Beruflich Dokumente
Kultur Dokumente
. (3)
SPE 115715 3
In Eq. 3,
r
is the shear stress that causes fracture sliding (reactivation), and
n
is the normal stress acting on the
fracture. The coefficient of static friction is equivalent to the slope of the fracture reactivation envelope in Fig. 1a and
typically falls in the range 1 6 . 0 . The shear and normal stresses acting on the fracture planes are functions of the
principal effective stresses,
H
and
h
, and the angle between the fracture planes and
H
. Thus some fractures are
more favorably oriented for reactivation than others.
In Fig. 3a (Zoback 2007), a series of randomly oriented fractures are shown. As pore pressure increases, the Mohr circle
moves toward the fracture reactivation envelop and a subset of pre-existing fractures begin to slip as soon as the fracture
reactivation envelope is exceeded (those shown in red in Fig. 3b). Fracture planes that form angles between 25 and 35
degrees with
H
are usually the most optimally oriented fractures for reactivation. Movement along these pre-existing
fractures manifests itself as microseismic activity.
The energy associated to microseismicity events generated by fracture reactivation radiates as a compressional (P) wave
traveling at the P-wave speed followed by a shear (S) wave traveling at the slower S-wave speed (Crampin et al. 1991).
Strategically located geophones can record signals that can be analyzed to locate the source of the emissions. Location is
determined by distance and direction from the receiver. Distance to the event can be obtained by knowing the velocities of
the P and S waves and the lag times between their arrivals. Direction is known from polarization of particle motion of the P
wave, which is along the path connecting the event and the receiver. When several receivers are deployed simultaneously,
event locations may also be determined by triangulation by knowing the velocity of the P waves.
The analysis of split shear waves from microseismicity has shown to be a valuable technique to detect the main
orientation and intensity of open fractures in the reservoir. The method is based on the fact that shear-wave propagating
through rocks with aligned fluid-filled fractures will split into two waves, a fast one polarized parallel to the predominant
open fracture direction, and a slow one, polarized perpendicular to it (Fig. 4). The time delayed between the fast and slow
wave is proportional to the fracture density, or number of fractures per unit volume (Rial 2005).
Cusiana-Cupiagua Surface Seismic Network
The CCSN was set up early in the development stage of the fields (1992) on a rugged topography and at the average
altitude in the network's area of approximately 1640 ft above sea level. The region average temperatures are about 26C and
relative humidity around 80%. These foothills, stretching all along the eastern Colombia, mark the limit between the stable
South America plate (Guyana shield) and the North Andean Block.
General Configuration
The basic shape of the network follows the shape of the fields, and thus the general geological strike. The dimensions are
approximately 50 miles (North-South) by 9 miles (East-West). The average distance between stations is 4 miles.
The configuration of the network, illustrated in Fig. 5, is certainly below the optimum, mainly in the extension of net
coverage towards East and West.
Hardware
The instrumentation of the CCSN consists of 18 surface seismological stations, each composed by a Digital Acquisition
System (DAS) and a three components sensor. The DAS units use 4 Gb compact flash memories for continuous microseismic
events recording storage, global positioning system time accuracy, Ethernet connectivity, among many other features that
allow a reliable network operation and maintenance
Processing Scheme
Data analysis
The scanning of all available data streams on screen for identification of event waveforms is usually a straightforward
activity. All local event waveforms are extracted and added to an event database, even if only recorded on a single station.
Phase readings
The phase reading is done visually, for both P and S waves and with all three components on screen. S-phases are read on
channels with clear onsets. Filtering is used as little as possible; very noisy seismograms are low-pass filtered, for events with
only few stations.
Magnitude calculation
Calculation of duration magnitudes is performed mainly to overcome the fact that occasionally S-waves are saturated;
thus the magnitudes are not calibrated for local conditions, but measured in a consistent way.
Hypocenter location
Routine locations are calculated with hypocenter location software. In some cases, single-station location routines for
polarization analysis of 3D-waveforms are used.
Advanced processing
Data from the CCSN, being recorded with three component sensors, allows determination of the basic shear-wave
splitting parameters, which are polarization angles and time delays between the two split S-waves. This information leads to
4 SPE 115715
models of the fracture directions in the rock mass, among other parameters, and can eventually also reflect changes in the
elastic properties of the media, due to injection or production of liquid or gas.
CCSN Objectives
At first, the CCSN was used to obtain data for seismic hazard models needed for the design of field infrastructure.
Because the region is seismically active, monitoring of local and regional seismicity in a risk reduction context has also been
a main objective. However, it has become increasingly evident that the network and its data is an invaluable asset for
evaluation of conditions relevant to understand reservoir performance.
Correlation between Microseismicity and Fluid Injection Tracking
Microseismicity monitoring can be used to map injected fluid movements in the reservoir. The mechanism connecting
microseiemicity with fluid movement within the reservoir can be summarized as follows: (i) fluid production/injection causes
changes in pore pressure; (ii) pore pressure changes cause variation in the in-situ stresses and deformation (sliding or
reactivation) of natural fracture networks; and (iii) reactivation on natural fracture networks induces microseismic events and
alters the permeability of the system. As mentioned earlier, the preferred direction for permeability enhancement by natural
fractures reactivation is between 25 and 35 degrees from the direction of
H
. Changes in production/injection conditions,
such as conversion of injectors to producers and new infill wells, may induce local changes in the magnitude and/or direction
of
H
and, therefore, temporal changes the preferential flow directions.
The case study of the effects of water injection into well Cusiana well TS26 demonstrates the strong correlation between
microseismicity and fluid injection in the reservoir (Garcia 2006). Over a period of 1032 days water was injected in the
reservoir at a wellhead average pressure of 4500 psi except for the interval starting on day 607 when, because of operational
problems, injection wellhead pressure decreased to 1250 psi. It then increased gradually until reaching the average pressure
of 4500 psi on day 826. Microseismicity was monitored before starting water injection. Figs. 6a and 6b show the injection
rate and the number of events per day, respectively, as function of time. The time scale in Figs. 6a and 6b include 120 days
before starting injection. A total of 2800 microseisic events were recorded during the test time. The epicenters of these local
events are shown in Fig. 7 which also includes the direction of
H
in the Cusiana Field.
Several evetns are observed from Figs. 6 and 7: (i) before starting injection, the intensity of microseismic events is
negligible, (ii) the increase in the number of events per day is immediate as soon as injection starts, and (iii) the intensity of
microseismic events is proportional to the injection rate. These observations clearly support the strong correlation between
microseismicity and fluid injection. Furthermore, the cloud of microseismic events exhibits an elliptical shape with the major
axis aligned with the
H
direction. This is consistent with the orientation of fracture planes that are optimally oriented for
shear reactivation (see previous section). This later observation indicates that: (i) it is actually the reactivation of natural
fracture networks that induces the microseismic events (the temporal and spatial proximity between injection and mircro-
earthquake generation also provides evidence that the injection triggered the microseismicity) , (ii) the preferred orientation
for permeability enhancement and, therefore, the preferential flow direction for injected fluids, is NW-SE, and (iii)
microseismicity can be used as a tool to determine the local orientation of
H
and its probable variation with time (this is
essential input information for any geomechanics model).
The distribution of shear-wave polarizations from the injection test time is plotted in Fig. 8. The polarization of the faster
split shear waves displays approximately parallel (red dots) and perpendicular (green dots) to
H
; however, the greatest
concentration of polarizations take place in the
H
direction indicating that during the injection period most of the fluid
moved in the NW-SE direction (the geometry of the swept area is elliptical with major axis aligned with
H
direction).
Current interpretation of the structure indicates that the flow component perpendicular to the
H
direction is generated by
fold-related fractures. A relevant fact from Fig. 8 is the gap of events observed in the injection period encompassed by days
607 and 826 corresponding to period of low wellhead injection pressure. This reveals once more the strong correlation
between microseismicity and reservoir dynamics.
Fig. 9 displays the distribution of shear-wave polarizations through rose diagrams. Each rose diagram refers to a two-
month period. The red and green lines represent the relative amount of open fractures sub-parallel and sub-perpendicular to
the
H
direction, respectively. This is a measurement of the relative amount of fluid moving in each one of these directions.
An important feature observed from Fig. 9 is that shear-wave polarizations change with time. These changes indicate that
local effective stresses are varying with time and induce temporal changes of preferential flow directions. Changes of
production/injection conditions in a well and its offset wells will alter the local pressure field which, in turn, changes the local
stresses magnitudes and, very probably, stress directions. Any change in stresses magnitudes and/or directions open pre-
existing natural fractures in the direction of the new local maximum stress and tend to close fractures in other directions.
Needless to say, this dynamic behavior has implications to reservoir management. In addition, these observations lead
reservoir engineers to move towards a more realistic modeling condition in which components of the reservoir static model
become dynamic variables. For example, the transmissibilities connecting adjacent cells could vary with time and pressure.
These changes in transmissibilities depend on production conditions such as well locations, perforated intervals, production
SPE 115715 5
rates, etc. A complete discussion of the effect of production conditions on potential permeability changes is presented
elsewhere (Osorio, et al. 1997, Osorio et al. 1998).
Fig. 10 presents a plot of the arrival time difference between the shear and compressional waves (S-P plot which to the
events distance). This plot provides information on preferential flow directions with depth. As observed from Fig. 10, for the
observation time selected from this study, injected water moved predominantly in the
H
direction (NW-SE) in the
intermediate and lower parts of the formation, while in the upper part of the formation, the preferential trajectory of flow was
in the direction of fold related fractures. Several facts could explain these differences in preferential flow trajectories between
units at different depths. Physical explanations include differences in fracture intensity and orientations, in pore pressure and,
therefore, in effective stresses, in local principal stress orientation and in fracture properties. Practical impact of this behavior
is interesting and is currently under further study.
Numerical Model Calibration
An additional strong correlation between microseimicity and reservoir dynamics is observed from the application of
microseismicity to numerical model calibration. A well calibrated reservoir model can be a tool to reliably predict future
reservoir performance if a set of reservoir description parameters can be incorporated to reproduce past production history as
outlined elsewhere (Calvin and Dalton 1990). Unfortunately, to arrive at a unique data set is almost impossible, and
alternatives to understand uncertainty impact of reservoir description on future performance using a reservoir model have
been proposed (Williams et al. 2004).
However, regardless of the approach reservoir engineers and geoscientists take to calibrate a reservoir model, subsurface
teams face the challenge of integrating data coming from different sources that most of the time is averaged information from
different scales (time, space). Most of the time, a good general picture about how the reservoir looks like can be drawn during
this data integration process and model calibration, but sometimes data remains outside of the picture waiting for a consistent
story that explains its existence.
In particular, the Cusiana field numerical model has gone through several efforts for developing a calibrated reservoir
model, reliable for production prediction. As a result, a history matched model helped to define and to successfully
implement a gas injection redistribution strategy where produced gas injection in the crest is going to be progressively moved
to the flank of the structure (Soto et al. 2006).
During the model calibration process, conventional strategies for history match were used with limited effectiveness.
However, the model was not able to reproduce GOR increase at wells located on the flank of the structure, and at the same
time, reservoir pressure at the crest was higher than measured. This observation was explained by the lack of communication
between the crest and the flank in spite of the improvement in the geological description.
A solution was to increase the transmissibility perpendicular to the structure crest allowing better movement of injected
gas towards the flank. Since the selection of reservoir areas to increase transmissibility is not unique, microseismicity was
used to guide the best areas for increasing injected gas flow. Cusiana reservoir depth and complex structure setting make the
use of time-lapse seismic for model calibration impossible although elaborated approaches to use 4D seismic are available
(Stephen 2006).
Microseismicity events collected over time exhibit alignment that can be correlated with the injection-production
imbalance. For a particular year, Fig. 11 shows a very strong correlation between the total volume injected and the
concentration of events in the area close to injectors. In fact, the energy of those events, expressed as the cumulative moment,
is proportional to the cumulative volume injected as illustrated in Fig. 12.
Based on these observations, the subsurface team chose the areas for increasing transmissibility using the cloud geometry
of production-induced microseismic events, improving production history with the numerical model (Fig. 13). The residual
model mismatch in GOR was attributed to the lack of vertical resolution of the numerical model, which creates excessive
fluid mixing not observed in the field.
Microseismicity event understanding as well as data from other sources were combined, where possible (Fig. 14), to
represent a much consistent view of the subsurface, which improved reservoir understanding and the creation of strategies to
increase oil recovery.
Conclusions
Mapping passive seismics in Cusiana and Cupiagua fields has revealed a strong correlation between reservoir dynamic
performance and production induced microseismicity. Fluid production/injection causes changes in reservoir pore pressure
and, therefore, in local effective stresses. The changes in effective stresses cause natural fracture deformations which, in turn,
change local transmissibilities and triggers microseismic events. The interplay of these two latter effects determines the
relationship between microseismicity and reservoir dynamics (pressures and fluid saturations among other factors). Several
highlights concerning this correlation and its great potential as a reservoir surveillance tool are noted here:
1. Cloud geometry of production-induced microseismic events is elliptical with major axis aligned with the maximum
stress direction, indicating that, at a given time, existing preferential flow directions are sub-parallel to the current,
local maximum stress directions. As a consequence, microseismicity is a potential surveillance technique to track
6 SPE 115715
fluid movement associated with changes in the magnitude and/or direction of local principal stresses induced by
local fluid production/injection operations.
2. Shear-wave polarizations changes with time confirm that production-induced changes of local effective stresses
(magnitudes and/or orientations) bring temporal changes in local preferential flow trajectories. This observation
suggests that preferential flow trajectories could be controlled by managing induced stresses, controlled by changes
in local production conditions. This performance may have important implications on reservoir management.
3. Arrival time difference between the shear and compressional waves (S-P plot) indicate that microseismicity also
represents a potential surveillance tool to track selective fluid movement with depth.
4. With respect to reservoir simulation, active tectonic areas present challenges to reservoir modelers from different
points of view. Static models are no longer static. Pore pressure changes influence more than material balance.
Rocks deform. Passive seismic appears revealing a relation between production/injection strategies and reservoir
performance. Based on those observations, we can conclude as outlined in this paper: (i) microseismic events can be
correlated with production/injection schemes that alter the natural state of rocks, (ii) understanding of
microseismicity leads to a more robust description of a reservoir although implementation of that description
requires much more modeling to be effectively used for prediction, and (iii) rocks react consistently to
production/injection induced perturbation. Pre-existing weak planes also observed at rock samples are reactivated
causing a measured wave response at surface. Therefore, data collected from different sources and scales should be
reconciled.
Finally, it should be emphasized that although field observations in Cusiana and Cupiagua, as describe in this paper,
clearly indicate that microseismicity is a potential surveillance tool to improve reservoir characterization and management.
Additional and more specialized applications will allow to obtain further benefits of this technology.
Nomenclature
p = pore pressure, m/Lt
2
, psi
f
p = fracture pore pressure, m/Lt
2
, psi
= Biots parameter
= coefficient of fracture static friction
= total stress, m/Lt
2
, psi
h
= minimum horizontal stress, m/Lt
2
, psi
H
= maximum horizontal stress, m/Lt
2
, psi
n
= normal stress, m/Lt
2
, psi
v
= vertical horizontal stress, m/Lt
2
, psi
= effective stress, m/Lt
2
, psi
h
= minimum horizontal effective stress, m/Lt
2
, psi
H
= maximum horizontal effective stress, m/Lt
2
, psi
n
= normal effective stress, m/Lt
2
, psi
= total stress, m/Lt
2
, psi
r
= fracture reactivation shear stress, m/Lt
2
, psi
Acknowledgments
The authors would like to thank SDLA Partners (BP Colombia, Tepma and Ecopetrol) for allowing publication of this paper.
Also we would like to show our gratitude to Corporacin OSSO for their continuous effort on micro seismic data acquisition,
processing and interpretation.
References
Avasthi, J.M., Nolen-Hoeksema, R.C. and El Rabaa, A.W.M. 1991. In-Situ Stress Evaluation in the McElroy Field, West
Texas. SPEFE (Sept 1991): 301-309. SPE-20105.
Calvin, C. M. and Dalton, R.L. 1990. Reservoir Simulation. Monograph series, SPE, Richarson, Texas 13.
Crampin, S. and Lowell, J.H. 1991. A Decade of Shear-Wave Splitting in the Earths Crust: What Does it Mean? What Use
Can We Make of it? Ans Should We Do Next?. Geophys. J. Int. 107: 387-407.
Cornet, F.H. and Jones, R. 1994. Field Evidences on the Orientation of Forced Water Flow with Respect to the Regional
Principal Stress Directions. Proc. First North American Rock Mechanics Symposium/The University of Texas at Austin,
Austin, Texas, 61-69.
SPE 115715 7
Fjaer, E., Holt, R.M., Horsrud, P. and Raaen, A.M.. 1992. The Effective Stress Concept. In Petroleum Related Rock
Mechanics. Elsevier Science, Chap. 1, 37-39. New York City, NY.
Garcia, A. 2006. Caracterizacin de una Cuenca Petrolera con Sismicidad Pasiva a partir de Anisotropa Ssmica y Anlisis
de Repetidores. PhD dissertation, Universidad del Valle, Cali, Colombia.
Osorio J. G., Chen H-Y, and Teufel L.W. 1997. Numerical Simulation of Coupled Fluid-Flow/Geomechanical Behavior of
Tight Gas Reservoirs. Paper SPE 39055 presented at the SPE Fifth Latin American and Caribbean Petroleum Engineering
Confrenece and Exhibition held in Rio de Janeiro, Brazil, 30 August 3 September.
Osorio J. G., Chen H-Y, and Teufel L.W. 1998. A Two-Domain, 3D, Fully Coupled Fluid-Flow/Geomechanical Simulation
Model for Reservoirs with Stress-Sensitive Mechanical and Fluid-Flow Properties. Paper SPE 47397 presented at the
SPE/SRM Eurock Conference held in Trondheim, Norway, 8-10 July.
Rial, J.A., Elkibbi, M. and Yang, M. 2005. Shear-Wave Splitting as a Tool for the Characterization of Geothermal Fractured
Reservoir: Lessons Learned. Geothermics 34: 365-385.
Salz, L.B. 1997. Relationship Between Fracture Propagation Pressure and Pore Pressure. Paper SPE 6870 presented at the
SPE Annual Technical Conference and Exhibition, Denver, Colorado, October 9-12.
Soto, L., Penuela, G., Benavides, I., Martinez, M., Castiblanco, I., and Chaves, I. 2006. Gas-Injection Redistribution
Revitalizes a Mature Volatile Oil Field: Cusiana Field Case Study. Paper SPE 103593 presented at the SPE Annual Technical
Conference and Exhibition in San Antonio, Texas, 2427 September.
Stephen, K. D.; Soldo, J., MacBeth, C., and Christie, M. 2006. Multiple-Model Seismic and Production History Matching:
A Case Study. SPEJ 11 (4): 418-430. SPE 94173 PA.
Teufel, L.W. and Farrel, H.E. 1990. Distribution of In Situ Stress and Natural Fractures in the Ekofisk Field, North Sea.
Proc. Third North Sea Chalk Symposium, Copenhagen, June 11-12.
Teufel, L.W. and Farrel, H.E. 1992. Interrelationship Between In Situ Stress, Natural Fractures, and Reservoir Permeability
Anisotropy: A Case Study of the Ekofisk Field, North Sea. Proc. Fractured and Joined Rock Conference, Lake Tahoe, June
3-5.
Williams J. J, G; Mansfield, M.; MacDonald, D. G.; Bush, M. D. 2004. Top-Down Reservoir Modelling. Paper SPE 89974
presented at the SPE Annual Technical Conference and Exhibition, Houston, Texas, 26-29 September.
Wright, C.A. et. al. 1995. Hydraulic Fracture Orientation and Production/injection Induced Reservoir Stress Changes in
Diatomite Waterfloods. Paper SPE 29625 presented at the SPE Western Regional Meeting, Bakersfield, March. 8-10.
Wright, C.A. and Conant, R.A. 1995. Hydraulic Fracture Reorientation in Primary and Secondary Recovery from Low-
Permeabilty Reservoirs. Paper SPE 30484 presented at the 1995 SPE Annual Technical Conference and Exhibition, Dallas,
Texas, October 22-25.
Zoback, M.D. 2007. Critically Stressed Faults and Fluid Flow. In Reservoir Geomechanics, Cambridge University Press,
Chap. 11, 340-377, New York City, NY.
SI Metric Conversion Factors
psi 6.894 757 E + 00 = kPa
ft 3.048* E - 01 = m
mile 1.609 344* E - 00 = km
lbf-ft 1.355 818 E - 03 = kJ
*Conversion factor is exact.
8 SPE 115715
Fig. 1. Location map of the BP Colombia fields in Colombia.
(a) (b)
Fig. 2. Mohr circle showing stress evolution with production/injection time and fracture and reactivation envelop. The stress state
required for fracture re-activation depends on the fractures orientation with respect to maximum stress orientation and the stress
path.
SPE 115715 9
(a) (b)
Fig. 3. Fractures occur with wide range of orientations. Not all fractures are permeable in todays local stress field. Those fractures
with a ratio of shear to normal effective stress between 0.6 and 1.0 are likely to slip during production/injection operations (modified
from Zoback 2007).
Fig. 4. Schematic view of shear wave splitting. A shear wave splits into two: one fast parallel and one slow perpendicular to the
open fractures in the medium (after Rial 2007)
10 SPE 115715
Fig. 5. Map of the Cusiana and Cupiagua Network showing its current stations distribution (squares) and geometry). Major faults
are represented by bold dark lines and lease areas by white lines.
(a) (b)
Fig. 6. Injection rate (part a) and the number of events per day (part b) as function of time. Before starting injection, the intensity of
microseismic events is negligible. Also, the intensity of microseismicity is proportional to the injection rate (modified from Garcia,
2006).
SPE 115715 11
Fig. 7. Epicenters of the local events recorded during the observation time selected to monitor microseismicity triggered by
injection in Well Cus TS26. The right side illustrates the average direction of
H
in Cusiana (part b, after Last et.al.).
Fig. 8. Distribution of shear-wave polarizations of the events presented in Figure 7. The greatest concentration of polarizations take
place in the
H
direction indicating that during the injection period most of the fluid moved in the NW-SE direction. The geometry
of the swept area is elliptical with major axis alight with
H
direction. (Modified from Garcia, 2006).
4 km 4 km 4 km
12 SPE 115715
Fig. 9. Rose diagrams representing the distribution of shear-wave polarizations. The red and green lines represent the relative
amount of open fractures sub-parallel and sub-perpendicular to the
H
direction, respectively. This is a measurement of the
relative amount of fluid moving in each one of these directions (modified from Garcia, 2006).
Fig. 10. Arrival time difference between the shear and compressional wave (S-P plot). For the observation time selected in this
study, injected water moved predominantly in the
H
direction (NW-SE) in the intermediate and lower parts of the formation, while
in the upper part of the formation, the preferential trajectory of flow was in the direction of fold related fractures (modified from
Garcia, 2006)..
SPE 115715 13
Fig. 11. Bubble map of Cusiana fluid production and injection. There are mainly three water injectors, which are responsible for
most of the microseismicity events in that year. In the map, there is a strong alignment of event related to Cusiana TS26 water
injector.
y = 0,1607x + 0,0876
R
2
= 0,9897
0,0
0,4
0,8
1,2
1,6
2,0
0 2 4 6 8 10 12 14
Cumulative water injection, million bbl
C
u
m
u
l
a
t
i
v
e
m
o
m
e
n
t
,
l
b
f
-
f
t
x
1
0
1
5
Fig. 12- Energy associated to microseismic events is proportional to the cumulative injection in the Cusiana TS26 area (modified
from: Jones, M. 2002. Consultancy Presentation, University of Portsmouth, Portsmouth, UK).
CSTS26
5 km
Production/Injection Microseismic events
5 km
Production/Injection Microseismic events
14 SPE 115715
Fig. 13. To allow gas in the crest of the structure to move towards the flank, permeability multipliers were used increasing
transmissibility from 2 to 20 times.
Fig. 14. X-direction transmissibility multipliers are located in areas of the reservoir using data from different sources such as core,
UBI and structural curvature analysis. In areas where data from different sources area available such as the area of RC2W well,
microseismicity confirms the extension in the reservoir of heterogeneity observed at the well level.