Journal of Petroleum Science and Engineering 178 (2019) 216–226

Contents lists available at ScienceDirect

Journal of Petroleum Science and Engineering

journal homepage: www.elsevier.com/locate/petrol

Forward modeling of electric and electromagnetic resistive logs with piston T

and annular invasion types in pre-salt carbonate reservoirs of the Santos
Basin, Brazil
Jéssica Lia Santos da Costaa,∗, Antonio Abel Gonzaléz Carrasquillaa,
Alfredo Moisés Vallejos Carrascob, Herson Oliveira da Rochac
a
Laboratory of Petroleum Engineering, Science and Technology Center, Northern Rio de Janeiro State University (UENF), Rodovia Amaral Peixoto, km 163, Av. Brenand
S/N, Imboassica, CEP, 27925-310, Macaé, RJ, Brazil
b
School of Engineering, Federal Fluminense University (UFF), Campus da Praia Vermelha, Rua Passo da Pátria, 156, CEP, 4210-240, São Domingos, Niterói, RJ, Brazil
c
Federal Rural University of Amazon (UFRA), Campus Parauapebas, Rodovia PA-275, Km 7, s/n, Gabinete 13, Rural Zone, CEP, 68515-970, Postal Box 3017,
Parauapebas, PA, Brazil

ARTICLE INFO ABSTRACT

Keywords: The invasion of drilling fluid is a highly complex issue in well logging interpretation and is not well understood.
Geoelectric modeling Choosing the appropriate geoelectric model that provides responses to the real data, is critical for elucidating
Invasion this process. For effective modeling, it is necessary to consider the resistivities of the fluids contained into the
Electric logging well and in the reservoir (washed zone) as well as the petrophysical properties of the geological formation
(virgin zone). Traditionally, to analyze the resistivities of the virgin zone (Rt) and the invasion diameter (Di), a
step profile or piston model is assumed, which is the simplest model of invasion. But This model does not
consider the presence of a transition zone or the gravitational effects of fluids, which can give rise to another
type of fluid distribution of fluids into the reservoir called annulus and affecting the electrical logging responses
(noise) causing significant error in the geophysical interpretation. In this work, we simulated the responses of
piston invasion and annulus models, using two algorithms for induction (ILD – deep and ILM – mean) and
laterolog (LLD– deep and LLS – shallow) responses. For such purpose, we performed forward modeling using
these algorithms based on the finite element technique for 2.5D symmetry. Finally, the real data obtained in
carbonate formations of the Santos Basin, Brazil, were compared with the responses, showing a good fit.

1. Introduction The invasion of the drilling fluid has a significant effect on the log
responses of the well logs, especially in the resistivity logs (Salazar and
During the drilling of a well, the drilling fluid is kept at a pressure Torres-Verdin, 2009). Depending on the depth of the invasion zone, the
greater than the pressure of the geological formation. Due to this dif- measured values in the virgin zone undergo changes, complicating the
ference, there is a tendency for the drilling fluid to infiltrate the porous evaluation of these values. Similarly, the values obtained by the tool are
medium in permeable layers and for the solid component to form a thin, also influenced by the resistivity of the mud filtrate, Rmf (Akinsete and
impermeable mudcake (Ribeiro and Carrasquilla, 2014). Adekoya, 2016) and its presence causes a noise type.
From the mudcake, the concentration of the mud filter decreases To model the resistivity of the non-invaded zone (Rt) and the ex-
with distance inside the reservoir, forming an intermediate zone of mud tension reached by the washed zone, which corresponds the mud in-
filtrate, which is subdivided into invasion and transition zones. Beside vasion diameter (Di), the piston profile is generally assumed due to its
this transition zone, there is a virgin or uninvaded zone, where oil, gas simplicity and the limited number of independent measurements of
and formation water are located (El-Wazeer and Haggag, 1999). In the radial resistivity. However, the invasion may not occur radially, and no
transition zone, an annulus zone could exist, wherein a displacement or extant model is considered accurate for the invasion of the drilling
mixture of fluids occurs. This zone determines the limit of invasion, fluid; furthermore, there are no direct methods to evaluate the invasion
known as the mud invasion diameter Di (Fig. 1). extent in situ, satisfactorily (Torres-Verdin et al., 2004).

Corresponding author.
∗

E-mail addresses: [email protected] (J.L.S.d. Costa), [email protected] (A.A.G. Carrasquilla), [email protected] (A.M.V. Carrasco),
[email protected], [email protected] (H.O.d. Rocha).

https://doi.org/10.1016/j.petrol.2019.03.038
Received 28 September 2018; Received in revised form 5 March 2019; Accepted 12 March 2019
Available online 19 March 2019
0920-4105/ © 2019 Elsevier B.V. All rights reserved.
J.L.S.d. Costa, et al. Journal of Petroleum Science and Engineering 178 (2019) 216–226

presence of annulus. Mathematical methods such as finite differences to

simulate triaxial induction equipment have already been used by Li
et al. (2003); Cozzolino and Silva (2007) presented some synthetic fo-
cusing techniques for electric tools; Podberezhny et al. (2016) worked
on one-dimensional (1D) simulation of the induction response for LWD
(logging while drilling) tools and Zeng et al. (2017) also used the
electric modeling using finite differences for inclined layers without
considering invasion. Although these general-purpose methods are
capable of dealing with arbitrarily complicated situations, none of these
works cited the MEF for modeling of annulus and none of the drilling
mud invasion studies were made for carbonates (rock considered to be
highly complex).
In this work, an initial model of geoelectric layers was constructed
based on the resistivity and porosity real log data corresponding to
carbonate formation. Next, the algorithms developed for the electric an
induction system were applied using two investigation depths (deep
and medium), in order to obtain responses close to those provided by
the real data. Adjustments in resistivity values and investigation depth
were carried out, in order to obtain a geoelectric model closer to the
reality, to eliminate the ambiguity in the answers. All this was made to
eliminate noise in the responses created by the invasion zone and an-
Fig. 1. Logging environment (modified from Anderson, 2001). nulus and to be able to know the true resistivity of the formation, which
is an important factor to be dimensioned for reserves evaluation, by
Geologically, there is great difficulty in the interpretation of car- using the Archie equation. Therefore, the described methodology pro-
bonate formations. The presence of the different types of porosity in this vides a way in the reservoir modeling and formation evaluation. The
type of rock causes that for each same carbonate layer there are several original contribution of this work is based on the fact of using two al-
depths of invasion, which alter the values of true resistivity, even more, gorithms (induction and laterog), with different investigations of pe-
when there is the presence of annulus. netration of mud drilling, when applied in the formation of complexes,
In order to solve this problem, thus, the forward modeling of geo- in this case the pre-salt carbonates.
physical well log responses is extremely important. In this work, we
used algorithms that perform the geoelectric modeling of electric and 2. Geological context
induction tools’ responses to simulate the effects of piston and annulus
models. The pre-salt covers a total area of approximately 227,000 km2,
Previous works were based on correction charts called ‘tornado’, which is 800 km long and, in certain regions, 200 km wide. The Santos
having as input data the resistivity values for three investigation depths, Basin is in this environment, 350 km from the coast (Fig. 2). It is bor-
however this procedure is limited for thin layers and when there is no dered to the north by the Cabo Frio High, which separates it from the
Campos Basin and to the south by the Florianopolis High, which

Fig. 2. Pre-salt location map in the Santos Basin (Adepted from Riccomini et al., 2012).

217
J.L.S.d. Costa, et al. Journal of Petroleum Science and Engineering 178 (2019) 216–226

separates it from the Pelotas Basin (Bruhn et al., 2003). Io (r) ( z zr)
j =
The geological petroleum system of the Santos Basin is restricted to 2 r (4)
the subsalt configuration (pre-salt range); it has, as potential petroleum-
where E is the electric field, j is the electric current density and V is
generating rocks, black shales rich in organic matter, interspersed with
the electric potential, considering a source of intensity Io in (r, z) =
carbonates, that were deposited in a paleolacustrine environment
(0, z T) . By inserting Equations (2) and (3) into Equation (4), the fol-
(formations Itapema and Piçarras of the Guaratiba Group) and; As re-
lowing is obtained:
servoirs, carbonates of the Itapema (coquinas) and Barra Velha (mi-
crobialites) formations, both of the Guaratiba Group, and may also ( V) =
I o (r) ( z zT)
occur in siliciclastic rocks (Formation Piçarras) and fractured basalts 2 r (5)
(Camboriú Formation). The migration of hydrocarbons generated in the where is the electrical conductivity of the geological medium.
rift section occurred through direct contact between the generating The differential equation of the scalar electric potential in a cy-
rocks and the reservoir rocks of the rift section. Nevertheless, the pre- lindrical coordinate system with azimuthal symmetry is as follows:
sence of an overlapping layer of salt (Ariri Formation) was likely re-
sponsible for an almost perfect sealant for this oil system (Papaterra, 1 V V I (r) ( z zT)
r (r, z) + (r, z) + = o
2010). r r r z z z 2 r (6)
The results of model simulations in this work were compared with where the potential V is the desired solution and is subject to the
real data from a well in the Santos Basin and the modeling section boundary conditions:
consists of a sequence of geoelectric layers represented by salt, carbo-
nate and shale formations. [V (r , z )]r + z 0. (7)
The equation (6) is treated applicating of the Galerkin method
3. Methodology (Cozzolino and Silva, 2007), using rectangular elements each of them
with four nodes (Fig. 5). In the simulations of the laterolog, the mesh
3.1. Data uses with 87 nodes in the radial direction and 251 nodes in the axial
direction, comprising a total of N = 21837 nodes and 21500 elements.
We used an open borehole dataset (A1) from a carbonate reservoir In the simulations of induction, the mesh uses with 51 nodes in the
drilled in the pre-salt section of the Santos Basin. The dataset includes radial direction and 301 nodes in the axial direction, comprising a total
geophysical well log data, such as caliper, gamma rays, sonic data, of N = 15351 nodes and 30000 elements.
resistivity, neutron values and density porosity values, as well as por- The induction logs’ responses were obtained by Cozzolino (2004)
osity core data obtained from the laboratory where the cores and the through the Maxwell equations in the frequency domain:
measurements (Fig. 3).
One of the most important factors in this type of modeling, which is ×E = µ H + Jm, (8)
sometimes neglected, is the contrast between the invaded zone (Rxo)
and formation resistivities (Rt). Induction measurements respond more ×H =( µ ) E+ Je, (9)
easily in conductive formations (Singer and Barber, 1988.), while la-
terolog measurements respond betterin resistive formations (Crary
×H =( µ ) E+ Je, (10)
et al., 2001).
(µH ) = 0, (11)

3.2. The forward problem where Jm and Je represent the magnetic and electric current densities,
respectively, and is the permissivity of the medium (rock containing
To perform the numerical simulations, the finite element (FE) ap- fluids). Since the source in question is usually considered a ring of
proach for 2.5 was used, which accurately accounts for the hetero- current (or magnetic dipole), the first two equations (Equations (8) and
geneities in a horizontally stratified medium (Fig. 4a). This approach (9)) are expressed in cylindrical coordinates; therefore,
works like a host rock for the heterogeneity being studied, which in this Ez E
case is the borehole itself and the mud invasion (Fig. 4b). For this r = µr Hr + rJm,r
z (12)
reason, it was necessary to first calculate the response of this 1D host
rock when using the responses of laterolog and induction tools Er Ez
r = µH + Jm,
(Anderson, 2001). z r (13)
Subsequently, we simulated models with a cylindrical invasion
around the wall using a cylindrical geometry or 2.5D model (Fig. 4c) set Er
(rE ) = µr Hz + rJm,z
in a horizontally stratified medium as the host rock. The response of the z (14)
host rock obeys the boundary conditions (considered as the primary Hz H
field), and the response of heterogeneities is modeled with the FE r =( ) Er + rJm,r
z (15)
technique (considered as the secondary field).
Hz Hz
r = ( ) Ef + rJe,
3.2.1. Electromagnetic modeling z r (16)
The semi-analytical solution of the electric potential uses Maxwell's Hr
equations (Equations (1)–(4)), which relate the electric field to the (rH ) =( )r E f + rJe,z
z (17)
electric potential generated by a stationary point current source as
follows: The densities of the electric current are disregarded if the source is
of the magnetic dipole type; keeping only Jm (magnetic transverse
×E =0 (1) polarization mode – TM). Otherwise, we disregard the magnetic
sources, retaining only the electric, Je (transverse electric polarization
j = E, (2) mode – TE). Due to the symmetry of the source, there are several null
components in the TE mode, Jm, r = Jm, = Jm, z = Je, r = Je, z and also
E = V (3) Er = Ez = H = 0 . In this way, the relations derived from the laws of

218
J.L.S.d. Costa, et al. Journal of Petroleum Science and Engineering 178 (2019) 216–226

Fig. 3. Data set used in the study, consisting of well logs (curves) and porosity and permeability laboratory measurements (red dots). (For interpretation of the
references to color in this figure legend, the reader is referred to the Web version of this article.)

Faraday and Ampère are reduced to the following set of equations:

E
= µrHr
z (18)

(wE ) = µr Hz
r (19)

Hr Hz
r = ( )E + Je, .
z r (20)
Fig. 4. Host geological media with three layers (a), 2D (b) and 2.5D mud in- Substituting the expressions of Hr and Hz and Equations (18) and
vasions (c). (19) in Equation (20), we have the following differential equation:

1 E 1 1
+ (rE ) = ( ) E + Je,
z µ z r µr r r (21)

219
J.L.S.d. Costa, et al. Journal of Petroleum Science and Engineering 178 (2019) 216–226

where Wi is the weight function. In the Bubnov–Galerkin method, the

weight functions Wi are equal to the base functions Ni used in the ap-
proximation of U:
Wj = Nj, i = 1, 2, …, n, (28)
and
n
Uˆ = Ni Uˆi
i=1 (29)
The functions Ni and Nj are generally chosen according to the geo-
metry of the element utilized to define the region. Substituting
Equations (28) and (29) into Equation (27), the following is obtained:
Fig. 5. Global coordinate system (r z ) and local system (ƞ; ξ) of a generic ele-
ment. The local system has origin in the centroid of each element and of the n
mesh (Cozzolino, 2004). L Ni Uˆi f Ni d =0 j = 1, 2, …, n,
i=1 (30)
As is commonly assumed, the magnetic permeability and the per- Considering the linear problem, by virtue of the operator L, equa-
missivity of the rocks are considered to be constant and equal to their tion (30) leads to:
10 7H
values in a vacuum (µ= 4 × m e = 36 × 10 9F / m ); therefore,
1
n
2E
1 [Ni Li Nj d ] Uˆ LNi d =0 j = 1, 2, …, n,
+ (rE ) + k2E = µJe, i=1 (31)
z r r r (22)
The solution of Equation (31) is simple and reduces to a system of n
This is the non-homogeneous Helmholtz differential equation,
unknowns in the form of [K] = f, where [K] is called a stiffness matrix
where k2 = µ( ) is the square of the complex wave number and
of the system and represents a composition of solutions in each element.
Je, , the source term, is defined by
The coefficient Kij is given by:
INT a (r rT ) (z zT )
Je, =
r (23) Kij = Nj LNi d
(32)
which represents the current density of NT rings of radius rT traveled by
a harmonic current of intensity I and positioned at z = zT . The ex- The Ni and Nj are defined in, respectively, 1 and 2 (parts of Ω)
pression is identical to that obtained by Moran and Kunz (1962) and implies that they are compact support functions. In this case, the
others when introducing the electric potential vector coefficient results:
= (0, A , 0) = E /( µ) , i.e., Kij = Nj LNi d = Kij = Nj LNi d
2A
1 1 2 (33)
+ (rA ) + k2A = Je,
z r r r (24) Where 1 2 are is the intersection of the supports of both functions.
Note that, if the intersection is null, the coefficient Kij is auto-
3.2.2. The finite element technique matically null. To the extent that the supports of the base functions
To understand how the Finite Element (FE) technique is used, let us intersect, the matrix K results in a matrix with few nonzero elements.
consider a simple application of this numerical approach to a 2D This matrix is defined as banded, sparse and positive definite, even after
boundary value problem, in which the simulated area is approximated the inclusion of the boundary conditions. Their solution is easily ob-
by triangular elements and the excitation source is provided as either an tained by the Cholesky factorization algorithm (Zimmerman, 2004).
electric or magnetic field (Silva et al., 2007). The local heterogeneity is Applying this approach, it was possible to represent the invasion zone
set near the center of the model, with the electric and magnetic fields in 2.5D geometry, with the modeling achieved using an algorithm de-
satisfying the respective partial differential equations. Generally, the veloped by Cozzolino (2004), which made it possible to evaluate, using
upper and lower boundaries of the simulated region, Ω, are located electrical resistivity values, the influence that the invasion zone has on
quite far from the well and the invasion zone and, per the Dirichlet the responses of laterolog and induction logs.
boundary conditions, are specified as F = 0 (F being either the electric
or magnetic field). In addition, the Neumann boundary conditions (δF/ 3.2.2.1. Well log of eletroresistivity through FE. Similarly, to Cozzolino
δn = 0) are specified for the vertical boundaries (Lovell, 1993). To (2004, 2007) the resistivity logs to be analyzed in this work are
show this mathematical formulation, let us consider a generic equation obtained with the numerical solution from the application of the FE
that controls any physical phenomenon in the determined region, ex- to equation (6). First the medium is divided into several cells
pressing the partial differential equation in a matrix form (elements). It was chosen by the use of rectangular elements and the
approximate solution for each element obtained by
LU = f (25)
4
where U is the target or unknown function, L is a differential operator V je (r , z ) = U je Nj (r , z ) ,
(generally, partial derivatives) and f is a known function with one or j=1 (34)
more variables. If U is approximated by a function U , an error is in-
where is the solution at node j of element e and Nj is the base
U je
troduced into the process, which is known as residue and is given as
function, defined in terms of local coordinates ( , ).
follows:
1 1
= LUˆ f (26) N1 = (1 )(1 ), N2 = (1 + )(1 ),
4 4
The approximated solution is reached through the minimization of (35)
1 1
Equation (26): N3 = (1 + )(1+ ), N4 = (1 )(1+ ),
4 4
Wi (LUˆ f ) = 0, i= 1, 2, …, n, (27) where,

220
J.L.S.d. Costa, et al. Journal of Petroleum Science and Engineering 178 (2019) 216–226

2
= (r + rm ),
r
withrm = (r2 + r1)/2 e r = r2 r1, (36)

2
= (z + z m),
z
withzm = (z 4 + z1)/2 e z = z4 z1. (37)

The dots (r1; z1); (r2 ; z2 ); (r3 ; z 3 ) e (r4 ; z 4 ) they are the coordinates of
the nodes of the generic element shown in Fig. 5.
The application of the Galerkin method, where the weight functions
are equal to the base functions, wi = N1, gives:

1 V V
r (r, z) Ni 2 rdrdz + (r, z) Ni 2 rdrdz Fig. 6. Piston-type model (modified from Singer and Barber, 1988).
r r r r z
Io (r ) (z zT )
= Ni 2 rdrdz ,
2 r (38)

for i = 1, 2, …, N , with N equal to the number of knots in the mesh.

Then, using the properties of the Dirac function and integration by parts
the integrals to the left of the equality signal, gives

V Ni V Ni
Int C r + 2 drdz = Io Ni (o, zT ),
r r z z (39)

where Int C represents the contour integrals

Fig. 7. Representation of resistivity behavior in the piston-type model (mod-
V V
Int C r Ni 2 d + Int C r Ni 2 d , ified from Asquith and Gibson, 1982).
r z
1,3 1,4 (40)

with 1,3 representing the boundaries of the mesh at r = 0 and r = rmax 3.3.2. Annulus type
and 1,4 the boundaries at z = zmax and z = zmax , respectively. The invasion of the mud filtrate into the oil-water transition zone
Substituting equation (34) into equation (39) gives, after assembly of changes the distribution of the densities in the lower reservoir range. A
the elementary matrices large difference between the salinity of the water and the drilling mud
is responsible for the presence of a ring of low resistivity near the
N
Nj Ni Nj Ni wellbore.
Uj r + 2 drdz = Io Ni (o, zT ) + Int C
r r z z The annulus develops when the mud cake has not yet been con-
j=1 (41)
stituted and the rate of invasion is high (Fig. 8). Thus, most of the oil
where N is the total number of nodes in the mesh. will be displaced in its upper part in a short period of time, with si-
The expression can still be rewritten in the matrix form multaneous miscible displacement of the original brine by the mud
=¯ filtrate with higher resistivity (Carrasco and Carrasquilla, 2011).
L U = F¯ (42)
This low resistivity is related to the salinity of the formation water
In this way, the matrices are solved by the use of four-point and the mud filtrate, whereas the geometric characteristics of the an-
Gaussian quadrature and the system solved with triangularization and nulus are dependent on the petrophysical parameters of the reservoir.
retro-substitution routines for symmetric and banding systems. This ring (Fig. 9) suppresses the sensitivity of the tool used to
measure the resistivity formation on virgem zone. The modeling of the
3.3. The models process of invasion of filtered mud is the only way to estimate the sa-
turation of hydrocarbons in situ from electric logs. Additionally, later-
Regarding the invasion model, it will be considered a piston-type olog measurements are only marginally affected by the comportment of
geometry in the upper part of the reservoir, while the annulus model a low-resistivity ring (Torres - Verdín et al., 2004).
will be applied in the lower section of the reservoir, where it includes
the presence of the aquifer. In both cases, the presence of the mudcake
will not be considered in the modeling since it is an impermeable ele-
ment and is of little thickness.

3.3.1. Piston type (step profile)

In the piston type or step profile (Fig. 6), the mud filtrate in the
washed zone is idealized as a cylinder around the hole. This model does
not consider the presence of a transition zone between the resistivity of
the flushed or washed zone (Rxo) and the virgin or uninvaded zone
(Rt). The radius of the well to be modeled has a value of 0.1 m.
According to Crary et al. (2001), in reservoirs with high resistivity
values, both induction and laterolog logs are necessary for an accurate
characterization of the geological formations. There are many wells
drilled where R < Rt in the zones of interest containing hydrocarbons.
In these cases, the interpretation employing laterologs is more ade-
quate. This is also true for wells where Rmf ≤ Rw, ensuring that
Rxo ≤ Rt at all intervals (Fig. 7). Fig. 8. Annulus-type model (modified from Singer and Barber, 1988).

221
J.L.S.d. Costa, et al. Journal of Petroleum Science and Engineering 178 (2019) 216–226

simulated values with the actual data. As carbonates have many var-
iations in porosity, this parameter that was also taken as reference in
the separation of the layers for the same lithological group (Fig. 11).
Thus, Figs. 12 and 13 show the responses to the piston-type model,
using laterolog and induction algorithms, respectively. The green line
represents the LLS response, black the LLD, cyan the response of the
ILM and blue the ILD, and for comparison purposes, the real data used
as the reference for direct modeling are plotted in red. In the piston
model, without annulus present, an adjustment in the model was made
Fig. 9. Representation of resistivity behavior in the annulus model (modified by comparing the real data and curves, predicted and can be easily
from Asquith and Gibson, 1982). verified by the R-squared values of the laterolog (LLS and LLD), whose
values are 0.7 and 0.8, respectively (Fig. 12).
The response of the models verifies that laterologs are strictly in-
4. Results
fluenced by invasion zones. Looking to the LLS laterolog at top layers
(500–595 m), we can see an increasing on the apparent resistivities
The portion of the reservoir used for modeling between the depths
values, if compared with true resistivity, Rt. This shows a sensibility of
of 500 and 800 m, was divided into two sections. The first part corre-
the electrical focused tools when the invasion zone is less resistive than
sponds to the upper interval 150 m that satisfies the model of the piston
the non-invaded zone. Already in the lower layers (595–650 m), the
type, and the second part corresponds to the model of the annulus type.
laterolog LLD increases in relation to laterolog LLS, due to the low in-
The depth of the oil-water contact (OWC) was defined by analyzing the
vasion of the fluid in the zone.
dataset provided, and the OWC was detected in the depth of the 723 m,
In the log of induction, there was a smaller adjustment compared to
where observed a reduction of the values resistivity, consequently, an
the laterolog; this occurs mainly in the sections where the resistivities
increase in water saturation, besides the separation between logs of the
are higher, as for example at depths ranges from 500 to 508 m and
porosity (density and neutron) (Fig. 10).
520–538 m. This adjustment occurs because the induction log has a
In the piston model, the resistivities vary from 1 to 2000 Ω m and
better response in areas of low resistivity, as verified by the R-squared
divided into 36 layers. In each layer, invasion of the drilling fluid was
value in the ILM induction being lower than that in the LLS laterolog
simulated. The depths of invasion depths vary from 0.1 to 0.45 m re-
(0.6 and 0.7, respectively) (Fig. 13). In these models is included the
presented by the color brown and color green for the formation re-
invasion of the drilling fluid according to the porosity of each layer. The
sistivity (Rt). These invasion depths were scaled based on the value of
depth on the vertical axis is referential.
the porosities of the layers and the adjustment of the responses of the
The annulus model has resistivity values ranging from 1 to

Fig. 10. Piston and annulus models (upper and lower left, respectively) and the processed dataset (right) of the stretch used in the modeling.

222
J.L.S.d. Costa, et al. Journal of Petroleum Science and Engineering 178 (2019) 216–226

Fig. 11. Distribution of layers in the piston-type model; the invasion is shown in Fig. 12. -Synthetic logs of LLS (green) and LLD (black) for the piston-type
brown and the virgin formation in green. (For interpretation of the references to model. (For interpretation of the references to color in this figure legend, the
color in this figure legend, the reader is referred to the Web version of this reader is referred to the Web version of this article.)
article.)

1000 Ω m over 48 layers. In the lower layers, the modeling includes the resistivities on log acquisition, mainly in the vicinity of the annulus (at
transition zone (between 737.5 and 777.5 m), the oil-water contact (at 735 m). Obviously, influence will not appear in the upper layers, as
777.5 m), the water zone (in blue at depth of 777.5–800 m), a washed occurred in the section of the piston model.
zone (brown) and, later, the annulus presence (blue zone between Fig. 16 shows the ILM response in cyan, the ILD in blue and the
737.5 and 777.5 m). In this case, the shape of the annulus is not regular actual data in red and used as the reference for the forward modeling.
and exhibits displacement in the vertical direction (Fig. 14). The aquifer The annulus model induction responses reveal that there was a better fit
represented here is 27.5 m thick. at the lower resistivities, except where the annulus is present, precisely
The responses to the annulus-type model are shown in Figs. 15 and because it affects the responses of the records, visible in the section
16, respectively. In Fig. 15, the green line represents the LLS curve and between the depths of approximately 740–760 m. Thus, the annulus has
in black represents the LLD curve. In the laterolog response (Fig. 15), it a strong influence on the responses of the logs, requiring a more careful
is possible to verify that when the resistivity decreases (depth interval adjustment in the modeling such as that presented in the present work.
between 760 and 770 m), there is a separation of the curves due to the In these induction responses for the annulus model, the R-squared re-
difficulty in reading of the laterolog in the regions of lower resistivities, sults were quite satisfactory (both 0.8) compared to those of the piston-
which can be seen in the fact that the values of R-squared (of LLS and type induction registers.
LLD) in the annulus model are smaller than in the piston model.
The LLS and LLD logs illustrate the influence of adjacent layers

223
J.L.S.d. Costa, et al. Journal of Petroleum Science and Engineering 178 (2019) 216–226

Fig. 13. Synthetic logs of ILM (cyan) and ILD (blue) for the piston-type model.
(For interpretation of the references to color in this figure legend, the reader is Fig. 14. Distribution of layers in the annulus-type model after a good fit with
referred to the Web version of this article.) forward modeling. The invasion is shown in brown, the aquifer in blue, and the
virgin formation in green. (For interpretation of the references to color in this
5. Conclusions figure legend, the reader is referred to the Web version of this article.)

This work shows a methodology for formation evaluation by using responses. Borehole correction was computed and applied to the logs
forward geoeletric modeling for two different geological reservoir en- and turns out to be most necessary when the formation resistivity is
vironments: the piston and annulus cases. Both cases influence in the lower than the mud resistivity.
responses of electrical and induction logging altering the desired re- The modeling of carbonate reservoirs is a complex task because in
sponses and given error in interpretation (noise effect). To solve this this type of reservoir, there is highly variable porosity, which resulted
problems, mathematical algorithms were used for galvanic (LLS and in the division of several zones of invasion into the same formation and
LLD) and electromagnetic (ILM and ILD) methods. In that sense, real suited for the piston modeling. In the annulus, the different zones of
data for complex geology (carbonate) was used. These records were invasion were considered in conjunction with the ring of low resistivity,
obtained from a pre-salt carbonate reservoir in the Santos Basin, Brazil. which allowed to model the annulus effect satisfactorily. Other models
The simulation of the responses of the resistive log of shallow (LLS), can be realized assuming other kinds of invasion models as ramp, slope
deep (LLD) and resistive log of induction was performed with sa- or smoothed transition zones and this methodology can be applied in
tisfactory results in modeling the different types of invasion. other kind of lithology. The main advantage of this methodology is its
The results show that it is possible to model with FEM and obtain a flexibility to model several 2.5D geological environments and its dis-
good fit using the piston model. In the characterization of the annulus, advantage is the need to have a prior knowledge of the environment to
the induction curves confirm its strong influence on the well log

224
J.L.S.d. Costa, et al. Journal of Petroleum Science and Engineering 178 (2019) 216–226

Fig. 15. Synthetic logs of LLS (green) and LLD (black for the annulus-type Fig. 16. Synthetic logs of ILM (cyan) and ILD (blue) for the annulus-type model.
model. (For interpretation of the references to color in this figure legend, the (For interpretation of the references to color in this figure legend, the reader is
reader is referred to the Web version of this article.) referred to the Web version of this article.)

be modeled through a preliminary interpretation and the successive References

adjustments to be made.
The relationships between induction and laterolog assume different Akinsete, O., Adekoya, D., 2016. Effects of Mud Filtrate Invasion on Well Log
values for different invasion radii and resistivities and they present distinct Measurements. Society of Petroleum Engineershttps://doi.org/10.2118/184308-MS.
response before the oil and water zones. These characteristics are very Anderson, B., 2001. Modeling and Inversion Methods for the Interpretation of Resistivity
Logging Tool Response. Western Connecticut State University, USA 386 pp.
important for the well log analysts as well as for log interpretation. Asquith, G., Gibson, C., 1982. Basic Well. Log Analysis for Geologists, Methods in.
Exploration Series, AAPG, Tulsa, Oklahoma USA.
Bruhn, C., Gomes, J., Lunches Jr., C., Johann, P., 2003. Campos Basin: reservoir char-
Acknowledgments acterization and management-historical overview and future challenges. In: Offshore
Technology Conference. OTC 15220, . https://doi.org/10.4043/15220-MS.
The authors acknowledge UENF for the computational support, SPE Carrasco, A., Carrasquilla, A., 2011. Three-dimensional numerical modeling of the re-
sponse of the induction profile in the presence of different geometries of the invasion
for providing articles for bibliographic consultation, PRH-226
zone. In: 12th International Congress of the Brazilian Geophysical Society, (In
Petrobras/UENF for the scholarship and Petrogal for the dataset. Portuguese). https://doi.org/10.1190/sbgf2011-430.
Crary, S., Jacobsen, S., Rasmus, J., Spaeth, R., 2001. Effect of resistive invasion on re-
sistivity logs. In: SPE Annual Technical Conference and Exhibition. Society of
Appendix A. Supplementary data Petroleum Engineers. https://doi.org/10.2118/71708-MS.
Cozzolino, K., 2004. Studies of the Transient Effects of Temperature and Invasion on Well
Supplementary data to this article can be found online at https:// Logs. Doctoral teses. UFRJ, Rio de Janeiro 146 pp. (In Portuguese).
Cozzolino, K., Silva, J.C., 2007. Synthetic focusing and simulation of dual laterolog tool in
doi.org/10.1016/j.petrol.2019.03.038.

225
J.L.S.d. Costa, et al. Journal of Petroleum Science and Engineering 178 (2019) 216–226

axisymmetric subsurface models. J. Appl. Geophys. 61 (2), 102–110. https://doi.org/ v32i2.453.

10.1016/j.jappgeo.2006.05.001. Riccomini, C., Sant, L., Colombo, C., Tassinari, G., 2012. Pre-salt: geology and explora-
El-Wazeer, F., Haggag, M., 1999. Formation evaluation masked by mud invasion. In: tion. USP Magazine 95, 33–42 (In Portuguese).
Middle East Oil Show and Conference, 20-23 February, Bahrain, SPE 53152, 6 pp. Salazar, J., Torres-Verdin, C., 2009. Quantitative comparison of processes of oil and
https://doi.org/10.2118/53152-MS. water-based mud filtrate invasion and corresponding effects on wellbore SER mea-
Li, C., Yang, W., Shen, J., 2003. Three-component Induction Logging Inhomogeneous surements. Geophysics 74 (1), E57–E73.
Modeling Using Finite Difference. SEG Technical Program Expanded Abstracts 2003. Silva, J., Carrasquilla, A., Priimenko, V., 2007. The two-dimensional GPR modeling for
https://doi.org/10.1190/1.1817989. near-surface investigation using the Dirichlet - Neumann boundary condition com-
Lovell, J., 1993. Finite Element Methods in Resistivity Logging. PhD. Thesis. Delft bination. Sib. J. Numer. Math. 10 (4), 385–399 Siberian Branch of Russian Academy
University of Technology, The Netherlands 206 pp. of Science, Novosibirsk.
Moran, J., Kunz, K., 1962. Basic theory of induction logging and application to study of Singer, J., Barber, T., 1988. The effect of transition zones on the response of induction
two-coil sondes. Geophysics 27 (6), 829–858. https://doi.org/10.1190/1.1439108. logs. In: SPWLA Symposium, 29th, San Antonio, Tex., 1988, Transactions, pp.
Papaterra, G., 2010. Pre-salt: Geological Conceptualization about on a New Exploratory L1–L13.
Frontier in Brazil. Masters dissertation. Institute of Geosciences, Federal University Torres-Verdin, C., George, B., Delshad, M., Sigal, R., Zouioueche, F., Anderson, B., 2004.
of Rio de Janeiro (In Portuguese). Assessment of in-situ hydrocarbon saturation in the presence of deep invasion and
Podberezhny, M., Polushkin, S., Marketz, F., Makarov, A., Sviridov, M., Mosin, A., 2016. highly saline connate water. Petrophysics 45 (2), 141–156.
Determination of Invaded Zone Properties from Time-Lapse Induction Logging Zeng, S., Chen, F., Li, D., Chen, J., Chen, J., 2017. 2.5D Finite difference modeling of
Measurements Based on Invasion Simulation and Electromagnetic Modeling. SPE- directional electromagnetic wave propagation resistivity logging in fully anisotropic
183103-MS. formation. SEG International exposition and 87th annual meeting.
Ribeiro, M., Carrasquilla, A., 2014. A new approach to calculate mud invasion in re- Zimmerman, W., 2004. Process, Modeling and Simulation with Finite Element Methods.
servoirs using well logs. Braz. J. Genet. 32, 215–223. https://doi.org/10.22564/rbgf. World Scientific, New Jersey 395 pp.

226