NOTIC CONCRNING COPYRIGHT RSTRICTIONS This document may contain copyrighted materials. These materials have been made available for use in research, teaching, and private study, but may not be used for any commercial purpose. Users may not otherwise copy, reproduce, retransmit, distribute, publish, commercially exploit or otherwise transfer any material. The copyright law of the United States (Title 7, United States Code) governs the making of photocopies or other reproductions of copyrighted material. Under certain conditions specified in the law, libraries and archives are authorized to furnish a photocopy or other reproduction. One of these specific conditions is that the photocopy or reproduction is not to be "used for any purpose other than private study, scholarship, or research." If a user makes a request for, or later uses, a photocopy or reproduction for purposes in excess of "fair use," that user may be liable for copyright infringement. This institution reserves the right to refuse to accept a copying order if, in its judgment, fulfillment of the order would involve violation of copyright law.
Geothermal Resources Council Transactions, Vol. 26, September 22-2, 2002 Deep Sub-vertical Structure at Soultz Hot Dry Rock Site stimated by Reflection Technique Using Multi-Component coustic mission vents Nobukazu Soma', Hiroaki Niitsuma*, and Roy Baria3 National Institute of dvanced Industrial Science and Technology (Japan) 2Graduate School of ngineering, Tohoku University (Japan) 3GI - xploitation Mini&re de la Chaleur (France) Keywords coustic emission, reflection, H DR/CS, Soultz, sub-vertical structure BSTRCT reflection method in which acoustic emission events are used as a wave source was applied to waveforms observed in 2000 at the Soultz hot dry rock test site, France, to extract subvertical structures around a deep artificial reservoir. We used 27 waveforms at wells 460 and OPS4 that had amplitudes large enough for the analysis. Wide-area imaging around the artificial reservoir was conducted frst, and reflectors with NW- S strike and high dip angles were detected. They were reasonably consistent with observations made at shallow depths in the wells or at the surface, such as from a reflection survey in sediments. More detailed estimates were also analyzed near the reservoir with a higher resolution. s a result, sub-vertical reflectors, which agreed well with each other at wells 460 and OPS4, were detected and their shapes were found to be similar to the side edge of the deep artificial reservoir. The existence of subvertical structure at a great depth at Soultz was verified, and furthermore, a strong influence of the pre-existing subsurface structure on the creation of the reservoir was also implied. Introduction In both hot dry rock (HDR) and hydrothermal geothermal development, it is very important to understand the pre-existing deep structure, including fault systems and layer boundaries, since the deep structure may play a large role in pressure and fluid propagation. However, in a geothermal field it is not an easy task to measure deep structure because of serious limitations on geological surveys due to high temperature, high pressure, and great depth. Therefore we have developed a type of reflection method using acoustic emission (, or, as it is often called microseismicity, MS) events. We call it the reflection method, and have already applied it to some geothermal fields (Soma and Niitsuma, 997; Soma et al., 2002). The uropean deep geothermal energy programme has been conducted by using HDR technology at Soultz-sous-For& lsace, France, since 987, supported mainly by the U, France and Germany (Baria et al., 99). t present, the program is going to commercial power generation for the future. large scale hydraulic stimulation was made in 993 by using the well GPK, and an artificial geothermal reservoir was created at a depth of around 300 m. From 994 to 99, a second deep well (GPK2) was drilled into the reservoir, which resulted in the successful development of a circulation system. In summer and autumn 997, a four months long term circulation test was performed that successfully demonstrated the strong possibility of the geothermal system serving as a future energy source (BaumgWner et al., 998). For the practical future commercial production of electricity, the well GPK2 was deepened to a depth of km in 999, and a temperature of 200 C was obtained (Baria, et. al., 2000). Next, a stimulation test in GPK2 was camed out in June-July 2000, and a large number of events were recorded during this period (sanuma, et. al., 200). t the Soultz site, the existence of deep sub-vertical structures inside and around the artificial reservoir can be presumed from observations inside boreholes GPKl and GPK2 at the depth of around 300 m, the fault system on the surface, the results of a reflection survey for shallow sediment, and tectonic stress. However, it was impossible to measure the structures directly by conventional techniques since they are very deep and inside the basement rock mass, although they are very important to the HDR system. Therefore, we applied the reflection method in an attempt to detect deep sub-vertical structure around the artificial reservoir. In this paper, we report the results of the reflection method using events detected in 2000, focusing on possible sub-vertical structure around the artificial reservoir. First, we provide general information about the events observed in 2000 at 2
Figure. plan view of the hot dry rock test site at Soultz-sous-For& France. 4C accelerometers were deployed in wells 40, 460, and OPS4. Hydrophones were in CPK and PS. Detector depths are also indicated. Figure 3. source distributions from 993 to 2000 at Soultz. Triangles mark the detector positions, * shows the events used in the reflection method, and broken lines indicate an arrangement of cross sections of detailed estimates in Figures 9 and 0. vents During Hydraulic xperiments in 2000 at Soultz L.,_...b 2.& 2. B TlmB {S) Figure 2. n example of the waveforms recorded by three accelerometers in 2000. Here, the y-component of well 40 is strongly contaminated by a periodic noise due to an electrical problem. Soultz, and briefly explain the reflection method in the time-frequency domain. Then we present wide-area and detailed estimates around the artificial reservoir. We also discuss the relationship between the detected sub-vertical reflectors and the shape of the reservoir on the basis of the source distribution. i The arrangement of wells and seismic detectors at Soultz is shown in Figure. The details of the observations made in 2000, which were conducted by a research team in the MTCI MURPHY international collaborative project, have already been reported by sanuma et al. (200). We used four-component (4C) accelerometers (Jones and sanuma, 997) in wells 40, 460, and OPS4, and two hydrophones in wells PS4 and GPKl. The waveforms were recorded by a specially designed digital data acquisition system with a sampling frequency of 0 khz, amplitude resolution of 6 bits, and data length of 60,000 words per trigger. These signal conditions were sufficient for using the reflection method. pproximately 40,000 events were recorded during the experiment (sanuma et al., 200). The quality of the detected signals was evaluated by a spectral matrix analysis and by the 4C-coherency technique. sanuma et al. (200) reported that all the 4C detectors, except those with electrical problems, could adequately detect 26
three-dimensional (3D) particle motion as a hodogram up to approximately 300 Hz. We adopted detectors only at wells 460 and OPS4 for examining waveforms with the reflection method since electrical problems were frequent at well 40 (Figure 2). It is clear from them source distributions accumulated from 993 to 2000 (Figure 3) that the new deep artificial reservoir created in 2000 is deeper than that created earlier. Therefore, the arrangement of source and detectors for the reflection method seems to improve the detection of sub-vertical reflectors around the reservoir because the detectable depth becomes deeper for reflectors with a high dip angle when the events of 2000 are used as the wave source. The Reflection Method in the Time-Frequency Domain The reflection method that we have developed is a type of reflection method that uses natural or induced events as a wave source. We can obtain deep images with reasonably high resolution from time-frequency domain analysis using wavelet transform (Soma, et. al., 2002). In the reflection method, reflected waves are distinguished from S-wave coda by the hodogram technique and 3D inversion is used for imaging with a non-systematic source distribution. We focus only on the S- to S-wave reflection in the analysis, since the S-wave energy usually dominates in our case, the P-wave reflection can be covered by the direct S-wave, and the energy of the conversion wave is generally much smaller. Because the reflected waves are covered by coda of the direct S-waves, conventional analysis based on wave energy or amplitude is useless for the analysis of reflected signals from a geothermal field. In the reflection method, the reflected waves are detected by investigating the shape of 3D hodograms. The shape of the hodogram is the trace of 3D particle motion during wave arrival, and is known to indicate the wave condition (Nagano, et. al, 989). The shape of the hodogram is spherical when an incoherent signal such as random noise is detected, but it becomes relatively linear upon the arrival of direct P- or S-waves, which are regarded as coherent signals. Tn our study, reflected waves are regarded as coherent signals, since we assumed that the reflector has some scale, and coda are normally regarded as incoherent signals (ki and Chouet, 97). Therefore, in the reflection method, we can distinguish reflected waves from coda by investigating the shapes of the 3D hodograms. For quantitative evaluation of the 3D hodogram shape, we used a covariance matrix method in the frequency domain, which is called a spectral matrix (Samson, 977). In this study, we regarded the wavelet transform as a time-frequency signal representation (Rioul and Vetterli, 99), which shows smoothed time-varying spectra similar to short-time Fourier transforms (STFT). Generally, a time-frequency signal representation allows us to compose a spectral matrix in the time-frequency domain (Moriya and Niitsuma, 996). Hence, we can define a spectral matrix SW (b, a) of the hodogram by using the wavelet transform as follows: Soma, et. a/. WJb, a) WJb, a) WXZ(b, a) S,(b,a)= Wyx(b,a) Wyy(b,a) Wyz(b,a) () WJb, a) WJb, a) WJb, a) where Wo(b, a)=wi(b, a)wj*(b, a), a is the frequency (scale), b is the time (shift), Wi(b, a) is the wavelet coefficient of the ith component, and * indicates the complex conjugate. We can treat this matrix as being in the time-frequency domain since scale and shift in the wavelet transform correspond to frequency and time, respectively. Then we can redefine Samson s global polarization coefficient, Cp(b, a), (Samson, 977) to be in the time-frequency domain as follows: where &=@, a) (i=,2,3;,>+3) are the eigenvalues of the matrix (eq. ) for each time (shift: b) and frequency (scale: a). Using this parameter, we can evaluate the shape of the 3D hodogram quantitatively in the time-frequency domain: Cp(b, a) = for an exactly linear hodogram and Cp(b, a) = 0 for a spherical hodogram. The arrival of coherent waves, such as reflected waves, should result in a high Cp value. Hence, we can use a high value of the parameter Cp(b, a) as an indicator of the arrival of reflected waves. For imaging the subsurface structure, we established a 3D inversion of the waveform, which shows the time-frequency distribution of the linearity of the 3D hodogram. The inversion concept is diagrammed in Figure 4. This inversion, which can generate a 3D estimate from a small number of detectors, is based on a type of diffraction-stack migration technique, except for using the linearity waveform Cp(b, a) in equation 2 of a 3D hodogram in the time-frequency domain. In addition, we enhanced the resolution in estimates by restricting the virtual reflected points by examining the orthogonality between the propagation direction and the S-wave polarization. The area in which the virtual reflected point occurs, where the strength of the hodogram linearity is plotted, can be reduced if we select the area where S-wave motion is perpendicular to the propagation direction from the reflected point when we focus on an S- to S-wave reflection. Furthermore, the effect of a heterogeneous source distribution is compensated by normalizing the wave density for a number of nearby events for each source. The heterogeneity of the source distribution causes the final image to have a dominant ellipsoidal artifact, which comes from the center of a source distribution. Normalization can reduce the effect of the ellipsoidal artifact (Soma and Niitsuma, 997). These two operations are effective when the inversion uses a natural distribution of induced seismicity. Wide-rea stimation Using vents in 2000 We can analyze the reflection method using events obtained in 2000 for a larger area around the Soultz site since the waveforms recorded by the MTC/MUR Y project have a sufficiently long data length (sanuma, et. al., 200). We used 27 waveforms that have relatively large amplitudes for the analysis. 27
T(a) T(a) Y 2 0 I v) -2000 '. (a)depth 200m GPK\ c -2000 (b)depth 300m 2000 L: distance ( - Detector) -Lr(al): path length for scale 'at' T 4 2Ooo =q$l,@- '- -I 3Ooo 3Ooo 3MK)- h!.. 4000.4, -4000- - -000.- vj - %-. O 0 / 0 6000 gom). v 6ooo 7000 7000 -(b)om - 7, (c)s400m I I I I I -2000 0 2000 (c)depth 4Wm I ' '. I (d)depth 00m 2000t - r r. Isodelay band of dt(al) of 3D hod0 ram.. restricted by S-wave polarization -2000 0 2000-2000 0 2000 Detector \ ~k-k *m. ^' [3 ~ction T(a) zimuth & Inclination of 3D hodogram at dt(a) t virtual reflect oint IT (Scattering -/ point [Depth 's-wave polarization fordt(a) Figure 4. The concept of 30 inversion in the reflection method. Figure shows estimates by the reflection method for slices at several depths using events collected at well 460 for a large area around the artificial reservoir at Soultz. In the figure, the darker color indicates a higher value of stacked hodogram linearity, which is supposed to be reflectivity in the analysis. The reflectors appear both northeast and southwest from well GPKl and seem to have a roughly NW-S strike. W- slices using events at well 460 are also shown in Figure 6, at the distances of (a) 400 m north from the well head of GPK, (b) across the GPKl, and (c) 400 m south from the GPK. The figure also shows that reflectors were detected around the depths of 200 m and 300-6000 m with a high dip angle. These reflectors correspond to the NW-S reflector in Figure. Wide-area estimates by the reflection method at well 460 of slices at depths of (a) 200 m, (b) 300 m, (c) 400 m, and (d) 00 m. Figure. Therefore, in spite of the great depth, this result implies the existence of sub-vertical structures around the artificial reservoir. Similar results were obtained by using events at well OPS4 (Figure 7) with compensation for the shallow strong virtual images that may be due to the setting condition of the detector. Since the bottom of OPS4 does not reach the basement granitic rock, the effect of the layer boundary between granite and sediments near the detector may not be negligible for a precise hodogram analysis. However, after removing the shallow virtual images, the arrangements of deep sub-vertical reflectors have the same tendency as those observed at well 460. The orientation of the detected sub-vertical reflectors looks similar to that of the fault system on the surface and to the results of a past reflection survey in shallow sediments. Figure 8 is one example of a past survey: an interpretation of a reflection survey in shallow sediments (Cautru, 986). The arrangement of faults is similar to that of the sub-vertical reflectors detected by the reflection method. We assume that some of the shallow faults ma) extend to such a great depth. Detailed stimation round the rtificial Reservoir For a more detailed interpretation of subvertical reflectors by the reflection method, we analyzed the area around the artificial reservoir more precisely with a higher spatial resolution. Figure 9 shows the estimates in the W- cross section eastern neighbor of the deep artificial reservoir by using events at wells 460 and OPS4 at various dis-
-3000-2000 GPKl -2000' 0 2000 2000-3000 -4OOo e000 0 Figure 7. Wide-area estimates by the reflection method at well OPS4. (a) Slice at a depth of 300 m, and (b) W- cross section at a distance of 700 m north from well GPKl. Ju -Jurassic Kp - Keuper t Bs - Buntsandstein 3000 I Km I L 4000 Figure 8. xample of interpretation of past reflection survey at Soultz (Cautru, 986). metres tances from and across well GPK. We also show the same estimates in a N-SW cross section through well GPK, which is almost perpendicular to the strike of the detected reflector, in Figure 0. The arrangement of cross sections in Figures 9 and 0 is superimposed on the horizontal distribution in Figure 3. These images are combinations of independent estimates using events at wells 460 and OPS4. The reflectors detected by using wells 460 and OPS4 have similar tendencies, even though we removed strong shallow virtual images when we used events from OPS4. Detected reflectors in these figures have relatively high dip angles, which may indicate sub-vertical structure. We suppose that the results imply the existence of sub-vertical faults near the artificial reservoir inside the almost uniform granite rock mass. Because we used S-wave reflection in the analysis, it is possible to detect thin structures such as faults, in contrast to a conventional P-wave reflection survey. In the figures, we also show projections of all the event locations from 993 to 2000 within a distance of k200 m from each cross section, which indicate a general view of the artificial reservoir. The general shape of the side edge of the deep artificial reservoir suggests a dip angle similar to that of the neighboring sub-vertical reflector, although the positions of the reflectors are not entirely definite since they are affected by the velocity model we assumed and because the estimated area is limited by the arrangement of the source and detectors. Particularly in Figure 0, we can see that the detected sub-vertical reflectors are located with almost the same dip angle as that of the side-edge of the clouds both in 2000 and in 993-6. If the existence of sub-vertical reflectors indicates a deep sub-vertical fault, the resemblance suggests that the creation of the deep artificial reservoir was strongly controlled by the pre-existing structure, which may play a serious role as an obstacle of fluid or pressure propagation during hydraulic stimulation. (a) n600 m (b) n400 m 2000 200t.. 400., 00 6000 I -4- l l -00 0 00 00000 200 3000 300. - I -... I I -00 0 00 00000 - T 200 20 300 30 400 40 00.._ -00 0 00 00000-00 0 00 00000 W - (m) W - (m) 29
N4 n W a 8-00 0 00 00000 SW - N (m) I GPKl Figure 0. N-SW cross section through well GPK of a detailed image near the artificial reservoir. Gray and black dots indicate the projection of the source distribution since 993 within a distance of f200 m from the cross section. Conclusion We applied the reflection method in the time-frequency domain to waveforms observed in 2000 at the Soultz HDR site, and tried to detect sub-vertical structures around a deep artificial reservoir. The 27 waveforms analyzed here were observed in wells 460 and OPS4 during hydraulic stimulation in 2000. These waveforms were recorded with high sampling frequency and a sufficient data length for use in the reflection method. The recorded signals of the waveforms were of good enough quality for the 3D hodogram analysis by spectral matrix analysis. The wide-area analysis by the reflection method extracted reflectors that had a NW-S strike and a high dip angle. They roughly agreed with observations of sub-vertical structure such as of the fault system at shallow depths. We also analyzed more detailed deep images around the artificial reservoir. Sub-vertical reflectors were again detected, and their shape was similar to that of the side edge of the deep artificial reservoir. These results suggest a strong influence by pre-existing structures such as faults on the creation of an artificial reservoir system, and they are the first indication of a sub-vertical structure around the deep artificial reservoir at the Soultz HDR site. cknowledgments This work was carried out as a part of the MTC & MURPHY international collaborative project supported by the New nergy and Industrial Technology Development Organization (NDO), Japan (International Joint Research Grant). We would also like to thank GI for providing the data from the uropean HDR site at Soultz-sous-For&, which is supported mainly by the uropean Commission, BMBF (Germany), and DM (France). Part of this study is also supported by the Industrial Technology Research Grant Program in 2000 from NDO. References ki, K. and Chouet, B., 97, Origin of coda waves: source, attenuation and scattering, J. Geophys. Res., 80,3322-3342. sanuma, H., Mochizuki, S., Nakazato, K., Soma, N., Niitsuma, H., and Baria, R., 200, Data acquisition and analysis of microseismicity from simulation of deep reservoir at Soultz by the MTCMURPHY international collaborative project, GRC Trans., 2 I6-66. Baria, R., Garnish, J., BaumgWner, J., GCrard,., and Jung, R., 99, Recent developments in the uropean HDR Research Programme at Soultz-sous-Forets (France): Proc. World Geothermal Congress, 99, Florence, Italy, International Geothermal ssociation, vol. 4, pp. 263-2637, ISBN 0-473-03 23-X. Baria, R., Baumgartner, J., Gerad,., and Garnish, J., 2000, The uropean HDR programme: main targets and results of the deepening of the well GPK-2 to 000 m, Proc. World Geotherm.Cong., 3643-362. Baumgiirtner, J., Gtrard,., Baria, R., Jung, R., Tran-Viet, T., Gandy, T., quilina, L., and Garnish, J., 998, Circulating the HDR reservoir at Soultz: Maintaining production and injection flow in complete balance - initial results of the 997 circulation experiment, Proceedings 23rd. Workshop Geothermal Reservoir ngineering, Stanford University, Palo lto, California, SGP-TR-8. Cautru, J. P., 986, Geological sections GPKl determined from seismic reflection profiles, lmrg document. Jones, R. and sanuma, H., 997, nalysis of the four component sensor configuration, MTC/NDO Internal Report, V0.2. Moriya, H. and Niitsuma, H., 996, Precise detection of a P-wave in low S/N signal by using time-frequency representations of a triaxial hodogram, Geophysics, 6 I, No., 43-466. Nagano, K., Niitsuma, H., and Chubachi, N., 989, utomatic algorithm for triaxial hodogram source location in downhole acoustic emission measurement, Geophysics, 4, No. 4,08-3. Rioul, 0. and Vetterli, M., 99, Wavelets and signal processing: I Signal Processing Magazine, 8, 4-38. Samson, J. C., 977, Matrix and Stokes velocity representations of detectors for polarized waveforms: theory, with some applications to teleseismic waves, Geophys. J. Roy. str. SOC., I, 83-603. Soma, N. and Niitsuma, H., 997, Identification of structures within the deep geothermal reservoir of the Kakkonda field, Japan, by a reflection method using acoustic emission as a wave source, Geothermics, 26,4344. Soma, N., Niitsuma, H., and Baria, R., 2002, Reflection technique in timefrequency domain using multicomponent acoustic emission signals and application to geothermal reservoirs, Geophysics, 67, (in print). 2 60