EA042140B1 - SEPARATION OF MULTIPLE SEISMIC SOURCES OF DIFFERENT TYPES USING INVERSION - Google Patents

Description

Links to related applications

The present patent application claims priority to U.S. Provisional Application Serial No. 62/475068, filed October 12, 2018, titled Separation of Large Numbers of Different Types of Seismic Sources Using Inversion, hereby incorporated by reference in its entirety in all senses.

The Research and Development Statement is not applicable.

State of the art

The purpose of a seismic survey is to create an image or map of the earth's subsurface by sending sound energy into the subsurface and recording echoes returning from the rock layers below. The source of downwardly propagating acoustic energy can be, for example, explosions or seismic vibrators on the earth's surface, or pneumatic sources in the marine environment. During a seismic survey, an energy source is placed at various locations near the earth's surface above the geological structure under study. Each time the source is activated, it generates seismic vibrations that travel down underground, interact with geological structures in the ground, and, upon return, are recorded at numerous points on the surface. A large number of source/record combinations are further combined to form a nearly continuous subsurface profile that can extend for many miles. In a 2D (2-D) seismic survey, record locations are typically distributed along a single line, while in a 3D (3-D) survey, records are distributed across the surface in a grid pattern. Simply put, a 2-D line seismic survey can provide a cross-sectional picture (vertical section) of the earth layers immediately below the survey sites. A 3-D survey produces a seismic data cube or volume that, at least theoretically, represents a 3-D picture of the subsurface zone below the survey area. In reality, however, both 2-D and 3-D surveys survey some volume of earth rock underlying the survey area.

The essence of the invention

In embodiments, in a method of seismic surveying over an area of a subsurface zone containing structural or stratigraphic features (features) indicative of the presence, migration or accumulation of hydrocarbons, at least a portion of a survey with a mixed seismic source (seismic source) is accessed, separated excitation of at least two interfering seismic sources using separation using seismic inversion, generating, based on the separation, a set (multiple) seismograms of individual sources, and using a set of source gathers for hydrocarbon exploration in a given area of the subsurface zone. A mixed source survey comprises excitations from at least two interfering seismic sources, and the seismic source excitations are produced by at least two seismic sources having different signatures (distinguishing feature sets) or frequency responses.

In an embodiment, the seismic survey method initiates a number of seismic source shots over a region of the subsurface zone containing structural or stratigraphic features indicative of the presence, migration, or accumulation of hydrocarbons, and conducts a mixed source survey containing data obtained from reflections generated by some number of excitations of seismic sources. A number of seismic source shots are produced by at least two seismic sources having different signatures or frequency responses, and a mixed seismic source survey includes shots from at least two interfering seismic sources. The excitations of the at least two interfering seismic sources are separated using seismic data inversion, and a collection of source gathers is generated based on the separation of the excitations of the at least two interfering seismic sources.

In an embodiment, in a seismic survey over an area of a subsurface zone containing structural or stratigraphic features indicative of the presence, migration or accumulation of hydrocarbons, at least a portion of a mixed source survey is accessed, the excitations of at least two interfering seismic sources are separated using seismic data inversion, generating a set of separate source gathers based on the separation, and using the set of source gathers to investigate hydrocarbons in a given area of the subsurface zone. A mixed source survey includes excitations from at least two interfering seismic sources, and the seismic source excitations are generated by at least a first seismic source and a second seismic source having different signatures or frequency responses. At least one of the first or second seismic sources is a low frequency source, and at least one of the collection of single source gathers contains low frequency data or data at other frequencies not found in conventional seismic sources.

- 1 042140

In an embodiment, in a seismic survey over a region of a subsurface zone containing structural or stratigraphic features indicative of the presence, migration, or accumulation of hydrocarbons, at least a portion of a mixed source survey is accessed and a set of generated source gathers is used to investigate at hydrocarbons in a given area of the subsurface zone. A mixed source survey comprises excitations from at least two interfering seismic sources, and the seismic source excitations are generated by at least two seismic sources having different signatures or frequency responses.

While several embodiments have been disclosed, other embodiments will be apparent to those skilled in the art upon reading the detailed description below. It will be understood that certain embodiments disclosed herein may be modified with respect to some of their features, within the spirit and scope of the claims presented herein. Accordingly, the following detailed description is to be regarded as illustrative only and not as limiting the invention.

Brief description of the figures

The attached figures illustrate embodiments of the subject matter disclosed herein. An understanding of the claimed object of the invention may be facilitated by the following description, together with the attached drawings, in which:

in fig. 1 shows a continuous seismogram trace for a single receiver spanning multiple shots from a number of sources;

in fig. 2 shows the combed data for the first source;

in fig. 3 shows the combed data for the second source;

in fig. 4 shows data from one of the earlier iterations for the first source after applying the smoothness constraint;

in fig. 5 shows data from one of the earlier iterations for the second source after applying the smoothness constraint;

in fig. 6 shows partially split overlapping data for the first source;

in fig. 7 shows typical results for a first simulation source that does not experience interference from a second source;

in fig. 8 illustrates the general configuration of the technical environment of the present disclosure;

in fig. 9 shows a seismic signal processing sequence suitable for use in the present disclosure;

in fig. 10 is a plan view of a mixed source seismic survey;

in fig. 11 schematically shows how different seismic bursts can be identified and isolated from a mixed source survey;

in fig. 12 shows a particular example of a workflow for a typical case, where source 1 is an impulse source, and sources 2 is a vibration source, with variable source signatures;

in fig. 13 shows an example of a method for extracting coherent signals from a noisy gather;

in fig. 14 shows an example of a method for replacing lost or contaminated tracks in a gather using Projection onto Convex Sets (POCS) interpolation;

In FIG. 15 is a graph of a common receiving point gather with three interfering seismic sources, each of which has its own uniform operating period different from the others.

Detailed description of the invention

Unless otherwise indicated, any use in any form of the terms connect, engage, bind, attach, or other term describing an interaction between elements is not meant to limit that interaction to only direct interaction between elements, and may also include indirect interaction between the described elements. In the following discussion and in the claims, the terms including and containing are used in a non-limiting manner, i.e. are to be interpreted as meaning including, but not reducible to.... The various characteristics mentioned above, as well as other features and characteristics described in more detail in this disclosure, will be well understood by those skilled in the art having read the foregoing detailed description of the embodiments with reference to the appended blueprints.

A seismic survey is made up of a very large number of individual seismic records or seismic traces. In a typical 2-D survey, several tens of thousands of seismic traces are typically obtained, while in a 3-D survey, the number of individual traces can reach many millions of traces. In Ozdogan Yilmaz Seismic Data Processing, Society of Exploration Geophysicist, 1987, chapter 1, p. 9-89 are general information related to conventional 2-D data processing.

- 2 042140

General information related to the acquisition and processing of 3-D data can be found in Yilmaz's book, Chapter 6, p. 384-427.

Traditionally, a seismogram track is a digital record of acoustic energy reflected from discontinuities or interfaces in the subsurface, where partial reflections can occur whenever there is a change in the elastic properties of subsurface materials (primary reflections). Reflected acoustic energy of other types is also increasingly used in geophysics. In addition to primary reflections, acoustic energy can, for example, be converted from vibrational waves of one type to vibrations of another type, refracted at layer boundaries, scattered by scattering objects, or turned back up due to velocity inhomogeneities (refracted waves). Waves can repeatedly interact with the structures of the subsurface zone inside the Earth before returning to the surface and being registered. Depending on the geophysical application, some waves are considered signal and some waves may be considered noise.

Samples are typically taken at 0.002 s (2 milliseconds or ms) intervals, although 4 ms and 1 ms sampling intervals are also common. Each discrete sample in a conventional digital seismogram is associated with a propagation time, and, in the case of reflected energy, a round trip time from the source to the reflector and back to the surface, assuming, of course, that the source and receiver are both located at the surface. In practice, many different variations of the conventional source/receiver scheme are used, such as VSP (vertical seismic profiles) surveys, sea bottom surveys, and so on. Further, the position of the gather trace on the Earth's surface is carefully tracked and is usually made part of the gather trace itself (as part of the header information). This allows the information contained in the seismogram tracks to be subsequently associated with specific surface and subsurface points, which is a means for placing and contouring the seismic data, and the attributes extracted from them, on a map (ie, displaying the data).

Displaying 3-D survey data can be done in various ways. First, horizontal slices can be extracted from a stacked or non-stacked volume of seismic data by collecting all discrete samples taken at the same travel time. This operation results in a horizontal 2-D seismic data plane. By creating a dynamic image from a sequence of 2-D planes, the interpreter can smoothly shift the displayed area through the volume, giving the impression that the layers are sequentially removed, exposing the information underlying them. Similarly, a vertical seismic data plane of arbitrary azimuth can be drawn through a volume by collecting and displaying seismograms lying along a defined line. This operation essentially extracts a single 2-D seismic profile from a 3-D dataset. It should also be noted that a 3-D dataset can be thought of as being made up of 5-D datasets that have been dimensionally reduced by stacking them into a 3-D image. The coordinate axes are typically time (z depth), x (eg, North-South direction), y (eg, East-West direction), source/receiver offset in the x direction, and source-receiver offset in the y direction. While the examples given here are mainly for the 2-D and 3-D cases, extending the process to four, five, or more dimensions is straightforward.

Seismic data, once acquired and properly processed, provides a wealth of information to the prospector, one of the people in an oil company who determines where to drill wells. For example, a seismic profile provides the explorer with a wealth of data on the rock layer structure of the subsurface zone and often reveals important features associated with hydrocarbon retention and accumulation, such as fractures, folds, anticlines, unconformities, and underground salt domes and veins, among others. During computer processing of seismic data, estimates of velocities in subsurface rocks are usually generated and inhomogeneities near the surface are detected and displayed. In some cases, seismic data can be used directly to estimate rock porosity, water saturation, and hydrocarbon content. Less explicitly, empirical correlations of seismic waveform attributes, such as phase, maximum amplitude, peak-to-minimum ratio, and a variety of other attributes, can often be established with known hydrocarbon deposits, and this correlation can be used in relation to seismic data collected from new targets. .

A well-known problem with seismic data is the relatively high cost of obtaining it. Indeed, in some cases, the cost of surveying can determine how cost-effective the development of the proposed site is. Therefore, technologies that reduce the cost of such surveys are important. One of the promising ways to reduce the cost of obtaining seismic data is to conduct closely spaced explosive excitations of two or more sources. The meaning of this approach is that a linear or areal array of receivers is used, and one or

- 3 042140 more sources during one recording period. In this case, reflections in the subsurface energy zone from one excitation source can be combined with reflections from later excitations, i.e. results from a mixed source survey. It should be noted that this contrasts sharply with the usual survey methods, when the waves returning from the subsurface zone from one source are not allowed to overlap with the returning waves from another source.

Mixing can occur when a reactivation of one source is performed before the stream of Earth-returning waves generated by the source activation has had time to decay (so-called self-timed data recording). This can also happen when several sources are activated during the same time period, when at some ranges or offset times their returning wave flows overlap. Both methods can be used simultaneously within the same registration.

Seismic surveys with two or more sources of the same type are possible for both land and sea surveys. For example, two simultaneously operating vessels with pneumatic sources can be used in marine seismic mode. In some methods, the two ships usually move in sync and the explosions they produce are usually carefully planned and coordinated (Spatially Separated Simultaneously Activated Source or Apparition method). In other ways, the sources move independently and work independently (the usual ISS methods are independent seismic source). However, the methods currently used assume that the shared sources are of the same type and, after separation, the returns from all sources are combined in a common data space in which it is not important for further processing which source a given bit of seismic energy data originates from. The return signals are usually reflections, but other types of return signals are also increasingly used, such as refracted waves, refractions, diffractions, multiply scattered reflections, etc.

Different types of sources have different advantages, and the industry is increasingly using sources of different natures to acquire data, thus benefiting from the use of different types of sources in a single survey. To save time and costs, the sources can work simultaneously. For example, in a marine survey, pneumatic sources can be used to cover the 2-80 Hz range, electric spark sources for the 70-120 Hz range, and low-frequency vibrators for the 1.5-3 Hz range, all of which can work simultaneously and be recorded by the same receivers.

In some embodiments, an attempt may be made to combine all of these recorded sources into a single broadband dataset, simulating a survey using a single source that combines the benefits of all different types of sources. The sources, however, cannot be considered identical, and they are usually consistent with each other, primarily so that they fit into the combined data set. In other cases, the sources may be of very different types, or used for completely different purposes, so that a clear distinction is maintained between them and the resulting datasets can be processed separately. In some embodiments, both approaches may be used simultaneously, with a combination of some overlapping sources, while separating other sources. For example, some embodiments may combine data from a pneumatic source and an electrospark source into a common 2-120 Hz range dataset used for both velocity model building and display, but leave a separate low-frequency 1.5-3 Hz vibrator for use. only at the initial stages of constructing a velocity model. In areas with complex geology, building a velocity model may be a prerequisite for successful imaging; mapping algorithms require a good velocity model of the Earth to get good results. The same data can be combined differently, depending on the application.

It also often happens that if two or more seismic surveys (which are assumed to be separate) are taken close enough to each other in time and space, sources from one survey are also recorded in the other. This seismic interference is usually perceived as unwanted interference. However, using the method described here, the interfered data may be separated and removed, or separated from each other and isolated as a separate data set with an ultra-long shot point, or may even be perceived as a desired additional signal and be included in the survey as another source.

With that said, one or more embodiments relate to a system and method for separating multiple seismic sources, which may have different source signatures, and where a number of seismic sources may be activated in a single acquisition of seismic data. Unlike previously used approaches, one or more embodiments of the proposed systems and methods allow the user to separate sources when returning seismic waves from different sources are received with overlap in time. Thus, in one or more embodiments, two or more separate sets of data from sources can be obtained, each of which contains different information from sources. Use over

- 4 042140 one set of sources will speed up seismic exploration, provided that the recorded seismic energy from different sources can be separated. This method can be particularly effective in wide azimuth reflection seismic surveys, since the kinematic characteristics of the reflectors will be significantly different for different shots, which makes it possible to use the continuity of reflections in several dimensions. In addition, the possibility of obtaining data sets from different sources will allow obtaining different types of information during the survey.

In short, systems and methods in accordance with one or more embodiments use a reversal technique to separate information obtained by acquiring seismic data in multi-source datasets. Seismic data acquisition is the recording of reflections from many seismic sources (eg Vibroseis® seismic vibrator, pneumatic sources, low frequency sources, etc.). Seismic records in which overlapping reflectors can be registered can be divided into datasets from a large number of sources. Random or pseudo-random delays between shots give rise to incoherent interference between different sources, while reflections associated with the same source give rise to coherent seismic wave arrivals. In some embodiments, one or more of the sources may be continuous or semi-continuous while the other sources are activated at random times. The separation can be accomplished by a numerical inversion process that uses the signatures of each source, the start times of each explosion of each source, and the coherence of reflected wave arrivals between nearby explosions. The signature of the source may correspond to the identity of the source (eg, sweep timing of each shot, frequency response, etc.). Potentially, the proposed method can speed up and reduce the cost of seismic surveys, obtaining a signal of better quality and more broadband than previously possible.

In some embodiments, the recorded data may be mathematically expressed as d=r m where d represents continuously recorded data containing overlapping signals, m represents the data tracks that would be recorded in a conventional seismic experiment with one source recorded at each time point, where the data from each source activations are represented by a separate seismogram track, and

G is a blend operator that sums together the tracks in m, using known source activation times, to form continuous blended data.

The systems and methods of one or more embodiments describe how to use mixed recorded data d to accurately estimate the required unmixed data m, which can then be used for further processing as if the data were recorded in separate, unmixed surveys. In the case of a continuous acoustic source, the moments of activation of the source can be set to convenient periodic repetition intervals, and G can contain a corresponding smooth decrease so that successive oscillation excitation points are sequentially added to each other, restoring the signal of a continuous source.

As represented in a computer system, d will typically be represented as a matrix whose dimensions are a time sample and a receiver number. In some embodiments, the implementation of the time axis may be divided into areas of record of a certain length. For example, continuous time may be divided into one-minute seismogram tracks, with the time axis divided into two dimensions - a minute, and a time sample within a minute. m typically represents a matrix whose dimensions are time, source activation number, source type, and sink number. The r will have dimensions of time (perhaps subdivided into two dimensions), source type, source number, and destination number. r will usually be a very large but very sparse matrix, and is therefore represented as a function call in computer software. It is usually not stored as a large static matrix, but instead its elements can be computed as needed from a table of source activation times.

In some embodiments, m may represent a complete survey containing a gather trace for each possible combination of source type, source location, and receiver.

Source-receiver combinations that do not actually occur in the survey (and therefore are not represented in the recorded data d) correspond to zeros in G. In this embodiment, m may contain tracks interpolated by the results of the process of separating signals from signals generated by nearby sources, which fill gaps in seismic surveys. Interpolation can also be used to replace recorded gather traces for which shot quality and/or receiver performance was so poor that the corresponding traces are best considered lost, or to replace traces clogged with transition spikes.

- 5 042140 interference. In other embodiments, m may only contain tracks for source-receiver combinations that actually participated in acquisition of the survey data. Intermediate cases are also possible, where m represents a complete survey for some types of sources, but only source-receiver combinations that were actually involved in the case of other types of sources. In any case, the problem can be viewed as an underdetermined problem, with more elements in m to compute than there are uniquely defined elements in d. To solve this problem, it is necessary to make some assumptions. The decision should take into account that the change in the recorded data should be a smooth function of the position of the source and sink for a given source, and use this to constrain estimates of the unblended data m, also called reference data. For this to work, the data must be acquired to ensure that wave arrivals from different sources are not coherent with each other, typically by introducing time blur at the source activation times. Information about the blurring of the moment of time is contained in the mixing operator G. To improve smoothness, in one or more embodiments, an additional constraint m = S m can be added where S is a matrix or operator describing the coherence criterion appropriate for each source.

The coherence criterion will then ensure that the data from the right sources are properly separated into m. The solution then reduces to solving a pair of equations d = Г m and m « S m

In one or more embodiments, a weight term W may also be introduced and the pair of equations solved

Wd"WrmHm"S where the weight member is a matrix W or, more generally, an operator.

This allows one or more embodiments to weight certain receivers over others to balance amplitudes from different types of sources, or to weight different time or frequency windows differently. Typically, these equations can be solved by a least squares algorithm (eg, the L2 algorithm) or a fast minimization algorithm (eg, the L1 algorithm), using one of the many available algorithms suitable for this purpose. The disclosed systems and methods offer an alternative iterative solution method that has shown excellent results. In some scenarios, there are two or more overlapping simultaneous sources of the same type, and the goal is to form an unblended data set equivalent to that obtained in a conventional survey where the same set of sources are fired sequentially and do not overlap. As for S, S can be any operator that limits the similarity or coherence of nearby explosions. S is not limited to unidirectional applications and is applicable to high dimensional data (eg, 2-D, 3-D, 4-D, etc.) or any suitable subset of reduced dimensional data. Also, it should not be deduced from the above equation that S must necessarily be a linear operator. Although in some applications S may be a linear operator, in other cases this variable may be represented by a non-linear operator or a linearized version of it.

In some embodiments, S is represented by a fast Fourier transform (FFT). For randomly placed explosions, S is best calculated by discrete Fourier transforms or some other method that provides a random sample of explosions. In some embodiments, S may include a 3D Fourier transform and a threshold operation to remove weak seismic arrivals that turn out to be noise and pass strong arrivals that are typically coherent signals. This allows S to improve the coherence of the recorded seismic signals.

The method may extract time windows from the data d corresponding to a source activation followed by an echo reception interval. This selection of time windows can be called data combing d (comb filtering). If there is no data overlap from individual source activations, then combing the data will directly result in m. This scenario corresponds to conventional single-source seismic processing. Mathematically,

Olcomb ⁼ Г* d, where the combing operator Г _t is the result of the transposition of the mixing operator Г.

If the sources overlap, then the gather traces in the mc _omb combed data can be considered contaminated with crosstalk from the other source(s). However, due to the blurring of the source activations, the crosstalk noise will be generally incoherent. This noise can therefore be attenuated by applying S, which removes non-smooth incoherent energy, yielding an improved unblended data estimate m _smooth . Mathematically, ^smooth ⁼ S m _combine - S Г d.

The algorithm then computes the mixed seismic data corresponding to this m _smooth , which

- 6 042140 ray represents an approximate estimate of the recorded data d.

The estimation of the data can be calculated as destimated ⁼ G ITIsmooth·

If the difference between the actual data d and d _estimated is small enough, then the process has successfully determined the coherent signals of the msmooth dataset, separated from signals from nearby sources, that match the field-recorded data d, and the process is terminated. Otherwise, the method can calculate the mismatch ^residual ^- d - destimated data and use it to improve m _smooth . This improvement is carried out by performing the following steps. First, the method can update the evaluation of the set

H1 _U pdated ⁼ Hlsmooth ⁺ G dresidual signal data, separated from signals from nearby sources. The method can then apply a coherence constraint to this adjusted estimate to compute a new m _smooth :

S IBupdated ITIsmooth

Finally, the method can now return to the beginning of this paragraph and repeat these steps until the residual mismatch (remainder) of the data is small enough. The definition of what can be considered small enough may be chosen as a decision criterion and/or selected based on an acceptable reduction or minimization of the residual data, and as a result may vary from analysis to analysis. In some embodiments, the data residual may be zero, while in other embodiments, the data residual may be considered sufficiently small if its values are less than the data noise floor.

In practice, on each iteration of the above loop, the method might store m _smooth from the previous iteration ( m _save ) and the proposed updated version ( m _proposed ), and then compute a weighted average of the two to use as the new updated version:

Hlsmooth ⁼ (1 - λ) resave ⁺ λ rnp _rO p _Osed .

The weighting factor λ must be positive, but not exceed the maximum mixing factor by more than one, i.e. the maximum number of overlapping sources in the processed data. Each track in m corresponds to a particular source-destination pair, so the method can simply use a weight term, making λ a function of source, destination, offset, time, and so on. In general, a weighted average can also be applied in the transformed domain so that λ can also be a function of frequency, and so on. For example, m _S mooth ⁼ FT [(1 - λ) FT(msave) ⁺ λ FT(mprOp _OS ed)], where FT and FT ^-1 denote the Fourier transform and the inverse Fourier transform, respectively. Weights, if necessary, can be applied in more than one area.

The method may include excitations of sources occurring within the same record, separated in time by random or pseudo-random time intervals. If the spatial arrangement of the recorded explosions is random, then it is unlikely that when recording data from each explosion, coherent energy from other explosions that occur with a small time interval from the recorded explosion will be recorded, which improves the separability of these explosions. When bursts are activated at random times, when these bursts are corrected for their individual times, they will give signals that are coherent from burst point to burst point, while interfering bursts will most likely be incoherent, and can be separated by the process described here. inversions. This improves the efficiency of using coherence in the separation process described here.

Preferably, the seismic survey itself will also emphasize the lack of coherence between different sources. For example, when firing at a position in a marine seismic survey, each source is activated when it reaches the next predetermined burst point. Target blast points are usually spaced at regular intervals. When the ship is moving at a constant speed, explosions will occur at regular intervals. As shown above, for mixed shots, time blurring is introduced to ensure that each explosion is not coherent with the others. However, in one or more embodiments, it can also be ensured that even without dithering, the different sources will still not be coherent when the survey is conducted in such a way that there is no formation of two ships in formation, i. e. when one ship is moving at a constant offset relative to the other. A simple way to achieve this is to assign different speeds to different vessels, where even if two vessels are moving in the same direction with parallel survey lines, their respective explosions will not fit into the general pattern. Different ship speeds correspond to different burst intervals, so the bursts will not be coherent even in the absence of blur.

The above description shows how to apply the method to sources of the same type when the sources are combined into a single volume m of unblended data. With the advent of seismic sources

- 7 042140 kov new type, for example, Wolfspar®, i.e. low-frequency vibration source, there was a need for systems and methods for processing data recorded while collecting data from sources of more than one type. Such systems and methods can process data recorded while collecting data from more than one type of source by applying:

(1) generalizations of the above algorithm to the case of many inputs and outputs, (2) modifications of the algorithm to balance data of different types so that one type of source does not dominate others in the inversion, (3) accounting for source signatures that can differ to a very large extent degrees in time duration, frequency content, etc., and may vary from oscillatory excitation point to oscillatory excitation point, and/or (4) sources that may be coherent in different regions.

The new system and method of one or more embodiments can be illustrated with a simple example of simultaneous data acquisition and processing using several different types of sources (eg, two sources - pneumatic and one low frequency source). The two pneumatic sources are of the same type and together provide one combined survey. The low frequency source is a different type of source with a different acquisition plan and provides a separate survey. In this example, the method records one combined dataset d as before, but now the method aims to define two different m's, one for a pneumatic survey and one for a low frequency survey.

Mathematically, the data in the method is now d = G _A t _A + r _w mw, where G _A is the mixing matrix fitted to the pneumatic source model data m _A , a

F _W is a mixing matrix fitted to the low frequency source model data m _W .

The goal is to find m _A and m _W that generate the recorded data d and also satisfy the conditions

GPa ~ Sa Sd and mw ⁼ Sw where S _A and SW define suitable smoothness constraints suitable for the respective source types and acquisition geometries.

As before, the method performs data combing. In FIG. 1 shows a continuous seismogram trace for a single receiver covering several shot points, both from pneumatic sources and from a low frequency vibratory source. The arrows indicate the places where it is clearly possible to see a low-frequency vibration source appearing between the arrivals of waves with a larger amplitude from a pneumatic source. The method further combs through the data from the pneumatics, with the calculation

WcombA ⁼ G ad (with the result shown in Fig. 2), and from a low-frequency source, with the calculation of 1Pcomb\L/ = r*wd (with the result shown in Fig. 3). The arrows in Fig. 2 note particularly strong interference from other sources, where A indicates interference from a pneumatic source, and W indicates interference from a vibrating low-frequency source. The high frequency noise shown in Fig. 3, is the interference from the pneumatic source. In one or more embodiments, smoothness constraints may be imposed, applicable to each case:

^m smoothA ⁼ Sa WcombA ⁼ Sa G*d d (the result is shown in Fig. 4) and nismoothW = Sw m _C ombW = Sw r*wd (the result is shown in Fig. 5).

The estimate of the data can be computed as ^estimated ⁼ l~A MsmoothA ⁺ ГW m _S moothW·

If the difference between the actual data d and d _estimated is small enough, then the method has successfully found a coherent data set m _smoothA , separated from a nearby data set m _smoothW , consistent with the data d obtained under operating conditions of the survey, and the method is completed. Otherwise, the method calculates the data mismatch dresidual ⁼ - d _es timated, and uses it to improve m _smoothA and m _smoothW . This is done with the following steps. First, the method updates the estimates of a dataset separated from the dataset from nearby sources, namely

ITIupdatedA ⁼ MsmoothA ⁺ G d Ciresidual I m _U pdatedW ⁼ UlsmoothW ⁺ G w ^residual

The method then imposes coherence constraints on the updated estimates to calculate new msmoothA and msmoothW:

- 8 042140

Sa ITIupdatedA m _S moothA ^and Sw m _up datedW ^smoothM

Finally, the method may repeat these steps until the residual data mismatch is small enough for the method to terminate. The method can be applied to sources of more than two types. As an alternative to obtaining an estimate of ^estimated ^- Gg ITIsmoothA + GW ITIsmoothW, the data estimate can be adjusted by G _a m _smooth A and G _W m _smoothW using alternative iterations that separate possible crosstalk between G _A m _smoothA and G _W m _smoothW . This path can also be extended to more types of sources.

In FIG. 6 shows the resulting data from a pneumatic source, separated from data from nearby sources, after several hundred iterations. After a sufficient number of iterations, this data approaches with great accuracy the exact unblended result shown in FIG. 7.

As in the case of data of the same type, at each iteration in the loop shown above, the method remembers m _smooth from the previous iteration and the proposed updated version, and then calculates the weighted average of these two matrices to use as a new update:

MsmoothA ⁼ (1 ^- Hell) mSaveA ⁺ Hell ITIproposedA, AND ITIsmoothW ⁼ (1 ^- Aw) mSaveW ⁺ Aw ITIproposedW.

To balance different types of sources, different weights are needed, and as before, λ can be used as a weight by source, source type, receiver, offset, time, etc. For example, if the pneumatic source signals are of significantly greater amplitude than the low frequency source signals, the method may start the inversion at a higher λ _Α than λ _W until the pneumatic source signals converge, then increase λ _W .

In practice, pneumatic sources can provide significantly higher amplitudes than low frequency source(s) above about 5 kHz, but low frequency source(s) can have amplitudes much higher than pneumatic sources, below about 2.5 kHz. Therefore, in some embodiments, weighting factors are applied in the frequency domain, with initial inversion iterations weighted to emphasize the correction of the strongest signals in each frequency range, after which, when stronger signals stabilize, in one or more embodiments, further iterations are corrected. weaker signals.

Pneumatic sources are generally assumed to have a consistent and repeatable source signature, and many efforts have been made to achieve this, but this may not be the case for all types of sources. A low frequency source in particular may not produce identical signals on every activation. However, the amount of water displaced is continuously recorded, and the signature of this source is therefore known in detail each time the source is activated. By designating the source signature as V, the method can compute the deconvolution operator of the stabilized source signature using any of the standard techniques. For example, the method may apply the traditional formula used in the frequency domain to compute a stable approximation of the inverse matrix:

V ¹ = V* / (VV* + e ² ), where the choice of the value of e is determined by the requirement to stabilize the inversion. If the source signature changes from one source activation to another, then the method will need to adjust for this change before the method can be applied to use coherence constraints.

Therefore, instead of

Sw m _U pdatedW ITIsmoothW, in the way

V Sw V' ¹ m _up datedw msmoothW-

Because the V ^-1 inversion is an approximation of the true inversion, there will be errors as a result of applying the inversion. However, the errors will be random and will be reduced by the iterative nature of the solution.

For some choices of S and/or V, it may be preferable to smooth and/or match the signature only when there is a change in m at each iteration, where the change will be called m _diff . If there is a need in the method to correct the signature of the low-frequency source, without changing, for example, the signature of the pneumatic source, then the update step in the algorithm becomes:

ITIdiffA ^- G A d _r esidual ^and l^diffW ^_ G W ^residual, followed by m _S moothA ⁺ Sa ITIdiffA m _S moothA AND ITIsmoothW ⁺ V Sw V ¹ ITIdiffW ITIsmoothW

Depending on the choice of operator S and the signature or matching operator V, this alternative form of the algorithm may lead to a better result. In particular, if the data residual is zero, the algorithm does not make any further adjustments, regardless of S and V.

- 9 042140

Although the above example uses two different types of sources, in other embodiments, data acquisition from more than two types of sources (eg, n types of sources) can be used at the same time. Mathematically, in this case, the method has d = Γι mi + G2 m2 + ... + G _p m _n where G 1 is the mixing matrix suitable for data m1 of the first type source, and G _p is the mixing matrix suitable for data m _n source n ^th type. The task is to find m1, m2, ... and m _n that generate the recorded data d and also satisfy the conditions _mi ~ Sim mi, m ₂ ~ S ₂ m ₂ , and m _n « S _n m _n , where S1, S ₂ , and S _n are appropriate smoothness constraints for the source of the appropriate types and acquisition geometry. In some embodiments, it is also possible to consider certain types of noise as a different source, and make decisions with that in mind. For example, if the noise is due to surface waves propagating at a slower phase velocity than any seismic arrivals of interest, and there is a sufficient sample of surface waves received by the receiving network (which will have a frequency below a sufficiently low frequency), then in some embodiments, there may be a transformed region is found where the surface wave data becomes sparse. Given this information, in some embodiments it is possible to create a smoothness constraint S _noise that operates in this region and uses this sparsity to isolate surface wave noise and attenuate the desired seismic signals. If the noise is generated by an explosion, in one or more embodiments, the timing of the explosion may be known. If the noise is continuous and background in nature, then in some embodiments evenly spaced windows with sloping transition zones at the edges can be used, overlapping the data arrival time windows for the corresponding burst times in the mixing/separating signals from signals from signals from nearby sources.

For a controlled source, such as a land or sea vibrator, in one or more embodiments, variation from one source to another activation during data arrival can be intentionally introduced, and this can also be a way to make different sources incoherent, which ensures their separation. In one or more embodiments, different sweep patterns for different vibrators may be used for this purpose. The current Wolfspar® low frequency source does not provide this flexibility, but it does allow the sign or phase in the source signature to be changed. In one preferred embodiment, sign/phase reversal can be used instead of time blur (or to increase time blur) to break unwanted coherence between sources.

In one or more embodiments, V may also be used not to remove the source signature, but instead to match the changed source activations to the standard signature. If M is the measured (or estimated) source signatures and D is the desired source signatures, then in one or more embodiments, for example,

V = M D7 (DD* + €1 ² ) AND V ¹ = V _o / (Vo V _o + e ₂ ² ), where V0 is calculated from M and D exactly like V, but possibly with a different ε1, for example, ε ₁ =1. Alternatively, in one or more embodiments, the implementation can be used

V ¹ \u003d D m7 (M M * + e ₂ ² ).

These formulas will be mainly applicable to the frequency domain.

The application of V ^-1 converts the actual source signature (M) to the desired source signature (D), which will introduce noise if the desired signature contains frequencies that are not present in the source signature, or are only present at a level not substantially greater than the noise. The stabilization parameter ε2 limits the amount of gain applied, and stabilizes the result. V converts the given signature (D) of the source back to the actual signature (M), which is usually a more stable operation. Typically, in one or more embodiments, therefore, more stabilization would be chosen for the inverse operator than for the forward operator, i. e. one or more embodiments would typically have ε ₂ >ε1>0.

It should be noted that in the case of a marine survey, one or more embodiments could select a desired source signature D that does not include the ghosting effects of the retrace source image, but includes this ghosting in M, and therefore uses a signature corresponding to ghost suppression. waves in the data. For this to work properly, in one or more embodiments, it must be ensured that there are changes in ghost image marks during data arrival, i.e. the depth of the source varies between explosions. Alternatively, in one or more embodiments, it can be ensured that the elements in the source matrix are not all at the same depth.

There may be a need to map the signature even for sources of the same type. Decades of work have been spent making source signatures based on pneuma

- 10 042140 source uniform and predictable within their main operating range. That is, approximately 5-40 Hz. Characteristics outside this frequency range have received less attention, and airstream streamers have been found to be significantly more variable in terms of their signatures over a wider range. Therefore, it would make sense, for example, to treat the port and starboard streamers for obtaining data from one vessel as two separate sources. The resulting two sets of data can then be compared with each other to quantify differences in source signatures that would ideally be identical sources. Or the same seismic streamers of pneumatic sources at different times can be considered as several separate sources, in order to fix and introduce correction of undesirable long-term changes in one source.

One or more embodiments may also split one source or source type into two or more by frequency. Continuing with the previous example, in one or more embodiments, bandpass filtering can be used to separate the data into two frequency ranges, eg below and above 5 Hz (with some overlap between them), such that d ⁼ d| _0W + dhigh In one or more embodiments, d _low and d _high can then be solved as two separate problems. For example, assuming that two airstream streamers operate identically at higher frequencies, but may differ below 5 Hz, then in one or more embodiments, solutions can be obtained for two different airstream streamers separately for d _low , but how for a single type source for d _high .

By measuring the changing signature of the air source in situ, for example, from hydrophone(s) placed near the streamer of the air sources, in one or more embodiments, the knowledge can be fed into the process and the measured signatures M can be used to correct the changing signatures of the air source to a predetermined a fixed signature D, just as in one or more embodiments could be done for a vibration source with a variable but measured signature.

While it may be preferable to match different types of sources to the same source signature before extracting coherent energy, such matching can create a situation where the signal-to-noise (S/N) ratio is low for one type of source. For example, a low frequency source such as Wolfspar® may be designed to supplement low frequency pneumatic sources. And to generate a minimum signal at higher frequencies. Thus, at frequencies above approximately 4 Hz, the pneumatic sources dominate the low frequency source, while below 2 Hz, the low frequency sources dominate the pneumatic sources. The combination of a low frequency source with pneumatic sources can create significant noise in the low frequency source signal at frequencies above 4 Hz, and in the pneumatic source signal at frequencies below 2 Hz. The problem as a whole is a problem with fuzzy constraints, since the sources of two different types usually overlap heavily in a narrow frequency range of about 2 to 4 Hz. Therefore, in one or more embodiments, the implementation of this problem can be solved in these two different frequency ranges separately. In particular, in one embodiment, the low frequency source above 5 Hz may be treated as noise first and not included in the higher frequency band inversion. Once the signals of the two pneumatic sources are separated from each other's signals, a low frequency source can be introduced and λ _A is set very small so that only very small corrections are allowed in the pneumatic sources. For the low frequency range, the reverse approach can be used, where the solution for the low frequency source is found first, followed by the addition of pneumatic sources. Finally, one embodiment may end up with a full range inversion. To generalize the concept to seismic sources of all types, in some embodiments, effective frequency bands may be defined that differ in acceptable S/N for each source, and very low weighting factors applied to other frequency bands in order to avoid crosstalk between different sources. .

Some embodiments have different types of sources, each optimized to cover a different frequency range so that different types of sources can be combined to form a single ultra-wideband source. In one or more embodiments, the signature of the desired UWB source can then be developed, and this signature can be partitioned over the optimal frequency ranges of different types of sources. So, for example, if D is the signature of the desired wideband source, and if the full frequency range is covered by three types of sources, in one or more embodiments, the source signature can be partitioned into overlapping frequency windows such that

D = D|ow ⁺ Dmjd + Dhigh,

Moreover, the frequency ranges are chosen so as to cover the optimal S / N ranges

- 11 042140 for each source type. The choice of frequency windows may differ from one analysis to another, as the choice may depend on the spectral content of the seismic sources. For example, a low frequency source may have a low frequency range of about 0 to 2.5 Hz, a mid frequency range of about 2.5 to 5 Hz, and a high frequency range of about 5 Hz to the Nyquist frequency (half the sample rate). Depending on the specific model and seismic sources used, other frequency ranges may be defined and used for source signatures. Then, in one or more embodiments, each type of source may be matched to a source signature in the frequency range of a suitable band pass filter using the matching operator V described above. After solving the problem for d ⁼ Г| _About w GP| _Ow + r _m jd ITImjd + Thigh ^highi in one or more embodiments, the desired ultra-wideband data set ^broadband ^- V |Ow ^low ⁺ V mjd ITlmid ⁺ V high HTIhigh can be further modeled

For implementation convenience, one or more embodiments may, for example, use the required data m' = V ¹ t as a variable adjusted before each iteration, so dbroadband ^{= m} 'low ⁺ ITl'mjd + m'high .

This requires direct adaptation of the rest of the formulas to compensate for the change in variables m to m'. One skilled in the art would recognize that there are other possible variations of the algorithm that change the calculations but leave the underlying mathematics unchanged.

In order to co-introduce different data, separated from data from neighboring sources, into a combined UWB dataset, all sources of different types must be represented on a common grid. In practice, for sources of various types, the reception and processing of data usually occur in different ways. In particular, at lower frequencies the sampling may be coarser, with the result that many points from the source may be lost in the case of low frequency sources. In one or more embodiments, these missing source points can be filled in during a process of separating overlapping signals by using a blending G operator to multiply these missing gather traces by zero. An appropriate selection of the smoothness constraint will have the effect of interpolating these lost source points so that all frequency ranges can be represented on a common grid, forming a combined dataset modeling a single broadband source.

Obviously, the correct choice of the smoothness constraint S is important for solving the problem. In general, any denoising algorithm used on a well-sampled seismic record can be used. One approach is to transform the data into a region where the desired signal becomes sparse and the noise non-sparse. All data below some percentage threshold can then be set to zero and the data can then be converted back to the original area. Typically, the data can be subdivided into overlapping pyramidal subsets, each of which can be smoothed, and the flattened sets can then be rejoined. This technique provides noise removal. Typically, a local 2-D, 3-D, 4-D, or even 5-D Fourier transform is chosen, representing the data as the sum of seismic wave arrivals at a linear time increment: on lines, planes, hyperplanes, etc. In the Radon transform, the data is represented as a sum of hyperbolas, so if the desired signals are assumed to be hyperbolic, then this representation area may provide the best result. Other applications use short wave trains (wavelets) of various types, and more recently other transform domains have been proposed, such as curvelet transform, seismlet transform, etc.

In one or more embodiments, a Fourier transform may be used and thresholding may start at nearly 100% while retaining the most coherent part of the data. For example, in FIG. 4 shows the inversion at the initial stage, with a very strong coherence constraint, giving an unnaturally smooth result. The high threshold used in the Fourier region only passes a narrow range of frequencies around the maximum frequency of the pneumatic source spectrum (about 8 Hz). With each iteration, the threshold is lowered, eventually allowing all or almost all of the data to pass in the final iteration(s).

The preferred program for this routine is threshold = 1 - (iter / niter) ² , where iter represents the iteration count and niter represents the number of iterations. A threshold of 1 lets only single samples with maximum amplitudes into the transformed data, a threshold of 0.5 lets most of the transformed data samples pass, and a threshold of 0 lets all of the transformed data samples pass.

- 12 042140

In one or more embodiments, different smoothing thresholds for different types of sources may also be used as another way to stabilize the inversion when stronger and weaker sources are mixed. For example, in one or more embodiments, a standard program may be used

- (iter / niter) ² thresholds for the stronger source, but for the weaker source the threshold stays high until the residual data mismatch falls to the signal levels of the weaker source. Only then, in one or more embodiments, the implementation begins to lower the threshold when smoothing a weak source.

Typically, in one or more embodiments, data weighting may also be performed before or after transformation, for example, to equalize amplitudes to account for offset, or to eliminate the dominance of one strong frequency band (as shown in FIG. 4). If the signal-to-noise ratio is sufficient, the weighting will be canceled after thresholding. However, if the signal is weak or absent, or the noise is high, in some embodiments, weighting could be used predominantly to remove noise. In this case, in one or more embodiments, the post-thresholding weighting is not canceled to avoid reintroducing previously removed noise. In one or more embodiments, the implementation may also use the knowledge of the features of the data acquisition. For example, in one or more embodiments, the missing data may be interpolated first using, for example, projected-to-convex sets (POCS) interpolation while leaving the non-missing seismic tracks unchanged, after which a smoothness constraint is imposed on a second pass.

In the case of, for example, marine nodal acquisition and processing of data from a combined low frequency source and a pneumatic source, the blast lines from pneumatic sources can be closely spaced (for example, at a distance of 50 m from each other). The streamer of marine bottom seismic survey nodes is usually placed along the explosive lines of pneumatic sources, but may be located somewhat ahead of the pneumatic sources. Thus, there is usually an interval near the central survey point when all (or almost all) nodes are installed at the same time, and the areal area with nodes waits for the installation of air sources to complete before retrieval begins. Around this point in the seismic survey, just before the nodes start pulling back instead of putting them in, a low frequency source arrives and crosses all lines of this combined source. This is one way to organize a survey to efficiently collect ultra-long offset data (which the low frequency source was designed to collect) at as many sites as possible. Because these frequencies are very low, and low frequency data collection is only for velocity model building and not imaging, the low frequency source lines can be much sparser than the pneumatic source lines, resulting in a much faster data acquisition rate. The excitation of low-frequency sources can be carried out at intervals of two or three weeks, while the excitations of pneumatic sources can be at intervals of several months. Thus, overlapping of excitations of pneumatic sources with signals of low-frequency sources can happen only in very rare cases.

As a result, in some embodiments, it is possible to consider airgun gather tracks contaminated by overlapping low frequency source signals as lost. In some embodiments, at the start of the inversion, one or more embodiments may treat contaminated bursts of pneumatic sources as lost tracks and interpolate them into data. If there are not too many of them, then this approach can be considered acceptable. Later in the inversion process, as data residuals decrease, in some embodiments, other smoothing techniques can be switched to, such as separating contaminated tracks from signals from nearby overlapping sources instead of discarding them.

Typical vibration sources produce a single fundamental frequency signal at any time and provide a wide bandwidth by sweeping the frequency over time. In contrast, pneumatic source signals are pulsed and broadband. Thus, at any given time, it is likely that only a narrow frequency range of the pneumatic source signal will be affected by interference from the vibrator signals. However, one or more embodiments can take advantage of this difference by marking the affected frequency domain samples as lost, instead of discarding the air source gather traces after the Fourier transform. Since the smoothing criterion is often applied in the frequency domain, this is a fairly straightforward solution. Vibrator signals usually include rest phases, i.e. the vibrator does not work continuously. These intervals can be long enough so that, for example, every third or fourth impulse burst can be completely free from interference interference from the vibration source. At low frequencies, the signals from these clean bursts can be sampled at a high enough, over-sampling rate.

- 13 042140 sampling and can be used to interpolate into polluted intermediate bursts, which will be considered as lost in the early stages of the inversion.

Similarly, the duration of signals from low frequency sources is often many times longer than the duration of signals from pneumatic sources (typically 60-300 s compared to 1215 s), and therefore only a part of the individual frequency sweeps of low frequency sources can be contaminated by a strong overlap of signals from pneumatic sources, and in some cases only for units located close enough to a working pneumatic source. In one or more embodiments, all low frequency source gather samples contaminated by pneumatic sources with an amplitude above a certain level may be flagged and may be considered as lost data that should be interpolated. The label considered lost can be individually sequentially marked samples in the time domain; it does not have to be a seismogram sequence. Consider lost marks can be used in both the time and frequency domains, or any other domain or domains. The label considered lost can be functionally related to arguments in many areas. In some embodiments, the label may take on values between 0 and 1, allowing lanes or samples to be reduced in size, but not entirely eliminated, and/or replaced with an interpolated value, but with a weighted average between the original value and the interpolated value.

In one or more embodiments, the known firing times of the pneumatic sources can be used to determine contaminated samples, but a simpler way to mark contaminated data samples from a low frequency source is to simply observe the energy level in a window around 8 Hz - at the frequency where the pneumatic sources energy is maximum. , and mark all samples where the energy in the window around the sample exceeds a certain level in that frequency range. Frequencies around 8 Hz are typical of existing pneumatic sources, but in general, in some embodiments, any frequency intervals most suitable for distinguishing an interfering source from a source whose signals are separated from signals from nearby sources are used. It also makes it possible to intercept air source excitations that were accidentally not recorded, or air source excitations belonging to a nearby survey not recorded in the source log.

In one or more embodiments, data collection and processing may be modified in order to ensure that the described method works. In some embodiments, in any time window where the low frequency source is operating, only the odd-numbered pneumatic source drive lines can be obtained in the method. This would ensure that only good quality uncontaminated data of the even source excitation lines would be obtained. Since signals from a low frequency source only affect the lowest frequencies, in one or more embodiments only lost data at the lowest frequencies will need to be interpolated. Doubling the cross profile sampling interval (eg 50x2=100m) at these frequencies still provides more than sufficient sampling. At higher frequencies, any contamination from the low frequency source will be small enough to be ignored, and in certain embodiments, all pneumatic sources excitation lines can be used. In some embodiments, it is also possible to avoid receiving source lines that are multiples of, for example, 3, 4, and other multiples during low frequency source operation, the possible conditions of which depend on overlapping frequency intervals of different types of sources and transverse distances between source lines. Experience has shown that conventional survey schemes tend to oversample at low frequencies, which can be used to advantage in certain embodiments.

The choice of data dimension and method of ordering data when applying S may depend on the specific conditions of data collection and processing, and in general may change with the type of source. Pneumatic sources are typically used with frequent sampling on a regular grid (eg, 50 m x 50 m), so in certain embodiments, there is a wide choice of data ordering paths that provide data smoothing. Therefore, for example, for a nodal survey, in certain embodiments, the smoothness constraint SA may be applied to 3-D common-receiver gathers. However, sampling of sea bottom knots is not as frequent (eg 400-800 m by 400 m). Anti-aliasing in a 3-D common shot gather at conventional air source frequencies can therefore not be used in certain embodiments to excite airguns at seabed nodes, at least without some processing to avoid aliasing distortions due to insufficient sample density.

In contrast, in a low frequency source (e.g. an experimental source such as Wolfspar®), data collection can only be done with a large step between the shot lines, so the low frequency source gathers collected at sea bottom nodes

- 14 042140 may be sampled fairly frequently only in 2-D but not in 3-D common-receiver gathers. For each node gather, the frequently sampled directions are position along the source line and time. The position of the sources across the line is not sampled frequently enough to be useful. However, because the source operates at such low frequencies, the same array of receiver nodes that provided coarse sampling at higher frequencies (above 2 Hz) can provide frequent sampling for low frequency source targets, which means that unlike pneumatic sources at conventional frequencies, in one or more embodiments, SW can be applied to a 3-D common shot gather. It should be noted that there is no requirement here that different sources or sources of different types necessarily have a coherence/smoothing constraint imposed in the same region; each must use the region and coherence constraint appropriate to that source.

This difference in what types of seismic records are considered to be sampled at sufficient frequency (by the Nyquist criterion) in different frequency ranges is another reason that it makes sense to divide the solution of the problem by frequency range. This could be done by simply splitting the original data into overlapping trapezoidal frequency bands, working on each individually, and then, if necessary, joining the results together. Or, splitting and subsequent joining can be done internally in the algorithm, for example, in the process of applying a smoothness constraint. Or, it can be done using frequency dependent weighting or limiting conditions within a single common inversion. The acquisition and processing density with respect to the frequency range will determine the possible grid choices to use for m, and what type of smoothness constraint should be applied. For example, this could either be an inversion for the full dataset interpolating some of the lost explosions, or only computing unmixed versions of the actual recorded explosions. There may be good reasons for different choices, not just for different types of sources, but also for different frequency intervals of the same source.

Pneumatic sources are generally reliable, and if there is any difficulty in locating the source at sub-optimal blast points, the blast can simply be repeated. In one or more embodiments, this will most likely not be possible using a new type of source, including a low frequency source of the Wolfspar® type. As a result, there will sometimes be places where there will be breaks or deviations in the lines of excitation. This can create problems during the smoothing step in the algorithm, as discontinuities will make even correct data look rough, which the smoothing step tries to fix.

This is similar to the problem that has arisen in wave migration, when offshore 2-D streamer seismic has typically been used. The towed streamer marine survey lines may not have been straight but curved due to changing transverse currents during data collection. Migration transformation algorithms assume that the data is strictly 2-D, and is not acquired along a tortuous data collection line. To solve this problem, a method called Azimuth Moveout (AMO is a migration operator that rotates the azimuth and modifies the data offset) was used. A simple head-on method would be used to flatten the reflectors in the data using known exact source and receiver positions. For example, to flatten the data, a reflected wave kinematic correction (NMO) calculated from a very simple approximate velocity model could be used. Next, the reverse operation is applied, but at the required positions of the source and receiver. If the data acquisition errors are not too large, the forward and backward operations almost cancel each other out. Even a very approximate correction (due to the very simple approximate velocity model) will work quite well in practice, allowing you to further work with the data as if it were 2-D for migration and mapping problems. In some embodiments, the same principles of separating signals from signals from nearby sources (overlapping signals), for example, by replacing S with (NMO actual grid NMO ordered grid) S (NMO ordered grid NMO actual grid)

In practice, in certain embodiments, the forward and reverse NMO operations will most likely not be applied separately, but will be combined into a single operation (as indicated by the combination of operations in parentheses). Some embodiments may also simply perform smoothing directly on the NMO data, such as by flattening the breaks, then smoothing and then returning the breaks to their original positions. Mathematically, this replaces S with (NMO actual mesh) S (MMO actual mesh)

Alternatively, in some embodiments, m may represent an ordered collection of data, and the formulas are adapted to include the mapping of data from the ordered grid to the actual source/sink positions in the mix operator G.

The method, in one or more embodiments, obviously depends on signal extraction using coherence, the ability to separate desired wave arrivals from arrivals originating from interfering sources due to their coherence (or lack of it) in some way.

- 15 042140 last. The data is obtained in a manner that makes such separation possible. In the literature, hardware random time dilution of the source excitation has been described as a generally preferred method for accomplishing this task, however, other methods can be successfully used in practice. The blurring of the excitation time may not be strictly random, but algorithmic, with a certain law of repetition of the displacement of the moments of excitation, programmed in the algorithm of the source controllers. The literature describes several good ways to develop an optimal dithering regimen. In other cases, such as during the Vibriseis™ terrestrial surveys, blurring can be achieved by simply allowing the operator of each source to choose when to trigger it. This method relies on the unpredictability of the actions of the human operator, providing sufficient randomness of choice to prevent unwanted coherence between different sources. In a marine survey, excitation according to position rather than time can provide sufficient randomness. The changing currents and motion of waves in the ocean will cause the surveying vessel to slightly lead or lag, effectively introducing environmentally generated random smearing at source firing times.

Depending on the acquisition geometry, source capabilities, and the choice of smoothing method, other methods of breaking the coherence between interfering seismic wave arrivals can be effective in practice. In one embodiment, all source excitations occur at periodically repeating times, but the choice of periods is made carefully so that wave arrivals generated by one source can be easily separated from wave arrivals generated by other sources. In FIG. 15 shows an example of a common receiver gather thinned using the first source shot times. The in-phase axis 1701 represents the arrival of a seismic wave generated by the first source. In this region, the in-phase axis 1701 has a sufficient sampling rate and has a horizontal roll-off in a certain region. The second source has a slightly longer period, such that the in-phase axes 1702 of the arrivals generated by the second source (but decimated according to the times of the first source) all have significantly steeper slopes. Although the event axes 1701 and 1702 both differ by a sufficient sampling rate within the coherence analysis window where they overlap (box 1710 shows one such possible coherence window), arrivals from two overlapping sources can be easily separated by their differing slopes. By adding a slope filtering step to signal coherence extraction, unwanted interference events can be identified by their unphysically steep slopes and discarded, and it will be sufficient for the algorithm to then arrive at the correct result, despite the high periodicity of all sources.

For other interfering source period selections, seismic arrivals (e.g., 1703) can become so mixed that within the window (box 1720) of the analysis, they essentially become a type of noise, as a result of which they are rejected in signal extraction by the coherence criterion even without additional slope filtering steps.

If interfering arrivals have markedly different signatures (eg, event axes 1701 and 1703 in FIG. 15), separation can be made by analyzing different sources in different areas. In FIG. 15, the in-phase axis 1701 is impulsive, and the axes 1703 are spread over time. After appropriate filtering for signature generation, in another analysis window, wave arrivals from the source that generated the event axis 1701 may be blurred in time, the arrivals from the source that generated the event axis 1703 may become impulsive, providing attenuation of one or the other arrival using the appropriate condition underpressure restrictions.

In practice, it may be preferable to combine several methods for ensuring the incoherence of interfering seismic wave arrivals. For example, in an embodiment, surveying vessels may fire sources according to position rather than time, thereby introducing randomness into source firing times. However, the specified speeds of the source carrier ships in this case should also be chosen different so that even if there is no blurring of time points due to natural causes (for example, if the sea is completely calm at some point in the survey area), the moments of excitation of the sources relative to each other were chosen so that within the analysis window only one source could be coherent at the same time.

For sources of different types, which may have very different nominal repetition intervals, a more general design challenge is to organize the survey so that there are no two different sources that could be excited at the same time, i.e. for any combination of different sources, no small integer multiples of the nominal repetition intervals of the sources can match. Vessel speeds will typically vary within a narrow operating range such that the unblurred source firing time intervals when firing at position also vary within a range, and this uncertainty usually needs to be taken into account.

- 16 042140

Suppose, for example, that pneumatic sources are excited at the position, and at some point in time a low-frequency vibration source is turned on. For nominal separation between firing points, the air source repetition interval can vary from 11 to 12 s, depending on the ship's speed. According to the survey plan, the low frequency source (excited at some point) should have a nominal unblurred repetition interval somewhere in the range of 100-120 s, with the exact value being a selectable design parameter. In one or more embodiments, it can further be ensured that the low frequency source cannot be coherent with pneumatic sources by selecting a repetition interval of 109 s. This increment is not an integer multiple of the air source repetition interval, which is any value between 11 and 12 s. Mathematically, a 9 interval from 11 to 12 s gives a value in the range of 99 to 108 s, and a 10 interval from 11 to 12 s gives a range of 110 to 120 s. Given an acceptable design range of 100 to 120 s, in one or more embodiments, a value of 108 to 110 s may be selected to avoid overlap with pneumatic sources. If nearby seismic equipment is fired in time rather than in position, the first option 109 s is also not a multiple of any integer second repetition interval less than 109 s. Thus, by choosing a repetition interval of 109 seconds, in one or more embodiments, the possibility of coherence between the low frequency source and any pneumatic sources can be minimized. Those skilled in the art should understand how to apply these principles to other seismic survey projects.

All the various options described in the previous paragraphs (both for the external signal separation algorithm and for the signal extraction algorithms using coherence reused within it) can be combined in different combinations. Obviously, different methods are possible, depending on the number and types of sources, their data collection geometry, the selection of the grid to represent the results, the frequency range used, the signature of the sources and the distance between them, and/or the applications required. In particular, one or more embodiments may use sources of any number of different types, and any number of sources of each type. The sources may be combined or may be used individually in any possible combination. In one or more embodiments, it is possible to combine all sources at one output, or have one output for each source, or one output for each type of source, or any combination of these options, with some sources combined at the output without others, etc. d.

Thus, in some embodiments, the application of the inversion process described here provides seismic data sufficiently free from interference at the receiver that can be used for display, pre-stack analysis, such as amplitude vs. offset (AVO) analysis. Offset) and building a velocity model, for example, inversion taking into account the shape of the pulse (full wave inversion, FWI - from the English. Full Waveform Inversion).

Explosive seismic data acquisition, where the recorded information from one explosion overlaps in time with other explosions, will greatly reduce the time (and cost) required to perform a seismic survey using explosions. This approach may also reduce the distance between explosion points (eg, in offshore surveys) so that better quality seismic images can be obtained, increasing the chances of finding economically significant quantities of oil and/or gas. The ability to use many types of sources while collecting data allows you to collect data over a wider frequency band with the same efficiency as data from a conventional pneumatic source.

To illustrate the operating conditions of the data acquisition and processing system, FIG. 8 depicts a general technical environment configuration in which the present disclosure would typically be used. A seismic survey project covering an economically significant area is developed by the prospector in step 110. Field acquisition parameters (e.g. shot spacing, shot line spacing, fold, source type, etc.) are typically selected according to this step. , although the ideal field design parameters are usually slightly (or significantly) modified to match actual survey conditions.

Seismic data (i.e., seismic traces) is collected in a field 120 above a subsurface zone of potential economic interest and then typically sent to a processing center 150 where it will be processed using various algorithms to make it more suitable for exploration. deposits. In some cases, some initial data processing may be performed while still in the field, which becomes more accessible and convenient as the computing power available to the field party increases. As shown above, within field 120 there may be two or more overlapping simultaneous sources of the same type, and/or there may be two or more overlapping simultaneously.

- 17 042140 operating sources of other types. As shown above, records from different sources can be combined into one set.

At the processing center, the seismic traces are typically subjected to the usual preparation processes 130 to prepare them for use by the prospector. The processed seismograms are then transferred for use by the systems and methods described, and may be stored, for example, in data storage devices such as a hard disk, magnetic tape, solid state drive, magneto-optical disk, DVD disk, or other mass storage device.

The methods disclosed herein may be implemented in the form of a computer program 140 that has been loaded into a programmable computer 150 where it is accessed by a seismic data interpreter or processor. It should be noted that a computer 150 suitable for use in the present disclosure would typically comprise, in addition to basic computing devices, servers and workstations, supercomputers, and more generally, a computer or network of computers that provides parallel and massively parallel computations in which the computational load is distributed between two or more processors. As also shown in FIG. 8, in some embodiments, a numbered area of interest in model 160 may be designated by the user and used to enter the processing computer program. In the case of a 3-D seismic section, the zone of interest model 160 would typically include specific parameters relating to the transverse extent and thickness (which may be variable and may be measured in units of time, depth, frequency, etc.) of the subsurface region of interest. The specific means by which these zones are formed, selected, digitized, stored and subsequently read during program execution is not important to the present systems and methods, and those skilled in the art will understand how this can be done in various ways.

The program 140 implementing the present process can be transferred to a computer executing it by means of, for example, a floppy disk, a magnetic disk, a magnetic tape, a magneto-optical disk, an optical disk, a read-only compact disk (CD-ROM), a rewritable compact disc, disc (DVD), memory card (RAM), flash memory card (RAM-flash), read-only programmable memory (PROM), or downloaded over a network (such as a wired or wireless network, etc.). In a typical seismic data processing environment, the methods of the present disclosure would be made part of a larger package of software modules designed to perform many of the processing steps listed in FIG. 9. After processing by the presented methods, the obtained seismograms are further distributed into seismogram layouts, summarized and displayed either on a high-resolution color computer monitor 170 or on paper in the form of a printed image of a seismic section or map 180. The seismic data interpreter then uses the presented images to identify subsurface signs indicating the generation, migration or accumulation of hydrocarbons.

As previously noted, the present disclosure will preferably be implemented as part of a conventional seismic data processing procedure of the type described in FIG. 9. Those skilled in the art will appreciate that the processing steps shown in FIG. 9 are just a general idea of the nature of the processes that can be applied to such data, and the choice and sequence of these processing steps and the algorithms used may depend largely on the individual seismic data processor, signal source (dynamite, vibrator, pneumatic source, low-frequency generator, etc.), data acquisition locations (land, sea, etc.), data processing company, etc.

As a first step, and as summarized in FIG. 9, a 2-D and 3-D seismic survey is taken over a defined volume of the earth's subsurface zone (step 210). The data collected during field work consists of non-stacked seismograms containing digital information representing the volume of rock located under the survey site. Methods for obtaining such data and processing them into a form usable by seismic processors and interpreters are well known to those skilled in the art. It should be noted that for the purposes of this disclosure, a seismic survey will be a mixed source survey in which reflections from a later source activation may interfere with (or coincide in time with) reflections from an earlier source activation. Once bursts or sources have been separated in accordance with the present disclosure, the non-stacked gathers resulting from this operation can be used as in any other gather. Thus, the systems and methods presented provide a large number of source gathers that are triggered to produce overlapping signals, where seismic data sets obtained are equivalent or substantially equivalent to seismic data sets acquired using fully separated source activations and seismic data acquisition (e.g., non-overlapping sources). and non-stacked seismograms).

The objective of a seismic survey is to obtain a set of spatially related seismograms over a target subsurface zone with commercial prospects. Data suitable for analysis by the methods disclosed herein may, in particular, consist of unstacked 2-D seismic

- 18 042140 a non-stacked 2-D seismic line extracted from a 3-D seismic survey or, preferably, a non-stacked 3-D portion of a 3-D seismic survey. The systems and methods disclosed herein are most effective when applied to a set of stacked gathers that are spatially correlated with some subsurface geological feature. Also, for the purpose of illustrating the disclosure only, the following is a discussion that involves the gathers contained in the 3-D survey (stacked or unstacked as the discussion requires), although any collected group of spatially related gathers may be used.

After the seismic data has been collected (step 210), it is typically transferred to a processing center at the survey site, where it undergoes some initial or preparatory processing steps. As shown in FIG. 9, the general first step 215 is to edit the input seismic data in preparation for further processing (eg, demultiplexing, gain recovery, wavelet shaping, bad seismic removal, etc.). Following this, the survey geometry can be determined (step 220) and the shot/receiver number and surface position stored as a seismic trace header. Once the geometry is determined, it is common to perform a velocity analysis and apply a kinematic reflection correction (NM0) to correct each seismogram track in time to account for time-of-arrival delays caused by settling. In some embodiments, the present embodiment may be used in connection with step 215, i. in conjunction with or instead of the Vibroseis® waveform/correlation steps, although of course this could be used elsewhere within this overall processing scheme.

After the initial pre-stack processing is completed, the seismic signal in the non-stacked seismic tracks is typically normalized before the stacked data volumes are generated (step 230). In FIG. 9, step 230 contains a typical signal processing/Scaling/Mapping process, but one skilled in the art will appreciate that many alternative processes can be used instead of those shown in the drawing. In any case, the ultimate goal from the prospector's point of view is to create a stacked seismic volume or, in the case of 2-D data, a stacked seismic line for use in hydrocarbon exploration in the subsurface zone of the earth.

As further shown in FIG. 9, any digital sample in the stacked seismic volume is uniquely identified by a trio of coordinates (X, Y, TIME), where the X and Y coordinates represent some position on the surface of the earth, and the time coordinate measures the recorded arrival time within the seismogram track (step 240). For definiteness, it is assumed that the X direction corresponds to the direction along the lines, and the Y coordinate corresponds to the direction across the line, as is conventional in the prior art. While time is the preferred and most common unit for the vertical axis, those skilled in the art will agree that other units are possible, including, for example, depth or frequency. In addition, those skilled in the art are well aware that it is possible to convert seismic track coordinates from one unit (eg, time) to another (eg, depth) using standard mathematical transformations.

The explorer can perform an initial interpretation 250 of the resulting stacked volume in which he locates and identifies major reflectors and faults if they occur in the data set. This may be followed by improving 260 the data quality of the stacked or non-stacked seismic data and/or generating (step 270) attributes based on them. In many cases, the explorer will revert to his original interpretation in view of the additional information gained from the data quality improvement and attribute generation steps (step 280). As a final step, the prospector will typically use the information gleaned from the seismic data, along with other data (magnetic surveys, gravimetric surveys, LANDSAT survey data, regional geosciences, well logs, cores, etc.), to locate subsurface structural and stratigraphic features indicating the formation, accumulation or migration of hydrocarbons (i.e. a set of 290 studies).

Referring to the systems and methods disclosed herein, it should be noted that the method may include separating two or more seismic sources fired during the same acquisition, where the similarity between closely spaced shots is used to limit the inversion process and create multiple separated seismic datasets.

As shown in FIG. 10 and 11, in order to collect mixed source survey data, a number of receivers 310 may first be distributed in a 2D array over the area to be surveyed. In some embodiments, only a few receivers, or up to several thousand receivers 310, may be used for surveying. The receivers 310 may be connected by cables to a central recorder, wireless communication may also be used for this, or each receiver may contain some amount of internal memory, in which to record received seismic signals. For professionals in the industry

- 19 042140 You are very familiar with these types of receivers.

In some embodiments, the signals of the receivers 310 will be recorded continuously for an extended period of time. In some embodiments, the recording from the receivers may last several hours, half a day, a whole day, many days, and so on. The only requirement is that the results of excitation of at least two sources must be received for recording. This is different from conventional seismic surveys where receiver signals are recorded for only a few seconds after the source is activated.

During the duration of the recording of the receiver signals, a number of seismic sources 320 will be activated within the survey area 300. In some embodiments, two or more sources will be used, and the sources may have different source signatures and/or frequency ranges, including any of those that are described here. In the case of a marine survey, it is likely that different types of sources will be used, but it is clear that this choice is at the discretion of the survey designer.

The two or more seismic sources 320 may include any number of discrete or continuous sources. In the case of a marine survey, seismic sources 320 may include one or more of pneumatic sources, spark sources, vibrators, or the like. In some embodiments, any number, volume, or type of seismic sources may be included in the array. In a particular case, the array may include one or more high volume air sources, one or more medium volume air sources, and/or one or more low volume air sources. Seismic sources can be placed in various layouts. For example, the seismic sources of a group can be placed close to each other and their activation should be part of a single seismic explosion. In the case of a marine survey, this may be a single seismic source array or multiple source arrays towed by one or more vessels. A sub-array (which may consist of one or more sources) may be considered as a sub-set of sources, with the source(s) assigned to each sub-array assuming simultaneous activation.

In a land survey, two or more sources 320 may include vibrators, chargers, or buzzing seismic sources. For example, two or more sources 320 may include surface vibrators (e.g., 10 or so), Mini-Sosie™ vibrators for seismic surveys, drop weight sources, etc. One type of seismic survey system uses a vibrator or a group of vibrators in as a source of acoustic energy (hereinafter referred to as a vibroseismic source). A vibroseismic source can generate (vibrate) waves of acoustic energy at predetermined vibrator installation points (VP - vibrator points). VPs may be marked with pickets placed by surveyors. Alternatively, a satellite positioning system (GPS) may be used to determine the location of the VP.

Other sources may include low frequency sources such as those used in marine surveys, including Wolfspar® sources. Each of the low frequency sources can operate at a single frequency (monochromatic low frequency sources), or cycle through two or more discrete frequencies (stepped low frequency sources), or scan within a narrow range of low frequencies to expand the range of frequencies produced by broadband sources (narrowband low frequency sources). Sources can generate waves with a constant amplitude, or the amplitude of the waves can change (smoothly increase and decrease). In this case, low frequency means frequencies less than about 6-8 Hz. In some embodiments, the implementation may be less than about 4 Hz, some may use frequencies up to about 2 Hz, or about 1.5 Hz, or about 0.5 Hz.

In some embodiments, two or more seismic sources 320 may have different signatures. In operations using Vibroseis, the vibrator(s) of this vibroseis usually generate vibrations according to the law that the frequency changes over a given time period. This law of vibration generation is called vibration frequency swing. A typical Vibroseis sweep law would be a linear frequency sweep from about 5 to 100 Hz, with a duration of about 5-30 seconds. Pneumatic sources can generate high frequency acoustic waves. The low frequency sources may generate low frequency signals for a longer period of time and, in some embodiments, the low frequency sources may operate continuously.

In some embodiments, the source activations may be separated in time by intervals of random duration. Also, the sources can be fired at short enough intervals so that there is some overlap or mixing between shots. When seismic records are corrected for time zero (fire time) of each source, the reflections associated with that source are usually coherent, but the energy from interfering sources is usually incoherent (i.e., reflections do not line up) because

- 20 042140 holding between explosions is random. In other words, in the case of a survey where each source 320 is a Vibroseis block, it can be assumed that the activations of the sources may in some cases be separated by an interval of a few seconds. It should be noted that the purpose of FIG. 10 is not to assume that all sources 320 will be activated at the same time, but to show that they are located at various points in the survey area 300. Some shoots may use ten or more different sources. An example of a survey method that produces data that would be suitable for use in the present disclosure is WO 2008/025986 (PCT/GB2007/003280) Howe's Seismic Survey Method, incorporated herein by reference in its entirety. Howe considers using checkerboard activation when there is some overlap in reflections coming from the subsurface.

While seismic sources 320 may contain discrete or semi-discrete signatures, one or more seismic sources may operate continuously with explosions occurring according to a certain law, or as a continuous source of radiation. In some embodiments, the method may include continuously occurring explosions, which can also be called continuous ordered explosions or explosions that occur continuously according to a certain law. In particular, instead of conducting explosions according to one scheme, and then switching to explosions according to another scheme, there is no time delay and a continuous sequence of explosions is performed, which, in the case of gratings of pneumatic sources, is a continuous sequence of pops. In this case, in some embodiments, there may be gaps of several seconds between the two patterns of explosions, or there may be no such gaps. As for the low frequency sources, they can operate continuously for a predetermined period of time to continuously generate low frequency energy.

As shown above, different types of sources may have different characteristics. Different sources having different signatures and/or frequency ranges allow the use of differences in source signatures, appropriate coherence criteria, and data acquisition and processing processes (e.g., 2-D, 3-D, etc.) to provide better separation in the collected data. Particular source combinations may include the use of air sources with vibrators, air sources with electric spark sources, air sources with a low frequency source (eg Wolfspar source, etc.), or a combination of three or more types of sources.

In FIG. Figure 11 summarizes what data from a mixed source might look like. Each receiver 310 outputs a signal to form a seismic track (eg, track 405) that can potentially last tens of minutes or several hours (or days, etc.). In this figure, track 405 is shown schematically as containing recorded signals from four different source excitations. While the signals in the drawing are shown to be similar to each other, the recorded signals may have signatures that match the source. When multiple sources having different source signatures are used, the signals may have corresponding different signatures. Each receiver 310 is associated with a location on the surface of the earth. When the signals that have been recorded from each receiver 310 are properly ordered and displayed, in some embodiments, a 3-D volume associated with x and y location coordinates will be generated for each receiver 310 to include locations based on latitude and longitude, and etc.

Preferably, during a mixed source seismic survey, the time that each source 320 is activated will be recorded, and it will record which sources may be located inside or outside the receiving field. In FIG. 11, T ₁ and T ₂ represent the known times (measured from an arbitrary zero time) when the two sources were activated, and the parameter N generally indicates the amount of time (and/or number of samples) after source activation during which reflections from subsurface zone from this source can be taken. In this particular example, as will be discussed in more detail below, the two source activations may originate from the same source (eg, two variable frequency vibroseismic generators), as a result of which they do not overlap in time. However, in other embodiments, the source activations may be from different sources. In this mode of operation, interference is likely to occur from other seismic sources that were active within the time interval shown, or that had subsurface reflections arriving during the same time interval. Based on the above, the considered processes could be applied in one way if two or more source activations from the same source were to overlap in time.

Generating multiple separate seismic datasets can generate seismic data suitable for various further processing methods, as described herein. In some embodiments, the resulting split seismic datasets may include different seismic trace data. For example, the first seismic dataset (eg, m1) may contain standard split seismic trace data suitable for conventional

- 21 042140 processing and display. Alternatively, or as a second set (eg, m ₁ , m2, ..., m _n , etc.) of split seismic data, the resulting split seismic data set may be used for other forms of processing. It should be appreciated that the inversion process disclosed herein may provide interpolation of some elements of separate seismic datasets based on the iterative use of the inversion process. This can provide some of the resulting split seismic datasets based on the overlap between the two source shots, which can create more complete split datasets.

In some embodiments, one or more separated seismic datasets may be used for waveform inversion (FWI) analysis. FWI is a method of processing seismic data in the time or frequency domain, providing a more general approach for imaging subsurface structures: instead of using only reflected or scattered waves coming from geological discontinuities in the Earth, FWI also considers transmitted/refracted waves propagating downward, then turning in a horizontal direction and, finally, heading upward in the form of ascending seismic waves (perhaps at a considerable distance from their source). The presence of subsurface structures advances, slows down and/or distorts these transmitted/refracted incident waves, and FWI determines their location and properties from the characteristic traces left by these structures in the data. It should be noted that without low-frequency, high-offset data that contains transmitted/refracted waves, full-wavelength inversion often fails and may not distinguish subsurface structures (i.e., may produce useless results). Unfortunately, conventional seismic sources do not provide the commonly required low frequency waves and, in particular, low frequency data that could be used for full wave inversion. Thus, the separation and generation of seismic seismic datasets can be used to provide the required seismic processing. While FWI is a natural application of this technique, other applications of partitioned data are possible that do not involve display or speed detection.

This description of the different types of sources, the general inversion model, and the types of seismic output required, allows us to proceed further to the processes used to implement the described models. In FIG. 12 is a processing flow diagram suitable for use when the mixed survey sources are one or more pneumatic sources and one or more seismic vibrators (eg, for a Vibroseis survey). It should be well understood by one skilled in the art how to extend the processing scheme of FIG. 12 for the case with three or more types of sources. The process can be seen as taking place in three nested layers, the outermost of which contains cells 1400, 1403, 1499, 1491, 1492, 1493, 1401, 1402, 1480, and 1485.

At step 1400, initialization is performed. Three data buffers are created, one holds the residual data (residual), another holds the split overlay data from the air source (signal 1), and the third holds the split overlay data from the vibration source (signal 2). The residual data is initially just copies of the recorded seismic data, and the two split signal buffers initially contain null data.

At step 1403, an inversion check is performed. This check may be necessary either because the residual data (the difference between the actual recorded data and the predicted data) is small enough, or because the maximum number of iterations has been reached. If this condition is met, split mixed overlapping data (consisting of two seismic track estimates) is output at step 1499. Otherwise, new residual data is entered in boxes 1401 and 1402, enclosing one iteration of the split source estimate for pneumatic and vibration sources (Vibroseis), respectively. Each of these boxes individually is analogous to the method described in US 8295124 B2 issued to Abma, in particular the steps described in FIG. 8 of this patent in steps 815 to 875. The subject matter of US 8295124 B2 is fully incorporated into the present disclosure. The outputs of each of 1401 and 1402 represent an updated estimate of the recorded data that would have been recorded if only the source of the respective type (pneumatic source for 1401 and Vibroseis for 1402) was active and the data for that source type was collected during conventional unblended data collection. Next, a logical negation is applied to these outputs, in cell 1480 - to the data from the pneumatic source, in cell 1485 - to the data from Vibroseis.

The logically negated predictions for data from one source are then summed in cell 1492 with the recorded data from cell 1491 to produce updated residual data. The newly updated residual data is stored in its data buffer (cell 1493) and passed on to cell 1403. The process is repeated until it is completed. It should be noted that if the prediction was correct, then the predictions for unblended data from the same source will be added to the data actually recorded, and cell 1492, i. the newly updated residual data will be identically zero. In practice, at the end of the process, the residual data will be

- 22 042140 will contain non-coherent noise, and the algorithm for separating data from data associated with nearby sources will still have the beneficial side effect of denoising the data. At least in theory, mixed data acquisition and processing should perform better than traditional unmixed data acquisition, since shorter acquisition times mean less noise for the same amount of signal.

The intermediate layer contains the inner cells of rectangles 1401 and 1402. First, the residual data is combed (at 1410 comb filtering using airgun activation times, and at 1415 comb filtering using Vibroseis activation times). Then, the current scores of the split overlapping data from the buffers (cell 1420 for airguns, and cell 1425 for Vibroseis) are each summed with the corresponding combed residual data in cell 1430 for airguns and cell 1435 for Vibroseis. The vibroseis data processing sequence includes an additional source signature regularization step (cell 1416), or the use of signature deconvolution, signature correlation, matching filter, etc. A work cycle, such as the cycle described by Abma in US 8295124 B2, would go directly to the data smoothing step to extract a coherent signal (cell 1440 for the air source, and cell 1445 for Vibroseis). When separating sources of different types, it may be necessary to equalize their amplitudes. Otherwise, the source of one type will override the source of another type, which will negatively affect the convergence of the method. In addition, it is useless to try to get a coherent signal that is not actually there, for example, trying to find a coherent signal above 4 Hz for a low-frequency source that radiates little or no energy in this range. Such an attempt will only accentuate the noise, and as a result of the poor convergence of the process, the expected result will not be obtained. Therefore, the modified method includes an additional time/frequency weighting step that Abma does not have, respectively, cells 1432 for the air source and 1437 for Vibroseis. This step is necessary to equalize the relative energy of sources of different types and attenuate energy at frequencies not generated by the corresponding source. The method then extracts the coherent signal (steps 1440 for the air source and 1445 for the Vibroseis source) as before. Optionally, additional time/frequency weighting may be performed in steps 1442 and 1447 to partially or completely remove the weights assigned in steps 1432 and 1437, respectively. If weighting was used for noise suppression, the weighting factors are retained. The result is an update of the separated overlapped source estimate, added to the previous estimate in steps 1450 and 1455, to provide updated estimates of the separated overlapped air sources and Vibroseis, respectively. The updated estimates of the split signals from the overlapping sources are stored in the respective data buffers, respectively, at steps 1460 and 1465. Next, at steps 1470 and 1475, the separated gather traces are further subjected to an inverse comb filtering operation, respectively, for pneumatic sources and Vibroseis. The Vibroseis workflow contains an additional step (box 1466) to remove the source signature entered in step 1416.

The innermost layer of the process is in cells 1440 and 1445, and consists of extracting the coherent signal. In FIG. 13, the block diagram 1500 illustrates the standard approach as described in the Abma patent for the case of sources of the same type, and this block diagram 1500 may represent the contents of cells 1440 or 1445. The data is organized into appropriate records samples, where the desired signal must be coherent in cell 1510. It is then subjected to a Fourier transform in cell 1520 so that the desired signal becomes sparse in this region. The smallest values in the transformed space are set to zero in step 1530, keeping the signal unchanged while attenuating the noise. The data is then returned to the original area in step 1540.

Also, as in the Abma patent, the data will be mostly split into overlapping N-dimensional sub-cubes, with a corresponding gradual reduction in the overlap zones, and doing this operation in each sub-cube, after which the sub-cubes are added together to restore the original version of the data, cleaned from noise.

An alternative method, which may be particularly useful in separating different types of data from signals generated by nearby sources, is illustrated by the block diagram in box 1600 of FIG. 14. This method may be used in place of the method represented by flowchart 1500, or it may be applied first followed by method 1500. In this method, polluted gather traces are marked as dead or lost, and standard methods of filling in the lost data by interpolation are used to replace the polluted tracks. The method shown in box 1600 is called POCS (Projected Convex Set) interpolation. Contaminated or lost seismic tracks have previously been identified and this information is stored in location 1650.

As in flowchart 1500, data is first grouped into suitable seismic record layouts (cell 1610). The data is then Fourier transformed (cell 1620) for their sparsity

- 23 042140 nii, while the smallest values are set to zero or reduced (cell 1630), after which the data is converted back to the original area (cell 1640). The difference is that at step 1660 the seismic tracks that do not require interpolation are replaced with their original values. At step 1670, the result is compared with the previous result, and if no change has occurred (or the maximum value of iterations has been reached), the process ends. Otherwise, the data containing the newly interpolated lost tracks is routed back to cell 1620 for the next iteration.

In FIG. 12-14 present a particular set of preferred embodiments. It should be well understood by one skilled in the art how to generalize the method to other cases, and how to introduce into the main operating cycle many of the alternative options described earlier in the disclosure.

It should also be noted that the Abma patent demonstrates two approaches: a constructive approach and a destructive approach. The same possibilities apply here. The above disclosure has described a generalization of a constructive approach to datasets from mixed sources of two or more types. One skilled in the art should be well aware of how to similarly generalize the methods described here for use in a destructive approach.

It should be noted that when operations are performed on seismic tracks of a particular type (eg, receiver samples), it is generally not necessary to collect those seismic tracks together in storage (eg, by sorting) in order to apply multitrack dataset processing to them.

Thus, in the above disclosure and claims further, when it is said that a seismic record assembly (shot point gather, receive point gather, etc.) has been collected or is available for further processing, these words should be taken in the broadest sense for coverage. cases when the seismic tracks that form the layout of seismic records are processed on site or operationally, in the course of work. In this case, there may not be any need for sorting or otherwise ordering the data.

In addition, in some embodiments, the systems or methods presented will be adapted for use in a vertical seismic survey (VSP) or similar downhole survey. By way of explanation, those skilled in the art will appreciate that data collection in a VSP survey can be expensive due to rig downtime. Conducting an accelerated VSP with overlapping sources of explosions can significantly reduce the cost of such surveys. However, the words mixed seismic survey used here should be understood in a broad sense, including both land and marine 2-D and 3-D surveys, as well as VSP surveys, cross-well seismic surveys, etc.

In addition, in the foregoing discussion, terminology related to working with conventional seismic data has been used. Those of skill in the art will appreciate that the disclosure herein may be applied to other fields of activity and be used to determine the location of other minerals in the subsurface, in addition to hydrocarbons. In particular, the described approach can potentially be used to process and analyze multicomponent seismic data, shear wave data, converted wave data, cross-well seismic data, VSP data, wave acoustic logs, controlled source or other electromagnetic data ( controlled source electromagnetic intelligence - CSFM, t-CSFM, etc.), or digital simulation of any of the processes mentioned. In addition, the techniques disclosed herein can then be applied to mathematically transformed versions of these same seismic tracks, including, for example: filtered data tracks, post-migration data tracks, frequency-domain Fourier transform data tracks, discrete orthonormal transforms, instantaneous phase data tracks, instantaneous frequency data tracks, quadrature tracks, analytic tracks, etc. In short, the processes disclosed herein can potentially be applied to a wide range of geophysical time sequences, however preferably applied to a set of spatially related time sequences.

According to the various systems and methods described, certain features may include, but are not limited to:

In a first aspect, in a method of seismic surveying over an area of a subsurface zone containing structural or stratigraphic features indicative of the presence, migration, or accumulation of hydrocarbons, at least a portion of a mixed source survey is accessed, the mixed source seismic survey comprising excitations at least two interfering seismic sources, and the seismic source excitations are produced by one or more types of seismic sources. Seismic sources of different types may have different signatures or frequency responses; separating excitations from at least two interfering sources using seismic data separation using inversion; generating one or more source gathers based on this split, and using one or more source gathers for hydrocarbon exploration in the given

- 24 042140 areas of the subsurface zone.

The second feature may include a method according to the first feature, in which, when separating the excitations of at least two interfering seismic sources: selecting a first coherence constraint for a first type seismic source from seismic sources of at least two types; selecting a second coherence constraint for the second type of seismic source from at least two types of seismic sources; and applying the first coherence constraint and the second coherence constraint in seismic inversion separation to a portion of the mixed source seismic survey in which acquisition of one or more source gathers is based on applying the first coherence constraint and the second coherence constraint .

The third feature may include a method according to the second feature, wherein the first coherence constraint is different from the second coherence constraint.

The fourth feature may include a method according to any of the first to third features, wherein, when separating the excitations of at least two interfering seismic sources: determining the activation times for each excitation of the interfering seismic sources from the excitations of at least two interfering sources; and using the activation times corresponding to the excitation of each seismic source to move the excitations of at least two interfering seismic sources at the initial time, wherein the generation of one or more source gathers is based on the movement of the excitations of at least two interfering seismic sources at the initial time.

The fifth feature may include a method according to any one of the first to fourth features, wherein separating shots from at least two interfering seismic sources using inversion seismic data separation solves an equation having the form:

d = Γι mi + Γί mi + ... + Γ _η m _n for matrices m1, ..., mi, ..., mn to obtain one or more source gathers, where n can be greater than or equal to 1, m ₁ - seismogram of the first source from seismograms of one or more sources, mi - seismogram of another source from seismograms of one or more sources; d is a matrix representation of the recorded seismic data, T 1 is a matrix defining the times of excitation of a source of the first type from seismic sources of one or more types, and T _i is a matrix defining the times of excitation of a source of another type from seismic sources of one or more types.

The sixth feature may include a method according to any of the first to fifth features, wherein each of the collection of source gathers has seismic data specific only to its frequency range (representing a different frequency range).

The seventh feature may include a method according to any one of the first to six features, wherein additionally: an inversion analysis is performed taking into account the waveform on at least one of the source gathers.

The eighth feature may include a method according to any one of the first to the seventh features, wherein the excitations of each of the at least two interfering seismic sources are separated in time by a time interval of random duration.

The ninth feature may include a method according to any of the first to eighth features, wherein the seismic sources are of at least two different types, and include a combination of at least two of: vibration sources, pneumatic sources, electric spark sources, and low frequency sources.

The tenth feature may include a method in accordance with any of the first to ninth features, in which at least one seismic source of at least two interfering seismic sources, when they are excited, changed the sign or phase of the excitation of this seismic source.

In the eleventh feature, in the method of seismic exploration: initiate a series of excitations of seismic sources over an area of the subsurface zone containing structural or stratigraphic features indicating the presence, migration or accumulation of hydrocarbons, and a number of excitations of seismic sources is caused in seismic sources of one or more types. Seismic sources of different types may have different signatures or frequency responses; and performing a mixed-source seismic survey comprising data obtained from reflections generated by a plurality of seismic source excitations, comprising excitations of at least two interfering seismic sources, wherein the excitations of at least two interfering seismic sources are separated using seismic data inversion, and one or more source gathers per

- 25 042140 based on the separation of excitations of at least two interfering seismic sources.

The twelfth feature may include the method according to the eleventh feature, in which the excitations of at least two interfering seismic sources are separated using seismic data inversion by solving an equation having the form:

d = P m-ι + P m, + ... + G _p m _n for matrices m1, ..., mi, ... mn to obtain one or more source gathers, where n can be greater than or equal to 1, m1 - first source gather from one or more source gathers, mi - other source gather from one or more source gathers, d - matrix representation of the recorded seismic data,

Г1 is a matrix defining the moments of excitation of the source of the first type from seismic sources of one or more types, and

Г _i is a matrix that determines the moments of excitation of a source of another type from seismic sources of one or more types.

The thirteenth feature may include a method according to the eleventh or twelfth feature, in which each of the collection of source gathers has seismic data specific to its frequency range only.

The fourteenth feature may include a method according to any of the eleventh to thirteenth features, wherein each of the excitations of the at least two interfering seismic sources are separated in time by a time span of random duration.

In a fifteenth aspect, when performing a seismic survey over an area of a subsurface zone containing structural or stratigraphic features indicative of the presence, migration, or accumulation of hydrocarbons, at least a portion of a mixed source seismic survey is accessed, the mixed source seismic survey comprising excitations at least two interfering seismic sources, and seismic source excitations are produced in seismic sources of one or more types. Seismic sources of different types may have different signatures or frequency responses, at least one type of seismic source being a low frequency source; separating excitations from at least two interfering seismic sources using seismic data inversion; generating one or more source gathers based on the separation, wherein at least one of the source gathers contains low frequency data or some lost frequencies; and using one or more source gathers for hydrocarbon exploration in the subsurface zone.

The sixteenth feature may include the method according to the fifteenth feature, wherein, when separating the excitations of at least two interfering seismic sources: selecting a first coherence constraint for a first type seismic source from seismic sources of at least two types; selecting a second coherence constraint for the second type of seismic source from at least two types of seismic sources; and applying the first coherence constraint and the second coherence constraint in seismic inversion separation to a portion of the mixed source seismic survey in which acquisition of one or more source gathers is based on applying the first coherence constraint and the second coherence constraint .

The seventeenth feature may include a method according to the fifteenth or sixteenth feature, in which, by separating the excitations of at least two interfering seismic sources using inversion of the seismic data, an equation is solved having the form:

d = P gs + P gs + ... + G _p m _n for matrices m1, ..., mi, ..., m _n to obtain one or more source gathers, where n can be greater than or equal to 1, m1 is the first source gather from one or more source gathers, mi is the other source gather from one or more source gathers, d is a matrix representation of the recorded seismic data,

Г ₁ is a matrix that determines the times of excitation of the source of the first type from seismic sources of one or more types, and

Г _i is a matrix that determines the moments of excitation of a source of another type from seismic sources of one or more types.

In an eighteenth aspect, in a method of seismic surveying over a region of a subsurface zone containing structural or stratigraphic features indicative of the presence, migration, or accumulation of hydrocarbons: at least a portion of a mixed source seismic survey is accessed, wherein the mixed source seismic survey comprises excitations along at least two interfering seismic sources, and seismic excitations

- 26 042140 seismic sources are produced in seismic sources of one or more types. Seismic sources of different types may have different signatures or frequency responses and use one or more generated source gathers for hydrocarbon exploration in the subsurface zone.

The nineteenth feature may include a method according to the eighteenth feature, in which, using a set of generated source gathers, an equation is solved having the form:

d = Π mt + Г, m, + ... + Г _p m _n for matrices m1, ..., mi, ..., mn to obtain one or more source gathers, where n can be greater than or equal to 1, m1 - first source gather from one or more source gathers, mi - other source gather from one or more source gathers, d - matrix representation of the recorded seismic data,

Г 1 is a matrix defining the moments of excitation of the source of the first type from seismic sources of one or more types, and

Г _i is a matrix that determines the moments of excitation of a source of another type from seismic sources of one or more types.

The twentieth feature may include the method according to the eighteenth or nineteenth feature, wherein at least one seismic source of the at least two interfering seismic sources has changed the sign or phase of the excitation of the seismic source during excitations.

The particular embodiments disclosed above are for illustrative purposes only, and the present disclosure may be modified and applied in various but equivalent ways that would be apparent to those skilled in the art having read the description. In addition, the device or construction described is not intended to be subject to any limitations other than those described in the appended claims. Therefore, it is obvious that the private embodiments disclosed above can be changed or modified, and such changes are considered to be within the scope and essence of the present invention. Alternatives that result from combining, combining and/or eliminating features of the variant(s) are also covered by the scope of the disclosure. While structures and methods are described in broader terms having, containing, accommodating, or including various components or steps, structures and methods may also consist essentially of or consist of various components and steps. The use of the term optional in relation to any element of a claim means that the element is required, or vice versa, the element is not required, both alternatives falling within the scope of the claims.

The numbers and ranges disclosed above are subject to change to some extent. In the case where a numeric range with a lower limit and an upper limit is used, any number and any included range within that range is more specifically disclosed. In particular, each range of values (in the form of from about a to about b, or equivalently from about a to b) in the present disclosure should be understood as defining any number or range covered by the wider range of values. In addition, the terms in the claims have their common, ordinary meaning, unless the patent applicant clearly and unambiguously indicates otherwise. Moreover, the indefinite articles used in the claims are intended to refer to one or more than one element that it introduces. If there is any conflict in the use of a word or term in the present description and in one or more patents or other documents, then the definitions corresponding to this description should be used.

Numerous modifications, equivalents, and alternatives will become apparent to those skilled in the art upon full reading of this disclosure. The following claims are intended to be construed as covering such modifications, equivalents and alternatives, where applicable. Accordingly, the scope is not limited to the above description, but only to the following claims, and this scope includes the equivalents of the subject matter according to the claims.

Claims (20)

CLAIM 1. The method of seismic exploration over an area of the subsurface zone containing structural or stratigraphic features indicating the presence, migration or accumulation of hydrocarbons, in which: accessing at least a portion of a mixed-source seismic signal survey comprising excitations of at least two interfering seismic sources, wherein the seismic source excitations are produced by at least two seismic sources having different signatures and different frequency responses; separating excitations of the at least two interfering sources using inversion seismic data separation based on different signatures and different frequency responses of the at least two seismic sources; generating a plurality of individual source gathers based on this separation, and using the plurality of individual source gathers to explore for hydrocarbons in said area of the subsurface zone. 2. The method according to claim 1, in which when separating the excitations of at least two interfering seismic sources: selecting a first coherence constraint for a first seismic source from at least two seismic sources; selecting a second coherence constraint for the second seismic source from at least two seismic sources; and applying the first coherence constraint and the second coherence constraint in seismic inversion separation to the mixed source seismic portion, the acquisition of the source seismogram collection based on the application of the first coherence constraint and the second coherence constraint. 3. The method of claim 2, wherein the first coherence constraint is different from the second coherence constraint. 4. The method according to claim 1, in which when separating the excitations of at least two interfering seismic sources: determining activation times for excitation of each interfering seismic source from at least two interfering sources; and using the activation times corresponding to the excitation of each seismic source to move the excitations of at least two interfering seismic sources at the initial time, and the generation of the set of source gathers is based on the movement of the excitations of at least two interfering seismic sources at the initial time. 5. The method according to claim 1, in which when separating the excitations of at least two interfering seismic sources using seismic data separation using inversion, an equation is solved having the form: d = Г ₁ t ₁ + Гг t ₂ for matrices m1 and m2 to obtain a set of source gathers, where m1 is the first source gather from the set of source gathers, m2 is the second source gather from the set of source gathers, d is a matrix representation of the recorded seismic data, Г 1 is a matrix defining the moments of excitation of the first source from at least two seismic sources, and Г ₂ is a matrix that defines the moments of excitation of the second source from at least two seismic sources. 6. The method of claim 1, wherein each of the plurality of source gathers has seismic data representing a different frequency range. 7. The method of claim 1 further comprising performing a full wave inversion analysis on at least one of the plurality of source gathers. 8. The method of claim 1, wherein the excitations of each of the at least two interfering seismic sources are separated in time by a time interval of random duration. 9. The method of claim 1, wherein the at least two seismic sources are distinct and form a combination of at least two of: vibration sources, pneumatic sources, electric spark sources, and low frequency sources. 10. The method according to claim 1, in which at least one seismic source of at least two interfering seismic sources during their excitations changed the sign or phase of excitation of this seismic source. - 28 042140 11. The method of seismic exploration, in which: initiating multiple seismic source excitations over an area of the subsurface zone containing structural or stratigraphic features indicative of the presence, migration or accumulation of hydrocarbons, said seismic source excitations being produced by at least two seismic sources having different signatures and different frequency responses; and performing a mixed-source seismic survey comprising data obtained from reflections generated by multiple seismic source shots and using shots from at least two interfering seismic sources, the at least two interfering seismic source shots being separated using seismic data inversion based on different signatures and different frequency responses of at least two seismic sources, and generating a set of source gathers based on separation of excitations of at least two interfering seismic sources. 12. The method of claim 11, wherein the excitations of at least two interfering seismic sources are separated using seismic data inversion by solving an equation having the form: d = Γι gn + G, m ₂ for matrices m1 and m ₂ to obtain a set of source gathers, where m1 is the first source gather from the set of source gathers, m2 is the second source gather from the set of source gathers, d is a matrix representation of the recorded seismic data, Г 1 is a matrix defining the moments of excitation of the first source from at least two seismic sources, and Г2 is a matrix that determines the moments of excitation of the second source from at least two seismic sources. 13. The method of claim 11, wherein each of the plurality of source gathers has seismic data representing different frequency bands. 14. The method of claim 11, wherein the excitations of each of the at least two interfering seismic sources are separated in time by a time interval of random duration. 15. The method of seismic exploration over an area of the subsurface zone containing structural or stratigraphic features indicating the presence, migration or accumulation of hydrocarbons, in which: access to at least a portion of a mixed-source seismic signal survey comprising excitations of at least two interfering seismic sources produced in at least a first seismic source and a second seismic source having different signatures and different frequency responses, wherein one of at least the measure of the first seismic source or the second seismic source is a low frequency source; separating excitations of at least two interfering seismic sources using inversion of seismic data based on different signatures and different frequency responses of at least two seismic sources from at least the first seismic source and the second seismic source; generating a plurality of individual source gathers based on the separation, wherein at least one of the plurality of individual source gathers contains low frequency data or some lost frequencies; and using a plurality of individual source gathers to explore for hydrocarbons in said area of the subsurface zone. 16. The method of claim 15, wherein when separating excitations from at least two interfering seismic sources: selecting a first coherence constraint for a first seismic source from at least two seismic sources; selecting a second coherence constraint for the second seismic source from at least two seismic sources; and applying the first coherence constraint and the second coherence constraint in seismic inversion separation to the mixed source seismic portion, the acquisition of the source seismogram collection based on the application of the first coherence constraint and the second coherence constraint. 17. The method of claim 15, wherein separating the excitations of at least two interfering seismic sources using seismic data inversion solves an equation having the form: - 29 042140 d = Γι m-ι + G, m ₂ for matrices m1 and m2 to obtain a set of seismograms of sources, where m1 is the seismogram of the first source from the set of seismograms of sources, m2 is the seismogram of the second source from the set of seismograms of sources; d is a matrix representation of the recorded seismic data, Г 1 is a matrix defining the moments of excitation of the first source from at least two seismic sources, and Г ₂ is a matrix that defines the moments of excitation of the second source from at least two seismic sources. 18. The method of seismic exploration over an area of the subsurface zone containing structural or stratigraphic features indicating the presence, migration or accumulation of hydrocarbons, in which: accessing at least a portion of a mixed seismic source survey comprising excitations of at least two interfering seismic sources produced by at least two seismic sources having different signatures and different frequency responses; separating excitations of at least two interfering seismic sources using seismic data inversion based on different signatures and different frequency responses of at least two seismic sources; generate a set of separate seismograms of sources on the basis of separation; and using a collection of separate seismograms of sources for exploration of hydrocarbons in the area of the subsurface zone. 19. The method according to claim 18, in which, using a set of generated seismograms of sources, an equation is solved having the form: d = P pc + G, t ₂ for matrices m ₁ and m ₂ to obtain a set of seismograms of sources, where m1 is the seismogram of the first source from the set of seismograms of sources, m ₂ is the seismogram of the second source from the set of seismograms of sources; d is a matrix representation of the recorded seismic data, Г ₁ is a matrix defining the moments of excitation of the first source from at least two seismic sources, and Г ₂ is a matrix that defines the moments of excitation of the second source from at least two seismic sources. 20. The method according to claim 18, wherein at least one seismic source of at least two interfering seismic sources, when they are fired, changed the sign or phase of the excitation of this seismic source.