Machine learning-based multi-domain processing method and device for seismic data

Abstract

The present invention provides a multi-domain processing method and device for seismic data based on machine learning. The multi-domain processing method for seismic data based on machine learning includes: obtaining seismic data of a target work area; determining the range of the feature domain in multiple domains of the seismic data according to a pre-generated machine learning model and the seismic data. The multi-domain processing method and device for seismic data based on machine learning provided by the present invention can expand the dimensional range of the machine learning model during seismic data processing, more comprehensively judge the characteristics of seismic data, and thus improve the accuracy of using machine learning to process seismic data.

Description

Technical Field

[0001] The present invention relates to the field of oil exploration, particularly seismic data processing technology, and specifically relates to a multi-domain processing method and device for seismic data based on machine learning. Background Art

[0002] Seismic data is excited by a seismic source on the surface and returns to the surface after propagating through the underground medium, and is divided into multiple domains according to different excitation and reception methods. In different domains, seismic signals are combined in different permutations, and with different permutation orders, the characteristics shown by seismic signals in different domains are also different. Therefore, during the seismic data processing, it is often necessary to perform processing in different domains according to the characteristics of the seismic signals to be processed. Machine learning is a research hotspot in the field of artificial intelligence and has been widely applied to data processing in various fields. Currently, machine learning mainly uses recursive neural networks, and some use multi-layer networks called deep learning networks, and those containing input and output layers are called feedforward deep learning networks. Through machine learning, automatic processing of seismic data can be achieved without manual intervention during the processing.

[0003] However, in the prior art, there are the following problems in using machine learning to process seismic data: The analysis and processing of seismic data in different domains are carried out independently, and no connection can be established between them, and they cannot complement each other. Summary of the Invention

[0004] Aiming at the problems in the prior art, the multi-domain processing method and device for seismic data based on machine learning provided by the present invention can expand the dimensional range of the machine learning model during seismic data processing, more comprehensively judge the characteristics of seismic data, and thus improve the accuracy of using machine learning to process seismic data.

[0005] To solve the above technical problems, the present invention provides the following technical solutions:

[0006] In a first aspect, the present invention provides a multi-domain processing method for seismic data based on machine learning, including:

[0007] Obtain seismic data of a target work area;

[0008] According to a pre-generated machine learning model and the seismic data, determine the characteristic domain range in multiple domains of the seismic data.

[0009] In one embodiment, the seismic data includes: common shot gather, common receiver gather, common midpoint gather, common offset gather, and common azimuth gather data;

[0010] The multiple domains include: common shot data domain, common receiver data domain, common midpoint data domain, common crossline data domain, common offset data domain, and common azimuth data domain.

[0011] In one embodiment, determining the range of the feature domain among multiple domains of the seismic data according to the pre-generated machine learning model and the seismic data includes:

[0012] Determining their respective multi-domain feature parameters and the domain range of the multi-domain feature parameters according to the seismic data and the machine learning model;

[0013] The multi-domain feature parameters include: dip angle, average multiple of the energy of a seismic trace and the energy of an adjacent seismic trace, and time difference between a seismic trace and an adjacent seismic trace.

[0014] In one embodiment, the machine learning model is a feedforward deep machine learning model; the steps of generating the machine learning model include:

[0015] Generating the machine learning model label according to the similarity of the dip angle, average multiple of the energy with an adjacent trace, and time difference with an adjacent trace of the seismic data;

[0016] Training the initial model of the machine learning model according to the seismic data with a signal-to-noise ratio greater than a preset value and the machine learning model label to generate the machine learning model.

[0017] In a second aspect, the present invention provides a multi-domain processing device for seismic data based on machine learning, and the device includes:

[0018] A seismic data acquisition unit for acquiring seismic data of a target work area;

[0019] A domain range determination unit for determining the range of the feature domain among multiple domains of the seismic data according to the pre-generated machine learning model and the seismic data.

[0020] In one embodiment, the seismic data includes: common shot gather, common receiver gather, common midpoint gather, common offset gather, and common azimuth gather data;

[0021] The multiple domains include: common shot data domain, common receiver data domain, common midpoint data domain, common cross-line data domain, common offset data domain, and common azimuth data domain.

[0022] In one embodiment, the domain range determination unit includes:

[0023] A feature parameter determination module for determining their respective multi-domain feature parameters and the domain range of the multi-domain feature parameters according to the seismic data and the machine learning model;

[0024] The multi-domain feature parameters include: dip angle, average multiple of the energy of a seismic trace and the energy of an adjacent seismic trace, and time difference between a seismic trace and an adjacent seismic trace.

[0025] In one embodiment, the machine learning-based multi-domain seismic data processing apparatus further includes a model generation unit for generating the machine learning model, and the model generation unit includes:

[0026] A model label generation module for generating the machine learning model label according to the dip angle of the seismic data, the average multiple of the energy of adjacent traces, and the similarity of the time difference with adjacent traces;

[0027] A model generation module for training the initial model of the machine learning model according to the seismic data with a signal-to-noise ratio greater than a preset value and the machine learning model label to generate the machine learning model;

[0028] The machine learning model is a feedforward deep machine learning model.

[0029] In a third aspect, the present invention provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the steps of the machine learning-based multi-domain seismic data processing method are implemented.

[0030] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the machine learning-based multi-domain seismic data processing method are implemented.

[0031] As can be seen from the above description, the embodiments of the present invention provide a machine learning-based multi-domain seismic data processing method and apparatus. First, seismic data of a target work area is obtained; then, according to the pre-generated machine learning model and the seismic data, the feature domain range in multiple domains of the seismic data is determined. The present invention can expand the dimensional range of the machine learning model during seismic data processing, more comprehensively judge the characteristics of seismic data, and thus improve the accuracy of using machine learning to process seismic data. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0033] Figure 1 Flow schematic of the machine learning-based multi-domain seismic data processing method in the embodiments of the present invention Figure 1 ;

[0034] Figure 2Schematic flowchart of step 200 in an embodiment of the present invention;

[0035] Figure 3 Schematic flowchart of the method for multi-domain processing of seismic data based on machine learning in an embodiment of the present invention Figure 2 ;

[0036] Figure 4 Schematic flowchart of step 300 in an embodiment of the present invention;

[0037] Figure 5 Schematic flowchart of the method for multi-domain processing of seismic data based on machine learning in the specific embodiment of the present invention;

[0038] Figure 6 Schematic diagram of shot domain seismic records in the specific embodiment of the present invention;

[0039] Figure 7 Schematic diagram of geophone point domain seismic records in the specific embodiment of the present invention;

[0040] Figure 8 Plan view of shot domain dip characteristic parameters obtained for the entire work area in the specific embodiment of the present invention;

[0041] Figure 9 Plan view of geophone point domain characteristic parameters in the specific embodiment of the present invention;

[0042] Figure 10 Plan view of shot domain dip similarity for the entire work area in the specific embodiment of the present invention;

[0043] Figure 11 Plan view of geophone point domain dip similarity for the entire work area in the specific embodiment of the present invention;

[0044] Figure 12 Plan view of dip characteristic parameters corresponding to the maximum similarity in the specific embodiment of the present invention;

[0045] Figure 13 Schematic diagram of the structure of the device for multi-domain processing of seismic data based on machine learning in an embodiment of the present invention Figure 1 ;

[0046] Figure 14 Schematic diagram of the structure of the domain range determination unit in an embodiment of the present invention;

[0047] Figure 15 Schematic diagram of the structure of the device for multi-domain processing of seismic data based on machine learning in an embodiment of the present invention Figure 2 ;

[0048] Figure 16 Schematic diagram of the structure of the model generation unit in an embodiment of the present invention;

[0049] Figure 17 This is a schematic structural diagram of an electronic device in an embodiment of the present invention. Detailed implementation manners

[0050] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0051] An embodiment of the present invention provides a detailed implementation manner of a multi-domain processing method for seismic data based on machine learning. Refer to Figure 1 This method specifically includes the following contents:

[0052] Step 100: Obtain seismic data of a target work area.

[0053] It can be understood that the seismic data in step 100 can be sorted according to the acquisition system to obtain the common shot gather, common receiver gather, common midpoint gather, common offset gather, and common azimuth gather data of the seismic data to be processed.

[0054] Step 200: Determine the characteristic domain range in multiple domains of the seismic data according to a pre-generated machine learning model and the seismic data.

[0055] Specifically, the seismic data adjacent to the same seismic record in different domains is different, and the seismic signal characteristics shown by the same seismic record in different domains are also different. Therefore, the signal of the seismic record can be effectively identified in different domains to obtain accurate parameters such as the dip angle and energy of the signal, so as to perform signal-noise separation.

[0056] From the above description, it can be seen that the embodiment of the present invention provides a multi-domain processing method for seismic data based on machine learning. First, seismic data of a target work area is obtained; then, according to a pre-generated machine learning model and the seismic data, the characteristic domain range in multiple domains of the seismic data is determined. The present invention can expand the dimensional range of the machine learning model during seismic data processing, more comprehensively judge the characteristics of seismic data, and thus improve the accuracy of using machine learning to process seismic data.

[0057] In one embodiment, the seismic data includes: common shot gather, common receiver gather, common midpoint gather, common offset gather, and common azimuth gather data;

[0058] The multiple domains include: common shot gather data domain, common receiver gather data domain, common midpoint data domain, common crossline data domain, common offset data domain, and common azimuth data domain.

[0059] For example, during seismic data acquisition, when the shot point excites and the receiver point receives, the received seismic data is shot gather data. By sorting it respectively according to the receiver point, midpoint, receiver line and shot line, and azimuth, the corresponding receiver gather domain, midpoint domain, crossline domain, and azimuth domain data can be obtained.

[0060] In one embodiment, referring to Figure 2 , step 200 includes:

[0061] Step 201: Determine their respective multi-domain characteristic parameters and the domain range of the multi-domain characteristic parameters according to the seismic data and the machine learning model;

[0062] When implementing step 201, the multi-domain seismic data can be first input into the seismic data machine learning model to obtain the multi-domain characteristic parameters of the seismic data; then, the obtained multi-domain characteristic parameters are input into the seismic data machine learning model to obtain the unique seismic data characteristic parameters, the domain range of the seismic data characteristic parameters, and the scores of the multi-domain characteristic parameters.

[0063] In addition, the multi-domain characteristic parameters in step 201 include: dip angle, average multiple of the energy of the seismic trace and the energy of the adjacent seismic trace, and time difference between the seismic trace and the adjacent seismic trace.

[0064] In one embodiment, the machine learning model is a feedforward deep machine learning model.

[0065] It can be understood that the goal of the feedforward network is to approximate a certain function. For example, for a classifier, y = f*(x) maps the input x to a class y. The feedforward network defines a mapping y = f(x; θ), and learns the value of the parameter θ to enable it to obtain the best function approximation. Inside the feedforward neural network, the parameters are propagated unidirectionally from the input layer to the output layer. Usually, the feedforward deep learning network includes a convolutional layer, a pooling layer, and a fully connected layer.

[0066] In one embodiment, referring to Figure 3 , the multi-domain processing method of seismic data based on machine learning further includes:

[0067] Step 300: Generate a machine learning model. Further, referring to Figure 4 , step 300 includes:

[0068] Step 301: Generate the machine learning model label according to the similarity of the dip angle, average multiple of the energy with the adjacent trace, and time difference with the adjacent trace of the seismic data.

[0069] Step 302: Train the initial model of the machine learning model based on the seismic data with a signal-to-noise ratio greater than a preset value and the machine learning model labels to generate the machine learning model.

[0070] Taking the shot gather seismic data as an example to illustrate Steps 301 to 302, select the shot gather seismic records with high signal-to-noise ratio as sample data, and manually calibrate the dip angle of the shot gather seismic records, the average multiple of the energy of adjacent traces, and the time difference with adjacent traces as labels (model training constraints). Use a feedforward deep learning network to train the samples and labels to obtain a configured common shot point domain machine learning model. Specifically, here a multi-task deep learning network is used to obtain three output parameters in sequence. The first task of the multi-task deep learning network, the adjacent trace time difference task, includes a convolutional layer, a pooling layer, and a fully connected layer. The second task, the dip angle of the seismic record, includes a convolutional layer, a pooling layer, and a fully connected layer, and its convolutional layer is connected to the output of the fully connected layer of the first task. The third task, the average multiple of the energy of adjacent traces, includes a convolutional layer, a pooling layer, and a fully connected layer, and its convolutional layer is connected to the output of the fully connected layer of the second task.

[0071] To further illustrate the present solution, the present invention provides a specific application example of the multi-domain processing method for seismic data based on machine learning. The specific application example specifically includes the following content. See Figure 5 。

[0072] S0: Configure a feedforward deep machine learning model.

[0073] Select seismic records with high signal-to-noise ratio as sample data, manually calibrate the dip angle of the seismic records, the average multiple of the energy of adjacent traces, and the time difference with adjacent traces as labels, and use a feedforward deep learning network to train the samples and labels to obtain a configured feedforward deep machine learning model.

[0074] Taking the shot gather seismic data as an example, select the shot gather seismic records with high signal-to-noise ratio as sample data, manually calibrate the dip angle of the shot gather seismic records, the average multiple of the energy of adjacent traces, and the time difference with adjacent traces as labels, and use a feedforward deep learning network to train the samples and labels to obtain a configured common shot point domain machine learning model. Similarly, configure a common geophone domain machine learning model, a common midpoint domain machine learning model, a common cross spread domain machine learning model, a common offset domain machine learning model, and a common azimuth domain machine learning model.

[0075] S1: Determine multi-domain feature parameters.

[0076] For example, during seismic data acquisition, seismic waves are generated at shot points and received by geophone points. The received seismic data is in shot domain. By sorting it according to geophone points, central points, geophone lines, shot lines, and azimuth angles respectively, corresponding geophone domain, central point domain, cross - array domain, and azimuth angle domain data can be obtained. For the same seismic trace, adjacent seismic traces are different in different domains, and the seismic signal characteristics shown in different domains are also different. Therefore, only by identifying the signals of seismic traces in different domains can accurate parameters such as the dip angle and energy of the signals be effectively obtained, so as to perform signal - to - noise separation.

[0077] Specifically, inputting the data of common shot gather, common geophone gather, common mid - point gather, common offset gather, and common azimuth gather into corresponding machine learning models respectively can obtain corresponding characteristic outputs. That is, obtain the dip angle of seismic signals in different domains, the average multiple of energy with adjacent traces, and the time difference with adjacent traces.

[0078] S2: Determine the domain range of multi - domain characteristic parameters.

[0079] Similarly, the machine learning model for processing multi - domain characteristic parameters of seismic data is also a feed - forward deep machine learning model. The input of the model is one or more of the characteristic parameters of seismic data in common shot domain, common geophone domain, common mid - point domain, common cross - array domain, common offset domain, and common azimuth domain. The output is the similarity between the multi - domain characteristic parameters and the accurate characteristic parameters of the signal, with a score range of 1 - 10 points, where 1 means dissimilar and 10 means exactly the same.

[0080] For example, select the multi - domain characteristic parameters of seismic records with high signal - to - noise ratio as sample data, manually calibrate the similarity of the dip angle, the average multiple of energy with adjacent traces, and the time difference with adjacent traces of this seismic record as labels, and use a feed - forward deep learning network to train the samples and labels to obtain the configured machine learning model for processing multi - domain characteristic parameters.

[0081] Similarly, here a multi - task deep learning network is still used to obtain three output parameters in sequence. The first task of the multi - task deep learning network, the adjacent - trace time - difference similarity task, includes a convolutional layer, a pooling layer, and a fully - connected layer. The second task, the dip - angle similarity task of the seismic record, includes a convolutional layer, a pooling layer, and a fully - connected layer. Its convolutional layer is connected to the output of the fully - connected layer of the first task. The third task, the average - multiple - of - energy - with - adjacent - traces similarity task, includes a convolutional layer, a pooling layer, and a fully - connected layer. Its convolutional layer is connected to the output of the fully - connected layer of the second task.

[0082] Output the similarity output by the model and the characteristic parameters corresponding to the highest similarity as the result.

[0083] For example, Figure 6 and Figure 7They are seismic records in the shot domain and the geophone domain respectively. There are differences in the linear signal dips between the seismic traces at the dashed line positions and the adjacent traces. Figure 8 and Figure 9 They are the shot domain dip feature parameters and the geophone domain feature parameters obtained after inputting the seismic records in the shot domain and the geophone domain of the entire work area into the machine learning model for multi-domain feature extraction of seismic data. Figures 10 to 12 They are the similarity and the dip feature parameters corresponding to the maximum similarity obtained after inputting the dip feature parameters in the shot domain and the geophone domain of the entire work area into the machine learning model for processing multi-domain feature parameters. Further signal-to-noise separation can be performed through the finally obtained dip features.

[0084] In this specific embodiment, first, a machine learning model for multi-domain feature extraction of seismic data is configured to obtain the multi-domain feature parameters of the seismic data; then, a machine learning model for processing multi-domain feature parameters of the seismic data is configured to obtain the scores of the unique seismic data features and the multi-domain feature parameters. Through this method, more accurate processing results can be obtained by considering information in multiple dimensions during the seismic data processing. The present invention can expand the dimensional range of the machine learning model during seismic data processing, more comprehensively judge the characteristics of seismic data, and thus improve the accuracy of using machine learning to process seismic data.

[0085] Based on the same inventive concept, the embodiments of the present application also provide a multi-domain processing device for seismic data based on machine learning, which can be used to implement the methods described in the above embodiments, as described in the following embodiments. Since the principle of the multi-domain processing device for seismic data based on machine learning to solve problems is similar to that of the multi-domain processing method for seismic data based on machine learning, the implementation of the multi-domain processing device for seismic data based on machine learning can refer to the implementation of the multi-domain processing method for seismic data based on machine learning, and the repeated parts will not be described again. As used below, the term "unit" or "module" can be a combination of software and / or hardware that can achieve a predetermined function. Although the systems described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware is also possible and contemplated.

[0086] An embodiment of the present invention provides a specific implementation manner of a multi-domain processing device for seismic data based on machine learning that can implement the multi-domain processing method for seismic data based on machine learning. Refer to Figure 13 , and the multi-domain processing device for seismic data based on machine learning specifically includes the following:

[0087] A seismic data acquisition unit 10, configured to acquire seismic data of a target work area;

[0088] A domain range determination unit 20, configured to determine the feature domain range in multiple domains of the seismic data according to a pre-generated machine learning model and the seismic data.

[0089] In one embodiment, the seismic data includes: common shot gather, common receiver gather, common midpoint gather, common offset gather, and common azimuth gather data;

[0090] The multiple domains include: common shot data domain, common receiver data domain, common midpoint data domain, common cross - line data domain, common offset data domain, and common azimuth data domain.

[0091] In one embodiment, referring to Figure 14 , the domain range determination unit 20 includes:

[0092] A feature parameter determination module 201, configured to determine their respective multi - domain feature parameters and the domain ranges of the multi - domain feature parameters according to the seismic data and the machine learning model;

[0093] The multi - domain feature parameters include: dip angle, average multiple of seismic trace energy and adjacent seismic trace energy, and time difference between seismic trace and adjacent seismic trace.

[0094] In one embodiment, a multi - domain processing device for seismic data based on machine learning, referring to Figure 15 , further includes a model generation unit 30, configured to generate the machine learning model, referring to Figure 16 , the model generation unit 30 includes:

[0095] A model label generation module 301, configured to generate the machine learning model label according to the similarity of the dip angle, average multiple of energy with adjacent traces, and time difference with adjacent traces of the seismic data;

[0096] A model generation module 302, configured to train an initial model of the machine learning model according to the seismic data with a signal - to - noise ratio greater than a preset value and the machine learning model label to generate the machine learning model;

[0097] The machine learning model is a feed - forward deep machine learning model.

[0098] As can be seen from the above description, the embodiment of the present invention provides a multi - domain processing device for seismic data based on machine learning. First, it acquires seismic data of a target work area; then, according to a pre - generated machine learning model and the seismic data, it determines the characteristic domain ranges in multiple domains of the seismic data. The present invention can expand the dimensional range of the machine learning model during seismic data processing, more comprehensively judge the characteristics of seismic data, and thus improve the accuracy of using machine learning to process seismic data.

[0099] The embodiment of the present application also provides a specific implementation manner of an electronic device capable of implementing all steps in the above - mentioned multi - domain processing method for seismic data based on machine learning, referring to Figure 17 , the electronic device specifically includes the following content:

[0100] A processor 1201, a memory 1202, a communications interface 1203, and a bus 1204;

[0101] Among them, the processor 1201, the memory 1202, and the communications interface 1203 communicate with each other through the bus 1204; the communications interface 1203 is used to implement information transmission between related devices such as server-side devices, detection devices, and client-side devices.

[0102] The processor 1201 is used to call a computer program in the memory 1202. When the processor executes the computer program, all steps in the above-mentioned method for multi-domain processing of seismic data based on machine learning in the embodiment are implemented. For example, when the processor executes the computer program, the following steps are implemented:

[0103] Step 100: Obtain seismic data of a target work area;

[0104] Step 200: Determine the range of the feature domain in multiple domains of the seismic data according to a pre-generated machine learning model and the seismic data.

[0105] An embodiment of the present application also provides a computer-readable storage medium capable of implementing all steps in the above-mentioned method for multi-domain processing of seismic data based on machine learning. A computer program is stored on the computer-readable storage medium. When the computer program is executed by a processor, all steps in the above-mentioned method for multi-domain processing of seismic data based on machine learning are implemented. For example, when the processor executes the computer program, the following steps are implemented:

[0106] Step 100: Obtain seismic data of a target work area;

[0107] Step 200: Determine the range of the feature domain in multiple domains of the seismic data according to a pre-generated machine learning model and the seismic data.

[0108] Each embodiment in this specification is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other. The key point of each embodiment is to illustrate the differences from other embodiments. In particular, for the hardware + program type embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiments.

[0109] The above description has been made of specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the acts or steps recited in the claims may be performed in a different order than in the embodiments and still achieve the desired result. Additionally, the processes depicted in the drawings do not necessarily require the particular order or sequential order shown to achieve the desired result. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0110] Although this application provides method operation steps as described in the embodiments or flowcharts, based on routine or non-creative labor, there may be more or fewer operation steps. The order of steps listed in the embodiments is only one way among many orders of step execution and does not represent the only order of execution. When the actual device or client product is executed, it may be executed in the order of the method shown in the embodiments or the drawings or in parallel (e.g., in an environment of parallel processors or multithreaded processing).

[0111] For convenience of description, when describing the above device, it is divided into various modules according to functions for separate description. Of course, when implementing the embodiments of this specification, the functions of each module can be implemented in the same or multiple software and / or hardware, or the modules implementing the same function can be realized by a combination of multiple sub-modules or sub-units, etc. The device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection to each other can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be in electrical, mechanical or other forms.

[0112] Those skilled in the art should understand that the embodiments of this specification can be provided as a method, a system or a computer program product. Therefore, the embodiments of this specification can take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware aspects. Moreover, the embodiments of this specification can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing computer-usable program code.

[0113] Embodiments of this specification can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. Embodiments of this specification can also be practiced in a distributed computing environment where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0114] The various embodiments in this specification are described in a progressive manner. For the same or similar parts among the various embodiments, reference can be made to each other. Each embodiment focuses on the differences from other embodiments. In particular, for system embodiments, since they are basically similar to method embodiments, the description is relatively simple, and for the relevant parts, reference can be made to the partial description of the method embodiments. In the description of this specification, the description with reference to terms such as "an embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the embodiments of this specification. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0115] The above is only the embodiments of the embodiments of this specification and is not used to limit the embodiments of this specification. For those skilled in the art, various changes and modifications can be made to the embodiments of this specification. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the embodiments of this specification shall be included within the scope of the claims of the embodiments of this specification.

Claims

1. A method for multi-domain processing of seismic data based on machine learning, characterized in that, it includes: Obtain seismic data of the target work area; According to the pre-generated machine learning model and the seismic data, determine the characteristic domain range in multiple domains of the seismic data; The seismic data includes: common shot gather, common receiver gather, common midpoint gather, common cross-line data domain, common offset gather, and common azimuth gather data; The multiple domains include: common shot data domain, common receiver data domain, common midpoint data domain, common cross-line data domain, common offset data domain, and common azimuth data domain; The step of determining the characteristic domain range in multiple domains of the seismic data according to the pre-generated machine learning model and the seismic data includes: Determine their respective multi-domain characteristic parameters and the domain range of the multi-domain characteristic parameters according to the seismic data and the machine learning model; The multi-domain characteristic parameters include: dip angle, average multiple of the energy of a seismic trace and the energy of adjacent seismic traces, and time difference between a seismic trace and adjacent seismic traces; Determining their respective multi-domain characteristic parameters according to the seismic data and the machine learning model includes: Input the common shot gather, common receiver gather, common midpoint gather, common cross-line data domain, common offset gather, and common azimuth gather data into the corresponding machine learning models respectively, and obtain the corresponding characteristic outputs to obtain the dip angle of seismic signals in different domains, the average multiple of the energy with adjacent traces, and the time difference with adjacent traces; Determining the domain range of the multi-domain characteristic parameters according to the seismic data and the machine learning model includes: Input one or more of the characteristic parameters of the seismic data in the common shot domain, common receiver domain, common midpoint domain, common cross-line domain, common offset domain, and common azimuth domain into a preset feedforward deep machine learning model to determine the domain range of the multi-domain characteristic parameters.

2. The method for multi-domain processing of seismic data according to claim 1, characterized in that, The machine learning model is a feedforward deep machine learning model; the steps for generating the machine learning model include: Generate the machine learning model label according to the similarity of the dip angle, average multiple of the energy with adjacent traces, and time difference with adjacent traces of the seismic data; Train the initial model of the machine learning model according to the seismic data with a signal-to-noise ratio greater than a preset value and the machine learning model label to generate the machine learning model.

3. A device for multi-domain processing of seismic data based on machine learning, characterized in that, it includes: A seismic data acquisition unit for acquiring seismic data of the target work area; A domain range determination unit for determining the characteristic domain range in multiple domains of the seismic data according to the pre-generated machine learning model and the seismic data; The seismic data includes: common shot gather, common receiver gather, common midpoint gather, common cross-line data domain, common offset gather, and common azimuth gather data; The multiple domains include: common shot data domain, common receiver data domain, common midpoint data domain, common cross-line data domain, common offset data domain, and common azimuth data domain; The domain range determination unit includes: A feature parameter determination module, configured to determine their respective multi-domain feature parameters and the domain ranges of the multi-domain feature parameters according to the seismic data and the machine learning model; The multi-domain feature parameters include: dip angle, average multiple of the energy of a seismic trace to the energy of an adjacent seismic trace, and time difference between a seismic trace and an adjacent seismic trace; Determining their respective multi-domain feature parameters according to the seismic data and the machine learning model includes: Inputting common shot gather, common receiver gather, common midpoint gather, common cross-line data domain, common offset gather, and common azimuth gather data into corresponding machine learning models respectively, and obtaining corresponding feature outputs, so as to obtain dip angles of seismic signals in different domains, average multiples of energy to adjacent traces, and time differences to adjacent traces; Determining the domain ranges of the multi-domain feature parameters according to the seismic data and the machine learning model includes: Inputting one or more of the feature parameters of the seismic data in the common shot domain, common receiver domain, common midpoint domain, common cross-line domain, common offset domain, and common azimuth domain into a preset feedforward deep machine learning model to determine the domain ranges of the multi-domain feature parameters.

4. The multi-domain processing device for seismic data according to claim 3, wherein, it further includes a model generation unit for generating the machine learning model, and the model generation unit includes: a model label generation module for generating the machine learning model label according to the similarity of the dip angle, average multiple of energy to adjacent traces, and time difference to adjacent traces of the seismic data; a model generation module for training an initial model of the machine learning model according to the seismic data with a signal-to-noise ratio greater than a preset value and the machine learning model label to generate the machine learning model; The machine learning model is a feedforward deep machine learning model.

5. An electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein, when the processor executes the program, it implements the steps of the machine learning-based multi-domain processing method for seismic data according to any one of claims 1 to 2.

6. A computer-readable storage medium, on which a computer program is stored, wherein, when the computer program is executed by a processor, it implements the steps of the machine learning-based multi-domain processing method for seismic data according to any one of claims 1 to 2.

Images