gp170/2001 #9 la cira case study: upscalingpangea.stanford.edu/~jack/gp170/gp170#9.pdf ·...

11
1 La Cira Case Study: Upscaling GP170/2001 #9 Rock physics reservoir characterization is based on applying rock physics relations, such as between velocity and porosity, to a volume of seismic velocity or impedance resulting from seismic inversion. The relevant rock physics laws are often derived from well log data. The spatial scale of log measurements is much smaller than the seismic scale. It is important to upscale rock physics relations used or to make sure that they hold at the seismic scale. Well 1803 in La Cira field (Colombia) is within a good-quality seismic coverage. We have used the seismic data to obtain velocity inversion in a strip that contains this and other wells (Figure 1), using Hampson-Russell software package with the well 1803 sonic log for initial velocity model. 1803 1863 1865 1783 1786 1787 STRIP 1 Xlines 1365-1375 Inlines 5650-5745 N S Sonic well used for the initial velocity model during inversion Control wells Figure 1. Geometry of the strip in La Cira Field where seismic inversion has been conducted. First consider well 1803 from where good quality velocity and shalines log data are available (Figure 2). It is clear from the graph that low velocity corresponds to low shale content and high velocity corresponds to high shale content. Let's relate one to the other in order to obtain a relation usable for reservoir characterization. The resulting crossplot is given in Figure 3. Apparent is the absence of a meaningful trend. One possible reason for the absence of a trend is the apparent difference in the frequency of data presentation in Figure 2: the shale content data has smaller spatial frequency than the velocity data. 3500 4000 4500 0 1 3 4 LC 1803 LOG DATA VSHALE Vp (km/s) DRKB (ft) Figure 2. Shale content and compressional-wave velocity versus depth in well 1803, La Cira. 0 0.2 0.4 0.6 0.8 1 2.8 3 3.2 3.4 3.6 3.8 VSHALE Vp (km/s) LC 1803 LOG DATA Figure 3. Shale content versus velocity as cross plotted from well log curves given in Figure 2.

Upload: others

Post on 02-Jan-2020

4 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: GP170/2001 #9 La Cira Case Study: Upscalingpangea.stanford.edu/~jack/GP170/GP170#9.pdf · 2001-01-30 · 3 La Cira Case Study: Upscaling GP170/2001 #9 The inversion results for well

1

La Cira Case Study: UpscalingGP170/2001 #9

Rock physics reservoir characterization is based on applying rockphysics relations, such as between velocity and porosity, to avolume of seismic velocity or impedance resulting from seismicinversion. The relevant rock physics laws are often derived fromwell log data. The spatial scale of log measurements is muchsmaller than the seismic scale. It is important to upscale rockphysics relations used or to make sure that they hold at the seismicscale.

Well 1803 in La Cira field (Colombia) is within a good-qualityseismic coverage. We have used the seismic data to obtain velocityinversion in a strip that contains this and other wells (Figure 1),using Hampson-Russell software package with the well 1803 soniclog for initial velocity model.

1803

1863

1865

1783

1786

1787

ST

RIP

1

Xlines 1365-1375Inlines 5650-5745

N

S

Sonic well used for the initial velocity model during inversion

Controlwells

Figure 1. Geometry of the strip in La Cira Field whereseismic inversion has been conducted.

First consider well 1803 from where good quality velocity and shalines log data are available(Figure 2). It is clear from the graph that low velocity corresponds to low shale content andhigh velocity corresponds to high shale content. Let's relate one to the other in order toobtain a relation usable for reservoir characterization. The resulting crossplot is given inFigure 3. Apparent is the absence of a meaningful trend. One possible reason for theabsence of a trend is the apparent difference in the frequency of data presentation in Figure2: the shale content data has smaller spatial frequency than the velocity data.

3500

4000

45000 1 2 3 4

LC 1803LOG DATA

VSHALE Vp (km/s)

DR

KB

(ft

)

Figure 2. Shale content andcompressional-wave velocityversus depth in well 1803, LaCira.

0

0.2

0.4

0.6

0.8

1

2.8 3 3.2 3.4 3.6 3.8

VS

HA

LE

Vp (km/s)

LC 1803LOG DATA

Figure 3. Shale content versusvelocity as cross plotted from welllog curves given in Figure 2.

Page 2: GP170/2001 #9 La Cira Case Study: Upscalingpangea.stanford.edu/~jack/GP170/GP170#9.pdf · 2001-01-30 · 3 La Cira Case Study: Upscaling GP170/2001 #9 The inversion results for well

2

La Cira Case Study: UpscalingGP170/2001 #9

By applying a running mean average to log data, we smooth them asshown in Figure 4. Now we can cross plot these smoothed data byplotting only selected (every 10th) data point. A clear trend appears inthe data that can be used to predict shale content from seismicvelocity inversion (Figure 5).

The main question to be addressed here is whether this shale contentversus velocity trend can be used for interpreting seismic inversionvelocity in terms of shale content for a volume of seismic data.

Figure 4. Running mean average applied to well logcurves given in Figure 1.

0 1 2 3 4

3500

4000

4500

LC 1803LOG DATA

SMOOTHED

VSHALE Vp (km/s)

DR

KB

(ft

)

2.8 3 3.2 3.4 3.6 3.80

0.2

0.4

0.6

0.8

1

Vp (km/s)

LC 1803LOG DATA

SMOOTHEDwith

SELECTED PLOTTING

VS

HA

LE

Figure 5. Cross plotting smoothed log data shown inFigure 3.

Page 3: GP170/2001 #9 La Cira Case Study: Upscalingpangea.stanford.edu/~jack/GP170/GP170#9.pdf · 2001-01-30 · 3 La Cira Case Study: Upscaling GP170/2001 #9 The inversion results for well

3

La Cira Case Study: Upscaling

GP170/2001 #9

The inversion results for well 1803 are shown at every 4 ms are plotted versus the two-way travel time in Figure 6a, together with the shale

content values given at every 4 ms.

Cross plotting the shale content versus the inversion velocity does not result in any meaningful trend (Figure 7a).

However, smoothing the shale and velocity data in Figure 6a (Figure 6b) shows that the log velocity and the inversion velocity are essentially the

same (as should be expected) in this case because the sonic well log serves as the initial velocity model.

Moreover, a meaningful shale content versus velocity trend appears if these smoothed data are cross plotted (Figure 7b). This trend is precisely

the same as obtained form the smoothed log data shown in Figure 5.

Figure 6. Well 1803. a. Shale content, velocity inversion (red) and velocityfrom well log shown at every 4 ms versus two-way travel time. b. Same data,smoothed.

750

800

850

0 1 2 3 4

LC 1803LOG DATA

andINVERSION

VSHALE Vp (km/s)

TW

T (m

s)

InversionWell Log

a0 1 2 3 4

LC 1803LOG DATA

andINVERSIONSMOOTHED

VSHALE Vp (km/s)b

0

0.2

0.4

0.6

0.8

1

2.8 3.0 3.2 3.4 3.6 3.8

VS

HA

LE

Vp (km/s)a2.8 3.0 3.2 3.4 3.6 3.8

Vp (km/s)b

Figure 7. Well 1803. a. Shale content versus velocity inversion based on theFigure 5a data. b. Shale content versus velocity inversion based on thesmoothed data from Figure 5b (black) with superimposed trend from Figure 4.

Page 4: GP170/2001 #9 La Cira Case Study: Upscalingpangea.stanford.edu/~jack/GP170/GP170#9.pdf · 2001-01-30 · 3 La Cira Case Study: Upscaling GP170/2001 #9 The inversion results for well

4

La Cira Case Study: Upscaling

GP170/2001 #9

Let us now consider well 1865 where a good shale content log is available, but the sonic data are poor (for this reason, well 1865 has not beenused as an initial velocity model for seismic inversion).

The inversion velocity for well 1865 is shown versus the two-way travel time in Figure 8a, together with the shale content at every 4 ms. Somecorrelation between the inversion velocity and the shale content is apparent. However, this correlation does not result in a clear shale contentversus velocity trend (Figure 9a). Smoothing the data with a running mean average improves the correlation (Figure 8b). Cross plotting placesthe resulting shale content versus velocity trend on top of the trends obtained for well 1803.

Figure 8. Well 1865. a. Shale content, velocity inversion (red) and velocityfrom well log shown at every 4 ms versus two-way travel time. b. Same data,smoothed.

Figure 9. Well 1865. a. Shale content versus velocity inversion based on theFigure 8a data. b. Shale content versus velocity inversion based on thesmoothed data from Figure 8b (black) with superimposed trends for well 1803from Figure 7b.

0 0.5 1 3 3.5 4

725

750

775

800

825

VSHALE Vp (km/s)

TW

t (m

s)

Well 1865Velocity

Inversion

a0 0.5 1 3 3.5 4

VSHALE Vp (km/s)

Well 1865Velocity

InversionSmoothed

b

2.8 3.0 3.2 3.4 3.60

.2

.4

.6

.8

1

Vp (km/s)

Well 1865

VS

HA

LE

a2.8 3.0 3.2 3.4 3.6

Vp (km/s)

1865

b

Simple smoothing can take care of carrying a rock physics relation from the log to seismic scale.

Page 5: GP170/2001 #9 La Cira Case Study: Upscalingpangea.stanford.edu/~jack/GP170/GP170#9.pdf · 2001-01-30 · 3 La Cira Case Study: Upscaling GP170/2001 #9 The inversion results for well

5

La Cira Case Study: Production Monitoring(1) Rock Physics Model

40 60 80

890

900

910

GR

Dep

th (m

)

1

2

#1880

2 3 4Vp (km/s)

0 0.1 0.2 0.3Porosity

Figure 1. Well LC-1880. Two up-finingcycles chosen for rock physics modeling.Cycle 1 on top and Cycle 2 below. Figure 2. Well LC-1880. Two up-fining cycles chosen for rock physics modeling, Cycle 1 (left)

and Cycle 2 (right). Velocity versus porosity, data and model lines. The data are color-codedby GR values. The model curves are for 100% water saturation (we assume that the sonicdata come from the mud-filtrate-invaded zone). The differential pressure used in the model is12 MPa. Mineralogy used in the model is given in the plot.

To develop a rock physics model that links porosity, mineralogy, pressure, and pore fluid bulk modulus to the elastic rock properties, use welllog data from La Cira well LC-1880. Consider an interval between 885 and 915 m (Figure 1) where two up-fining depositional cycles arehighlighted in the GR track.

Velocity is plotted versus porosity for Cycle 1 and Cycle 2 in Figure 2. The curves in these two plots come from the uncemented (friable) sandmodel. We can see that as the clay content in the rock increases, the data points move from the clean-sand line to the shaly-sand line.Although the GR values are about the same in Cycles 1 and 2, the model line that describes the shaly part of Cycle 2 requires more clay asinput than in Cycle 1. This modeling result is consistent with the core description that states that the shaly part in Cycle 1 is siltstone whereasthat in Cycle 2 is claystone.

GP170/2001 #9

Page 6: GP170/2001 #9 La Cira Case Study: Upscalingpangea.stanford.edu/~jack/GP170/GP170#9.pdf · 2001-01-30 · 3 La Cira Case Study: Upscaling GP170/2001 #9 The inversion results for well

6

La Cira Case Study: Production Monitoring(2) Rock Physics Model for Fluid Identification

Figure 3. Well LC-1882. Vp versus Vs from log data, color-coded by water saturation. The two theoretical lines are for100% water saturation (red) and 100% oil saturation (blue).Also superimposed are five empirical lines.

The well log Vp is plotted versus Vs in Figure 3. The data points arecolor-coded by water saturation. The data that correspond to oil-saturated reservoir sand fall in between two theoretical curvescomputed for 100% water and 100% oil saturation, respectively. Thefive different empirical Vp versus Vs lines also separate the water-saturated from oil-saturated data. The advantage of the theoreticaltreatment of the fluid detection problem is that one can explicitlyspecify the reservoir conditions and pore-fluid properties that is notstraightforward in empirical relations. A similar plot for core data fromwell LC-1882 is given in Figure 4. The friable sand model allows forpore-fluid detection in LCI.

GP170/2001 #9

Figure 4. Well LC-1882. Vp versus Vs from core data.The original velocities have been measured on room-drysamples (black). The 100% water saturation and 100%oil saturation data points have been computed usingGassmann’s fluid substitution.

1

3

5

7

9

11

13

1 2 3 4 5 6 7

Clean GasClean OilClean BrineClay Brine

S-I

mped

ance

P-Impedance

Friable Sand ModelLa Cira

HYDROCARBONS

Page 7: GP170/2001 #9 La Cira Case Study: Upscalingpangea.stanford.edu/~jack/GP170/GP170#9.pdf · 2001-01-30 · 3 La Cira Case Study: Upscaling GP170/2001 #9 The inversion results for well

7

Figure 5. Production rate versus time forwell LC-1882. Large increase inproduction in 1995 is due to theopening of the La Cira Sands interval.

La Cira Case Study: Production Monitoring(1) Production History in Producing Intervals

GP170/2001 #9

Well LC-1882 was completed in ZoneC with open-hole well logmeasurements, including monopolesonic, collected in 1988. Seven yearslater, Zone C in this well was closed,and new cased-hole monopole anddipole sonic logs were recorded. Anew interval (La Cira Sands) aboveZone C was open for production(Figure 5).

0

40

80

120

160Oil (bbls/d)Water (bbls/d)Gas (scf/d)

Year

Pro

du

ctio

n R

ate

89 90 91 92 93 94 95 96

ProductionfromC zone

ProductionfromLa Cira Sands

1995Monopole and DipoleAcquisition

1988MonopoleAcquisition

In Figure 6 we plot shale content, the Vp/Vs ratio, and Vp versus depth for Zone C sands from the early(1988) and recent (1995) measurements. Superimposed on the Vp/Vs track are two model curves, one for100% water and the other for 100% oil saturation. The intervals where the Vp/Vs data curve falls in betweenthe two theoretical model curves (for 100% water and 100% oil saturation, respectively) correspond to thecompleted hydrocarbon-bearing intervals from which oil was produced between 1988 and 1995.

Figure 6. Well LC-1882, Zone C,interval open to production. (a) Shalevolume versus depth with barsindicating completed intervals. (b)Vp/Vs ratio versus depth. The red andblue curves are from our model for100% water and 100% oil saturation,respectively. The black data curve isfrom the 1995 data. The part of thedata curve that falls below the 100%water saturation curve is shown in gray.(c) Vp versus depth. The black and thered curves are for the 1988 and 1995data, respectively.

In the Vp track, a drop in the velocityis observed between the years 1988and 1995. The observed difference,and, in general, small velocity values,also tie in well with the intervals openfor production. The observed largeVp decrease in some of the intervalsis most likely due to the addition offree gas to the original fluid systemas the reservoir pressure falls belowthe bubble point. Notice that theintervals with free gas correspond tothose where the 1995 Vp/Vs data fallbelow the 100% oil saturation curve.

Page 8: GP170/2001 #9 La Cira Case Study: Upscalingpangea.stanford.edu/~jack/GP170/GP170#9.pdf · 2001-01-30 · 3 La Cira Case Study: Upscaling GP170/2001 #9 The inversion results for well

8

La Cira Case Study: Production Monitoring(1) Production History in Producing Intervals

GP170/2001 #9

The same hydrocarbon indicator technique is applied to the La CiraSands interval that was closed to production (Figure 7). As expected, therepeated Vp data show no temporal changes. At the same time, theVp/Vs ratio in the data clearly delineates (by falling between the twomodel lines) the intervals that were chosen for completion independently,based on the open-hole resistivity data (see water saturation in the firstframe of Figure 8). The large production rate increase shown in Figure 6after 1995 is due to oil production from these intervals.

Figure 7. Well LC-1882, La Cira Sands,interval initially closed to production.Same as Figure 6 with water saturationcurve added to the shale content track.

CONCLUSION

The case study based on LCI datashows that first-principle-basedtheoretical rock physics modelingcan be used to interpret P- and S-wave well log data for hydrocarbondetection behind the casing influvial Tertiary sands. The samerock physics approach should bevalid for identifying hydrocarbonsin LCI and other fields with similargeologic setting from surfaceseismic if pre-stack or offset-stackdata are available. The rockphysics interpretation of dipolesonic logs appears to be apromising methodology for thedetection of bypassed and untappedoil, and pay intervals behind thecasing in mature fields such as LCI.

Well log data also show that anoticeable change in Vp can beexpected in produced intervals.Such changes, combined with rockphysics interpretation can be abasis for seismic reservoirmonitoring.

Page 9: GP170/2001 #9 La Cira Case Study: Upscalingpangea.stanford.edu/~jack/GP170/GP170#9.pdf · 2001-01-30 · 3 La Cira Case Study: Upscaling GP170/2001 #9 The inversion results for well

9

APPENDIX: Vp/Vs Relations for Fluid DetectionTypically, these relations calculate Vs from Vp in water-saturated rock. It is assumed

that if the measured Vp falls below the Vp versus Vs line, rock has hydrocarbon

GP170/2001 #9

The shear-wave velocity (or the shear modulus) is a necessary input parameter for full waveform elastic modeling

of the seismic response. Some of the rock physics models discussed here provide both bulk and shear modulus.

However, our experience shows that the shear-wave velocity theoretical curves often do not match the data (given

that the rock physics diagnostic has been done using the compressional modulus). For this reason, we p resent here

Vp/Vs relations to be used for calculating Vs from the compressional-wave data (or theoretical curves).

1.1. Castagna et al. (1993) Mudrock Equation:

VS = 0.862VP − 1.172 km / s .

This equation has been derived from in-situ data for water-saturated shales. The common saturation fluid here is

formation water.

1.2. Castagna et al. (1993) Equations for Limestones and Dolomites:

Limestone: VS = −0.055VP2 +1.017VP − 1.031(km/s);

Dolomite: VS = 0.583VP − 0.0789 (km/s).

Both equations are based on laboratory measurements conducted on water-saturated samples. The common

saturation fluid has to be selected accordingly.

1.3. Krief et al. (1990):

VP − sat2 − VFl

2

VS −sat2 =

VP−mineral2 − VFl

2

VS− mineral2 ,

where VP− sat a nd VS− sat are the compressional- and shear-wave velocity in the saturated rock, respectively;

VP− mineral and VS− mineral are the compressional- and shear-wave velocity in the mineral phase of the rock,

respectively; and VFl is the velocity in the fluid.

Page 10: GP170/2001 #9 La Cira Case Study: Upscalingpangea.stanford.edu/~jack/GP170/GP170#9.pdf · 2001-01-30 · 3 La Cira Case Study: Upscaling GP170/2001 #9 The inversion results for well

10

Vp/Vs Relations forFluid Detection

GP170/2001 #9

1.4. Greenberg and Castagna (1992) Method: This empirical formula (below) is based on a n umber of datasets

where ultrasonic velocity measurements were conducted on pure water (the common saturation fluid) saturated rocks.

To calculate Vs from Vp for other fluids use Gassmann's equation.

VS =

1

2{[ Xi

i =1

4

∑ aijVPj

j =0

2

∑ ] + [ Xii =1

4

∑ ( aijVPj )−1

j= 0

2

∑ ]−1 }, X ii =1

4

∑ = 1;

where Xi is the volume fraction of pure mineral constituents in the mineral phase with "1" standing for sandstone

(quartz); "2" for limestone (calcite); "3" for dolomite; and "4" for shale. The empirical coefficients aij are tabulated

below:

i Mineral ai 2 ai1 ai 01 Sandstone 0 0.80416 -0.85588

2 Limestone -0.05508 1.01677 -1.03049

3 Dolomite 0 0.58321 -0.07775

4 Shale 0 0.76969 -0.86735

1.5. Williams (1990) Method:

VP /VS = 1.182 + 0.00422∆tS Sands,

VP /VS = 1.276 + 0.00374∆tS Shales.

where ∆tS is in µs/ft.

To convert the Williams equations into another form, we use ∆tS = 304.8 / VS, where VS is in km/s. Then

we have

VS = 0.846VP −1.088 km / s Sands,

VS = 0.784VP − 0.893km / s Shales.

Compare this to Castagna's mudrock line: VS = 0.862VP − 1.172 km / s.

These equations can be also re-written to express the Poisson's ratio through Vs. For example, for the Williams

Sandstone equation we have:

VP

VS

= 1.182 + 1.29

VS

, ν = 1

2

(VP / VS )2 − 2

(VP / VS )2 − 1

.

Page 11: GP170/2001 #9 La Cira Case Study: Upscalingpangea.stanford.edu/~jack/GP170/GP170#9.pdf · 2001-01-30 · 3 La Cira Case Study: Upscaling GP170/2001 #9 The inversion results for well

11

Vp/Vs Relations forLithology PredictionEXAMPLE: MARLS

GP170/2001 #9

2

3

3 4 5 6

Vs

(km

/s)

Vp (km/s)

MudrockLine

LimestoneCurves

Well A

Well B

WilliamsShale

2.0 2.2 2.4 2.6RHOB

0 0.2 0.4 0.6Porosity

PhiRHO

1 2 3 4 5Velocity (km/s)

VpVs

2 4 6 8 10 12 14P-Impedance

0.2 0.3 0.4Poisson's Ratio

Well A

25 50 75100GR

Dep

th (m

)

100 m

~ 1800 m

MARL

1 10Resistivity

2.0 2.2 2.4 2.6RHOB

0 0.2 0.4 0.6Porosity

PhiRHONPHI

1 2 3 4 5Velocity (km/s)

VpVs

2 4 6 8 10 12 14P-Impedance

0.2 0.3 0.4Poisson's Ratio

20 40 60 80GR

Dep

th (m

)

100 m

~ 1800 m

MARL?

Well B