During the last 10 years, the oil and gas industry has seen significant developments in seismic acquisition strategies, pushing the boundaries imposed by the technical limits of conventional 3D marine acquisition using narrow-azimuth towed streamers.
New acquisition geometries, including ocean-bottom nodes, coil, wide- and multi-azimuth shooting, can provide considerable improvements in data quality, but acquisition costs can be higher. There is therefore a need to develop new strategies that reduce costs, without imposing a noticeable deterioration in data quality.
A joint industry project (JIP) to ease the detection of by-passed hydrocarbons and help identify satellite fields has recently been concluded. The project, titled BLISS (Blind Identification of Seismic Signals), began in 2010 and is focusing on the identification of advanced signal processing techniques that can be applied to seismic processing problems.
The latest phase of the project received support from two operating companies, with assistance from ITF, the UK-based technology facilitator for the global oil and gas industry.
An interdisciplinary collaboration between the Department of Physics at the University of Alberta, and the Jean Kuntzmann Laboratory in Grenoble, France, united geophysicists and experts in statistics and signal processing. The goal was to identify techniques in advanced signal processing that have been largely ignored by the geophysical community, but carry promise for a step-change in seismic processing algorithms.
The BLISS consortium investigated how the detection of by-passed hydrocarbons and new satellite fields can make an impact on field life extension in mature areas, such as the North Sea and the Gulf of Mexico, particularly with respect to increasing the signal-tonoise ratio of seismic data, and extracting geologic features of interest.
The ease with which these additional reserves can be detected depends predominantly on the quality and resolution of the seismic data available for interpretation.
The two main approaches within the BLISS project relate to: local signal enhancement and robust wavelet estimation for blind deconvolution, in order to achieve the highest quality and resolution feasible.
Blind deconvolution is a technique that can be used to separate the seismic source signature, i.e. the explosion or air gun signature, from the true geology, without having any knowledge of either signal.
This technique is of interest to the communication industry, as it enables more effective use of bandwidth by separating voice patterns from the spoken words prior to compressing both. In exploration seismology, the voice patterns would be equivalent to the source signature and the spoken words—the geology.
Phase 1 of the project led to new statistical methods for random and coherent noise attenuation, in both pre-stack and stacked sections, blind PP/PS wavefield separation, and non-minimum phase wavelet estimation from surface seismic.
The aim of Phase 2 was to evaluate and adapt these techniques for advanced data processing and noise reduction, to render them suitable for increasing the quality and resolution of conventional seismic data, as well as developing novel strategies to reduce acquisition times.
Defined projects for Phase 2 include: simultaneous shot acquisition and post acquisition separation, and nonlinear sparse deconvolution.
Simultaneous shot acquisition and post-acquisition separation
One possibility to reduce total acquisition times is to fire several shots simultaneously at different locations. This is particularly interesting for wide-angle, towed-streamer acquisition geometries, since they involve multiple ships. The challenge then becomes how to then separate simultaneous shots to recover the individual shot gathers.
Independent component analysis (ICA) is an emerging technology in the field of advanced signal processing. It separates a set of observed signals into the statistically, most independent components by appealing to higherorder statistics.
ICA retrieves the original source signals blindly if they are statistically independent, without the need of further information. Blindly, in this context, means that no information is available about waveforms or polarizations of the desired source signals.
Independent component analysis holds promise for post-acquisition separation of simultaneously acquired shot records. Successful applications could have a far-reaching impact on future seismic acquisition strategies by all major operators.
Nonlinear sparse deconvolution
Predictive deconvolution is possibly the most commonly applied, statistical, signal-processing technique for any commercially acquired, seismic reflection survey. This determines a linear combination of observations that yields the optimal trade-off between recovery of the underlying signal structure and noise amplification.
Any linear deconvolution filter (such as Wiener filtering) has to find a compromise between recovery of the reflectivity series and noise amplification. It cannot solely recover the reflectivity series, given a linear combination of noisy past observations.
Predictive deconvolution also makes a minimum phase assumption for the wavelet, which is not generally valid. Nonlinear deconvolution is significantly more computation intensive than its linear counterparts; yet it offers the opportunity to create a step change in current processing strategies for seismic data by providing impedance sections with limited to no well control. The resulting deconvolved seismic sections are always sparse by construction and therefore ideal for creating blocky seismic impedance sections. The latter are routinely used in seismic interpretation.
Statistical wavelet estimation
Recent developments in advanced signal processing may make it possible to estimate accurate enough wavelets in exploration settings from surface seismic alone, with limited to no well control to produce meaningful interval attenuation maps.
Robust wavelet estimation is also crucial for other important applications, such as seismic inversion, guided interpolation of well-tie misfits, analysis of amplitude variations with offset, and for quality assessment of amplitude and phase-controlled processing.
Deterministic wavelet estimation is often done by means of seismic-towell ties, that is, by finding the digital filter that matches best synthetic seismograms created from well logs to observed data.
In these cases, the well logs act as ground truth. Unfortunately, this procedure only works at well locations and different wells can give different wavelet estimates, thus raising the question how to predict best wavelet variations away from the well.
In addition, no wells are available in virgin exploration areas. Statistical methods developed in the BLISS project enable accurate wavelet estimation directly from the observed seismic data, and have been shown to be robust even for relatively low-quality recordings.
Finally, the developed wavelet-estimation methods provide the information needed by the sparse deconvolution methods to enhance the resolution of seismic data.
Seismic attributes are commonly used by the oil and gas industry to facilitate interpretation of large 3D datasets and to effectively communicate subtle structural and stratigraphic features. Attributes can lead to simplified images of the subsurface and can detect many features at or below the resolution of seismic data.
The BLISS project set out to develop robust interpretation attributes, based on spectral analysis by empirical mode decomposition.
This decomposition was developed by researchers at NASA to facilitate analysis of complicated nonlinear signals. Methologies developed led to high-resolution seismic attributes. This is not only relevant for the detection of structural and stratigraphic features in controlled-source seismic data, but developed applications also have implications for data analysis in fields such as earthquake seismology, biomedicine, telecommunications, and remote sensing.
These combined projects should produce novel statistical signal processing tools for high-resolution imaging of small-scale structures, leading to an increased confidence in the interpretation of current and new exploration prospects.
As the oil and gas industry has developed, so have its requirements on information quality and detail that can be extracted from the vast data collected by seismic surveys. The output from this project, some of which is already being used by the sponsors, will result in processing software that can be applied on a range of datasets, including legacy data, new 3D seismic and repeat 4D surveys for exploration, appraisal and exploitation purposes.
As such, it may contribute to the discovery of new satellite fields, as well as enabling the progress of oil recovery from mature fields to be tracked. OE
|Mirko van der Baan is the principal investigator of the Blind Identification of Seismic Signals JIP, and the Microseismic Industry Consortium, a collaborative venture with the University of Calgary, dedicated to research in microseismicity. He is also a co-founder of the Centre of Integrated Petroleum Engineering and Geosciences, a joint initiative of several departments at the University of Leeds, UK, to foster and promote multidisciplinary petroleum-related research and teaching.|
|Colin Sanderson has been with ITF since its formation in 1999. During this time, as a senior technology analyst, he has collaborated with researchers, technology developers, and industry domain specialists across a wide range of technology areas within the oil and gas industry. Recently, he has concentrated on subsurface activities.|