Using multicomponent seismic for reservoir ...

This paper shows part of the effortin Venezuela during the last eightyears to record, process, and interpretmulticomponent seismic data. Wepresent the main results of three pro-jects aimed mainly to help other stud-ies in fracture, lithology, and heavyoil characterization in different typesof geologic environments.

Maporal Field. This field, in westernVenezuela in the Barinas-Apure Basin,produces primarily from a fracturedcarbonate reservoir at a depth of 10 000ft. The area is structurally simple. Inearly 1994, Pdvsa recorded in this fieldthe first 3-C/2-D multicomponent sur-vey in Venezuela to obtain informationabout fracture orientation and densityat the reservoir level that could helpdesign trajectories of horizontal wells.

The survey was designed to inter-sect existing well control and maxi-mize data quality with respect toresolution and signal-to-noise ratio.Three 10-km 2-D multicomponentlines were centered over the reser-voir along three azimuths. Lines 1and 3 were parallel to the field’s twomain fault systems. Line 2 bisects theangle between lines 1 and 3, formingan angle of approximately 41° withline 1 (Figure 1). To avoid aliasing ofsurface waves, the geophone groupinterval was 17 m with a linear arraybetween stations. The far offset wasextended to 3600 m to allow con-verted-wave events to arrive outsidethe surface waves cone.

Figure 1 shows the behavior ofmigrated horizontal componentsaround points 7 km from the begin-ning of each line. At the blue pointon line 3, the radial componentarrives faster than the transversecomponent. At the red point on line1, the radial component arrives later.Similarly, at the green point on line2, energy arrives faster in the trans-verse component. These observationssuggest that the orientation of fastshear waves is roughly in the samedirection of line 3, which is also par-allel to the orientation of maximumhorizontal stress.

We performed rotation analysisalong the migrated horizontal com-

1036 THE LEADING EDGE SEPTEMBER 2001 SEPTEMBER 2001 THE LEADING EDGE 0000

Using multicomponent seismic for reservoircharacterization in Venezuela

REINALDO J. MICHELENA, MARÍA S. DONATI, ALEJANDRO A. VALENCIANO, and CLAUDIO D’AGOSTO, Petróleos deVenezuela (Pdvsa) Intevep, Caracas, Venezuela

Figure 1. Examples of migrated horizontal components for selected loca-tions along the 3-C lines. Rotation analysis of these data yielded the frac-ture orientation and density map in Figure 2.

Figure 2. Map view of fracture orientation and density at the top of the car-bonate reservoir. Maximum horizontal stress is parallel to line 3 (blue).Fracture orientations closely follow the alignment of major faults. Changes infracture density are related to structural variations and changes in rock type.

ponents to estimate the orientation ofthe faster shear arrival and the trav-eltime difference between the twoarrivals at each point. The analysiswas based on the ratios of amplitudesbetween the two horizontal compo-nents. After analyzing the energyratios around various significantreflections, we obtained maps of ori-entation of the faster shear wavearrival which correlated very well thealignment of the major faults thatcross the field. Figure 2 is a map gen-erated at the top of the carbonatereservoir where the effect of upperlayers has been removed by layerstripping. Orientations were esti-mated only along the 3-C lines.Angles for areas between lines wereestimated by bidimensional splinesinterpolation. Lengths of red seg-ments are proportional to differencesin traveltime between fast and slowshear-wave arrivals. Figure 3 showshow orientations obtained from rota-tion analyses are also consistent withorientations obtained from FMS logs.They reveal that the dominant frac-ture system in the field strikes north-west-southeast, making an angle ofapproximately 30° with line 3.Because of the good correlation withthe alignment of major faults in thefield and the good agreement withthe orientation of the dominant frac-ture set observed at the wells, weinterpreted Figure 2 as the map offracture orientation of the densest setof fractures in the field. This map wasused to design trajectories of hori-

zontal wells.VP/VS ratio and lithology of for-

mations right above the target are wellcorrelated in this field. Figure 4 com-pares the VP/VS ratio estimated from adipole sonic log with a gamma-ray logfrom the same well. VP/VS ratiosobtained after a careful interpretationand correlation of PP and PS data fol-low the same trend of the VP/VS ratiosobtained from the dipole sonic log.Ratios estimated from the seismic dataare well correlated with net sand thick-ness estimated from gamma-ray logsat different locations. Seismic-derivedVP/VS ratios were later used to gener-ate maps of net sand thickness acrossthe reservoir.

Figure 5 shows the relationsamong structural variations along line1, changes in lithology indicated bychanges in seismic-derived VP/VSratio,and changes in anisotropy (propor-tional to fracture density) indicated bychanges in traveltime differencesbetween fast and slow shear-wavearrivals along the same line. The greenline indicates the variations in depthof the top of the target. Notice how theintensity of anisotropy is not sym-metric around the zone of the structurewith strongest curvature. The reasonfor this asymmetry is because theflanks of the zone of strongest curva-ture contain different rock types thatrespond differently to the same defor-mation. Because shales are more plas-tic than sands, they tend to fracture lesswhen a similar stress is applied to both.

Lagunillas Field. This field is east ofLake Maracaibo Basin. The targethorizon is a Miocene Formation char-acterized by interbedded series ofshales, clays, and highly unconsoli-dated sandstones at a depth of 2000-2600 ft (300-500 ms). Heavy oil servesas the “glue” that holds the rocktogether in the producing area. Thereservoir is in a monocline structurewith a mild dip to the southwest.

In 1997, Pdvsa recorded a 3-D/3-C baseline data set simultaneouslywith a conventional 3-D volume aspart of a project to monitor steaminjection. The purpose of the 3-D/3-C experiment was to evaluate thepotential of using multicomponentdata to quantify changes in reservoirconditions with time. Only the base-line survey has been recorded so far.

The 3-D/3-C survey covers 0.33km2. Sixteen geophones were perma-nently buried at a depth of 30 m; per-manently burying the geophonesensures the required repeatability fora time-lapse study. Spacing betweenreceivers and receiver lines was 196m. The 4104 explosive sources wereburied at a depth of 15 m and had abrick distribution. The same sourceswere used to acquire a conventional P-wave data set that covered a largerarea. These P-wave data have goodsignal-to-noise ratio, high-frequencycontent, and good continuity along thetarget horizon.

The most significant resultsderived from processing this multi-component survey were:

• Unconsolidated sands produceazimuthal anisotropy. As evident inFigure 6, there is coherent energy inboth horizontal components alongall time sections. We believe seismicdata are responding in this case to


Figure 3. Comparison of fracture orientation from FMS logs and from mul-ticomponent seismic data. Two fracture sets intersect well 23. Seismic datatend to follow the densest.

Figure 4. Correlation betweengamma-ray log and VP/VS ratio.

the maximum horizontalstress in the area. No bire-fringence analysis has beenperformed so far in thesedata.

• Resolution of horizontal com-ponents is higher than reso-lution of vertical component(Figure 6), suggesting thatmuch more detail about thereservoir can be obtainedwith PS data than with con-ventional PP data.

• Horizontal componentsshow lateral variations inamplitude not observed inthe vertical component. This can beseen in Figure 6 and more clearly inFigure 7, which compares three timeslices of PP and PS data. The reasonsfor these changes are unknown butmay be related to changes in reser-voir conditions due to steam injectedinto the reservoir 18 months beforethe acquisition, lateral changes inazimuthal anisotropy, or lateralchanges in lithology. Confirmation ofany of the hypotheses will requiremore modeling and calibration andcareful removal of azimuthal

anisotropy effects.• Time migrations using PS migra-

tion velocities yielded betterimages than migrations using PSrms velocities.

Orinoco Oil Belt. This area, thelargest heavy oil accumulation in theworld, is in Eastern Venezuela.Reservoirs consist of closely interbed-ded series of heavy-oil sand bodiesat very shallow depths. In August1997, one operator in the arearecorded a 3-D/3-C multicomponent

data set to test how 3-C data couldhelp improve shallow resolution ofreservoirs with little or no contrast ofacoustic impedance between sandsand shales. The 3-D/3-C data wereprocessed by two contractors. Thisstudy uses one of their final P-P andPS migrated cubes. In addition toseismic data, Pdvsa was providedwith dipole sonic, density, gamma-ray, and spontaneous potential logsfrom three wells (although not alllogs were available for all wells).

Figure 8 shows how the gamma-


Figure 6. Time-migrated seismic sections for vertical,radial, and transverse component PP time. The energy inboth horizontal components is clear evidence ofazimuthal anisotropy in this field. The target is between300 and 500 ms (PP time).

Figure 7. Comparisonof horizon slices(from 3-D P-wavedata) and times slices(from PP and PS seis-mic data). There isgood correlationbetween both P-wavevolumes. Strongchanges in amplitudeare observed in con-verted waves. Thethree red points arewells used in theprocessing.

Figure 5. Relations among structural deformations, frac-ture density (proportional to strength of anisotropy),and lithology along line 1. The anisotropy vertical axisin the top figure indicates the difference in ms betweenfast and slow shear-wave arrivals. More plastic shales(VP/VS greater than 2, in pink) tend to fracture less thansands around the dome.

Horizon slice Time

ray log correlates with different seis-mic properties such as velocities, den-sities, and impedances. As expectedin this area, there is little correlationbetween gamma ray and acousticimpedance—the reason why com-pressional seismic data have not beenreliable to detect changes in lithol-ogy. The correlation with wave veloc-ities and lithology is also poor. Thecorrelation with shear-wave veloci-ties is better than with compressional-wave velocities. VP/VS and gammaray are also poorly correlated.However, the correlation betweengamma ray and density is high,which suggests that medium densitycan be used to help differentiatelithologies in this area. When the twoparameters that show the highest cor-relation with gamma ray (density andshear-wave velocity) are crossplot-ted and each point is colored withthe corresponding gamma-ray value,we observe better separation betweensands and shales (Figure 9). This sug-gests that elastic parameters of themedium estimated from seismic datacan be used to differentiate shalesfrom sands after training a classifi-cation algorithm at the well.

An arbitrary line that interceptedthe three wells where logs were avail-able was extracted from the 3-D/3-CPP and PS cubes. We used a com-mercial algorithm to perform strati-graphic inversion of these twopoststack lines. The results were com-bined to obtain the elastic parametersof the medium (Valenciano andMichelena, 2000). Figure 10 shows adensity section obtained after invert-ing PS poststack data and combiningthis result with the acoustic imped-ance and estimates of the product

VP/VS and ratio VP/VS derived fromthe dipole sonic log.

Figure 11 shows the lithology dis-tribution over the arbitrary lineextracted from the 3-D/3-C data set.This was obtained after extrapolatingalong the section the relationsbetween velocities and density at thewell in the middle. The extrapolationof relations among elastic parame-ters was performed by using aneural-net-based classification algo-rithm on the elastic parametersderived from the seismic data. Sandsare blue, shales are red, and transi-tion zones are yellow.

These encouraging results helpedjustify the current reprocessing ofthese data with more emphasis onstatic corrections, wavelet process-ing, and preservation of true relativeamplitudes.

Final remarks. Pdvsa faces manyreservoir characterization challengesin Venezuela, particularly on land, inwhich multicomponent seismic canmake enormous contributions. ManyVenezuelan reservoirs have complexstratigraphy and are in structurallycomplex areas. Fracture, lithology,and fluids characterization in com-plex areas require better seismicimages. For this reason, geophysicistsin Pdvsa will focus in the next fewyears on the use and enhancement of3-C seismic processing and imagingalgorithms, and the development ofnew analysis/interpretation methodsfor these data. Particular emphasiswill be on estimation of petrophysi-cal parameters and interpretation of3-C seismic attributes. As part of itsexploration and production strategy,Pdvsa plans to use 3-C seismic to help


Figure 8. Correlations among gamma-ray, density,VP/VS ratio, S-wave velocity, P-wave velocity, P-waveimpedance, and S-wave impedance logs. Numbers atthe bottom are the correlation coefficients betweengamma ray and the log above. Figure 9. Plot of density versus S-wave velocity that is

color-coded using the gamma-ray log. Sands andshales are clearly separated.

Figure 10. Density section estimated after combining PP and PS inversions.There is good agreement between estimated and expected densities exceptbetween 0.2 and 0.4 s, where the seismic derived density shows lateral vari-ations around the well.

delineation and characterization ofoffshore reservoirs in the near future.

Suggested reading. “Multicomponent3-D seismic pilot study in the Orinocoheavy oil belt,” by Lansley et al. (TLE,1998). “Mapping distribution of fracturesin a reservoir with PS converted waves,”by Ata and Michelena (TLE, 1995).

“Birefringence study on 3-C/2-D:Barinas Basin (Venezuela)” by Donatiand Brown (SEG 1995 ExpandedAbstracts). “Quantifying errors in fractureorientations estimated from PS con-verted wave” by Michelena (SEG 1995Expanded Abstracts). “Stratigraphic inver-sion of poststack PS converted wavesdata,” by Valenciano and Michelena

(SEG 2000 Expanded Abstracts).“Tomography + prestack depth migra-tion of PS converted waves” byD’Agosto et al. (SEG 1998 ExpandedAbstracts). “P- and S-wave separation ata liquid-solid interface,” by Donati andStewart (Journal of Seismic Exploration,1996). “Making AVO analysis for con-verted waves a practical issue” by Donatiand Martín (SEG 1998 ExpandedAbstracts). LE

Acknowledgments: This paper is the result ofeight years of effort by many individuals withinPdvsa. Eulogio del Pino initially championedwithin Corpoven (a former affiliate of Pdvsa)using 3-C data to characterize fractures.Subsequently, many people have been involvedin the different Pdvsa multicomponent projects.Elías Ata, Manuel González, Jesús Sierra, JavierPérez, Rafael Banchs, Germán Pérez, EdgardoPadrón, Alberto La Cruz, Martin Carry,Mariangela Capello, Hermes Malcotti, AntonioCaldera, Carlota Pereira, and José Luis Pérezare just a few. Many results in this paper, andmany more not shown, are in one way oranother the result of their work. We processedthe Lagunillas 3-C data using SeisUp fromGeocenter Inc. Thanks also to Sincor for per-mission to present results from its area and toPdvsa Intevep for permission to publish thispaper.

Corresponding author: [email protected]


Figure 11. Lithology distribution after classifying the estimated elasticparameters using a neural-net-based algorithm. There is good agreementbetween seismic-based predictions and well logs, in particular below 0.44 swhere the target is located. The initial model was built using only informa-tion from well A.

Using multicomponent seismic for reservoir ...

Documents

Transcript of Using multicomponent seismic for reservoir ...