Crack identification method based on MVMD frequency division and competitive neural network multi-attribute fusion

Abstract

The present application relates to a kind of fracture identification method based on MVMD frequency division and competition neural network multi-attribute fusion, it includes: determining the effective frequency band range of seismic data, based on MVMD method, low, middle, high three frequency band extraction processing is carried out to seismic data;Combining well logging interpretation information, respectively calculate multiple attributes based on the frequency division seismic data, select optimal sensitive attribute from multiple attributes, and calculate the eigenvalue of sensitive attribute, obtain the orthogonal principal component component;Different principal component components are clustered based on competition learning neural network algorithm, and the attribute probability body of different frequency bands is obtained, the frequency fusion of three frequency band probability bodies is carried out, and the comprehensive attribute body is output, the fine identification of study area fracture is realized.The present application can improve the precision and accuracy of reservoir fracture identification.The present application can be applied in the field of geophysics, machine learning and the like.

Description

Technical Field

[0001] This invention relates to the field of machine learning technology in geophysics, and in particular to a crack identification method based on MVMD frequency division and competitive neural network multi-attribute fusion. Background Technology

[0002] With the deepening of geophysical exploration, fractured oil and gas reservoirs with huge development potential are emerging and have become one of the main areas of oil and gas exploration. Fractures can not only store oil and gas, but also transport and migrate oil and gas, which has a significant impact on the oil and gas production capacity of reservoirs. Accurate prediction of the spatial distribution of fractures is of great importance for studying fractured oil and gas reservoirs, improving exploration success rate, and improving development efficiency. Seismic attribute analysis is one of the commonly used techniques for fracture prediction. Attributes such as coherence, curvature, and ant-body are often used in fracture interpretation. However, for complex reservoirs with strong homogeneity and large variations in physical parameters, single attribute analysis has strong ambiguity and cannot accurately reflect the fracture development. In addition, previous studies have confirmed that frequency-division attributes can show the characteristics of geological bodies at different scales of the reservoir, and provide a more refined reservoir description.

[0003] However, traditional time-frequency analysis methods suffer from low frequency division accuracy and modal aliasing, failing to meet the demands of high-resolution seismic interpretation. Furthermore, most conventional attribute fusion techniques are based on simple pixel stacking, lacking clear physical meaning and thus ill-suited to complex geological environments. In conclusion, effectively improving the accuracy and precision of fracture identification remains a critical technical challenge in fractured reservoir exploration. Summary of the Invention

[0004] To address the aforementioned problems, the purpose of this invention is to provide a fracture identification method based on MVMD frequency division and competitive neural network multi-attribute fusion, which can improve the accuracy and precision of reservoir fracture identification.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a fracture identification method based on MVMD frequency division and competitive neural network multi-attribute fusion, comprising: determining the effective frequency band range of seismic data; performing frequency division extraction processing on the seismic data into low, medium, and high frequency bands based on the MVMD method; calculating multiple attributes based on the frequency-divided seismic data in conjunction with well logging interpretation information; selecting the most sensitive attributes from the multiple attributes and calculating the eigenvalues ​​of the sensitive attributes to obtain orthogonal principal component components; clustering different principal component components based on a competitive learning neural network algorithm to obtain attribute probability volumes for different frequency bands; frequency fusion of the three frequency band probability volumes to output a comprehensive attribute volume, thereby achieving precise fracture identification in the study area.

[0006] Furthermore, spectral analysis is used to decompose the seismic data volume within the effective frequency band into representative single-frequency data sub-volumes to describe geological information at different scales and levels.

[0007] Furthermore, the MVMD method includes:

[0008] Based on the common frequency components existing between all channels of the input data, a variational model is introduced into the signal decomposition process to determine the constrained optimization problem and solve the optimal solution of the variational model; the IMF components of each channel simultaneously iteratively update the center frequency and finite bandwidth to adaptively obtain k finite IMF components.

[0009] Furthermore, the constrained optimization problem is:

[0010]

[0011]

[0012] In the formula, For the corresponding channel c and mode vector u k The analytical representation of each element in (t). ω represents the time-dependent partial derivative. k Where C is the center frequency, and u is the total number of data channels. k，c Let x be the k-th mode vector of channel c, where t is time. c (t) represents the input data for channel c.

[0013] Furthermore, the selection of the most sensitive attribute from multiple attributes includes:

[0014] Based on the frequency division results, eight seismic attributes were selected for calculation: coherence, curvature, ant volume, instantaneous phase, root mean square amplitude, variance, first derivative, and sweet spot. Combined with known well logging interpretation results, four seismic sensitive attributes, namely coherence, variance, ant volume, and curvature, were selected.

[0015] Furthermore, the acquisition of the comprehensive attribute body includes:

[0016] After standardizing the attributes, principal component analysis is performed to obtain orthogonal principal component components.

[0017] The principal component of the sensitive attribute is used as an input sample set, the clustering result of the output classification is set, and the mapping relationship between the input and the output is established.

[0018] By using the Sigmoid activation function to optimize the mapping relationship, the clustering result variables are mapped to the range [0, 1] according to the mapping relationship. Then, the clustering result is converted into a geologically significant fracture development probability volume, and the clustering fusion results of the coherence, variance, ant volume and curvature attributes of the low, medium and high frequency band sub-frequency data volumes are obtained.

[0019] Furthermore, the optimized mapping relationship is as follows:

[0020]

[0021] In the formula, C' represents the clustering result, and W i Let A represent the weight matrix. j Let b represent the j-th principal component. i This represents the bias term, where n represents the number of clusters in the clustering result, and m is the number of sensitive attributes.

[0022] A crack identification system based on MVMD frequency division and competitive neural network multi-attribute fusion, comprising:

[0023] The first processing module determines the effective frequency band range of the seismic data and performs frequency extraction processing on the seismic data in three frequency bands (low, medium, and high) based on the MVMD method.

[0024] The second processing module, in conjunction with well logging interpretation information, calculates multiple attributes based on frequency-division seismic data, selects the most sensitive attributes from among the multiple attributes, calculates the eigenvalues ​​of the sensitive attributes, and obtains orthogonal principal component components.

[0025] The output module clusters different principal component components based on a competitive learning neural network algorithm to obtain attribute probability volumes of different frequency bands. It then fuses the three frequency band probability volumes to output a comprehensive attribute volume, enabling precise identification of cracks in the study area.

[0026] A computer-readable storage medium storing one or more programs, the one or more programs including instructions that, when executed by a computing device, cause the computing device to perform any of the methods described above.

[0027] A computing device includes: one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include instructions for performing any of the methods described above.

[0028] The present invention has the following advantages due to the adoption of the above technical solutions:

[0029] This invention improves the accuracy and precision of fracture identification in complex reservoirs by using multi-channel variational mode decomposition (MVMD) frequency division and competitive learning neural network attribute fusion clustering for fracture identification. Attached Figure Description

[0030] Figure 1 This is a framework diagram of a crack identification method based on MVMD frequency division and competitive neural network multi-attribute fusion in one embodiment of the present invention;

[0031] Figure 2This is a flowchart of MVMD in one embodiment of the present invention;

[0032] Figure 3 This is the frequency division result in one embodiment of the present invention; wherein, Figure (a) is the original profile, Figure (b) is the 20Hz low-frequency profile, Figure (c) is the 40Hz mid-frequency profile, and Figure (d) is the 60Hz high-frequency profile.

[0033] Figure 4 This is a preferred sensitive attribute in one embodiment of the present invention; wherein, Figure (a) is a 20Hz low-frequency sensitive attribute, Figure (b) is a 40Hz mid-frequency sensitive attribute, and Figure (c) is a 60Hz high-frequency sensitive attribute.

[0034] Figure 5 This is the principal component contribution distribution in one embodiment of the present invention;

[0035] Figure 6 Figure (a) shows the attribute clustering fusion result in one embodiment of the present invention. Figure (b) shows the probability slice of low-frequency cracks, Figure (c) shows the probability slice of medium-frequency cracks, Figure (d) shows the probability slice of high-frequency cracks, and Figure (d) shows the slice of comprehensive prediction result. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention are within the scope of protection of the present invention.

[0037] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0038] To improve the accuracy and precision of reservoir fracture identification, and addressing the shortcomings of conventional frequency division and attribute fusion methods, this invention proposes a fracture identification method based on multi-channel variational mode decomposition (MVMD) frequency division and competitive neural network multi-attribute fusion. This invention performs multi-attribute analysis based on frequency division data to improve the accuracy of fracture identification in complex reservoirs.

[0039] In one embodiment of the present invention, a crack identification method based on MVMD frequency division and competitive neural network multi-attribute fusion is provided. In this embodiment, as... Figure 1As shown, the method includes the following steps:

[0040] 1) Determine the effective frequency band range of the seismic data, and perform frequency extraction processing of the seismic data into low, medium and high frequency bands based on the MVMD method;

[0041] 2) Based on the frequency-division seismic data, various attributes such as coherence and curvature are calculated by combining well logging interpretation information. Sensitive attributes are selected from the multiple attributes, and the eigenvalues ​​of the sensitive attributes are calculated to obtain orthogonal principal component components.

[0042] In this embodiment, principal component analysis is used to calculate the eigenvalues ​​of the sensitive attributes.

[0043] 3) Based on the competitive learning neural network algorithm, clustering is performed between different principal component components to obtain attribute probability volumes of different frequency bands. The three frequency band probability volumes are fused to output a comprehensive attribute volume, thereby realizing the fine identification of cracks in the study area.

[0044] Step 1) above also includes: using spectral analysis to decompose the seismic data volume within the effective frequency band into representative single-frequency data sub-volumes to describe geological information at different scales and levels.

[0045] In this embodiment, determining the effective frequency band range can highlight the effective information needed in the seismic data. Spectral analysis is used to decompose the seismic data volume within the effective frequency band [f0, f1] into representative single-frequency data sub-volumes, thereby achieving separate descriptions of geological information at different scales and levels. The accuracy of the frequency-divided data mainly depends on the resolution of the time-frequency analysis algorithm. Addressing the algorithmic shortcomings of existing time-frequency analysis methods, this invention introduces the multichannel variational mode decomposition (MVMD) method.

[0046] In step 1) above, MVMD extends the traditional VMD algorithm from one dimension to multiple dimensions, improving the lateral stability of the frequency division data. The method of obtaining intrinsic mode function (IMF) components based on MVMD differs from traditional empirical mode decomposition (EMD). Specifically, the MVMD method is as follows:

[0047] Based on the common frequency components existing between all channels of the input data, a variational model is introduced into the signal decomposition process to determine the constrained optimization problem and solve the optimal solution of the variational model; the IMF components of each channel simultaneously iteratively update the center frequency and finite bandwidth to adaptively obtain k finite IMF components.

[0048] The constrained optimization problem is as follows:

[0049]

[0050]

[0051] In the formula, For the corresponding channel c and mode vector u k The analytical representation of each element in (t). ω represents the time-dependent partial derivative. k Where C is the center frequency, and u is the total number of data channels. k，c Let x be the k-th mode vector of channel c, where t is time. c (t) represents the input data for channel c.

[0052] Specifically, for input data X(t) containing C data channels, denoted as [x1(t), x2(t), ... x... c The specific steps of the MVMD algorithm are as follows: (t)]

[0053] 1.1) Assume k multivariate modulated oscillators u k (t), so that:

[0054]

[0055] In the formula: u k (t)=[u1(t),u2(t),...u c (t)].

[0056] 1.2) Obtain vector u using the Hilbert transform algorithm. k The analytical representation of each element in (t) is denoted as Calculate the one-sided spectrum, then combine it with the exponential term e. -jwkt Multiply to adjust its corresponding center frequency ω k (t), so that the spectrum of each mode is modulated onto its corresponding fundamental frequency band, and after harmonic conversion The L2 norm of the gradient function is used to estimate the mode u. k The bandwidth of (t).

[0057] Due to the single frequency component ω k Used as the entire vector Harmonic mixing is required, therefore, it is necessary to perform multi-element oscillation u k Find the common frequency component ω in multiple channels in (t). k ,consider Given all channels, the constraint-based optimization problem becomes: The sum of the IMF components decomposed by each channel is sufficient to reproduce the input signal, while minimizing the sum of the intrinsic mode function bandwidths.

[0058]

[0059] In the formula, For the corresponding channel c and vector u kThe analytical representation of each element in (t). This represents the partial derivative with respect to time.

[0060] 1.3) Solving the above variational problem, we construct the augmented Lagrange representation as follows:

[0061]

[0062] 1.4) To solve this transformed unconstrained variational problem, the Alternate Direction Method of Multipliers (ADMM) is applied to alternately update and sum, and then u is calculated. k (t) and center frequency ω k In order to obtain the decomposed signal components. Similar to the mode update method of the VMD algorithm, we can obtain the mode update as:

[0063]

[0064] The center frequency is updated as follows:

[0065]

[0066] By adaptively decomposing the signal's frequency band through updating relationships, k finite-bandwidth IMF components are obtained. Furthermore, since the MVMD method can simultaneously compute data from multiple channels, it ensures frequency consistency across channels, making signal analysis more stable.

[0067] Based on the above theoretical foundation, this embodiment decomposes the seismic data volume into three frequency bands: low, medium, and high. Seismic attributes related to fractures, such as coherence, curvature, ant-like volume, instantaneous phase, and root-mean-square amplitude, are then calculated on these frequency-divided data volumes. Referring to known well logging interpretation results and combining them with existing geological knowledge, m (m≥2, m∈N) seismic sensitive attributes with good correlation and high correspondence to well logging interpretation are selected.

[0068] In step 2) above, the sensitive attribute is selected from multiple attributes, specifically as follows:

[0069] Based on the frequency division results, eight seismic attributes were selected for calculation: coherence, curvature, ant volume, instantaneous phase, root mean square amplitude, variance, first derivative, and sweet spot. Combined with known well logging interpretation results, four seismic sensitive attributes, namely coherence, variance, ant volume, and curvature, were selected.

[0070] In this embodiment, since a single seismic attribute often only reflects partial information, fusing multiple sensitive attributes plays a crucial role in reducing the ambiguity of tectonic interpretation. Conventional attribute fusion is mostly based on simple linear pixel overlay. While the fusion result increases the probability of larger-scale cracks, different attributes may interfere with each other, leading to the loss of some detailed information and affecting the accuracy of crack identification. Therefore, this embodiment uses an unsupervised competitive learning neural network for multi-attribute clustering fusion. Through competitive learning, the learning weight coefficients of different attributes are dynamically updated, improving the accuracy and resolution of attribute fusion.

[0071] Step 3) above, obtaining the comprehensive attribute body, includes the following steps:

[0072] 3.1) After standardizing the attributes, perform principal component analysis (PCA) to obtain orthogonal principal component components;

[0073] In this embodiment, since there are differences in amplitude and physical meaning between different attributes, the selected sensitive attributes are first standardized with a value range of [0,1].

[0074]

[0075] Principal component analysis (PCA) is used to output the standardized eigenvalues ​​of m sensitive attributes, thereby removing redundant information and providing the principal component contribution rates for attribute fusion. The characteristic equation can be expressed as:

[0076] R=λI m (7)

[0077] In the formula, R is the covariance matrix of the input data. The eigenvalues ​​are obtained by solving equation (7), and the cumulative principal component contribution rate is calculated based on the eigenvalues:

[0078]

[0079] 3.2) Treat the principal component components of the sensitive attributes as a set of input samples, set the clustering result of the output classification, and establish the mapping relationship between the input and the output;

[0080] In this embodiment, the principal component components output by principal component analysis are regarded as a set {A1, A2, ... A...} m Assuming the output clustering result C has n categories, then for each category of output result Ci, there exists a weight matrix Wi and a bias term bi, which are iteratively updated according to the Kohonen learning rule:

[0081] ΔW i =α(A i-1 -W i-1 (9)

[0082] Where α is the learning rate function that decays over time, we can further establish a mapping relationship between the input and the output:

[0083]

[0084] 3.3) Clustering results often lack clear geological meaning and have inaccurate boundary delineation, making it difficult to achieve precise identification of complex reservoir structural details. To address this, the Sigmoid activation function is used to optimize the mapping relationship, mapping the clustering result variables to the range [0, 1] according to the mapping relationship. This transforms the clustering results into a geologically meaningful fracture development probability volume, thereby obtaining the clustering fusion results of the coherence, variance, ant-like volume, and curvature attributes of the low, medium, and high frequency band sub-data volumes.

[0085] In step 3.3) above, the optimized mapping relationship is as follows:

[0086]

[0087] In the formula, C' represents the clustering result, and W i Let A represent the weight matrix. j Let b represent the j-th principal component. i This represents the bias term, where n represents the number of clusters in the clustering result, and m is the number of sensitive attributes.

[0088] In the attribute fusion process, Euclidean distance is used as a competition mechanism. The smaller the Euclidean distance between unit vectors, the higher the similarity between the two vectors and the greater the probability of winning the competition, so as to achieve higher accuracy seismic multi-attribute clustering fusion.

[0089] Example: In this example, a seismic data volume of a certain work area is selected as an example. A longitudinal section is randomly selected as follows: Figure 3 As shown in Figure (a), the seismic profile has a time range of 2550-3200 ms and contains 450 data channels. The study layer is located in the strong reflection area in the middle of the profile. Time-frequency analysis revealed that the effective frequency band of the data volume is 10Hz-60Hz. The low, mid, and high frequencies of this data were determined to be 20Hz, 40Hz, and 60Hz, respectively. High-precision frequency division was performed using MVMD.

[0090] 1) Assume k multivariate modulated oscillators u k (t), so that:

[0091]

[0092] In the formula, u k (t)=[u1(t),u2(t),...u c (t)].

[0093] 2) Obtain vector u using the Hilbert transform algorithm. k The analytical representation of each element in (t) is denoted as To calculate the one-sided spectrum, and then with the exponential term e -jwkt Multiply to adjust its corresponding center frequency ω k (t), so that the spectrum of each mode is modulated onto its corresponding fundamental frequency band, and after harmonic conversion The L2 norm of the gradient function is used to estimate the mode u. k The bandwidth of (t).

[0094] Due to a single frequency component ω k Used as the entire vector Harmonic mixing is required, therefore, it is necessary to perform multi-element oscillation u k Find the common frequency component ω in multiple channels in (t). k ,consider Given all channels, the constraint-based optimization problem becomes: The sum of the IMF components decomposed by each channel is sufficient to reproduce the input signal, while minimizing the sum of the intrinsic mode function bandwidths.

[0095]

[0096] In the formula: For the corresponding channel c and vector u k The analytical representation of each element in (t). This represents the partial derivative with respect to time.

[0097] 3) Solving the above variational problem, we construct the augmented Lagrange representation as follows:

[0098]

[0099] 4) To solve this transformed unconstrained variational problem, the alternating direction method of multipliers is applied to alternately update and sum, and then u is obtained. k (t) and center frequency ω k In order to obtain the decomposed signal components. Similar to the mode update method of the VMD algorithm, we can obtain the mode update as:

[0100]

[0101] The center frequency is updated as follows:

[0102]

[0103] After completing the above steps (e.g.) Figure 2 As shown), the seismic data volume is decomposed into the following: Figure 3The data volumes in Figures (b) to (d) represent the low, medium, and high frequency bands. The low-frequency profile reveals large-scale structural information, while the high-frequency profile contains more detailed geological information. Based on the high-precision frequency division results, eight seismic attributes—coherence, curvature, ant-body, instantaneous phase, root mean square amplitude, variance, first derivative, and sweet spot—were selected for calculation. Combined with known well logging interpretation results, four seismically sensitive attributes—coherence, variance, ant-body, and curvature—were selected as the most suitable. Figure 4 Figures (a) to (c) show the extracted sensitive attribute slices of the frequency-division seismic data.

[0104] To achieve information fusion of the above four earthquake sensitivity attributes, a competitive learning neural network is used for attribute clustering. First, the attributes are standardized:

[0105]

[0106] Principal component analysis was performed on the standardized data to obtain four orthogonal principal component components, and the cumulative contribution rate was calculated based on the eigenvalues, such as... Figure 5 As shown, component 1 has the largest contribution rate, components 2 and 3 have the largest contribution rate, and component 4 has the smallest contribution rate. Considering the principal component components of the four sensitive attributes as an input sample set {A1, A2, A3, A4}, and setting the output as a clustering result C|(crack development area and non-development area), for each class of output result C... i Each has a weight matrix Wi and a bias term bi, and is iteratively updated according to the Kohonen learning rule:

[0107] ΔW i =α(A i-1 -W i-1 (17)

[0108] Where α is the learning rate function that decays over time, and further, a mapping relationship between input and output can be established:

[0109]

[0110] Clustering results often lack clear geological meaning and have inaccurate boundary delineation, making it difficult to achieve precise identification of complex reservoir structural details. Therefore, a Sigmoid activation function is introduced to map the clustering result variables to the range [0, 1]. This allows the clustering results to be converted into a geologically meaningful fracture development probability volume, and equation (18) is optimized as follows:

[0111]

[0112] According to Equation 19, the clustering and fusion results of the coherence, variance, ant volume, and curvature attributes of the low, medium, and high frequency band sub-data volumes are obtained, such as... Figure 6 The crack development probability volume shown in Figures (a) to (c) is sliced ​​along the layer. The competitive learning neural network effectively combines the features of the four attributes and reflects them in the attribute fusion probability volume according to a certain contribution value.

[0113] Finally, the low, medium, and high frequency attribute probabilities are fused using frequency fusion to obtain a comprehensive attribute body. Figure 6 Figure (d) shows a slice along the layer of the comprehensive attribute body, which enables accurate identification of the crack distribution in the study area.

[0114] In one embodiment of the present invention, a crack identification system based on MVMD frequency division and competitive neural network multi-attribute fusion is provided, comprising:

[0115] The first processing module determines the effective frequency band range of the seismic data and performs frequency extraction processing on the seismic data in three frequency bands (low, medium, and high) based on the MVMD method.

[0116] The second processing module, in conjunction with well logging interpretation information, calculates multiple attributes based on frequency-division seismic data, selects the most sensitive attributes from among the multiple attributes, calculates the eigenvalues ​​of the sensitive attributes, and obtains orthogonal principal component components.

[0117] The output module clusters different principal component components based on a competitive learning neural network algorithm to obtain attribute probability volumes of different frequency bands. It then fuses the three frequency band probability volumes to output a comprehensive attribute volume, enabling precise identification of cracks in the study area.

[0118] The system provided in this embodiment is used to execute the above-described method embodiments. For specific processes and details, please refer to the above embodiments, which will not be repeated here.

[0119] A computing device provided in one embodiment of the present invention can be a terminal, which may include: a processor, a communication interface, memory, a display screen, and an input device. The processor, communication interface, and memory communicate with each other via a communication bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and a computer program. When executed by the processor, the computer program implements a crack identification method based on MVMD frequency division and competitive neural network multi-attribute fusion. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The communication interface is used for wired or wireless communication with external terminals. Wireless communication can be achieved through Wi-Fi, a management network, NFC (Near Field Communication), or other technologies. The display screen can be a liquid crystal display or an e-ink display. The input device can be a touch layer covering the display screen, or buttons, a trackball, or a touchpad mounted on the casing of the computing device, or an external keyboard, touchpad, or mouse. The processor can call logical instructions stored in the memory.

[0120] Furthermore, the logical instructions in the aforementioned memory can be implemented as software functional units and sold or used as independent products, and can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0121] In one embodiment of the present invention, a computer program product is provided, the computer program product including a computer program stored on a non-transitory computer-readable storage medium, the computer program including program instructions, and when the program instructions are executed by a computer, the computer is able to perform the methods provided in the above-described method embodiments.

[0122] In one embodiment of the present invention, a non-transitory computer-readable storage medium is provided, which stores server instructions that cause a computer to perform the methods provided in the above embodiments.

[0123] The computer-readable storage medium provided in the above embodiments has a similar implementation principle and technical effect to the above method embodiments, and will not be described again here.

[0124] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0125] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0126] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0127] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A crack identification method based on MVMD frequency division and competitive neural network multi-attribute fusion, characterized in that, include: The effective frequency band range of the seismic data was determined, and the seismic data was processed by frequency extraction in three frequency bands (low, medium, and high) based on the MVMD method. Based on well logging interpretation information, multiple attributes are calculated from frequency-division seismic data. Sensitive attributes are selected from these attributes, and their eigenvalues ​​are calculated to obtain orthogonal principal component components. Based on the competitive learning neural network algorithm, clustering is performed between different principal component components to obtain attribute probability volumes of different frequency bands. The three frequency band probability volumes are fused to output a comprehensive attribute volume, thereby achieving precise identification of cracks in the study area.

2. The crack identification method as described in claim 1, characterized in that, Spectral analysis is used to decompose the seismic data volume within the effective frequency band into representative single-frequency data sub-volumes to describe geological information at different scales and levels.

3. The crack identification method as described in claim 1, characterized in that, The MVMD method includes: Based on the common frequency components existing between all channels of the input data, a variational model is introduced into the signal decomposition process to determine the constrained optimization problem and solve the optimal solution of the variational model; the IMF components of each channel simultaneously iteratively update the center frequency and finite bandwidth to adaptively obtain k finite IMF components.

4. The crack identification method as described in claim 3, characterized in that, The constrained optimization problem is: In the formula, For the corresponding channel c and mode vector u k The analytical representation of each element in (t). ω represents the time-dependent partial derivative. k Where C is the center frequency, and u is the total number of data channels. k，c Let x be the k-th mode vector of channel c, where t is time. c (t) represents the input data for channel c.

5. The crack identification method as described in claim 1, characterized in that, The selection of sensitive attributes from multiple attributes includes: Based on the frequency division results, eight seismic attributes were selected for calculation: coherence, curvature, ant volume, instantaneous phase, root mean square amplitude, variance, first derivative, and sweet spot. Combined with known well logging interpretation results, four seismic sensitive attributes, namely coherence, variance, ant volume, and curvature, were selected.

6. The crack identification method as described in claim 1, characterized in that, The acquisition of the comprehensive attribute body includes: After standardizing the attributes, principal component analysis is performed to obtain orthogonal principal component components. The principal component of the sensitive attribute is used as an input sample set, the clustering result of the output classification is set, and the mapping relationship between the input and the output is established. By using the Sigmoid activation function to optimize the mapping relationship, the clustering result variables are mapped to the range [0, 1] according to the mapping relationship. Then, the clustering result is converted into a geologically significant fracture development probability volume, and the clustering fusion results of the coherence, variance, ant volume and curvature attributes of the low, medium and high frequency band sub-frequency data volumes are obtained.

7. The crack identification method as described in claim 6, characterized in that, The optimized mapping relationship is as follows: In the formula, C' represents the clustering result, and W i Let A represent the weight matrix. j Let b represent the j-th principal component. i This represents the bias term, where n represents the number of clusters in the clustering result, and m is the number of sensitive attributes.

8. A crack identification system based on MVMD frequency division and competitive neural network multi-attribute fusion, characterized in that, include: The first processing module determines the effective frequency band range of the seismic data and performs frequency extraction processing on the seismic data in three frequency bands (low, medium, and high) based on the MVMD method. The second processing module, in conjunction with well logging interpretation information, calculates multiple attributes based on frequency-division seismic data, selects the most sensitive attributes from among the multiple attributes, calculates the eigenvalues ​​of the sensitive attributes, and obtains orthogonal principal component components. The output module clusters different principal component components based on a competitive learning neural network algorithm to obtain attribute probability volumes of different frequency bands. It then fuses the three frequency band probability volumes to output a comprehensive attribute volume, enabling precise identification of cracks in the study area.

9. A computer-readable storage medium for storing one or more programs, characterized in that, The one or more programs include instructions that, when executed by a computing device, cause the computing device to perform any of the methods described in claims 1 to 7.

10. A computing device, characterized in that, include: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs including instructions for performing any of the methods described in claims 1 to 7.

Images