P068 Case Study 4D Processing OBC versus Streamer Example of OFON filed, Block OML102, Nigeria

Transcription

1 P068 Case Study 4D Processing OBC versus Streamer Example of OFON filed, Block OML102, Nigeria T. Castex* (Total SA), P. Charrier (CGG), M.N. Dufrene (Total SA) & C. Orji (EPNL) SUMMARY This case study describes the time-lapse or 4D processing between 3D OBC data (2004) and 3D streamer data (1992) acquired on the OFON field Block OML102 offshore Nigeria. From the interpretation point of view the aim of this project was to define the limits of the RIII reservoir. The two seismic acquisitions were very different in terms of data quality (azimuths, Signal to Noise ratio, fold coverage, radio-navigation and positioning quality). Because of this, special attention was paid, during the processing, to the cross equalization and the 4D binning steps. A significant 4D signature was extracted from the final difference despite a high level of remaining noise mostly due to the poor quality of the baseline survey positioning (streamer data) and to the presence of remaining footprints. Conclusions were drawn about the possible delineation of the gas accumulation.

2 Introduction: Time lapse or 4D processing techniques have evolved rapidly during these last years. It has been shown that the matching of repeated 3D surveys can be very helpful to detect production effects within hydrocarbon reservoirs and very useful for injection wells setting. The 4D attributes can also provide information for the reservoir extension and the fault delineation. The purpose of this abstract is to highlight some of the main steps related to the time-lapse processing and to describe the methodological approach developed according to legacy data [1] in the case study of OFON field, block OML102 offshore Nigeria. As the two surveys were very different in terms of acquisition and quality we studied the more appropriate sequence to be applied for both seismic datasets. The cross-equalization (not developed here) and the simultaneous 4D binning were used to optimize the 4D seismic matching. For the main steps of the processing, the cross-plotting NRMS versus Predictability was used to assess the 4D repeatability. Acquisition parameters: The base survey was a 3D marine dataset recorded between 1990 and 1992 by DIGICON with a dual source and one streamer configuration, a nominal stacking fold of 30, a cable length of 3000m, and an acquisition bin of 6.25m x 25m. Positioning anomalies were encountered due to missing information in trace headers. Shot point numbers and gun mask were rebuilt using observer logs information. So confidence in the streamer navigation data was questionable. The repeated survey was the 3D OBC 2C seismic recorded in 2004, employing 4 ocean bottom cables per swath, shot along the cables with maximum offsets of 4000m and nominal stacking fold 120 for a 12.5m x 25m bin size. Fold s OBC STREAMER 2 km Figure 1: Comparison of Raw Stacks (OBC and Streamer)

3 Main steps in the OFON D processing: A) Amplitude correction and De-stripping: Non-repeatability was severely affected for both surveys, on one hand for the streamer data by the poor quality of positioning, the presence of footprints and the irregularity of the fold of coverage, and on the other hand for the OBC data by consistent amplitudes anomalies. A surface consistent amplitude correction was computed for OBC acquisition to compensate the variation of coupling between receivers. First an offset correction was computed, then a decomposition in shot and receiver was performed. The corrections for shot were very small due to the fact that there were no differences of coupling between shots. For the marine dataset a 3D standard de-stripping method [2] by filtering was applied. From the measured average amplitude over a large window a first step of amplitude correction between shots and channels was computed (1 st and 2 nd terms). Then a sail line by sail line amplitude correction (3 rd term) was calculated for the whole spread by applying a geostatistical filtering. B) PZ summation versus Deconvolution: The usual way for time lapse processing is to process simultaneously the two surveys through the same processing sequence. It is quite normal when the two surveys have the same acquisition scheme (streamer versus streamer or OBC versus OBC). However, in this particular case 3D streamer and 3D OBC processing sequences were obviously different. For example, the PZ summation processing for OBC introduces a difference with the streamer data because the water layer peg-leg attenuation is implicitly included in the processing. In order to compensate for this peg-leg effect we applied to the streamer data set a deconvolution designed to obtain as much as possible a similar effect. Different de-convolution parameters were tested to obtain the most appropriate sequence. Thus a gapped de-convolution in the Tau-p domain (operator 160ms, gap 24ms) was applied to the streamer data to produce an equivalent effect to the OBC de-peg-leg. Then a second deconvolution in X,T domain (operator 160ms, gap 24ms) was applied with the same parameters on both datasets, to remove the residual multiples(see figure 2). 1 km 0,5s OBC : Depegleg option in PZ summation + Gapped de-convolution in TX domain STREAMER : Gapped de-convolution in Tau-P domain + Gapped de-convolution in TX domain Figure 2: Comparison between Depegleg versus Tau-P deconvolution

4 C) Offset regularization and 4D Binning : These two surveys were not acquired specifically for 4D analysis. Therefore it was an important challenge to optimize the 4D signal. First we took particular care of the binning and the offset regularization. It is currently accepted in the 4D processing industry that the best criteria for 4D binning is to select traces with closely repeated geometry. It is also acknowledged that it is preferable to perform 4D depopulation rather than two 3D depopulations separately. Finally a combination of double 4D binning including one interpolation of missing traces using the streamer dataset as a reference was worked out. First a duplicated traces rejection (based on shot azimuth criteria) and 3D missing trace interpolation (in FXY domain) were performed on streamer data to have a single trace per offset class in a bin. Then the OBC data was depopulated by taking to the streamer data as a reference in order to obtain the highest repeatability according to the source and receiver location criteria. D) Bin Centring : Bin centring impacts the migration and the 4D difference after migration, by increasing the signal to noise ratio after migration. Two methods have been tested for bin centring: a) high resolution Radon transform b) DMO followed by reverse DMO. The Quality Control of the 4D repeatability was done by by cross-plotting Predictability versus NMRS in a window situated above the reservoir. It shows that DMO + reverse DMO processing gave the best result (see figure 3) because the scattering of points is reduced. The normalized RMS amplitude difference (NRMS) between two cubes gives the amplitude repeatability (abscissa values). The predictability (Y axis) based on the correlation between traces from base and monitor trends towards one when the traces are very similar. If two traces are perfectly identical, the Predictability equals 1 and the NRMS 0. Bin centring HR Radon Transform method Bin centring before migration DMO DMO-1 method Figure 3: QC of Predictability versus NRMS for two different bin centring methods Results: The frequency bandwidth of the 4D final data has been reduced (5-45 Hz) due to the fit to streamer data. We lose 20Hz in comparison with the 3D OBC processed data. The presence of footprints and noise on the 4D difference cube due to streamer data and the loss of fold on OBC data (from 120 to 30, in order to fit streamer data) have decreased the quality of the 4D signal [3]. However, despite the non-repeatable geometry of the two surveys this 4D processing allows a better definition of the GOC within the RIII reservoir. A fault boundary can be interpreted at the northern limit of the gas accumulation and a limit to the south appears at the GOC (see figure 4). These features are estimated to be valuable information for a better understanding of the reservoir dynamic behavior.

5 Figure 4 : Existing faults interpreted on 4D differences 3D Section : Minimum Amplitude Full OBC - possible fault at the northern limit of 4D anomaly - limit of the amplitude to the South corresponding to the GOC 4D Section : Difference between Minimum Amplitude OBC - Streamer Conclusions: The conditions of geometrical repeatability were not a priori very good between OBC and streamer 3D surveys on the Ofon field. A great effort was made during the 4D processing to homogenize the two data sets, especially for amplitude corrections and binning methodology., In spite of numerous difficulties, a meaningful 4D signature was finally extracted from the 4D seismic data set with a positive contribution to the reservoir interpretation. Acknowledgments The authors thank Elf Petroleum Nigeria Limited (EPNL) for permission to show this work. References: [1] Bertrand, A. and Bannister, D. [2005] Enhanced 4D processing and quantitative analysis on Troll West using multiple vintages of legacy data, First Break 23, [2] Hoeber, H. [2003] On the use of geo-statistical filtering techniques in seismic processing, SEG 2003 [3] Orji, C. [2006] Challenges of Streamer and OBC Time Lapse seismic matching : OFON field, Eastern Offshore, Niger Delta. NAPE 2006 conference (Nigerian Association of Petroleum Explorationist ).