Professional Documents
Culture Documents
3D simulation of transient electromagnetic field for geosteering horizontal wells
3D simulation of transient electromagnetic field for geosteering horizontal wells
3D simulation of transient electromagnetic field for geosteering horizontal wells
www.elsevier.com/locate/rgg
Abstract
The article discusses numerical simulation of the transient electromagnetic field. The field source is an induction coil. We consider the
situation when a logging tool is in a horizontal well in a medium with horizontal and vertical boundaries. The specific features of this problem
are the metallic mandrel of the tool, 3D geometry, and distant boundaries. The method of separate computation of normal and anomalous
fields is proposed. The finite element method is used for spatial approximation of the field, and the implicit finite-difference scheme is used
for approximation of the field in time. Correctness and advantages of the method are shown. Some numerical results are demonstrated. The
method proposed can be used when designing tools for geosteering.
© 2011, V.S. Sobolev IGM, Siberian Branch of the RAS. Published by Elsevier B.V. All rights reserved.
* Corresponding author.
E-mail address: Elizaveta.Onegova@bakerhughes.com (E.V. Onegova)
1068-7971/$ - see front matter D 201 1, V . S. S o bolev IGM, Siberian Branch of the RAS. Published by Elsevier B.V. All rights reserved.
doi:10.1016/j.rgg.2011.06+.005
Downloaded from http://pubs.geoscienceworld.org/rgg/article-pdf/52/7/725/5224841/v52i7_s1068797111001088.pdf
by CNOOC user
726 E.V. Onegova and M.I. Epov / Russian Geology and Geophysics 52 (2011) 725–729
1 ∂E0 ∂J
rot rot E0 + σ0 =− , (2)
µ ∂t ∂t
∂J
−∫ ⋅ ψ dΩ ∀ ψ ∈ H0 (rot, Ω0),
∂t
Ω0
where
H0 (rot, Ω0) = υ ∈ (L2(Ω0))3 rot υ ∈ (L2(Ω0))3,
υ × n|∂Ω = 0
0
is the space of test functions, ψ = (0, ψϕ, 0)T.
The variational equation for the 3D problem has the
following form:
1 ∂E+
∫ µ rot E+ ⋅ rot ψ dΩ + ∫ σ ∂t
⋅ ψ dΩ = Fig. 2. The electromotive force in the homogeneous medium. 1, with pipe; 2, no
pipe.
Ω Ω
∂E0
(σ0 − σ) ∫ ⋅ ψ dΩ ∀ ψ ∈ H0 (rot, Ω), scheme developed, let us consider some computations. Fig-
∂t ure 2 shows the electromotive force induced in the receiver
Ω
coil as a function of time. Two results are presented: with pipe
where ψ = (ψx, ψy, ψz)T. and without pipe. Formation resistivity was 100 Ohm⋅m; steel
To approximate the time derivative of the solution, the pipe resistivity was 7.14 × 10–7 Ohm⋅m; its magnetic perme-
completely implicit three-layer scheme is used (Marchuk ability was 100µ0 (µ0 = 4π × 10−7H/m); the outer and inner
1980) radii of the pipe were 0.04 and 0.07 m, respectively; coil radius
∂E was 0.085 m; the distance between the transmitter coil and
≈ γ0 Ej − γ1 Ej−1 + γ2 Ej−2,
∂t t = t receiver coil was 5 m; the electric current was 5 A; the number
j of loops in the transmitter coil was 100; the number of loops
∆01 + ∆02 ∆02 ∆01 in the receiver coil was 10.
where γ0 = , γ1 = , γ12 = , ∆12 =
∆01 ∆02 ∆12 ∆01 ∆12 ∆02 In the graph, the pipe affects the transient process greatly,
tj−1 − tj−2, ∆02 = tj − tj−2, ∆01 = tj − tj−1, ti is the time layer, and both quantitatively and qualitatively. In the range between
5 × 10−4 and 5 × 10−2 s, the signal from the pipe decreases
Ei is the solution at the ith layer.
quite slowly, and one can even observe some increase at the
For the finite-element discretization, rectangles with piece-
wise-bilinear basis functions are used for the 2D problem, and time of 2 × 10−2 s. After the time of 0.1 s, the signal from the
hexahedrons with edge basis functions of the first order, for pipe decreases greatly, and after the time of 2 s, it attenuates
the 3D problem (Nedelec, 1980). as t−5 / 2. Physically, this means there are no more currents in
the pipe, and the other currents are far from the initial source
that the situation becomes equivalent to that of a magnetic
Testing dipole in the homogeneous medium. On the other hand, the
signal level at these times is much lower than the measured
The results of 3D modeling using the bedded medium were values. Generally, it should be noted that the computational
compared to known solutions (Tabarovsky and Sokolov, scheme can model the signal in the dynamic range around
1982). The magnetic dipole was used as the field source. In 200 dB.
the 3D formulation, a coil with the radius of 0.01 m was used
as the field source. Relative discrepancy between solutions did
not exceed 1% in the time range between 10–7 and 10–2 s. Numerical experiments
Besides, the results of solving the 2D problem were
compared with computations obtained using FEMAX (Be- Let us consider a geoelectrical model typical of the west
spalov, 2002) and COMSOL (Pryor, 2009). The transient Siberian oil and gas province. The horizontal interval of the
process in the homogeneous medium with a metal pipe of a well is in the oil-saturated reservoir (ρ1 = 15 Ohm⋅m, Fig. 3).
finite length was modeled. A coil coaxial with the pipe was
The reservoir has a clay cap (ρ2 = 4 Ohm⋅m). Under the
used as the field source. Pipes with diverse electric conduc-
tivity and magnetic permeability (magnetic steel, nonmagnetic reservoir, there is a water-saturated interval (ρ3 = 8 Ohm⋅m).
steel, and copper) and diverse containing media were consid- In front of the tool, orthogonally with respect to the well, there
ered. Average relative discrepancy between solutions obtained is a clay layer (ρ4 = 2.0–3.5 Ohm⋅m). The drilling mud with
using three different methods was 2–3% in the time range a high clay content has electric resistivity of 2 Ohm⋅m.
between 10–7 and 10–2 s. Resistivity of the pipe made of nonmagnetic steel is
To understand the degree of the impact the pipe might have 7.14 × 10−7 Ohm⋅m. The outer and inner radii of the pipe were
upon the signal and show the possibilities of the computational 0.04 and 0.07 m, respectively; coil radius was 0.085 m for
Fig. 4. Cross section via the xy plane with the hexahedron grid.
both transmitter coil (T) and receiver coil (R); the well radius
was 0.108 m. The electric current in the transmitter coil was
5 A; the number of loops in the transmitter coil was 100; the 0.3% when the grid of 606 thousand finite elements is used.
number of loops in the receiver coil was 10. The object of Without dividing the total field into the normal field and the
measurements was the electromotive force induced in the anomalous field, this error reached 2% even though the grid
receiver coil. comprised 1035 thousand finite elements.
Figure 4 shows a grid of hexahedrons for the computational Time dependence of the modulus of the anomalous com-
domain. Inside the well, the grid is radial in the xy plane, and ponent of the electromotive force for different resistivities of
in the neighborhood of the horizontal boundaries, it becomes layers ρ2, ρ3, and ρ4 are shown in Fig. 5. The time interval
rectangular. Because the computational domain is symmetrical shown is the one where the curves are especially dissimilar.
with respect to the axis y = 0, the problem was solved in a We consider anomalous fields, because the relative difference
half of the domain only (y ≥ 0), and the homogeneous Dirichlet between the total fields in different models does not exceed
boundary conditions were applied at the line of symmetry. 9%. The magenta curve corresponds to the model with the top
As it was mentioned, the total field was divided into the layer only (ρ4 = ρ1, ρ3 = ρ1); green, with the bottom layer
normal and anomalous fields, and these components were only; black, with horizontal layers (no vertical interlayering);
computed separately. The medium for the normal component red and blue are the curves corresponding to the models with
consisted of the pipe and the well penetrating the homogene- all boundaries with ρ4 = 3.5 Ohm⋅m and 2.0 Ohm⋅m, respec-
ous formation (ρ1 = 15 Ohm⋅m). The normal field was com- tively. The top horizontal layer gives a greater contribution to
puted with the relative error of 0.1%. This estimate was made the signal than the bottom one, as it is closer to the source
on the basis of comparing two solutions obtained at two nested and has higher conductivity. All transient curves differ with
grids. The relative error of computing the anomalous field, respect to the moment of time when the sign change takes
on the average, was about 2%. It should be noted that the place.
anomalous component does not exceed 10% of the total field. Figure 6 shows the ratio between the modules of the
Thus, the error of computing the total field does not exceed anomalous component of the electromotive force in the model
with and without the vertical boundary, in the presence of the
horizontal boundaries. Results are given for different distances
Fig. 5. Anomalous component of emf. 1, top layer only (ρ4 = ρ1, ρ3 = ρ1,
h1 = 4 m); 2, bottom layer only (ρ4 = ρ1, ρ2 = ρ1, h2 = 9 m); 3, horizontal layers Fig. 6. Anomalous component of emf relative to the signal in model with
only (ρ4 = ρ1, h1 = 4 m, h2 = 9 m); 4, with all boundaries (ρ4 = 3.5 Ohm⋅m, ∆ = no vertical boundary for different distances to the vertical boundary,
11 m); 5, with all boundaries (ρ4 = 2.0 Ohm⋅m, ∆ = 11 m) ρ4 = 2.0 Ohm⋅m.
Conclusions
Fig. 7. Anomalous component of emf relative to the signal in model with no References
horizontal boundary for different distances to the horizontal boundaries,
ρ4 = 2.0 Ohm⋅m. Anderson, B., Chew, W.C., 1989. Transient response of some borehole
mandrel tools. Geophys. 54 (2), 216–224.
Bespalov, A., 2002. FEMAX—Software for simulation of magnetic induction
∆. Two time intervals can be set apart, based on the curve tools in vertical wells. SEG Expanded Abstracts 21, 708–711.
behavior: less and greater than 10–5 s. For the first interval, Epov, M.I., Morozova G.M., Antonov, E.Yu., 2002. Electromagnetic Method
the relative value of the electromotive force reaches its of Finding Faults in Casing of Oil and Gas Wellbores (Theoretical Basis
and Method) [in Russian]. NITs OIGGM, SB RAS Publishing House,
maximum, which corresponds to the sign switch of its
Novosibirsk.
anomalous component. For the second interval, the electromo- Kaufman, A.A., Sokolov, V.P., 1972. Theory of Inductive Logging Using
tive force in the model with a vertical boundary differs from Transient Electromagnetic Method [in Russian]. Nauka, Siberian Branch,
the corresponding model with no vertical boundary by the Novosibirsk.
factor of 1.5 at ∆ = 11 m, by the factor of 1.2 at ∆ = 15 m, Marchuk, G.I., 1980. Methods of Computational Mathematics [in Russian].
Nauka, Moscow.
and by the factor of less than 1.1 at ∆ = 20 m. That is, Nedelec, J.C., 1980. Mixed finite elements in R3. Numer. Math. 35, 315–341.
sensitivity to the boundary is observed when it is located Onegova, E.V., 2010. Effect of multicoil electromagnetic tool eccentricity on
11–15 m away from the source. The vertical boundary at 20 m measured signals. Russian Geology and Geophysics (Geologiya i Geofiz-
does not seem to have any effect on the signal. Thus, there ika) 51 (4), 423–427 (540–545).
Plyusnin, M.I., Vilge, B.I., 1969. Foundations of inductive logging using
are limitations on determining the distance to a remote transient electromagnetic method. Izvestiya Vysshikh Uchebnykh Zave-
boundary ahead of the tool. denii. Geologiya i Razvedka, No. 5, 158–165.
Let us consider sensitivity of the signal to the horizontal Potapov, A.P., Kneller, L.E., 2000. Mathematical modeling and interpretation
boundaries. Figure 7 shows the ratio between the modules of of borehole pulsed electromagnetic wall thickness measurements. Geofiz-
ika, No. 5, 27–30.
the anomalous component of the electromotive force in the
Pryor, R., 2009. Multiphysics Modeling Using COMSOL: A First Principles
model with and without the horizontal boundaries, in the Approach. Jones & Bartlett Learning, Sudbury.
presence of the vertical boundary. As one can see, the red and Sidorov, V.A., 1996. Borehole fault analyzers and wall thickness measurement
black curves, corresponding to a fixed distance to the top tools for logging multistring boreholes. NTV Karotazhnik, No. 24,
boundary h1 and different distances to the bottom boundary pp. 83–94.
Tabarovsky, L.A., Sokolov, V.P., 1982. Code for computing transient field
h2, agree quite well. This means that sensitivity to the lower of dipole sources in horizontally layered medium (ALEKS), in: Elektro-
boundary is weak. At the same time, the tool retains sensitivity magnetic Methods in Geophysics [in Russian]. IGiG SO AN SSSR,
to the top boundary located at h1 = 20 m. Novosibirsk, pp. 57–77.