Vous êtes sur la page 1sur 6

Uncertainty in geosteering and interpretation of horizontal

wells The necessity for constraints and geometric models


John Zhou 1

The Leading Edge 2015.34:492-499.


Downloaded from library.seg.org by Piyush Pasari on 05/31/15. For personal use only.

Abstract
While geosteering or interpreting horizontal wells, engineers are constantly faced with the challenge of correctly dening the well position with respect to the target reservoir and
other geologic markers from an often incomplete set of data.
Traditional petrophysical evaluation can be performed only
after the geometric relationship between the well and the target reservoir is interpreted. Measurements made while drilling
range from a simple gamma-ray (GR) log to resistivity, apparent density and neutron porosities, borehole images, azimuthal deep resistivity, and beyond. With one or a combination of
these logs and other peripheral data such as well logs from oset/pilot wells and seismic images, the question remains as to
how accurate the interpreted relationship is between the well
trajectory and the target reservoir. In a vertical well, an explicit
geometric model of the formation is often not necessary. There is
a 1D layered-earth model, either consciously or subconsciously,
in peoples minds to aid the interpretation. On the other hand,
because of the lateral variation of properties and thus the geologic complexity encountered in a horizontal well, it becomes
critical to explicitly construct a formation cross-sectional model,
validated in terms of its correctness and uniqueness against the
available measurements and other known information. To properly construct, update, and validate the formation model, the
completeness of information for solving the geometric relationship between the well and the target reservoir is investigated. It
is found that a data set with logs from the horizontal well alone
is not adequate for the task. Proper constraints based on depositional environments must be introduced. Well logs from oset/
pilot wells dene the formation sequence and add the high-resolution details to the interpretation. Three-dimensional reservoir models from seismic images guide the interpretation of the
geologic trends along the well trajectory and the extrapolation of
the model property into the volume that is not sensitive to well
logs. Modeling the tool responses and understanding the underlying response characteristics help to mitigate the interpretation
uncertainty by extracting more geometric information from the
physical measurements. One should also be aware of the possibility that the interpretation might not be unique even though a
model ts the data.

resistivity, density/neutron, and so forth to estimate petrophysical properties such as porosity, water saturation, and permeability along the well path and nishes with reserves evaluation
and completion/production recommendations. On the other
hand, in a horizontal well, the rst priority is to understand the
whereabouts of the well trajectory versus the target reservoir.
The geometric interpretation answers the questions of whether
the well is inside, above, or below the target and the distance
to boundaries (structural and/or uid contacts), wherever feasible. It is thus imperative to understand the limitation of the
available information for that purpose. It is equally important
to introduce proper constraints to mitigate the lack of information from well logs.
Well-placement decisions when drilling horizontal wells
also require that the interpretation be done in real time. Therefore, visualization, processing, and interpretation of well logs
require an ecient approach. This article reviews the workow
and illustrates proper ways of constraining the interpretation.
Before starting the analyses, let us dene the data-visualization
convention used here. Data and formation cross section will be presented on the attened vertical curtain dened by the well trajectory to facilitate the derivation of a plausible and realistic formation
model. Figure 1a shows the 3D view of the well path and the vertical curtain (the curved yellow surface) that the well denes. Figure
1b displays the 2D cross section attened from the vertical curtain
shown in Figure 1a. Well logs along the well path are projected to
true vertical depth (TVD) and lateral distance (LD) (Figure 1b).

Introduction
Well-log interpretation in horizontal wells still faces challenges only a limited amount of information is acquired
because of cost constraints and/or availability of technology.
Furthermore, well logs from a horizontal well have dierent response characteristics than those from a vertical well, even inside the same formation (Zhou, 2007, 2008).
Traditional interpretation, performed mostly in vertical
wells, uses physical measurements such as gamma ray (GR),
1

492

Maxwell Dynamics, Inc.

THE LEADING EDGE

Figure 1. Convention used in visualizing well logs acquired in a horizontal well. (a) Well and vertical curtain dened by the well in three
dimensions. (b) The attened vertical curtain shown in a 2D crosssectional display, along with its projected logs.
http://dx.doi.org/10.1190/tle34050496.1.

May 2015

Special Section: Well geosteering

The Leading Edge 2015.34:492-499.


Downloaded from library.seg.org by Piyush Pasari on 05/31/15. For personal use only.

typically with the borehole penetrating each layer only once. The
measurements at each depth point are sensitive to the physical
property around that point and reect the true value around
the point, assuming that environmental eects are either negligible or correctable. These measurements are extrapolated laterally in ones mind, along with the consideration of well logs
available in neighboring wells.
In a horizontal-well environment, far fewer measurements
(only GR or GR plus resistivity, and rarely with porosities and
others) are acquired in the majority of the wells while drilling.
The geometric complexity one faces, on the other hand, is orders
of magnitude greater (Figure 2).
To further complicate the task, compared with the reservoir dimension, the depths of investigation (DOI) are relatively shallow (within the dashed lines in Figure 2d), ranging
Completeness of measurements
from borehole wall to less than a few meters into the forFor any interpretation tasks, the rst question is whether the mation. In other words, any formation volume a few meters
data acquired contain the information sucient for solving the from the borehole is not investigated by logging tools, and
problem posed or the completeness of the data.
its properties cannot be determined solely based on available
To interpret a horizontal well, the problem posed is to dene well logs.
the geometric relationship between the well and the target resAnother fact about the lack of the needed geometric inervoir. Let us start with the well data-acquisition and interpre- formation (e.g., up and down) in a horizontal well is that contation practices to examine the dierences between vertical and ventional measurements (e.g., GR and resistivity logs) are
horizontal well environments.
omnidirectional or the measurements cannot tell whether the
In a typical vertical well, wireline triple-combo measure- contribution is from above or below the sensors. For example, if
ments (formation density, neutron-porosity, deep/intermedi- the wellbore exits a target sand accidentally, one cannot tell from
ate/shallow resistivity, natural gamma radiation, hole size, and omnidirectional data whether it exits from the upper or lower
uid temperature, all in a single logging pass) commonly are boundary, assuming that the physical properties of the formaacquired. The geometric model, subconscious in many peo- tion above and below are about the same.
ples minds, is more or less a 1D layered formation (Figure 2a),
To overcome omnidirectionality, borehole image logs
from GR or other physical parameters
thus are recommended highly, and in
some cases, they are absolutely necessary because of their geometric information content. New generations of
logging-while-drilling (LWD) tools
such as deep azimuthal propagation
tools might also help in dierentiating the directions. Nevertheless,
these sensors still are constrained by
their shallow depth of investigation.
Furthermore, because of their high
cost, image and/or other azimuthally sensitive measurements are implemented less commonly in practice.
When seismic images are available, they provide guidance on geologic trends and the approximate
location of the target. However, seismic data cannot give the precise position because of their relatively poor
spatial resolution and their uncertainFigure 2. Well and formation environments used by many well-log interpreters and geosteering engity in time-depth conversion.
neers. (a) Vertical well environment where a simple 1D model is often sucient to interpret measureThe above discussion emphasizes
ments. (b) Horizontal well with consideration of formation heterogeneity in lateral direction, a 2D
that
well logs (and other data) for well
model. (c) A more challenging condition in interpreting a horizontal well in which geologic events such
steering
or interpreting the horizontal
as faulting, pinch-outs, unconformities, and so forth might be present. (d) The volume of formation
sensed by well data is a small tubular space, indicated by the dashed lines along the wellbore.
well are likely to be incomplete. To sort

For simplicity, two types of wells are mentioned in this article, namely, the vertical well and the horizontal well. In this
article, vertical well refers to any well that is vertical or close to
vertical. Oset wells and/or pilot wells typically are vertical and
are used as reference to facilitate the interpretation of the horizontal well. In the category of horizontal well, we include wells
that are horizontal as well as those at high angle. The horizontal
well is the one to be interpreted.
Last, because of diculty with eld data releases, the examples used in this article are numerically constructed based
on real-world cases to illustrate and convey the points about
uncertainty in geosteering and interpreting horizontal wells
and the ways of reducing interpretation ambiguity through
geologic constraints.

Special Section: Well geosteering

May 2015

THE LEADING EDGE

493

the uncertainty caused by using an incomplete set of well data,


an interpretation model should be introduced, and proper geologic constraints have to be applied.

The Leading Edge 2015.34:492-499.


Downloaded from library.seg.org by Piyush Pasari on 05/31/15. For personal use only.

Necessity for an explicit formation model


In interpreting a vertical well, people seldom explicitly construct a formation model. In many cases, eyeball interpretation
is adequate for that purpose. For example, from the GR and resistivity logs in Figure 3a, one has in mind already a formation
model, illustrated in Figure 3b, without having to explicitly construct the model. The model consists of two parts, one for physical parameters (gamma ray and resistivity in red lines in this
case) and the other for geometry (the parallel-layer formation
on the right-hand side of Figure 3b). In short, people dont have
to build an explicit formation model to start the petrophysical
evaluation in vertical wells.
On the other hand, when a horizontal well is under consideration, an explicit formation model becomes a necessity. Figure 4a displays a set of GR and resistivity logs. The interpreted
model in Figure 4b consists of sand and shale layers, which
might be from a natural extension of the vertical-well interpretation practice in which layering is dened based on high and
low values in the logs. Varying the dip angle and thickness of the
formation layers, one might nd a series of models tting the
logs. The grayed-out area in Figure 4 indicates the volume that
is not sensitive to the logs.
Naturally, the question is whether the model in Figure 4b is
acceptable or consistent with available well logs. For those who
know the resistivity tool-response characteristics when a borehole crosses a layer interface at a high angle (Wu et al., 1991),
it becomes apparent that the model in Figure 4c is more appropriate than that in Figure 4b. The telltale sign is the so-called
resistivity horn.
Resistivity logs vary smoothly across a bed boundary when
the borehole is perpendicular to the bed boundary. The logs
have a high-resistivity horn anomaly if the borehole intersects
the boundary at an angle close to parallel, assuming the neighboring beds are also of high-resistivity contrast. The lack of
a horn anomaly at LD 715 ft indicates that the borehole is
closer to perpendicular to bedding than parallel to bedding.
Thus, the model in Figure 4c is more appropriate than that
in Figure 4b. The horn anomaly illustrated here serves as an
example, and often, there are many other factors (e.g., anisotropy) to be considered carefully.
If one can aord a borehole image log (e.g., GR image)
and/or an azimuthal propagation-resistivity image (Figure 5), the model in Figure 4c can be further validated or
conrmed. Through extrapolation away from the wellbore,
the model in Figure 4c can be made into a 2D cross section,
shown in Figure 5a.
Whatever model is derived, it also has to be veried to assess
whether the model is geologically feasible. Figure 5b depicts a
geologic cross section with a fault as what the model in Figure
5a might represent.
Even after the above steps, the model in Figure 5a is only
one of the likely realizations because any volume a few meters
from the borehole is not sensed by well logs.

494

THE LEADING EDGE

May 2015

Figure 3. Well logs acquired in a vertical well and the typical model
in the interpreters mind. (a) Well logs acquired in a vertical well. (b)
A typical model derived from the well logs in (a).

Figure 4. The need for a formation model consistent with the available measurements and geologic understanding. (a) Logs from a horizontal well. (b) One of the solution models. (c) Another formation
model more consistent with tool-response characteristics, although
still not necessarily unique. The grayed-out area indicates the volume
that is not sensitive to logs.

Special Section: Well geosteering

The Leading Edge 2015.34:492-499.


Downloaded from library.seg.org by Piyush Pasari on 05/31/15. For personal use only.

Figure 5. A formation model needs to be checked against the geologic


understanding when one interprets a horizontal well. (a) The interpreted
cross section is validated further if image and/or azimuthal propagationresistivity image are acquired. (b) Geologic realization corresponding to
the logs measured and the model interpreted in (a).
Figure 6. In most cases, the formation sequence observed in the oset

As emphasized above, in a horizontal well, the geologic


complexity along the well path becomes hard to comprehend
without a formation model and validation against the data. The
question now becomes, What formation model is a true or acceptable representation of the geology encountered?

well also is observed in the horizontal well after accounting for the
structural variation such as the up and down in true vertical depth.
(a) This shows good continuity and is relatively easy to correlate between
wells. (b) A more challenging environment for correlation.

Constraints based on depositional environments


One of the main constraints comes from the depositional
environment for sedimentary rocks. In the majority of sedimentary rocks, stratication or lateral continuity occurs at the time
of deposition. The depositional sequence is maintained mostly
over geologic time, although geologic activities might give rise
to more complex structures such as anticlines, faults, erosion and
unconformity surfaces, and so forth.
To understand the whereabouts of the well path versus an
expected geologic structure, one relies on correlating well logs
between the horizontal well and its oset/pilot wells, as illustrated in Figure 6. For good geologic continuity (Figure 6a),
it is relatively easy to correlate and to determine the horizontal well position. However, when missing layers (because of
pinch-outs or erosion surfaces, for example) and faults with
large vertical oset (larger than the DOI) are present, the uncertainty in determining the whereabouts of the horizontal
well increases (Figure 6b).
Let us illustrate a workow in deriving a geologic model
around and along the horizontal-well trajectory in question. When a horizontal well is designed and being drilled,

496

THE LEADING EDGE

May 2015

Figure 7. Geometric relationship among horizontal well, pilot well,


and oset well.

there are already good position controls from seismic images


and oset wells. Figure 7 shows the geometric relationship
among geologic surfaces and wells (horizontal, pilot, and oset wells). Pilot and oset wells oer high-resolution constraints, whereas surfaces will guide the correlation over a
large lateral extent. Intersecting lines between the 3D surfaces and the vertical curtain dened by the horizontal well
are referred to here as 3D markers.
Because of the close proximity, the pilot well and its well
logs are chosen to construct a formation model to guide the

Special Section: Well geosteering

The Leading Edge 2015.34:492-499.


Downloaded from library.seg.org by Piyush Pasari on 05/31/15. For personal use only.

interpretation in the horizontal well. Note that a pilot well


is not always available because of its added cost, and in that
case, one typically chooses the most representative oset well
to start with.
Figure 8 illustrates the well logs and model built from
the available well logs in the pilot well. In addition, Figure
8 shows the markers that indicate the target reservoir location and other signicant layers. These markers are named as
well markers, to dierentiate them from the 3D markers mentioned earlier.
The horizontal well should encounter the same or a similar
sequence of formation layers as observed in the pilot well, although inevitably at dierent vertical depths. Figure 9a plots the
GR and resistivity logs in the horizontal well (for clarity, only a
subset of logs is displayed). In Figure 9a, there are also the GR
log projected from the oset well and the 3D markers from Figure 7 for reference.
For interpretation, the formation model and well logs (Figure 8) from the pilot well are brought into the horizontal well
(Figure 9b) as the starting model and one of the constraints. The
red lines (gamma ray and resistivity) identify the model, and the
green lines (GR-p and ILD-p) identify the well logs from the
pilot well.
The model from the pilot certainly does not directly t the
data in the horizontal well (Figure 9b) without further editing. By maintaining the layer thickness as more or less constant and exing the model up and down to achieve a good t
between the model and the well logs (Figure 9c) while honoring the trends indicated by the 3D markers and the oset
well, we arrive at the cross-sectional interpretation shown in
Figure 9c.

The correctness of the interpretation is checked by the


closeness in tting the well logs in the horizontal well with
those from the pilot/oset wells. The model can be validated
further by performing a tool-response modeling with the logging tools moving along the well path. In Figure 9c, the well
logs (GR-m and RAD2-m) are the modeled logs which overlay
closely with the measured logs (GR and RAD2), especially notable around the lateral-distance interval of 2400 ft to ~ 2600
ft. The dierence between the pilot-well induction log ILD-p
and the horizontal-well propagation resistivity RAD2 is caused
mainly by the dierence in response characteristics in a vertical
well environment versus a horizontal well environment, among
other contributing factors.
Eective and ecient modeling capabilities enable realtime log processing and interpretation that subsequently impact
well-placement decisions. Modeling helps to extract the desired
information such as the distance to uid contacts, distance to
bed boundaries, resistivity anisotropy, and apparent dip from
well logs acquired in horizontal wells.

Figure 9. Horizontal well interpretation aided by information from


Figure 8. Well logs and formation model built in the pilot well. Markers
indicate the layers of signicance and the target reservoir location.

498

THE LEADING EDGE

May 2015

pilot/oset wells and 3D surfaces. (a) Well logs measured in the horizontal well. (b) Starting model from the pilot well. (c) Interpreted
cross-sectional model.

Special Section: Well geosteering

Even with good data tting and modeling validation, one


should remember that the interpreted model in Figure 9c might
not be unique. The volume a few meters from the borehole is derived through extrapolation from pilot/oset wells. No information is measured directly from that volume.

The Leading Edge 2015.34:492-499.


Downloaded from library.seg.org by Piyush Pasari on 05/31/15. For personal use only.

Conclusions
To successfully interpret and geosteer a horizontal well, a
formation model must be constructed to account for the geologic
complexity encountered along the well path. Because of cost and
technology limitations, the set of logs from the horizontal well
alone is generally incomplete in answering the whereabouts of
the well path versus the target reservoir. Geologic constraints
based on depositional environments must be applied by using
the information from oset/pilot-well logs and the seismic images and large-scale geologic understanding. The visualization
of well logs/reservoir in a multidimensional space also becomes
essential to include the geometric information between the horizontal well and the target formation.
Tool-response modeling/inversion and the recognition of toolresponse characteristics also will help to extract the geometric

Special Section: Well geosteering

information that otherwise is not readily available to enable a


more unique interpretation. It also should be noted that a model
that ts the measured logs might not always be unique.

Acknowledgments
The author thanks Maxwell Dynamics for permission to publish this article. He is also grateful to Carlos Torres-Verdn, who
reviewed and suggested improvements to the article.
Corresponding author: JohnZhou@MaxwellDynamics.com

References
Wu, J.-Q., M. Al Wisler, and W. C. Barnett, 1991, Bed boundary detection using resistivity sensor in drilling horizontal wells:
Presented at the 32nd Annual Symposium, SPWLA.
Zhou, Q., 2007, Interpreting resistivity logs from deviated wells:
69th Conference and Exhibition, EAGE, Extended Abstracts,
http://dx.doi.org/10.3997/2214-4609.201401794.
Zhou, Q., 2008, Log interpretation in high-deviation wells through
user-friendly tool-response processing: Presented at the 49th
Annual Symposium, SPWLA.

May 2015

THE LEADING EDGE

499

Vous aimerez peut-être aussi