Deblending Seismic Data

CROSS REFERENCE TO RELATED APPLICATIONS

[0001 ] This application claims priority and benefit from U.S. Provisional Patent Application No.62/145,518, filed on April 10, 2015, for "Simultaneous Source Separation" the content of which is incorporated in its entirety herein by reference.

BACKGROUND TECHNICAL FIELD

[0002] Embodiments of the subject matter disclosed herein generally relate to deblending data acquired with receivers detecting simultaneously reflections due to distinct signals, in particular, to methods that perform first an inversion to select a transform sub- domain including the underground formation's response, and a least square inversion to obtain a solution in the sub-domain.

DISCUSSION OF THE BACKGROUND

[0003] Structure of underground formations is customarily explored with seismic surveys to generate images used, for example, to locate gas and oil reservoirs. The seismic surveys acquire and study reflections of seismic signals injected in the surveyed formations. The signals are reflected, refracted and/or transmitted when encountering variations of propagation velocity. Receivers detect and record these reflections as seismic data. In time, the amount of seismic data and the complexity of data processing have increased tremendously due to the increased data processing capacity (both hardware and software) and development of survey equipment (seismic signal sources, receivers, etc.). These improvements have yielded sharper images of the underground formations, for bigger volumes and based on a higher density of information. The time necessary to acquire the survey data has continued to remain an important limitation to the cost-effectiveness of this type of geological prospecting.

[0004] One way to shorten the survey time is using a technique known as "simultaneous source acquisition." In this type of acquisition, time intervals between source activations (i.e., generating signals incident to the surveyed underground formation) are shorter than a listening time necessary to record all the reflections after one source's activation. Simultaneous source acquisition is now performed on land and in marine environments (with ocean bottom receivers or towed streamers), with continuous or non-continuous recording. Using simultaneous source acquisition yields blended data (i.e., generated by receivers detecting overlapping reflections due to different incident signals), and, therefore, an additional data pre-processing (known as "deblending") becomes necessary to extract datasets for each incident signal.

[0005] Numerous deblending algorithms have been developed in the last 10 years. However, these algorithms usually exploit particular data acquisition features related to geometry (source and receiver positions) and/or emitted signals (randomness of one source shooting relative to other shots, phase encoding, etc.).

[0006] Accordingly, it is desirable to develop efficient deblending methods usable for data gathered in more (or most) simultaneous source acquisition scenarios.

SUMMARY

[0007] Deblending of simultaneous source acquisition data in receiver gathers is performed in two phases: in a first phase, the support (relevant solution components) in a transformed domain is identified, and, in a second phase, the solution is sought by a least square inversion in the restricted support.

[0008] According to an embodiment, there is a deblending method for seismic data recorded by a receiver detecting simultaneously reflections due to distinct signals. The method includes receiving the seismic data and emitted signal information, and performing a first inversion of the seismic data, the first inversion being formulated to minimize a number of non-zero coefficients of a first inversion result in a transform domain. The method then includes performing a second inversion of the seismic data seeking an underground formation response in a sub-domain of the transform domain, the sub- domain being defined by vectors of a transform domain basis for which the first inversion result has the non-zero coefficients. The method then includes using the underground formation response obtained in the second inversion, to extract at least one deblended dataset corresponding to one of the distinct signals, from the seismic data. [0009] According to an embodiment, there is a data processing apparatus having an interface configured to receive seismic data recorded by a receiver detecting simultaneously reflections due to distinct signals, and emitted signals information, and a data processing unit. The data processing unit is configured to perform a first inversion of the seismic data, the first inversion being formulated to minimize a number of non-zero coefficients of a first inversion result in a transform domain, and a second inversion of the seismic data seeking an underground formation response in a sub-domain of the transform domain, the sub-domain being defined by vectors of a transform domain basis for which the first inversion result has the non-zero coefficients. The data processing unit is also configured to extract at least one deblended dataset corresponding to one of the distinct signals, from the seismic data using the underground formation response obtained in the second inversion.

[0010] According to another embodiment there is a computer-readable recording medium storing executable codes which, when executed by a data processing unit make the data processing unit to perform a seismic data deblending method. The method includes receiving the seismic data and emitted signal information, and performing a first inversion of the seismic data, the first inversion being formulated to minimize a number of non-zero coefficients of a first inversion result in a transform domain. The method then includes performing a second inversion of the seismic data seeking an underground formation response in a sub-domain of the transform domain, the sub-domain being defined by vectors of a transform domain basis for which the first inversion result has the non-zero coefficients. The method then includes using the underground formation response obtained in the second inversion, to extract at least one deblended dataset corresponding to one of the distinct signals, from the seismic data.

BRIEF DESCRIPTION OF THE DRAWINGS

[0011 ] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate one or more embodiments and, together with the description, explain these embodiments. In the drawings: [0012] Figure 1 illustrates a data processing flow of a deblending method according to an embodiment;

[0013] Figure 2 illustrates another data processing flow of a deblending method according to another embodiment;

[0014] Figure 3 is a flowchart of a deblending method according to yet another embodiment;

[0015] Figure 4 is a block diagram of a data processing apparatus according to an embodiment;

[0016] Figure 5 is a graphic illustration of blended data in a receiver gather; and

[0017] Figure 6 is a graphic illustration of de-blended data obtained from the receiver gather illustrated in Figure 5, using a method according to an embodiment.

DETAILED DESCRIPTION

[0018] The following description of the exemplary embodiments refers to the accompanying drawings. The same reference numbers in different drawings identify the same or similar elements. The following detailed description does not limit the invention. Instead, the scope of the invention is defined by the appended claims. The following embodiments are discussed, for simplicity, with regard to land seismic data acquisition. However, similar embodiments and methods may be used for a marine data acquisition system and for surveys using electromagnetic waves.

[0019] Reference throughout the specification to "one embodiment" or "an embodiment" means that a particular feature, structure or characteristic described in connection with an embodiment is included in at least one embodiment of the subject matter disclosed. Thus, the appearance of the phrases "in one embodiment" or "in an embodiment" in various places throughout the specification is not necessarily referring to the same embodiment. Further, the particular features, structures or characteristics may be combined in any suitable manner in one or more embodiments.

[0020] In various embodiments detailed in this section, deblending data in receiver gathers employs two minimizations seeking underground formation's response in a transform domain. A sparse inversion minimizes a two-term cost function including a data matching term and a weighted norm of the solution (i.e., a vector) in the transform domain. The solution of this first inversion is used to select a sub-domain including the underground formation's response (i.e., the solution's restricted support). Then, a least square inversion seeks the underground formation's response in this sub-domain, such that a data matching term to be minimized.

[0021 ] Figure 1 is a processing data flow corresponding to a deblending method according to an embodiment. Seismic data recorded by one receiver 1 10 and emitted signal information 120 for the signals whose reflections were detected by the receiver are the starting point (i.e., the method's input). Although the following description refers to a single receiver, the following methods can be similarly applied for each of plural receivers.

[0022] The seismic data carries information about the distinct signals causing the reflections, and the underground formation's response (known as "seismic traces") between the receiver and the sources.

[0023] The emitted signals information may include the source signatures, the source activation times and the source positions when activated, etc. Some of the emitted signal information may be captured via measurements. For example, if the source emitting the signal is vibroseis, the source signature may be determined based on forces measurements.

[0024] A source operator A is built using the emitted signal information. Continuous record J of the receiver can be expressed as the result of applying operator A to the seismic traces u :

b = Au = MRu . (1 )

[0025] Operator A can be decomposed in a restriction operator R and a multiple convolution operator M. The restriction operator maps the volume of seismic traces to the acquired seismic traces. The multiple convolution operator computes, for each seismic trace, a convolution between the trace and the source signature emitted at the respective position and time. The summation of all these contributions forms (theoretically) the continuous record b .

[0026] At 130, a sparse inversion of seismic data 120 seeks a solution (i.e., a version of the underground formation's response) in the curvelet domain. [0027] Data can be transformed in curvelet domain by applying a multi-scale and multi-dimensional transform as described in the article "The curvelet representation of wave propagators is optimally sparse" by E. J. Candes and L. Demanet published in Comm. Pure Appl. Math., No. 58, pp.1472-1528, 2005, and the article "Fast Discrete Curvelet Transforms," E. J. Candes, L. Demanet, et al., published in SIAM Multiscale Model. Simul., No. 5, pp. 861 -899, 2006. For example, for a two dimensional seismic sample D(t, s) a\ time t and position s , a curvelet coefficient c(j, k, l) for a frequency band j , dip / and time-space displacement k , in a curvelet basis cp . _{i l} is :

C(J ^' , ) = j D(t, s^. _{i x} {t, s)dtds

(2)

[0028] In contrast to the time-space or frequency basis, a curvelet atom (unit vector of the transform domain basis) is localized in both frequency and time-space. Seismic data is represented by few non-zero coefficients in the curvelet domain. Although this embodiment refers to the curvelet domain, other domains may be considered. Other transforms usable to represent the seismic data in other domains may be: wavelet transforms (which were the first multi-scale transforms and are frequently used in signal processing, a Radon transform (which is based on dip decomposition), beamlets, shearlets, seislets, etc. All these transforms divide the Fourier domain in distinct entities, so that seismic data may be represented by coefficients associated with these entities.

[0029] Thus, the sparse (first) inversion seeks a vector x in the curvelet domain, vector x being related to the seismic traces u as:

where ^ is an inverse curvelet transform. Vector x \s sparse because seismic data is represented by few non-zero coefficients in the curvelet domain.

[0030] Ideally, one would want to find vector x ^' \n the curvelet domain, which has fewest non-zero coefficients, and minimizes

- ΐ < σ , (4) where A is a source operator built using the emitted signal information and σ represents the noise level in raw records. However, optimization problem (4) is not solvable in a reasonable time over the curvelet domain. Therefore, the first (sparse) inversion seeks vector x , which minimizes a two-term cost function:

[0031 ] The value of balance coefficient λ (i.e., constraint's weight) is controlled by the operator running the software, and it is selected such that to achieve a reasonable number of coefficients (the higher the value the fewer coefficients).

[0032] This inversion may be solved using a Fast Iterative Shrinkage-Thresholding Algorithm, FISTA (which is described in the article "A fast iterative shrinkage-thresholding algorithm for linear inverse problems" by Amir Beck and Marc Teboulle, published in SIAM J. Img. Sci., No 2(1 ), pp.183-202, March 2009) and the Uniform Discrete Curvelet Transform (which is described in the article "Uniform discrete curvelet transform" by T.T. Nguyen and H. Chauris, published in IEEE Transactions on Signal Processing, No. 58(7), pp. 3618-3634, July 2010).

[0033] In one embodiment, weights are applied to curvelets atoms (the weights discriminating the curvelets atoms based on their position or angular orientations in the volume). In this case, the inversion seeks vector xjn the curvelet domain, which minimizes:

where W is a diagonal matrix of the weights.

[0034] As any iterative process, the first inversion stops when a predetermined criterion is met. The criterion may be performing a predetermined number of iterations, or achieving a target minimization (i.e., residual below a predetermined threshold). At the end of the inversion, a sub-domain of the curvelet domain (known as "restricted support") is defined by curvelets corresponding to the non-zero coefficients of the first inversion's solution. [0035] Further, at 140, a least square inversion of the seismic data is performed on the restricted support. This least square inversion aims find a vector x _s in the sub-domain that minimizes distance between data { Αφ ^~ 'x _s ) reconstructed using vector _s and seismic data b :

where ^ is the inverse curvelet transform in the sub-domain. Vector x _s is an estimate of the underground formation response.

[0036] This second inversion may be solved using a conjugate gradient algorithm. The second inversion also ends when a predetermined criterion is met. This criterion may also be performing a predetermined number of iterations, or achieving a target minimization (i.e., residual below a predetermined threshold). The predetermined number of iteration for the first and the second inversion may be the same or different. The predetermined threshold for the second inversion has to be less than for the first minimization.

[0037] At 150, the solution vector x _s \s used to compute a noise model. When extracting a deblended dataset related to one signal, the overlapping reflections due to another signal are noise. The solution vector may be used to estimate the reflections due to the other signal(s) than a targeted signal.

[0038] Meanwhile (parallel or sequentially with operations 130-150), the seismic data 120 is correlated with the emitted signal information 1 10, to obtain correlated blended common receiver data 160. The computed noise model is subtracted from the correlated blended common receiver data 160, at 170, to obtain deblended datasets 180. These deblended datasets are then further processed to obtain images of the explored underground formation.

[0039] According to one embodiment, the deblending may then be enhanced by using the same double inversion strategy for one of the deblended datasets as illustrated by data flow in Figure 2. A debleded dataset 220 focuses on one of the distinct signals whose reflections are recoded, and may be represented as a seismic trace volume b'of the receiver. Deblended dataset 220 may be one of the deblended datasets 180 or may have been obtained using another deblending method.

[0040] At 230, a sparse inversion of deblended dataset 220 seeks a vector x' in the curvelet domain vector x' minimizing

[0041] A new sub-domain of the curvelet domain ("updated restricted support") is defined based on the non-zero coefficients of this first inversion's solution.

[0042] Then, at 240, least square inversion of the deblended dataset is performed on the updated restricted support. The solution of this second inversion is an updated vector x in the sub-domain that minimizes distance between data { ) simulated using this vector and seismic trace volume b' .

[0043] At 250, the updated vector x _s ' is used to compute an updated noise model that is then, at 270, subtracted from the correlated blended common receiver data 160, to obtain updated deblended datasets 280.

[0044] The methods according to various embodiments have been successfully tested for land seismic data acquired with simultaneous vibroseis.

[0045] Figure 3 is a flowchart of a deblending method 300 according to an embodiment. Method 300 includes receiving seismic data recorded by a receiver detecting simultaneously reflections due to distinct signals, and emitted signal information, at 310. The emitted signal information may include source signature, source activation time and source position for each signal. A source operator used in the first inversion (and also in the following second inversion) is built using the emitted signal information.

[0046] Method 300 further includes performing a first inversion of the seismic data in a transform domain, at 320. The first inversion is formulated to minimize a number of non- zero coefficients of the result. The transform domain may be the curvelet domain. In one embodiment, weights may be applied to discriminate between components. [0047] A sub-domain of the transform domain is defined by vectors of a transform domain basis for which the first inversion has yielded the non-zero coefficients. Method 300 then performs a second inversion of the seismic data in this sub-domain at 330.

[0048] At 340, the solution of the second inversion is used to extract deblended seismic datasets corresponding to each distinct signal from the seismic data.

[0049] The methods described in this section have been successfully tested for land seismic data acquired with simultaneous vibroseis.

[0050] Figure 4 illustrates a block diagram of a seismic data processing apparatus 400 usable to perform these methods, according to an embodiment. Hardware, firmware, software or a combination thereof may be used to perform the various steps and operations. Apparatus 400 includes a computer or server 402 having one or more central processing units (CPU) 404 in communication with a communication module 406, one or more input/output devices (I/O) 410 and at least one storage device 408. Hardware, firmware, software or a combination thereof may be used to perform the various steps and operations of the methods described in this section.

[0051 ] Communication module 406 may be used to obtain the seismic datasets. Communication module 406 may intermediate wired or wireless communication of server 402 with other computing systems, databases and data acquisition systems across one or more local or wide area networks 412.

[0052] I/O devices 410 may be used to communicate with a user or to display any images or models of the surveyed underground formation. I/O devices 410 may include keyboards, point and click type devices, audio devices, optical media devices and visual displays.

[0053] CPU 404, which is in communication with communication module 406 and storage device 408, is configured to perform the first and second inversion, and to extract deblended datasets from seismic datasets as in any of the methods described in this section.

[0054] Storage device 408 may include magnetic media such as a hard disk drive (HDD), solid state memory devices including flash drives, ROM and RAM and optical media. The storage device may store data as well as software code for executing various functions including the deblending methods described in this section.

[0055] Figure 5 is a graphic illustration of blended data in a receiver gather. The vertical axis is time from the emission (increasing from the top to bottom) and the horizontal axis is one line of shots, the traces corresponding to 12.5m spacing. Figure 6 is a graphic illustration of de-blended data obtained from the receiver gather illustrated in Figure 5, using a method according to an embodiment (e.g., method 300). The x and y axes in Figure 6 have the same significance as in Figure 5. A comparison of Figures 5 and 6 reveals the disappearance of the cross-talk noise in the de-blended data.

[0056] The disclosed exemplary embodiments provide methods and systems for deblending seismic data. It should be understood that this description is not intended to limit the invention. On the contrary, the exemplary embodiments are intended to cover alternatives, modifications and equivalents, which are included in the spirit and scope of the invention as defined by the appended claims. Further, in the detailed description of the exemplary embodiments, numerous specific details are set forth in order to provide a comprehensive understanding of the claimed invention. However, one skilled in the art would understand that various embodiments may be practiced without such specific details.

[0057] Although the features and elements of the present exemplary embodiments are described in the embodiments in particular combinations, each feature or element can be used alone without the other features and elements of the embodiments or in various combinations with or without other features and elements disclosed herein.

[0058] This written description uses examples of the subject matter disclosed to enable any person skilled in the art to practice the same, including making and using any devices or systems and performing any incorporated methods. The patentable scope of the subject matter is defined by the claims, and may include other examples that occur to those skilled in the art. Such other examples are intended to be within the scope of the claims.