How single-sensor seismic improved image of Kuwait s Minagish Field

Transcription

1 How single-sensor seismic improved image of Kuwait s Minagish Field Ayman Shabrawi 1, Andy Smart 1 and Boff Anderson 2 of WesternGeco, with Ghassan Rached 3 and Adel El-Emam 3 from Kuwait Oil Company, describe how single-sensor, digital seismic recording applied onshore in an environmentally and seismically noisy area delivered a highresolution, high-frequency reservoir image. The Kuwait Oil Company (KOC) and WesternGeco conducted the first point-source/point-receiver (singlesensor) onshore seismic survey in the Middle East in early 2004 as a pilot study under a Joint Technology Agreement. The study investigated and subsequently determined that Q-Land single-sensor acquisition and processing techniques could improve seismic imaging and reservoir characterization of the onshore Minagish Field, for which previous attempts to derive reservoir properties from seismic had been largely unsuccessful. The key to success was the ability to effectively remove noise and preserve signal fidelity and high frequencies in the pre-stack data. Minagish Field setting Kuwait contributes approximately 9% of the world s daily output of oil. Minagish Field, discovered in 1959 in southwestern Kuwait, is one of the country s main producing fields. It was selected for the pilot study to address multiple development and exploration objectives, one of which was to improve the resolution of the seismic time-lapse response across the producing oolitic reservoir. Fluid movements in the reservoir are complex with water influx in higher permeability layers overriding oil. Acquiring seismic in an operating oilfield requires detailed planning, hazard identification and mitigation to ensure workforce and equipment safety. Operational safety is especially challenging in the Minagish Field. Recent military activity in the area has left the desert strewn with unexploded ordnance that ranges from minefields to cluster bombs and hand grenades (Figure 1). To mitigate these hazards MineTech International was employed to secure the source and receiver lines, ensuring that any ordnance found were marked and reported to the Kuwaiti Ministry of Defence, who either destroyed in situ or removed them. Also within the survey area were two oil gathering stations, numerous flares, a water injection plant, producing wells, a quarry, pipelines, roads and overhead power lines. In addition to contributing to the hazardous environment, these sites served as major non-repeatable noise generators. Further, unusually heavy rain while surveying flooded the prospect area, adding to the other operational challenges. Geophysical background KOC acquired a 48-fold 3D survey over the Minagish Field in The seismic crew returned in 1998 to acquire a trial 4D survey over a 24 km area centred on the main injector well. The baseline seismic data for the 4D study was extracted from the 1996 volume and both surveys were reprocessed using the day s most advanced seismic data processing techniques. Core sample and log analyses had predicted that a 4D effect should be observable with a 5% difference in acoustic impedance, resulting from a 5% to 95% change in water saturation. However, the D study found that while possible 4D effects could be observed around the injector well, they had similar amplitude to the ambient noise. The key contributors to the 1998 survey s inability to detect a clear time-lapse signal were limitations on resolution within the reservoir, inferior noise attenuation during acquisition, and low signal-to-noise ratios. Figure 1 The 2004 Q-Land survey area (blue box) measured approximately 8 km x 3 km. The red swath indicates the extent of the minefields in the area, and the infrastructure to the east is the gathering station that was built after acquiring the 1996 and 1998 conventional surveys. 1 P.O. Box 1106, 2nd Floor Munira Commercial Complex, Salem Al-Mubarak Street, Salmiya 22011, Kuwait 2 Schlumberger House, Buckingham Gate, Gatwick Airport, RH6 ONZ, UK 3 Kuwait Oil Company, P.O. Box 9758 Ahmadi, Ahmadi, Kuwait 2005 EAGE 63

2 first break volume 23, February 2005 The study concluded that the data acquired in the 1996 and 1998 seismic programmes could not properly detect 4D effects. Even additional reprocessing would not enhance the ability to detect 4D effects from these datasets. It was evident that future, repeat 3D (time-lapse) studies would require changes in acquisition methodology to reduce non-repeatable signal and coherent noise. Within the survey area, environmental, non-repeatable noise generators, such as gas flares and water injection plants, coupled with seismic-generated noise, such as air blast, ground roll and multiples, contribute to the severe contamination of the raw seismic records. The spatial source and receiver arrays used during conventional acquisition imperfectly attenuated the coherent and ambient noise, and adversely affected signal fidelity, limiting the vertical and lateral resolution at the principal reservoir targets and providing poor imaging of the deeper prospects within the field. Tests have shown how increasing array length degrades signal fidelity and limits bandwidth (Figure 2).This is caused by the aliasing of signal and noise energy. The conventional data previously acquired over the Minagish Field with 50 m long receiver arrays provided an image sufficient for structural interpretation, but not for reservoir characterization, due to aliasing effects in the higher frequencies. When processing seismic data for reservoir properties it is important to ensure that the data used are not corrupted by aliasing and that true amplitude and phase are maintained. Single sensor solution Having determined that monitoring the Minagish waterflood using conventional technology would be a challenge, the Kuwait Oil Company asked WesternGeco to propose possible solutions. A joint technology agreement was established specifically to determine the suitability of WesternGeco s Q- Technology service to solve the dilemma. A new baseline for the new 4D study was acquired using single-sensor land acquisition and processing techniques. This dataset delivered better signal-to-noise ratios and higher vertical and lateral resolution than the conventional datasets. The study also investigated the non-repeatable noise seen in the D dataset that was possibly due to positioning uncertainties. Figure 2 The point-source/point-receiver test illustrated was carried out with single sensors spaced 2m apart and a single vibrator. By using a 2-m array length wave types were recorded without aliasing. As the array length increased, first the airwave, then the ground roll, and finally the first breaks became aliased, which manifested itself as cross-banded signal areas in the xt-domain and wrap-around of the noise energy in the fk-domain. In addition to the wrap-around, an overall stronger attenuation of the high-frequency signal was observed EAGE

3 Four new technologies were used during the joint technology agreement pilot study: Q-Land single-sensor acquisition and processing system, which can record 20,000 live channels (at 2 ms sample rate) and applies proprietary noise suppression techniques (Ozbek, 2000) within the digital group forming (DGF) process. 80,000 lb peak force Desert Explorer vibrators, which provide an ultralow distortion seismic source. Multilevel, single-sensor Q-Borehole Versatile Seismic Imager (VSI) tool acquires zero-offset and walkaway vertical seismic profiles (VSPs). Well-Driven Seismic technology, which uses information from VSPs and well logs to calibrate, at every stage, preand post-stack processing sequences to the borehole. Successful project management using these new technologies included forming a multidisciplinary team of client and service company experts that followed the structured and integrated workflow detailed in Figure 3. The team determined the data required to evaluate multiple targets and designed the survey accordingly. It was expected that the reduced noise and preserved signal made possible by these new technologies would allow improved imaging not only of the primary Cretaceous Minagish Formation, but also of other reservoir targets within the Cretaceous, Jurassic and Permian systems. The primary Cretaceous targets are shown in Table 1. Success was measured on a target-by-target basis with a set of geological Table 1 Primary Cretaceous reservoir objectives and geophysical metrics, which included signal-to-noise ratio, bandwidth, correlation of well synthetic to Q-Land seismic trace, and measurement of heterogeneity lengths of structural elements (Table 2). Secondary project objectives included structural imaging and reservoir characterization of the Jurassic Najmah, Sargelu, and Marrat formations, as well as Permian Khuff and Pre-Khuff formations.the results for these objectives are not covered in this article. Integrated project design and execution Establishing an integrated overall design was the most important phase of the project. It was critical that the acquisition layout, source and processing parameters were optimally defined to achieve the objectives. With single-sensor acquisition and processing, particular emphasis is paid to source parameter testing as the quality of the seismic source has a greater impact on the returned seismic signal, which can be more faithfully preserved. The Q-Land pilot study consisted of four separate phases, each designed to provide information for use in the subsequent phase. The joint technology agreement called for some of the phases to include multiple objectives. Figure 3 The multidisciplinary team followed the structured and integrated workflow shown to achieve optimum results EAGE 65

4 first break volume 23, February 2005 Table 2 Objectively derived and mutually agreed upon metrics for primary objectives Phase 1 evaluated the signal and noise characteristics of the survey area. This was achieved with two tests. A box test used 1600 single-sensor channels positioned every 2.5 m, with a source line of 3.0 km in length and vibrator points recorded at 100 m intervals away from the spread. The objective was to characterize the coherent source-generated noise as a prerequisite for designing an optimum geometry for phases 2 and 3. The second test of phase 1 was a cross-spread test consisting of laying a 4.7 km by 45 m wide line of single sensors spaced every 5 m. Two source lines were acquired perpendicular to the 2D receiver line with 5 m source interval. The results were used to analyze the source-generated noise propagating in the cross-line direction. The cross-spread tests were necessary for a better understanding of the 3D nature of the coherent noise generated by the vibrators, as well as the area s gathering centres and water injection plant (Figure 4, top panel). The analysis of the coherent noise consisted of measuring the velocity, frequency and wavelength of each mode and then testing different Q-Land sensor layout designs and DGF schemes that would effectively attenuate the coherent noise while leaving the underlying signal intact (Figure 4, bottom panel). Phase 2 was designed to evaluate the seismic source effort required for the new project and the ability of the system to detect fractures. It consisted of planning the location of three finely sampled (5 m single-sensor spacing) 2D lines intersecting in a star array over a Minagish Field well that previously had been logged to surface with a full suite of logs. Each line was 7000 m long and rotated 60 0 to one another. The fracture orientation at the interest target depth for the area also was considered when designing the orientation of these test lines. Figure 4 The raw single-sensor, cross-spread noise record shows the various coherent noise trains present in the Minagish survey area (top panel). Digital Group Forming (DGF) effectively removed the noise in the same record while preserving the signal (bottom panel) EAGE

5 Figure 5 The multiple generators that were identified from the VSP (left-hand panel) were used in the Interbed Multiple Predictor (IMP) processing stage. Results are shown before (centre panel) and after (right-hand panel) applying IMP. Involvement of Kuwait Oil Company interpretation staff throughout the operation allowed the data acquired at all stages to be calibrated against borehole data. The source parameters used for the 3D surface seismic acquisition were determined by recording sweep parameter tests into the single-sensor Q-Borehole VSI tool, provided by Schlumberger, which was positioned at the main reservoir level. The interpreters then compared inverted surface seismic parameter test data to well data at the reservoir level. This approach, performed onsite, provided confidence in the process and allowed team members to make acquisition parameter decisions by an objective analysis of signal-tonoise, wavelet stability and spectral bandwidth. A shaped sweep with emphasis on the higher frequencies was chosen. Once the sweep parameters were decided from the borehole seismic, a further test into a 2D surface line was conducted to determine how many vibrators to use. Proprietary M30/625 Desert Explorer vibrators were used as they have a peak force capability of 80,000 lb. However, during the course of the project, the fundamental force output of the vibrators was limited to 48,000 lbs which is a drive level of 60%. The reduced drive level on a larger vibrator delivered a far-field signal with high amplitude and low distortion. Average distortion figures of less than 10% were observed for all programme phases. Due to the absorption and high levels of flare noise observed in phases 1 and 2, and the VSP, a box of four tightly grouped vibrators was chosen to resemble a point source and maximize the amount of energy being input. While a group of vibrators produces more ground roll than a single vibrator, single-sensor recording allows effective ground roll removal within the DGF process. In addition to source testing, Schlumberger recorded two walkaway VSPs on a well in the centre of the survey area using the VSI tool. These data were acquired to image the subsurface adjacent to the borehole and provide calibration of data processing parameterization as part of the Well-Driven Seismic process. The latter included information on amplitude gain recovery and absorption Q factors, velocity and velocity anisotropy, AVO effects, and the modelling, characterization and identification of multiple energy (Morice et al., 2002, 2003). Figure 5 shows the results of identifying the main multiple generators from the VSP data and removing them in the final processing. Phase 3 involved acquiring a new 24 km 2 3D pilot Figure 6 A visual comparison reveals the superior noise attenuation of the 2004 Minagish survey results. The left-hand panel shows a raw conventional shot record from the 1996 conventional survey, and in the centre panel an F-K filter has been applied (offsets are limited to match the Q-Land record). The right-hand panel shows a 2004 Q-Land section after Digital Group Forming. During the 2004 survey additional oilfield infrastructure was generating significantly more noise than in EAGE 67

6 first break volume 23, February 2005 baseline survey, using the optimum source and receiver parameters established in Phases 1 and 2. Since the 1998 survey, an oil gathering station with three large flares and a water injection plant had been built within the survey area, adding significantly to environmental noise. The new data were higher resolution with lower noise relative to the previous conventional 3D surveys, despite the gathering station, which is a testament to the power of single-sensor acquisition and processing techniques. Figure 6 shows a comparison of a 1996 shot record before and after an F-K (frequency-wavenumber) filter had been applied, and a recent 2004 digitally group-formed Q- Land shot record. Both shots were taken within 100 m of one another. The location where the two spreads intersected is marked with a red tick mark. Clearly, there is significantly more signal and less noise prevalent in the 2004 shot record, despite increased environmental noise in the survey area. Phase 4 has involved a repeatability test in the central area of the Phase 3 acquisition to investigate the value of the point-source/point-receiver Q-Land methodology in reducing the effect of non-repeatable noise, which was found to be an issue in the 1998 trial survey. The results of this phase are presently being analyzed. Survey results It was a relatively easy task to determine that the survey had been successful based on the predetermined metrics (Table 2), especially because the main metric, increasing the bandwidth at the Minagish target, increased from 44 Hz to 70 Hz. Comparison stacks of the previously acquired conventional data and Q-Land final results are shown in Figure 7. An additional highlight of the survey was the availability of an infield migrated cube, which was processed on the crew and delivered only six days after acquisition ended. This comprised 16 terrabytes of 384 million uncorrelated single-sensor seismic traces, which equates to a single-sensor trace density of over 16 million traces per km 2 - a densely sampled wavefield by any definition. After DGF, the output trace density was 304 thousand traces per km 2 and the data volume was reduced to 80 gigabytes. This was equivalent to the conventional 1996 data volume of 48 gigabytes for the same 24 km 2 survey area. The total integration of data processing and acquisition into a single integrated workflow was the key to delivering the migrated field volume in such a short time frame. Data were transferred from the recorder to the infield processing center via removable hard disks. A proprietary recording format was used to ease data input to the processing sequence. Simultaneous processing and acquisition were thus achieved. Some noise attenuation processes were performed on the uncorrelated data before having perturbation corrections and spatial resample filters applied during the DGF process. A normal processing sequence, without Well-Driven Seismic, was then applied to the group-formed data. The quick availability of a migrated field cube allowed the Kuwait Oil Company interpreters to immediately investigate the data volume. In parallel to this processing, the data were sent to WesternGeco s regional data processing centre in Cairo. A Well-Driven Seismic processing sequence was applied with input from Kuwait Oil Company geoscientists. Figure 7 A 1996 conventional seismic section (top left) is compared to the same section with the Q-Land N-S section interleaved and denoted by the vertical dotted lines (top right). The sections below are blow-ups from the rectangular boxes on the main sections. The Minagish formation appears at about 1500 ms. Note the increased vertical and lateral resolution at the reservoir and in the deeper sections EAGE

7 Outlook The industry has only just started to explore the true possibilities of single-sensor acquisition and processing, with new single-sensor capabilities in data processing applications continuously evolving. What the Kuwait Oil Company Minagish pilot study has demonstrated is that Q-Land technology and methods can efficiently and properly sample the 3D wavefield in order to effectively remove noise and apply perturbation corrections before group forming. This resulted in increased bandwidth and the preservation of high frequency and signal fidelity in pre-stack data significant data quality improvements. While the Minagish project was a high-density development survey, single-sensor methods are equally applicable to low-density exploration projects. The ability to provide raw data input records with increased signal-to-noise ratios permits the recording and processing of lower fold exploration data with better data quality than a conventional survey. The benefits of this approach are that the data can be integrated later in the field life with the results from subsequent singlesensor acquisition campaigns, providing high-density volumes for reservoir characterization and a baseline for timelapse reservoir monitoring. The industry has really only begun to explore the possibilities for single-sensor acquisition and processing. WesternGeco will continue to invest in and develop point source/point receiver technology, convinced that its superior noise attenuation and perturbation correction capabilities preserve the seismic signal more faithfully than conventional methods. Acknowledgements The authors thank Stephen Pickering of WesternGeco for editing this article and the following individuals for their invaluable contributions: King Lau of Kuwait Oil Company, and WesternGeco s Peter van Baaren, Mark Daly and Dr Andreas Laake. References Morice, S., Volterrani, S., Nafie, T. and Shabrawi, A. [2002] Well-Driven Seismic Processing and Reservoir Characterization. AAPG International Petroleum Conference and Exhibition, Cairo.. Morice, S. [2003] Integrated borehole and surface seismic: New technologies for acquisition, processing and reservoir characterization, Hassi Messaoud Field. 14th SPE Middle East Oil & Gas Show and Conference (MEOS). Ozbek, A. [2000] Adaptive Beam Forming with Generalized Linear Constraints. SEG 70th Annual International Meeting EAGE 69