Sei sulla pagina 1di 29

Monitoring Coal Bed Methane Production: A Case Study from

the Powder River Basin, Wyoming, United States of America.

Olusoga Martins Akintunde


Department of Geophysics
Stanford University

Abstract

The growing significance of the Powder River Basin’s Coal Bed Methane (CBM) to
United States domestic energy supplies has heightened interest in its exploration and
exploitation. Systematic and continued development of this resource would require
adequate characterization of the CBM reservoir, better reservoir management practices
and monitoring strategies that would ensure optimum and/or efficient gas production in
an environmentally sound manner. The primary objective of this study is to monitor time-
variant changes in the velocity of seismic waves associated with dewatering during
methane production. Dewatering causes a decrease in pore pressure resulting in an
increase in gas saturation. Gas lowers the velocity of seismic waves whereas a decrease
in pore pressure increases the velocity. Knowledge of the spatial distribution of gas
between wells might be helpful in identifying both the source of produced water and the
spatial efficiency of the dewatering process.

This study is based on three cross-well seismic surveys executed at a test site in the
Powder River Basin. The surveys were run between a pair of monitoring wells that
straddle a completed CBM production well. The monitoring wells were provided for
these surveys with the help of CononcoPhillips, Bureau of Land Management (BLM),
and the Western Resources Project (WRP). The first survey was completed in December
2002 before dewatering began and thus provided the baseline image or tomogram. The
second and third surveys were run in July 2003 and June 2004 respectively. Changes in
the P-wave travel times can be seen in unprocessed seismograms though tomography was
used to image the velocity changes between wells from the traveltime differences.

From the baseline tomogram, I have delineated four major lithologic units that are typical
of the Upper Cretaceous sediments in the Powder River Basin. And from the time-lapse
surveys, I estimate that the spatial distribution of the velocity changes approximates 6%
above the coal as well as inside the coal layer. This difference can be attributed primarily
to changes in pore-fluid saturation occasioned by the dewatering process. The observed
changes demonstrate feasibility to monitor the CBM production process and that the
changes are consistent with theoretically derived coal-physics models and observations
from production data.

1
Introduction

Globally, coal constitutes a major portion of the world’s mineral energy resources. It
serves as a source for electricity generation, heat and coking coal for iron and steel
making. Coal is formed by the physico-chemical alteration of peat (coalification) by
processes involving bacterial decay, compaction, heat and time. It is a readily
combustible rock containing more than 50% by weight of carbonaceous material. Also, it
contains about 90% of gas mostly methane, hence the name Coal Bed Methane (CBM).
The types of coal, in increasing order of alteration (or rank) are lignite (brown coal), sub-
bituminous, bituminous and anthracite. Porosity in coal varies considerably (1.6% to
25%) depending on the coal’s rank (King and Wilkins, 1944). Also, coal exhibits low
density, low velocity, low magnetic susceptibility and low radioactivity. Its electrical
resistivity is higher when compared with surrounding sediments.

The coal structure contains pores with different pore sizes. Methane is stored in the coal
by being adsorbed to the micro pores and cleat internal surfaces as free gas in the micro
pores and cleats (figures 1 and 2) and/or through dissolution in water. The coal structure
and cleat system (figure 2) play vital roles in the coal gas storage system. Coal gas
reservoirs are dual storage reservoirs consisting of primary and secondary storage
systems. The primary storage system makes up 98% of the reservoir volume and contains
organic matter, inorganic material, inherent water and gas stored within very small pore
spaces. Primary system gas storage is dominated by sorption phenomena because of the
small size of the pores. During sorption, the molecules that are within very close
proximity to solid surfaces are attracted to the solid and are packed closer than expected
from the pressure conditions. The primary porosity is relatively impermeable and mass
transfer is dominated by diffusion. Commercially productive coal reservoirs contain a
well-developed secondary storage dominated by natural fractures. Natural fractures are
very significant to commercial production of CBM. Flow through the secondary storage

Figure 1: Reflection behavior in coal (after Prasad, 2001).

2
system is due to pressure gradients between the fracture system and production wells.
The majority of gas in a coal gas reservoir diffuses through the primary storage system,
desorbs at the interface between the primary and secondary systems, and then flows
through the secondary systems to wells (Mavor et al. 2002).

Free gas in the micropores

Free gas in the cleats

Cleats

Methane dissolved in water


Pore space filled with water

Adsorbed gas

Figure 2: Principal mechanisms for primary storage of methane within the coal’s cleats and micropores

Located in the Central part of the PRB of Wyoming and Montana (figure 3), the Big
George is currently one of the world’s most established coal fields. Along with the
Wyodak, the Big George is a CBM producer in the PRB.

Figure 3: Map of geologic age of coal of United States showing the location of the Powder River Basin

3
PRB accounts for 800 billion tons of coal and is estimated to produce 25 trillion cubic
feet of gas - about 20% of the CBM being produced in the United States. The PRB coals
are biogenic, permeable, shallow and low-rank coals. They are characterized by higher
moisture content and lower carbon content when compared with high-rank bituminous
coals. The common practice for primary CBM recovery at the PRB is to depressurize the
coal, usually by pumping water from CBM production wells installed in coal beds for
several months. The wells are completed by installing well casing to the top of the coal
bed, reaming the coal, and then leaving the coal bed open to the hole. Water is then
removed from the well by installing a submersible pump in the open coal bed; pumping
reduces hydrostatic pressure in the coal that allows the methane to be released. The gas is
transported to the surface in the space between the tubing attached to the submersible
pump and the well casing and the produced water is discharged to the surface via the
tubing (figure 4).

At the initial stage of dewatering (figure 5), gas production is negligible, but increases
and stabilizes as dewatering progresses over an appreciable period. Richardson and
Lawton (2002) showed that the dewatering process affects the acoustic and elastic
properties of the CBM reservoir, causing appreciable acoustic impedance discontinuity
within the coal-bearing layer and the surrounding strata. These changes in turn affect the
amplitude and travel times of reflected and transmitted waves, paving the way for cross-
well seismic mapping of changes caused by the dewatering process. Time-lapse seismic
monitoring depends on production-related changes in the acoustic velocity and density of
the reservoir rocks. Most potential monitoring applications rely on saturation changes that
dominantly affect the P-wave velocity of the reservoir. This is because the bulk modulus
of the rock is related to both the bulk modulus of the rock matrix and the bulk modulus of
the reservoir fluids.

Historically, the first quantitative data set on time-lapse reservoir monitoring emanated
from rock physics measurements at Stanford University in the mid 1980s. The results of
the study show that laboratory measurements on heavy-oil saturated core samples showed
large decreases in seismic rock velocity when the viscous oil was heated (Nur et al. 1984;
Wang and Nur, 1986, and Nur, 1989). Other notable time-lapse reservoir monitoring
studies in the last two decades are: steam injection by Pulin et al. 1987, Lumley 1995a, b
and Jenkins et al. 1997. Others include fireflood EOR process by Greaves and Fulp 1987,
a North Sea gas cap expansion project by Johnstad et al. 1995 and West Texas CO2
injection project by Harris et al 1995. It is, however, pertinent to note that recent and
current effort on time-lapse seismic reservoir monitoring has focused mostly on
hydrocarbon reservoirs (Tura and Lumley, 1998, 1999, Landro et al. 2003, Landro 2001
and Harris et al. 1995). This underscores the need to apply the same tool to the CBM
reservoir process. Continued and systematic development of the CBM resource for
optimum gas production would require understanding the effectiveness of the efficiency
of the dewatering process and knowledge of the spatial distribution of gas between wells

In this study, cross-well seismic surveys were designed to image production-induced


changes in P-wave velocity associated with primary CBM production at a site in the
Powder River Basin (PRB). The pre- and post- CBM production cross-well surveys were

4
executed in December, 2002, August, 2003 and June 2004 respectively in the vicinity of
the Big George formation.

Figure 4: Schematic diagram of Coal Bed Methane production at the Powder River Basin

Figure 5: Conceptual CBM Production at the PRB (modified from Ayers, 2002)

Geologic Background

Geologically, the PRB is an intermontane foreland arch basin formed during the
Laramide Orogeny. It is in southeast Montana and northeast Wyoming, characterized by
rocks of late Cretaceous to early Tertiary age (figure 3). The basin is flanked by the

5
Bighorn Mountains on the west, the Casper Arch-Laramie Range on the southwest and
south, the Hartville Uplift on the southeast, the Black Hills on the east, and the Miles City
Arch on the northeast, all formed during the Laramide Orogeny (Flores 2004). Also, it is
a thick sequence of sedimentary rock formed in a large down warp within the
Precambrian Basement. Sediments range from Paleozoic at the bottom through Mesozoic
to Tertiary at the top. The Basin’s started evolving about 60 million years ago, when it
was just a flat, sea-level plain, mudding from a retreating sea that had covered Wyoming
for 40,000,000 years. Over the next 14 million years, the Basin rose to its present 3000ft
with the Big Horns Mountains and the Black Hills rising thousands of feet higher.
Following were episodes of geological processes favoring sediment accumulation into
pure peat. Further effects of temperature, erosion and deposition transformed the peat into
some of the thickest, low ash coals in the world. Presently, the PRB has fully matured
geologically as an energy source and a potential factor in global climate and energy
change (Harris et al 2004, Akintunde et al 2004) .Details about the geology is contained
in Larson1989, Debruin et al. 2000, Randall 1991, Rice et al. 2000 and Flores R.M. 2004).

Methodology

This study attempts to address two fundamental issues that are germane to the problem of
seismically monitoring the CBM recovery process. They are:

(i) Map production-induced changes in P-wave velocity caused by dewatering


(ii) Interpret velocity changes in terms of gas saturation and pressure changes.

To actualize the above-stated goals: I did coal physics feasibility analysis, processed
three crosswell field seismic data sets, and carried out time-lapse image analysis and
interpretation. These methods of study and their results are discussed as follows.

Coal Physics Feasibility Analysis (Numerical Modeling)

Prior to the field experiments, I carried out a modeling study to predict changes in P-
wave velocity in dry coal when saturated with fluids (water and methane). I extracted P-
wave velocities for dry coal and their pressure dependence from laboratory measurements
on bituminous, Permian coal sample contained in Yu et al. 1993. The sample has a
porosity of 2.9% and density of 1.35g/cc. I modeled the data using Gassmann’ equation
(Gassmann, 1951), appropriate fluid properties and coal physics relationships for P-wave
velocity, effective fluid modulus and density of a fluid-saturated rock (Mavko et al 1998).
Gassmann theory is a physical expression that relates the fluid-saturated moduli to the
known moduli and fluid properties of a reservoir process undergoing changes in pore
fluid. It permits estimation of velocity of fluid-saturated rocks from dry rocks and vice
versa. Also, it is equivalent to the low-frequency limit of Biot’s theory (mavko et al
1998). Additionally; the Gassmann theory is strictly valid for low (quasistatic)
frequencies, implying that the calculated fluid-saturated velocities are of low frequency.
We used dry data because the most plausible way for applying ultrasonic core data to

6
field conditions is to use velocities measured on (nearly) dry cores and then employ
Gassmann for the fluids substitution. This Gassmann’s analysis assumes free gas,
homogeneous saturations and does not account for gas adsorption and/or desorption.
Figure 6 shows the results of the fluid substitution based on the famous Gassmann’s
equation. The green curve represents the dry data set extracted from Yu et al 1993. It
shows that P-wave velocity (Vp) in dry coal increases as differential pressure (confining
pressure minus pore pressure) increases. The influence of increasing differential pressure

Figure 6: Gassmann-predicted changes in P-wave velocity in fluid-saturated coal

is to close the thin cracks and penny-shaped pores and to make better contact between
particles in the rock. With addition of 100% water (H2O) as depicted by the blue curve,
the Gassmann-derived P-wave velocity increases considerably. Coal when fully saturated
with water tends to exhibit a large increase in P-wave velocity perhaps due its large bulk
modulus (2.25 GPa) and zero shear modulus. It is pertinent to remark that the Gasmann-
derived curve for the 100% water saturated coal compares favorably with the laboratory
measured P-wave velocity data for the water saturated coal, represented by the black
curve (Yu et al. 1993). The observed variations especially at lower differential pressure
might be due to dispersion effects and some numerical uncertainties.

Addition of methane (CH4) into the water-coal mixture decreases the P-wave velocity
(the red curve) due to the high compressibility of methane and density effect.
Quantitatively, the magnitude of the changes in P-wave velocity due to CH4 flooding
would depend partly on the degree of water/methane saturations and partly on pore

7
pressure. For shallow reservoir conditions (lower differential pressures) typical of the
Powder River Basin, the Gassmann predicted changes in P-wave velocity range from 5%
to 15% depending on pore pressure and the degree of methane/water saturations.
Furthermore, the observed changes in P-wave velocity are largest at low differential
pressure. As pointed out by Wang et al 1989, seismic velocities are more sensitive to
pore-fluid saturations at low differential pressures. To quantitatively and qualitatively
relate changes in P-wave velocity to changes in saturation and pore pressure, I extended
the Gassmann coal physics analysis to a sensitivity study. Figure 7 relates the computed
fluid density (RHOF), fluid bulk modulus (Kf), density of the saturated fluid (RHOBsat)
and saturated bulk modulus (Ksat) to water saturation (Sw) for the dry Vp when saturated
with CH4 and H2O. We used these parameters to estimate the saturated velocities (Vpsat)
at an assumed reference pressure of 5Mpa as shown in figure 7. Subsequently, we
calculated changes in Vp (Vp) as functions of changes in pressure (P) and water
saturation (Sw) at an assumed reference pressure of 5Mpa and water saturation of 0.5,
and plotted the results as shown in figure 8. The estimated Vp due to CH4 and H2O
saturations show sensitivity to both P and Sw (figure 8). The observed reduction in
Vp can be attributed to the high compressibility of methane.

While the Gassmann-predicted changes in P-wave velocity of the examined coal from Yu
et al 1993 provide necessary proof of concept and validation for the results of field
experiment, it might not be strictly valid for the Powder River Basin’s Big George coal.
Powder River Basin coals are shallow, of lower-rank and sub-bituminous. Coal
depending on its rank and geology exhibits significant variations in porosity and structure.

Figure 7: Plots of computed fluid properties versus Sw for Vpdry at reference pressure of 5Mpa

8
Figure 8: Estimated Vp as functions of P and Sw due to CH4 and H2O saturations (upper left and right
hand plots show P-wave velocity in the dry rock versus differential pressure and P-wave velocity in the
saturated rock as a function of water saturation. The reference pressure and saturation values are shown.

Crosswell Seismic Data Acquisition

Two observation wells, spanning 150ft and straddling a production well (CP 23-35), were
used for the crosswell surveys. The source well is labeled in Figure 9 by 23-35BG, while
the receiver well is labeled 23-35W. The baseline or reference survey was acquired
shortly after the production well was completed and before dewatering began in
December, 2002, while, the first monitoring survey was run in July, 2003. Each of the
surveys covered about 900ft to the total depth of the wells at about 1400ft. Shots were
fired from a down-hole piezoelectric source every 1.25 ft in the source well. The seismic
waves propagated between wells were picked up by an array of hydrophones positioned
inside the receiver well. Crosswell seismic profile is generally recorded either in form of
common source or common receiver “fans”. In the common receiver mode of recording,
the receiver is positioned at depth and the source is moved in the other borehole to create
the fan. The receiver is then repositioned and the source scan repeated. A typical common
receiver profile (CRP) from the CBM data is shown in figure 10, illustrating direct P-
wave and tube waves. In the common source profile (CSP), the source is positioned at
depth and the receiver is moved in the other observation well to create the fan. The
source is then repositioned and the receiver scan repeated. As shown in figure 11, the
CSP equally permits identification of direct P-wave. In both common source and

9
common receiver sorts; the direct arrival travel path increases hyperbolically with
increasing source to receiver offset, and as such the direct wave traveltime has hyperbolic
moveout. For large offsets, the direct and reflected events asymptote a straight line that
has moveout of the medium’s velocity (figure 12).

Figure 9: Crosswell Tomography Survey Geometry

Direct P-waves

Tube
Waves

Figure 10: Data plot (about 100 traces) in common receiver sort showing direct P-waves and tube waves.

10
Tube
Waves

Figure 11: Data plot (about 120 traces) in common source gather showing the direct arrival travel path as it
increases hyperbolically with increasing source to receiver offset. Tube waves are also present.

1/V

Figure 12: Data plot (about 90 traces) in a common source profile showing weak, far-offset direct P-waves
obscured by tube waves. The tube waves are recognized by their high amplitude, low propagation speed.

11
The crosswell seismic profile can also be viewed in common mid-point (CMP) and
common vertical offset (CVO) for ease of analysis and processing. In the CVO
arrangement, the direct arrival travel path is constant and therefore its moveout is zero. In
the CMP sort, direct arrivals have twice the hyperbolic moveout. The [CMP, CVO]
domain offers an advantage for both picking traveltimes and wavefoeld separation. In
picking, the CVO direct arrivals are maximally flat, while in CMP direct arrivals have
twice the moveout. The [CSP, CRP] and [CMP, CVO] comprise the two major domains
or four major data sets for analysis and processing crosswell seismic profiles. In each of
the three surveys that we used for this study, the data were acquired by first lowering the
source and detectors to their deepest positions. The detector array was kept stationary as
the source was raised and fired in motion every 1.25 feet with a sweep spectrum of 200-
2000 Hz. After the source reached its shallowest depth, it was again lowered as the
detector array is raised to its next position. Acquisition continued in this fashion until a
complete survey of 182 source positions and 116 detector positions were completed. The
survey took approximately 24 hours to complete.

Data Processing

A zero-offset gather from the data, shown in figure 13, provides useful stratigraphic
information. The coal has a lower seismic velocity (~7500 ft/sec) and is easily delineated
(1150’–1235’) by its larger travel time relative to the surrounding zones. Strong tube
waves are present in the data, though not easily seen in this gather. The same zero-offset
gathers (though rotated 1800) are shown for all the three data sets in figure 14 to illustrate
what I refer to as zero-offset repeatability test. This involves geophysical examination of
the differences between the surveys on a horizon basis. Intuitively; we expect geologic
events outside the reservoir level to be static (time-invariant), while seismic events within
the coal reservoir are expected to be dynamic (time-variant). I performed this zero-offset
repeatability analysis by putting red skeletal markers on horizons that are both within and
outside the reservoir level in survey 1 as shown in figure 14. We then projected same
onto surveys 2 and 3 respectively. As displayed in the resultant images, there is a
traveltime delay (a positive traveltime shift) within the reservoir level in survey 2 relative
to survey 1. And in survey 3, there is a negative traveltime shift within the reservoir level
relative to survey 2. The scenario is such that there is an increase in traveltime changes
between surveys 1 and 2 within the coal reservoir. In survey 3, the observed increase in
traveltime in survey 2 has disappeared and tends to approximate the traveltime observed
in survey 1 within the reservoir. Though these production-induced changes are small,
they are real and statistically significant for time-lapse processing and interpretation. It is
interesting to note that observed differences in horizons outside the reservoir are quite
negligible, showing that events outside the reservoir are repeatable.

I pre-processed the data to suppress the tube waves and conditioned the direct arrivals
for travel time picking of the P-wave first arrivals (figure 15). More than 21,000 seismic
traces were picked and processed with travel time tomography to produce the velocity
image between wells. The traveltime picking was systematically done and same part of
the first arriving wavelet was picked for different offsets to minimize picking errors and

12
ensure good convergence. I manually picked the P-wave first arrivals from each of the
data set (figure 15). The low-velocity coal layer can also be identified from the traveltime
maps (figure16). The traveltime picking was executed in four domains: common receiver,
common source, common offset and common mid-point gathers to ensure consistency.
Evidence of production-associated changes in traveltimes within the coal reservoir can be
seen in the traveltimes difference maps as shown in figure 17. P-wave travel-times
increase in survey 2 relative to survey 1 especially at the reservoir level due to production
effects (figure 17). There are also changes in P-wave traveltimes between surveys 1 and
3, and 2 and 3 respectively. Although these changes are not uniform within the
heterogeneous reservoir, they tend to be primarily controlled by pore-fluid saturation
effects occasioned by dewatering. Estimating velocities from these crosswell traveltimes
constitute traveltime tomography or velocity inversion. The basic, traditional traveltime
processing steps involve: picking of traveltimes; calculating traveltimes for an assumed
slowness model and inverting a matrix equation where the knowns are given by the
traveltime residuals (calculated minus observed traveltimes) to obtaining corrections to
the assumed slowness model (Harris et al. 1990, 1995).

Figure 13: (a) The zero-offset gather includes only sources and detectors at the same depth. (b) The coal
zone (1150’-1240’) is clearly delineated by its low velocity, i.e., higher travel time.

13
Figure 14: A display of zero-offset sections for the three data sets showing evidence of production-induced
changes within the coal reservoir and repeatability of events outside the reservoir.

Figure 15: Data plot of a common receiver gather showing typical direct P-waves’ picks (in green)

14
Figure 16: Maps of P-waves’ traveltime picks (in msec) for the three surveys. The presumably coal zone is
easily delineated through rays traveling from the soirce locations to the receiver locations.

15
Figure 17 (A to C): Maps of traveltime picks difference for the time-lapse surveys.

16
Theoretically, an observed data vector t (observed traveltimes) is a discreetly sampled
function of some specific model parameter vector S (the slowness or the reciprocal of the
velocity). And for a ray lm in a medium where the slowness is S(x,y), the traveltime along
the ray is traditionally given as

tm =  S(x, y )dlm m = 1, ..., n,


lm
(1)

Where dlm is the incremental distance along the raypath lm. In general the raypath and
traveltimes depend on the slowness distribution. In geophysics unlike medical imaging,
straight rays are rarely found and as a result, the problem in equation (1) becomes highly
Non-linear since the unknown S(x,y) is also implicitly present in the path (Nolet, 1987;
Peterson et al 1985). If the medium is perturbed to SI(x,y) = S(x,y) + S (x,y), the new
traveltime calculated along the new raypath lI is

t 1 =  1 S 1 ( x, y )dl 1 (2)
l

Using Femat’s principle, it can be shown (Aki and Richards, 1980) that the difference in
traveltimes between the two media is

Δt =  ΔS ( x, y )dl (3)
l

where Δt = t1 − t . The nonlinear problem is then solved as a sequence of linearized steps


that seeks to minimize the difference between real and calculated traveltimes.

Prior to the 2-D processing, I considered 3-D deviations of the wells by applying
necessary deviation corrections that allow for a projection of the 3-D deviations (in a N-E
coordinate) on to a 2-D plane (figures 18 and 19). This is necessary to ensure realistic
velocity estimates from the 2-D inversion. The well deviations can cause artificial
anisotropy that can impact on the fidelity of the 2-D tomography result for reliable
geologic interpretation of velocity model. For this study, the picked traveltimes were
imported into TOMOXpro where 2D tomography was performed. The algorithm
performs a non-linear traveltime inversion and reconstructs velocity models for any
survey geometry. Through an iterative process, it can obtain a reliable velocity model
(Zhang and Toksoz, 1998). A 1-D velocity model was used to start the 2-D inversion.
This 1-D initial model was constructed from zero-offset data (Figure 13). The resulting 1-
D starting model for the inversion is shown in Figure 20. This approach was used in
processing the baseline survey. For the crosshole tomography of the second and third
surveys, I used the output of the 2-D inversion for the baseline survey as the starting
model with a view to minimizing time-lapse noise and subsequently used the same
inversion algorithm to get their respective tomograms. The resulting 2-D tomogram for
the baseline survey is shown in figure 21, where we plotted the gamma ray logs for the
observation wells alongside to validate qualitative geologic interpretation in conjunction
with the baseline tomogram.

17
Figure 18: Schematic showing three dimensional deviations of the source (in red) and receiver (in blue)
wells. Their orientations or azimuths are in a North-East geographic coordinate system.

Figure 19: Schematic of the survey coverage and geometry of the two wells after deviation corrections
(thickly colored sections of the red and blue lines illustrate survey coverage within the subsurface).

18
Figure 20: 1-D velocity model (blue line) from zero-offset data. The red dots show the
picked zero offset points used to create the 1-D starting model for the inversion.

Figure 21: Baseline tomogram with gamma ray logs (insets are tomogram-derived
velocity logs).

19
A cross-plot of the baseline velocity data with the gamma ray data for each of the
observation wells further shows consistency in geologic interpretation in which with the
low-velocity and low-radioactive coal is easily identified (figure 22). The baseline and
repeat tomograms exhibit the same geologic characteristics (figure 23) and provide
images of the vertical and lateral changes in velocity. To permit quantitative estimation of
changes in P-wave velocity due to CBM production, I computed the differences between
baseline and repeat tomograms and obtained the results shown in figure 24.

Figure 22: Cross-plots of baseline velocity and gamma ray in the two wells showing subsurface
heterogeneity & spatial distribution of velocity within the coal layer and other subsurface strata

Data Interpretation

Four distinct geologic units can be identified from the baseline tomogram. They include
shaly-sand, sandy-shale, coal, and sandstone. Of interest to us is the low-velocity,
biogenic and low-rank Big George coal zone at a depth of around 1150ft to1240ft. The
coal aquifer is confined above by the low permeability sandy-shale. Qualitative
interpretation of the gamma ray logs plotted alongside the baseline tomogram (Figure 21)
shows geological features similar to the ones observed on the baseline and repeat
tomograms (figure 23). This good correlation corroborates the quality of the tomographic
inversion process. The difference tomograms are shown in figure 24. The 1st difference

20
tomogram (Vp1) was computed from the derived tomograms for the baseline survey of
December 2002 and 1st repeat survey of July 2003 (after about 8 months of dewatering).
The 2nd difference tomogram (Vp2) was computed from the derived tomograms for the
baseline survey of July 2002 and second repeat survey of June 2004 (after about 19
months of continuous dewatering).

Figure 23: Baseline and Repeat Tomograms (from left: survey 1, survey 2 and survey 3).

Also, I calculated the 3rd difference tomogram (Vp3) from the repeat tomograms of the
2nd survey (July 2003) and 3rd survey (June 2004). The 1st difference tomogram (Vp1)
shows a reduction in Vp of about 4% to 5% within the coal zone, perhaps due to partial
gas saturation and/or methane desorption. Both the 2nd and 3rd tomograms (Vp2 and
Vp3) also show quantifiable reductions in Vp (between 2% to 5%) in some portions of
the coal zone due to gas saturation. However, continued depressurization and subsequent
increase in differential pressure (confining pressure/overburden pressure minus pore
pressure) coupled with gas saturation might have influenced the observed changes in P-
wave velocity of about 0.5% to 6% in other portions of the 2nd and 3rd difference
tomograms. These changes are not uniform due to the heterogeneous nature of the
confined coal aquifer. Interestingly, the observed changes in pore pressure (figure 25) are
so small to have had any considerable effect on the P-wave velocity in the saturated coal.
As shown in figure 25, the pore pressure changes by 77 Psi (0.5Mpa) between the 1st and

21
2nd surveys, and by 7 psi (0.09Mpa) between the 2nd and 3rd surveys. These changes are
small and appear not to have much effect on changes in P-wave velocity.

Figure 24: Measured difference tomograms from the time-lapse surveys. From left is the 1st difference
tomogram, Vp1 (survey2 minus survey 1); 2nd difference tomogram, Vp2 (survey 3 minus survey 1) and
3rd difference tomogram, Vp3 (survey 3 minus survey 2). Arrow shows the location of the production well.

On the other hand, the measured gas pressure data in figure 26 shows that gas production
increases between the 1st and 2nd surveys, justifying the observed reduction in P-wave
velocity as observed on the 1st difference tomogram. And between the 2nd and 3rd surveys,
gas production reduces, corroborating the observed increase in P-wave velocity observed
on the 2nd and 3rd difference tomograms. This shows that the changes in P-wave velocity
in the saturated coal depend primarily on gas saturation occasioned by the dewatering
process. With continued dewatering, pore pressure reduces causing the closing of
presumably layer cavities or air-filled cracks (Terry 1959) in the coal. This allows the
signals to travel more efficiently through the coal with greater velocity and less
attenuation, more so with the observed slight reduction in gas production between the 2nd
and 3rd surveys.
There are also changes both above and below the coal zone. These changes which are
approximately 6% might be due to pressure-related anomalies and fluid drainage and /or
migration through hydraulically active fractures (Colmenares and Zoback, 2003) to
surrounding porous and permeable beds. Also, the observed changes above the coal zone
might be due to water coming from the overlying, low-permeability sandy-shale due to

22
depressurization effects. There is no gas in this caprock, suggesting that the observed
increases in P-wave velocity above the reservoir level are due to water saturation.

Figure 25: Water Head Pressure as a function of time

The scenario is such that besides the gas that is migrating from the reservoir to the
surface through the production well, water is also flowing in through the overlying sandy-
shale. This phenomenon tends to undermine the spatial efficiency of the dewatering
process. Because the changes are so small, there is also some possibility for inversion
artifacts in the difference tomograms. Besides, residual differences in the repeat time-
lapse surveys that are independent of changes in subsurface geology could cause time-
lapse noise and impact the desired repeatability of the time-lapse surveys. These
differences can arise from geometry variations in the repeat surveys, hydrophone
positioning and recording fidelity differences. The observed time-lapse changes agree
reasonably well with predictions based on Gassmann’s equation (figures 27 and 28),
underscoring the effectiveness and applicability of high resolution crosswell seismic for
monitoring.

In creating the model shown in figures 27 and 28, I remodeled the laboratory data from
Yu et al. (1993) to match the prevailing geo-reservoir conditions at the PRB and the
observed baseline coal velocities. Prior to dewatering, the P-wave velocity in the water-

23
Figure 26: Measured Gas Pressure as a function of time

Figure 27: Velocity in saturated coal as a function of gas saturation. Fully water saturated velocity at
around 7500ft/s drops to around 6900ft/s due to gas saturation (remodeled data from Yu et al 1993)

24
saturated coal seam (average baseline velocity) was around 7500ft/s and about 8% to
15% higher than the velocity in the dry coal. At the initial stage of depressurization, the
P-wave velocity increases due to a reduction in pore pressure and an increase in
differential pressure. As differential pressure goes up, the number of cracks and the thin
gaps contributing to attenuation decreases, leading to a rise in Vp. Also; with methane
desoption, the coal matrix shrinks and the velocity drops due to partial gas saturation and
more so that methane has high compressibility (low bulk modulus). This initial gas

Figure28: P-wave velocity’s behavior in saturated coal as a function of differential pressure.

drainage at a pore pressure change of about 77Psi or 0.5Mpa produces a negative velocity
difference of about 6% (i.e. a velocity of about 7050ft/s) when compared with the
baseline velocity (of 7500ft/s) as shown in figures 27, 28 and 32.. However, with
continued dewatering and depending on the degree of pore-fluid saturation (gas), the
velocity tends to increase again (about 7850ft/s) with a change in pore pressure of around
7Psi or 0.09Mpa. Consequently, the velocity differences between the fully saturated
water condition and the subsequent stages of gas-saturated state might show a positive
increase following several months of dewatering or depressurization (figures 21, 23 and
24). And this observation is true for our case in that a reduction in gas production level in
survey 3 as shown in figure 26 impacted on the P-wave velocity. This provides a physical,
corroborative evidence for the observed time-lapse changes in the 2nd and 3rd difference
tomograms.

25
Conclusions

I have imaged and quantitatively characterized production-induced changes in P-wave


velocity at the Powder River Basin’s Big George Coal using time-lapse crosswell seismic
surveys. The results show that the spatial distribution of the P-wave velocity changes
approximates 6% both within and outside the coal layer. This difference is due to a
combination of gas saturation and changes in pore pressure. Besides, we observed that
the magnitudes of the P-wave velocity changes depend largely on the degree of changes
in gas production. Though these changes are small and non-uniform due to the
heterogeneous nature of the coal aquifer, they are statistically significant for time-lapse
monitoring. The observations further suggest that produced water comes from the
overlying sandy shale and that pressure drawdown in the coal is not spatially efficient.

References

AKI, K., and RICHARDS, P.G. 1980, Quantitative Seismology: Theory and Methods,
Methods, W.H. Freeman and Co., San Francisco, 932 pp.

AKINTUNDE, O.M., HARRIS, J.M. and QUAN, Y. 2004, Time-lapse seismic


monitoring of Coal Bed Methane production: A case study from the Powder River Basin.
A paper presented at the 2004 SEG International Meeting, Denver, USA.

AYERS, W.B. 2002, Coalbed gas systems, resources, and production and a review of
contrasting cases from the San Juan and Powder River Basins. AAPG Bulletin, V.86,
No.11, PP.1853-1890

COLMMENARES, L.B. and ZOBACK, M.D.2003, Review of wellbore completion


methods of CBM wells in the Powder River basin.

DEBRUIN, R.H., LYMAN, R.M., JONES, R.W. and COOK, L.W. 2000, Information
Pamphlet 7. Wyoming State Geological Survey.

FLORES, R.M. 2004, Coal Bed Methane in the Powder River Basin, Wyoming and
Montana: An Assessment of the Tertiary-Upper Cretaceous Coalbed Methane Total
Petroleum System. USGS Digital Data Series DDS-69-C Version 1.0.

GASSMAN, F. 1951, Elastic waves through a packing of sphere. Geoph.16, 673-685.

GREAVES, R.J., and FULP, T.J., 1987, Three-dimensional seismic monitoring of an


enhanced oil recovery process: Geophysics, 52, 1175-1187.

HARRIS, J.M, LAZARATOS, S. and MICHELENA, R., 1990, Tomographic string


inversions, 1990 SEG Meeting, San Francisco.

26
HARRIS, J.M, NOLEN-HOEKSEMA, R.G., LANGAN, R.T., VAN-SCHAACK, M.,
LAZARATOS, S.K., and RECTOR, J.W., 1995, High Resolution Cross Well Imaging of
a West Texas Carbonate Reservoir: Part1- Project Summary and Interpretation:
Geophysics 60, 667-681.

HARRIS, J.M., AKINTUNDE, O.M, and URBAN, J., 2004, A feasibility study for CO2
monitoring in coal. A paper presented at the 2004 AAPG International Conference,
Dallas, USA.

JENKINNS, S.D., WAITE, M.W., and BEE, M.F., 1997, Time-lapse monitoring of the
Duri steamflood: A pilot and case study: The Leading Edge, 16, 1267-1273.

JOHNSTAD, S.E., SEYMOUR, R.H., and SMITH, P.J., 1995, Seismic reservoir
monitoring over the Oseberg field during the period 1989-1992: First Break, 13, 169-183.

KING, J.G. and WILKINS 1994, Proc. Conf. Ultrafine Struct. Coals and Cokers,
BCURA, London. P.46.

LANDRO, M. 2001, Discrimination between pressure and fluid saturation changes from
time-lapse seismic data: Geophysics 66, 836-844.

LANDRO, M., VEIRE, H.H., DUFFAUT, K and NAJJAR, D. 2003, Discrimination


between pressure and fluid saturation changes from marine multicomponent time-lapse
seismic data. Geophyiscs, Vol. 68, No.5. P.1592- 1599.

LARSON, V.E. 1989, Preliminary evaluation of Coal Bed Methane Geology and
Activity in the Recluse Area, Powder River Basin, Wyoming, Quarterly Review of
Methane from Coal Seams Technology.

LUMLEY, D.E., 1995a, Seismic time-lapse monitoring of subsurface fluid flow: PhD
thesis, Stanford University.

LUMLEY, D.E., 1995b, 4-D seismic monitoring of an active steamflood: 65th Ann.
Internat. Mtg., Soc. Expl. Geophys., Expanded Abstracts, 203-206.

MAVKO, G., MUKERJI, T. and DVORKIN, J. 1998, The Rock Physics Handbook:
Tool for Seismic Analysis in Porous Media.

MAVOR, M.J., GUNTER, W.D., ROBINSON, J.R., LAW, D.H., and GALE, J., 2002,
A Paper presented at the Society of Petroleum Engineers (SPE) Gas Technology
Symposium held in Calgary, Alberta, Canada.

NOLET, G. 1987, Seismic tomography: With applications in global seismology and


Exploration geophysics: D. Reidel Publishing Co.

27
NUR, A., 1989, Four-dimensional seismology and (true) direct detection of
hydrocarbons: The petrophysical basis: The Leading Edge, 8, no. 09, 30-36.

NUR, A, TOSAYA, C.and THANH, D.V., 1984, Seismic monitoring of thermal


enhanced oil recovery processes: 54th Ann. Internat. Mtg., Soc. Expl. Geophys.,
Expanded Abstracts, session RS.6.

PETERSON, J.E; PAULSON, B.N.P; and MCEVILLY, T.V. 1985, Application of


Algebraic reconstruction techniques to crosswell seismic data: Geophysics, 1566 – 1580.

PRASAD, M., 2001, Mapping impedance microstructure in rocks with acoustic


microscopy: The Leading Edge, Vol.20, no.2, PP. 172-179

PULLIN, N., MATTHEWS, L., and HIRSCHE, K., 1987, Techniques applied to obtain
very high resolution three-dimensional seismic imaging at an Athabasca tar sands thermal
pilot: The Leading Edge, 6, no.12, 10-15.

RICHARDSON, S.E. and LAWTON, D.C. 2002, Time-lapse seismic imaging of


enhanced coal bed methane production: a numerical modeling study: CREWERS
Research Report 14: 17.1 – 17.13.

RANDALL, A.G. 1991, Shallow Tertiary Gas Production, Powder River Basin of
Wyoming’s Coal Bed Methane.

RICE, C; ELLIS, M. and BULLOCK, J 2000, Water Co-Produced with Coal Bed
Methane in the Powder River Basin, Wyoming: Preliminary compositional data: U.S.
Geological Survey Open- File Report. P.20

TERRY, N.B.1959, Dependence of elastic behavior of coal on microcrack structure. Fuel


3, 125-146.

TURA, A., and LUMLEY, D.E., 1998. Subsurface fluid flow properties from time-lapse
elastic wave reflection data: Mathematical Methods in Geophysical Imaging, 5, 125-138.

TURA, A., and LUMLEY, D.E. 1999, Estimating pressure and saturation changes from
time-lapse AVO data: 69th Ann. Internat. Mtg., Soc. Exp. Geophys., Expanded Abstracts,
1655-1658.

YU, G., VOZOFF, K. and DURNEY, D.U. 1993, The influence of confining pressure
and water saturation on dynamic elastic properties of some Permian coals. Geophysics
58: 30-38

WANG, Z., and NUR, A. 1986, Effect of temperature on wave velocities in sands and
sandstones with heavy hydrocarbons: 56th Ann. Internat. Mtg. Soc. Expl. Geophys.,
Expanded Abstracts, session BHG1.2

28
WANG, Z., MICHAEL, E.C. and LANGAN, R.T. 1998, Seismic monitoring of a CO2
flood in a carbonate reservoir. Geophysics Vol.63, No.5. PP. 1604-1617.

ZHANG, J. and TOKSOZ, M.N. 1998, Non-linear refraction travel time tomography.
Geophysics 63: 1726-1737.

29

Potrebbero piacerti anche