June 2022 /// Vol 343 No. 6)

Features

Processing of a large offshore 3DVSP DAS survey in a producing well

DAS fiber deployed on several wells’ completion strings offshore recorded a 3DVSP survey with ongoing production, but with significant noise encountered. A novel, two-step approach on one well addressed fiber directivity response on recorded data to recover relative amplitudes. The final 3D VSP image showed improved spatial resolution over target intervals.

Francesco Miranda, Emmanuel Spadavecchia, Shujaat Ali, Roberto Castañeda-Aguilar

Borehole seismic acquisition over fiber-optic cables has gained popularity in recent times. For conventional wireline surveys, it avoids the need for a dedicated VSP run, resulting in significant time and cost-savings.1,2,3 (Borland et al., 2016; Useche et al., 2020; Martinez et al., 2020). It also allows cost-effective recording of large receiver aperture 3DVSPs in fields, where wells are already instrument with optical fiber.4 (Mateeva et al., 2013). 

Fig. 1. Source and receiver geometry.
Fig. 1. Source and receiver geometry.

Fiber deployment over tubing, such as the case in the present study, often results in noisier data, due to suboptimal receiver coupling conditions. The noise issue is further exacerbated if data are recorded while wells are under flowing condition. Moreover, fiber deployment in highly deviated wells also causes severe receiver directivity footprint, which needs to be addressed in processing. All these factors impose considerable challenges to processing of the data acquired under such conditions, as the case with the present study. 

DATA ACQUISITION 

The survey consists of over 10,000 shots recorded by fiber in four wells. The data are acquired in a grid pattern with shot spacing of 50 m and line spacing of 125 m. Figure 1 shows the source and receiver locations. The colored dots are receivers, and empty circles are sources. In the present study, we show the results from a single well with receivers represented by green dots. 

Quality of acquired data is quite variable. The well deviation improves fiber coupling, resulting in higher signal-to-noise ratio than observed in vertical wells with similar fiber deployment. Production noise is present on all wells; however, amplitude of this noise varies over time, as well as over receiver depths. 

DATA PROCESSING 

A comprehensive data processing chain, comprising multi-domain denoise, source designature, predictive deconvolution, receiver directivity compensation, and 4D interpolation, was applied to process the data. Reverse time migration (RTM) was carried out for imaging, and some post- migration conditioning—including receiver directivity correction, dip guided selective stacking, residual alignment, nonlinear stacking, and ray-based illumination compensation—was applied. The breakdown of each processing step is discussed in the following sections. 

SHOT DOMAIN PROCESSING 

Fig. 2. Near offset common source gather before (left) and after (right) source domain processing.
Fig. 2. Near offset common source gather before (left) and after (right) source domain processing.

Production-related noise was dominant in lower frequencies (up to 35 Hz) and was spread over a range of velocities in the shot domain. Tau-p and an FX infinite impulse response dip filter were used to remove the very low velocity components of this noise. A tau-p domain denoise was also done to attenuate constant velocity noise by algorithmically identifying and suppressing the anomalous amplitude values in this domain. 

A shot-by-shot source signature deconvolution was applied to debubble, deghost and zero-phase the signal. Far-field signature for this operation was computed from the near-field hydrophone for each record. 

To suppress the repetitive ringing noise, due mainly to coupling conditions, a predictive deconvolution operation was used. Figure 2 compares one such gather after shot domain processing. Downgoing wavefield is also removed in shot domain processing by a Tau-p domain filter. 

A shot domain scaling by 1/p values was applied in Tau-p domain.5 (Sayed et al., 2020). This step effectively provides amplitude-only DAS to velocity scaling. For example, common receiver gather, before and after scaling, is compared in Fig. 3. It can be seen from this figure that scaling by 1/p improves amplitude balancing on either side of the minimum offset location. 

Fig. 3. Near well line, common receiver gather before (left) and after (right) shot domain scaling by 1/p.
Fig. 3. Near well line, common receiver gather before (left) and after (right) shot domain scaling by 1/p.

RECEIVER DOMAIN DENOISE AND INTERPOLATION 

The data are NMO-corrected by ray tracing through a 1D-velocity function, extracted from an isotropic PSDM velocity model at the wellhead location. The NMO-corrected data are then sorted in receiver domain. When such data are analyzed along or orthogonal to the shooting direction, several parabolic events are identified with their origin at the zero-offset location. A three-dimensional parabolic Radon transform (3D PRT) is used to model these events. 

Fig. 4. Near well line, common receiver gather after NMO correction, before (left) and after (right) 3D PRT noise estimation and removal processing.
Fig. 4. Near well line, common receiver gather after NMO correction, before (left) and after (right) 3D PRT noise estimation and removal processing.

The modeled coherent noise through this process is further isolated by curvelet transform before adaptively subtracting from the data. The parameters for this 3D PRT transform are carefully selected for signal protection and are varied with receiver depth. For example, NMO- corrected common receiver gather from the near-well location line, before and after 3D PRT denoise is shown in Fig. 4.

Fig. 5. Example mid-offset shot record raw (left) and processed upgoing wavefield.
Fig. 5. Example mid-offset shot record raw (left) and processed upgoing wavefield.

The data after this step are regularized to a 50-m × 50-m grid through a 4D matching pursuit Fourier interpolation (4D MPFI) algorithm. This step further attenuates the production noise that, due to its time-variant nature, remains incoherent in receiver domain. Following this step, NMO correction is removed. Examples of common shot and common receiver gathers of raw and processed data after this step are compared in Figs. 5 and 6, respectively. As a final spectral shaping step, a bandwidth extension algorithm is used to enhance amplitudes, both on lower- and higher-end of input spectra. 

Fig. 6. Near well line, common receiver raw (left) and processed upgoing wavefield after complete denoise.
Fig. 6. Near well line, common receiver raw (left) and processed upgoing wavefield after complete denoise.

REVERSE TIME MIGRATION 

The processed data are migrated through reverse-time migration (RTM) of receiver sub-groups. Each sub-group is comprised of 10 consecutive receivers (~50-m span). These receiver-group image volumes are scaled by the local fiber directivity correction field. This correction term is computed by taking the inner product of receiver slowness vectors (3D ray tracing) and fiber direction vector (well trajectory). An example receiver directivity correction section, as well as images before and after correction, are shown in Fig. 7. The directivity correction display is scaled between 1 and 3. 

Fig. 7. Ray tracing-based receiver directivity correction for a receiver in highly deviated part of the well (top); receiver gather RTM volume before (middle) and after (bottom) ray tracing-based receiver directivity correction; each vertical bar is 100 m apart.
Fig. 7. Ray tracing-based receiver directivity correction for a receiver in highly deviated part of the well (top); receiver gather RTM volume before (middle) and after (bottom) ray tracing-based receiver directivity correction; each vertical bar is 100 m apart.

POST MIGRATION ENHANCEMENT 

The fiber directivity-corrected receiver subgroup image volumes are processed through an interpretation-guided selective stacking process, where each image volume is decomposed into several 3D dip components. These dip components are weighted with respect to the background dip field and restacked. This reduces migration smiles while retaining spatial resolution. 

Fig. 8. Surface seismic (top) and DAS 3DVSP (bottom) image. Circles show the receiver locations near the cross-section.
Fig. 8. Surface seismic (top) and DAS 3DVSP (bottom) image. Circles show the receiver locations near the cross-section.

The image volumes are then sorted into VSP common image point gathers (VSP-CIP gathers), and residual or trim statics are applied before stacking by a nonlinear stacking process. To reduce the amplitude differences, due to illumination, a weight cube produced by ray tracing is applied to the RTM stack. Near-well PSDM and DAS 3DVSP images are compared in Fig. 8. Similarly, two depth slices close to the reservoir interval are shown in Fig. 9. 

CONCLUSIONS 

A DAS 3DVSP survey recorded in very challenging data acquisition conditions is handled through a rigorous processing sequence to yield high-resolution, relative amplitude friendly subsurface image volume. It is demonstrated that through careful processing, DAS data acquired in a high-noise, producing well environment, can yield excellent-quality images with improved resolution in comparison to surface seismic. This case study is relevant to the oil and gas industry because most wells completed with fibers are either injectors or producers. The ability to handle the high flow noise is key for successful seismic monitoring projects using DAS data.  

Fig. 9. Two depth slices near the reservoir interval surface seismic (left) and 3D VSP (right). Overall 3D VSP shows higher spatial resolution of terminations and fault planes.
Fig. 9. Two depth slices near the reservoir interval surface seismic (left) and 3D VSP (right). Overall 3D VSP shows higher spatial resolution of terminations and fault planes.

ACKNOWLEDGMENTS 

The authors would like to thank Eni and Schlumberger for permission to publish the results, and Eni for the data release. 

REFERENCES 

  1. Borland, W., R. Hearn and A. Sayed, “Improving the acquisition efficiency of acquiring borehole acoustic data on a wireline hybrid electrooptical cable,” Expanded Abstracts, 800–804, 86th Annual SEG International Meeting, 2016, doi: https://doi.org/10.1190/segam2016-13951115.1. 
  2. Useche, M. A., A. Martinez, S. Silva, P. Heredia and L. F. Albarracin, 2020, “From checkshot to VSP using the same acquisition footprint enabled by fiber optics,” Expanded Abstracts, 90th Annual SEG International Meeting, doi: https://doi.org/10.1190/segam2020-3428169.1. 
  3. Martinez, A., M. Useche and R. Guerra, “Use of fiber optic acoustics to improve drilling efficiency and well placement,” OTC paper 30866-MS, Offshore Technology Conference, Houston, Texas, 2020. 
  4. Mateeva, A., J. Lopez, J. Mestayer, P. Wills, B. Cox, D. Kiyashchenko, Z. Yang, W. Berlang, R. Detomo and S. Grandi, “Distributed acoustic sensing for reservoir monitoring with VSP,” The Leading Edge, 32, pg. 1278–1283, 2013, doi: https://doi.org/10.1190/tle32101278.1
  5. Sayed, A., S. Ali and R. R. Stewart, 2020, Distributed acoustic sensing (DAS) to velocity transform and its benefits: 90th Annual International Meeting, SEG, Expanded Abstracts, 3788–3792, doi: https://doi.org/10.1190/segam2020-3424622.1

The Authors ///

Francesco Miranda has more than 20 years of oil and gas experience in the subsurface domain. He is the technical leader for borehole seismic for Eni. Throughout his career, he has held various technical and leadership positions, specializing in borehole seismic feasibility, processing, and interpretation and surface seismic processing, interpretation, seismic attributes, and seismic inversion. Mr. Miranda also has experience with R&D studies in seismic while drilling and geosteering. He holds a Doctor of Philosophy in geophysics from UC Swansea, UK.
Emmanuel Spadavecchia is the borehole seismic manager for Eni. He has nearly 10 years of oil and gas experience in the subsurface domain. Prior to his current role, he was a senior geophysicist at Eni, where he was project leader for in-house projects, and he also supervised outsourced projects. Mr. Spadavecchia holds an MS degree in Digital Signal Processing and a BS degree in Telecommunication Engineering from Politecnico di Milano.
Shujaat Ali has more than 20 years of borehole seismic experience in both technical and leadership positions, and is an expert in 3D VSP processing. He is the team lead for the Schlumberger North America Land borehole seismic processing group. Prior to his current role, Mr. Ali held several positions, including geophysicist for the 3D VSP Center of Excellence in Houston, and lead geophysicist for Kuwait, UAE and Pakistan. He specializes in borehole seismic survey design, acquisition, processing and imaging, and has experience in advanced analysis of borehole seismic data for anisotropy, AVO, and full waveform inversion. Mr. Ali holds an MS degree in geophysics from Quaid-e-Azam University, Pakistan.
Roberto Castañeda-Aguilar has nearly 20 years of oil and gas experience in the subsurface domain. He is a geophysicist for Schlumberger, specializing in seismic interpretation and borehole seismic processing. From 2013 to 2017, Mr. Castañeda was involved with the Schlumberger Innovation Center in Salt Lake City, Utah, where he conducted research into new workflows for unconventional reservoir characterization. His main technical interest is quantitative analysis in exploration and reservoir characterization. He holds a BS degree in geophysical engineering from the National Polytechnic Institute of México (IPN).

Related Articles ///

FROM THE ARCHIVE ///