr = 0.00042786 * GR _ C 2 − 0.045725 * GR _ C + 1.0094 (25) Conclusions and recommendations


1. Limestone Facies presents higher fracture intensity with
LOGFRACtr = −4.846 * LOGFRAC2 + 4.9430 * LOGFRAC − 0.26519 (26) 4.8 fractures/ft (539 fracture in 111 ft) followed by
PHIEtr = −0.000017641 * PHIE 2 − 4.4695 * PHIE + 0.067903 (27) Sandstone Facies with 2.1 fractures/ft (323 fractures in
153 ft), and lastly the Clay Facies with 0.38 fractures/ft
LogK tr = 0.49178 * sumtr 2 + 2.1638 * sumtr + 0.75411 (28)
(115 fractures in 301 ft).
2. Fracture aperture (width) observed on cores does not
With these equations, it was possible to calculate the
exceed, in most of the cases, the millimeter. It was proved
logarithm of fracture permeability and to compare it with the
the importance (theoretically and in practice) that this
direct measurements of laboratory to see its correspondence
parameter has on the permeability and productivity of
(Figure 10). The calculated correlation coefficient (R2) is close
the NFR.
to 0.8 and that the model works well for almost the whole
3. Comparison of the different porosity models obtained
range of values of fracture permeability.
from log interpretation show the effective porosity highly
Blind test validation. To confirm the validity of the proposed correlates with the secondary porosity. Therefore, little
model, a blind test was carried out. The correlation coefficient contribution exists from the matrix to the system
(R2) obtained for this well validation is shown in Figure 11. It total porosity.
can be noticed that spite having model with a high correlation 4. The conditions that better identify fractured intervals are:
coefficient, more information is required to build a more the photoelectric potential (PEF), the relationship
robust model. between resistivities, the relationship between waves of
the sonic log, and the secondary porosity. The PEF was
Decision Tree Application also important in the discrimination of areas with
Tree based modeling is considered an exploratory technique presence of gas.
for uncovering structures in the data. It is a way to present 5. The weighting method of each one of the conditions
rules to predict or explain responses for a categorical or selected for the definition of fractured intervals was a
continuous variable16. valid and reliable alternative for the solution of the
Basically, the tree graph is a series of nodes or vertices and a problem. The use of cutoffs for area and well represents a
set of edges. The tree graph has three kinds of nodes. The hierarchical approach for the established conditions.
node at the top of the tree is called a root node. In addition, 6. It is recommended to extend the fracture intensity
there are internal and terminal nodes. The terminal or leaf estimation to reservoir scale using
node has no descendant or in other words it has the final geostatistical techniques.
classification for a path; the internal nodes are nodes that have 7. Modeling of fracture permeability by direct and inverse
descendants or are susceptible to more splits or rules. approaches have restrictions related to the limited
The building of decision tree for this case used three wells for quantity of data and to the use of theoretical models that
obtaining the tree graph. The fractured intensity was the value represent in a very simplistic way a real fractured system.
of cores, and it was the parameter for estimating. The Figure Each one of these used approaches in an independent way
12 shows the decision tree chart with 20 terminal nodes, or does not guarantee the proper reservoir characterization
ways for obtaining the fractured intensity. According to this for a NFR. The combination of both of them allows the
chart the indicator which has the more impact is the PEF reduction of the associated uncertainty.
indicator because it is on the top the chart17. The next 8. We developed an analytic expression that demonstrates
indicators are RR, DCAL, and GRC. the dependence of the fracture permeability on the
According to this tree, in conditions of missing well logs, the fracture intensity.
PEF indicator is absolutely necessary for estimating the 9. It was generated a model to determine the fracture
fractured intensity, but PHIS indicator is not necessary. permeability departing from indirect determinations
(FACIE, GR, LOGFRAC and PHIE) and the use of the
Conceptual validation by means of dynamic data of pressure GRACE algorithm. The model developed presents a
and production correlation coefficient of 0.78.
Spivey and Economides22 and Aguilera3 evaluated the build up 10. It was demonstrated, in the wells of the area, the relative
tests in five producing wells. The studies concluded about the effect of the fracture intensity on the fracture permeability
presence of a dual porosity reservoir conformed by a matrix and productivity for a NFR.
system (micro fractures) and another of fractures (macro
fractures). The pressure tests confirmed the high permeability Acknowledgments
associated with the best oil producers wells and they We would like to thank the Empresa Colombiana de Petróleos
correspond in good proportion to those that exhibit higher S.A for supporting this work.
fracture intensity and higher value of fracture permeability
seen on cores. This confirms the strong impact of the fracture Nomenclature
intensity on fracture permeability of a NFR and its incidence VSHmin = Minimum Volume Shale, fraction
on well productivity. The Figure 13 shows the relationship GRlog = Gamma Ray Log, API
existent among fracture intensity, fracture permeability, GRarena = Clean Sand reading, API
and production. GRarcilla = Shale Sand reading, API
DPHI = Apparent porosity, fraction
SPE 86935 7

RHOB matriz = Matrix density, gr/cc 6. Baker, R. and Kuppe, F.: “Reservoir Characterization for
RHOB log = Density log, gr/cc Naturally Fractured Reservoirs,” paper SPE 63286, Annual
RHOB fluido = Fluid Density, gr/cc Technical Conference and Exhibition held in Dallas, Texas 1-4
October, 2000.
NPHI_LS_C = Neutron Log, limestone matrix,
7. Ohen, H., Daltaban, S., and Enwere, P.: “A systematic and
fraction detailed Approach to Fractured Reservoir Petrophysical
NDPHI = Neutron-Density Porosity, fraction Modelling for Reservoir Simulation,” Scott Pickford Group,
SPHI = Sonic Porosity, fraction Core Laboratories Company, 2000.
DT log = Sonic Log, msec-ft 8. Verga, F.M., Carugo, C., Chelini, V., Maglione, R., and De
DT matriz = Matrix transit time, msec-ft Bacco, G.: “Detection and Characterization of Fractures in
DT fluido = Fluid transit time, msec-ft Naturally Fractured Reservoirs,” paper SPE 63266, SPE Annual
PHIE = Effective porosity, fraction Technical Conference and Exhibition held in Dallas, Texas,
PHIA = Apparent porosity, fraction October, 2000.
PHIS = Fracture Porosity, fraction 9. Olarewaju, J., Ghori, S., Fuseni, A., and Wajid, M..: “Stochastic
“Simulation of Fracture Density for Permeability Field
DCAL = Differential between caliper log Estimation,” paper SPE 37692, SPE Middle East Oil Show and
and bit size, inches Conference held in Manama, Bahrain, March 1997.
CAL = Caliper log, inches 10. Ashish, B. and Schechter D.: “Characterization of the Naturally
BS = Bit size, inches Fractured Spraberry Trend Shaly Sands Based on Core and Log
RR = Relat. between resistivities, ohm-m Data. Permian Basin Oil and Gas Recovery,” paper SPE 35224,
RT_C = Corrected Deep Resistivity Log, Conference held in Midland, Texas, March 1996.
ohm-m 11. Aguilera, R., Naturally Fractured Reservoirs, Second Edition:
RXO = Corrected Micro Spherical PennWell Publishing, Tulsa, Oklahoma, 1995.
Resistivity Log, ohm-m 12. Sibbit, A. M.: “Quantifying Porosity and Estimating Permeability
from Well Logs in Fractured Basement Reservoirs,” paper SPE
DML = Micro-Resistivity Differential, 30157, SPE PetroVietnam held in Vietnam, March 1995.
ohm-m 13. Iverson, W. P.: “Fracture Identification from Well Logs,” paper
HMNO = Micro-normal log, ohm-m SPE 24351,Rocky Mountain Regional Meeting held in Casper,
HMIN = Micro-inverse log, ohm-m Wyoming, May 1992.
RDT = Acoustic relation between shear 14. Boyeldieu, C and Winchester, A.: “Use of the Dual Laterolog for
and compresional transit time. the Evaluation of the Fracture Porosity in Hard Carbonate
DT shear = Transit time shear wave, Formations,” Offshore South East Asia Conference,
microsec/ft February 1982.
DT compresional = Transit time compresional wave, 15. Breiman, L., Friedman, J.H., Olshen, R.A., and Stone, C.J.:
Classification and Regession Trees, Wadsworth and
microsec/ft Brooks/Cole, Monterrey, CA., 1984.
φ = Fracture Porosity (percent) 16. Perez, Hector H., "Permeability characterization and Spatial
IF = Fracture Intensity, # fractures/ft Modeling in Complex Reservoir: Use of Tree Classfiers and
Hf = Fracture Aperture, mm Markov Random Field," M.S. Thesis, Texas A&M
H = Core Length, ft University, 2002.
Q = Flow Rate, Bbl/d 17. Perez, Hector H., Datta-Gupta, A., and Mishra, S.: "The Role of
K = Matrix Permeability, md Electrofacies, Lithofacies, and Hydraulic Flow Units in
A = Sectional Flow Area, ft2 Permeability Prediction from Well Logs: A Comparative
Analysis Using Classification Trees," paper SPE 84301
∆P = Pressure Differential, psi presented at the 2003 SPE Annual Technical Conference,
µ = Viscosity, cp Denver, 5-8 October.
L = Distance, ft 18. Suau, J. and Gartner, J.: “Fracture Detection from the Logs,”
Evaluación y Detección de Fracturas, 212.75 E814, 1986.
References 19. Jennings, J. W. Jr. and Lucia, J. F.: “Predicting Permeability From
1. Nelson, R. A.: Geologic Analysis of Naturally Fractured Well Logs In Carbonates With a Link to Geology for Interwell
Reservoirs. Gulf Publishing Company, Houston, Texas (1985) Permeability Mapping,” paper 71336, SPE Annual Technical
2. Bushteiner, H. H., Warpinski, N. R., and Economides, M. J.: Conference and Exhibition held in New Orleans, October,2001.
“Stress-Induced Reduction in Fissured Reservoirs,” paper SPE 20. Breiman, L and Friedman, J.H.: “Estimating Optimal
26513 presented at the 68th Annual Technical Conference and Transformations for Multiple Regression and Correlation,”
Exhibition of the Society of Petroleum Engineers held in Journal of the American Statistical Association, 1985.
Houston, Texas, 3-6 October 1993. 21. Lee, S. H. and Datta-Gupta, A.: “Electrofacies Characterization
3. Aguilera, R.: “Reservoir Study of the Cimarrona Naturally and Permeability Predictions in Carbonate Reservoirs: Role of
Fractured Reservoir, Guaduas Field,”. Servipetrol Ltda, Multivariate Analysis and Non-Parametric Regression,” paper
February 1999. SPE 56658, Annual Technical Conference and Exhibition held
4. Beck, J., Shultz A., and Fitzgerald, D.: “Reservoir Evaluation of in Houston, Texas. October, 1999.
Fractured Cretaceous Carbonates in South Texas,” paper M, 22. Spivey, J. and Economides C. E.: “Data Review and Pressure
SPWLA Logging Symposium Transactions (1977). Transient Analysis and Design,” Holditch and Associates,
5. Arango, S.: Determinación del Indice de Intensidad de June 1998.
Fracturamiento y su incidencia sobre la permeabilidad de los
Yacimientos Naturalmente Fracturados, M.S. Thesis,
Universidad Industrial de Santander, February 2003.
8 SPE 86935

Table 1. Number of conditions used by well Table 4. Correlation matrix to fracture permeability inverse model.
Variables with a higher value than 0.4 were used in the modelling
WELLS #CONDITIONS WEIGHTS
1, 2 7 0.14 VARIABLE LOGKf FACIE GR_C LOGFRAC PHIE
3, 4, 5 6 0.16
LOGKf 1
6, 7 5 0.20
FACIE -0.41296275 1

Table 2. Relationship between fracture intensity and fracture GR_C -0.4839679 0.81473606 1
porosity
LOGFRAC 0.51309753 -0.79951719 -0.84633053 1
Hf = 0.2 mm Hf = 0.4 mm
H=1 ft H =1 ft PHIE 0.45361774 -0.82466873 -0.89099461 0.87074546 1
IF φ (%) IF φ (%)
(# fract/ft) (# fract/ft)
5 0.33 5 0.66
10 0.66 10 1.31
15 0.98 15 1.97
20 1.31 20 2.62
25 1.64 25 3.28
30 1.97 30 3.94
35 2.30 35 4.59
40 2.62 40 5.25
45 2.95 45 5.90
50 3.28 50 6.56

Figure 1. Types of fractures seen on the cores.


Table 3. Relationship between fracture intensity and fracture
permeability

Hf = 0.1 mm Hf = 0.2 mm
FRACTURE INTENSITY
IF Κf (D) IF Κf (D)
(# fract/ft) (# fract/ft)
1 0.28 1 2.22
FRACT/FT

2 0.55 2 4.43
3 0.83 3 6.65
4 1.11 4 8.86
5 1.39 5 11.08
6 1.66 6 13.30 1 2 3 4 5 6

7 1.94 7 15.51 WELLS


8 2.22 8 17.73 Figure 2. Fracture Intensity seen on the cores.

9 2.49 9 19.94
10 2.77 10 22.16
SPE 86935 9

PHICORE Vs.PHIN
0.20
0.18

PHICORE (fraction)
A B
0.16
0.14
D 0.12
0.10
C
0.08
0.06
A'
Figure 3. Thin section. Porosity is only associated to the fracture.
E 0.04
0.02
0.00
0.2 0.18 0.16 0.14 0.12 0.1 0.08 0.06 0.04 0.02 0
PHIN (fraction)

A Figure 6 Theoretical representation of fracture porosity. Points D


and E correspond to fractured intervals.

Figure 7. LOGFRAC curve (last yellow curve).


Figure 4. Typical interpreted log.

2.5
High Fracture Intensity
with hydrocarbons
2.0 Low Fracture Intensity
with hydrocarbons

1.5 Fracture Intensity


Fractured with water
LLD/LLS Shales
1.0

0.5 Non Fractured Zone

0
0 100 200 300 400
LLD (ohm-m)
Figure 8. Comparison and validation of LOGFRAC curve (yellow)
Figure 5 Relationship between deep and shallow resistivity logs with COREFRACT curve (blue) and FMIFRAC curve (green).
to identify fractured intervals.
10 SPE 86935

PEF<4.41
SUPOSICIONES
RR<2.39 RR<20.17

Q DCAL<1.34 GRC<44.68 DCAL<0.42 0.57


h PEF<3.48 GRC<40.16 PEF<5.05 GRC<17.44
0.09

RDT<1.56 0.06 DCAL<1.80 0.07 0.78 0.21 DCAL<1.44 GRC<37.04

DP w L 0.07 0.09 0.37 0.14 PEF<5.16 PHIS<-0.06 0.12

0.14 0.15 0.15 DCAL<0.60

0.59 PEF<4.93

GRC<29.88 0.17

DCAL<1.26 0.40
Q
Figure 9. Idealistic flow model to represent fracture permeability. 0.19 0.78

Figure 12. Behavior of the Fracture Intensity by Decision Tree


Analysis.

Measured Log K vs. Calculated Log K Relative effect of fracture intensity on permeability and well
productivity
5
4
W-1 Permeability
Calculated Log K

3
2 Fracture Intensity
1 W-2 Production
0
y = 0.7791x + 0.5878 Wells
-1 W-3
R2 = 0.7985
-2
-3
W-4
-3 -2 -1 0 1 2 3 4 5
Measured Log K 0.0 0.2 0.4 0.6 0.8 1.0

Relative Effect
Figure 10. Correlation between log fracture permeability
measured on cores and the computed by GRACE.
Figure 13. Relative effect of fracture intensity on permeability and
well productivity.

Measured LOGK Vs. Calculated LOGK


5.0

4.0
y = 0.71x + 0.2714
R2 = 0.6321
3.0
Comp. LOGK

2.0

1.0

0.0
0.0 1.0 2.0 3.0 4.0 5.0
Measured LOGK

Figure 11. Correlation between log fracture permeability


measured on cores and the computed by GRACE – Blind Test.

You might also like