A multi-frequency attribute fusion sand body interpretation method, device and readable storage medium based on t-distributed stochastic neighbor embedding

Abstract

The present invention discloses a multi-frequency attribute fusion sand body interpretation method, device and readable storage medium based on t-distribution stochastic neighbor embedding. Among them, a multi-frequency attribute fusion sand body interpretation method based on t-distribution stochastic neighbor embedding includes calculating the Fourier transform spectrum of seismic data to be analyzed and determining the number of frequency components to be extracted; calculating the time-frequency spectrum of the seismic data to be analyzed; extracting multi-frequency attributes of the seismic data according to the number of frequency components and the time-frequency spectrum; determining parameters for fusing multi-frequency attributes according to the size of the seismic data; and fusing the multi-frequency attributes by combining the parameters for fusing multi-frequency attributes and the t-distribution stochastic neighbor embedding algorithm to interpret the sand body thickness and spatial distribution. The t-distribution stochastic neighbor embedding algorithm can better fit the complex non-linear relationship between the sand body thickness and multi-frequency attributes in the absence of prior knowledge.

Description

Technical Field

[0001] The present invention relates to the technical field of oil and gas exploration and development, and particularly relates to a multi-frequency attribute fusion sand body interpretation method, device and readable storage medium based on t-distribution stochastic neighbor embedding. Background Art

[0002] In the field of oil and gas extraction technology, sand body interpretation, especially the interpretation of sand body thickness and spatial distribution characteristics, is crucial for oil and gas extraction and well location deployment. However, the relationship between sand body properties and their seismic responses is usually complex and non-linear, which increases the difficulty of accurately delineating the sand body boundary and interpreting the sand body thickness.

[0003] In sand body interpretation methods, seismic attribute analysis technology is widely used. Some seismic attributes, such as amplitude, frequency, phase and texture, etc., are applied to predict reservoir thickness and interpret the geological structure of the reservoir. Due to the non-stationary nature and wide-band nature of seismic data, it is difficult to accurately delineate the sand body distribution and interpret the sand body thickness only using a single seismic attribute. To solve the limitations of using a single seismic attribute analysis, exploration geophysicists have developed and designed many technologies and methods based on multi-attribute analysis. The red-green-blue (RGB) mixing technology can fuse three seismic attributes and can be used to interpret reservoirs with different thicknesses. However, the RGB mixing technology is a linear model and can only mix three seismic attributes. Therefore, it is difficult for the RGB mixing technology to solve complex and non-linear sand body interpretation problems. Summary of the Invention

[0004] Based on this, it is necessary to provide a multi-frequency attribute fusion sand body interpretation method, device and readable storage medium based on t-distribution stochastic neighbor embedding to solve the complex and non-linear sand body interpretation problems in the prior art.

[0005] A multi-frequency attribute fusion sand body interpretation method based on t-distribution stochastic neighbor embedding provided by the present invention includes the following steps:

[0006] Calculate the Fourier transform spectrum of the seismic data to be analyzed and determine the number of frequency components to be extracted;

[0007] Calculate the time-frequency spectrum of the seismic data to be analyzed;

[0008] Extract the multi-frequency attributes of the seismic data according to the number of frequency components and the time-frequency spectrum;

[0009] Determine the parameters for fusing the multi-frequency attributes according to the size of the seismic data;

[0010] Combine the parameters for fusing the multi-frequency attributes and the t-distribution stochastic neighbor embedding algorithm to fuse the multi-frequency attributes to interpret the sand body thickness and spatial distribution.

[0011] Further, the steps of calculating the Fourier transform spectrum of the seismic data to be analyzed and determining the number of frequency components to be extracted include:

[0012] Read the three-dimensional seismic data u(t, inline, xline);

[0013] Calculate the Fourier spectrum of the three-dimensional seismic data;

[0014] Determine the number of frequency components M to be extracted according to the bandwidth range of the Fourier spectrum.

[0015] Further, the steps of calculating the time-frequency spectrum of the seismic data to be analyzed include:

[0016] Calculate the three-dimensional seismic data through the S transform;

[0017] For the given three-dimensional seismic data u(t, inline, xline), the time-frequency spectrum result obtained by its S transform is denoted as ST(τ, f), and the definition of the S transform is

[0018]

[0019] wherein, inline and xline represent the single-channel seismic signals to be processed in the three-dimensional seismic data, t represents the time variable, f represents the frequency variable, and τ reveals the position information of the Gaussian window function w(t).

[0020] Further, the steps of extracting the multi-frequency attributes of the seismic data according to the number of frequency components and the time-frequency spectrum include:

[0021] Extract M multi-frequency attributes ST(τ, f m , inline, xline), m = 1, 2... M, where f m represents the frequencies of the M multi-frequency attributes to be extracted.

[0022] Further, the steps of fusing the multi-frequency attributes by combining the parameters of the fused multi-frequency attributes and the t-distributed stochastic neighbor embedding algorithm include:

[0023] Fuse the multi-frequency attributes ST(τ, f m , inline, xline), m = 1, 2... M using the t-distributed stochastic neighbor embedding algorithm according to the parameter σ of the fused multi-frequency attributes;

[0024] Calculate the similarity p of the mapping of x i to the data point x j ; j|iand x j is mapped to the data point x i with similarity p i|j ;

[0025] From p j|i and p i|j calculate the joint probability distribution p ij , calculate the joint probability q i between the points y j and y ij in the low-dimensional space distribution;

[0026] Use the KL divergence to evaluate the accuracy of q ij mapped to p ij , and minimize the KL divergence through the gradient descent method;

[0027] Among them, p i|j and p j|i represent the conditional probability of the similarity between data points in the high-dimensional space distribution, p ij represents the joint probability of the similarity between data points in the high-dimensional space distribution, q ij represents the joint probability of the similarity between data points in the low-dimensional space distribution.

[0028] Furthermore, the step of calculating the similarity p i when x j is mapped to the data point x j|i and the similarity p j when x i is mapped to the data point x i|j also includes:

[0029] Based on a set of three-dimensional seismic data {x1, x2, x3,..., x n} where x i represents a feature vector at a time point of a seismic signal, and the number of elements in the feature vector is M, the similarity p i when the data point x j is mapped to the data point x j|i is expressed as

[0030]

[0031]

[0032] The similarity p j when the data point x j is mapped to the data point x j|i is expressed as

[0033]

[0034] Among them, The data point x i The variance of the Gaussian function centered at .

[0035] Furthermore, the j|i and p i|j Calculate the joint probability distribution p ij , calculate the midpoint y of the low-dimensional space distribution i and j The joint probability q between ij The steps also include:

[0036] Joint probability distribution p ij Expressed as

[0037]

[0038] Given a dataset in a low-dimensional space {y1, y2, y3, ..., y n}, data point y i and j The joint probability q between ij It can be calculated as

[0039]

[0040] Where n is the number of sampling points of 3D seismic data.

[0041] Furthermore, the KL divergence is used to evaluate q ij Mapping to p ij The steps to minimize the KL divergence by gradient descent also include:

[0042] The t-distributed random neighbor embedding algorithm uses KL divergence to evaluate q ij , mapping p ij Accuracy:

[0043]

[0044] The KL divergence in the above formula is minimized by the gradient descent method:

[0045]

[0046] Among them, P represents the joint probability distribution in the high-dimensional distribution, and Q represents the joint probability distribution in the low-dimensional distribution.

[0047] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the above-mentioned multi-frequency attribute fusion sand body interpretation method based on t-distribution random neighbor embedding are implemented.

[0048] The present invention also provides a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the steps of the above-described method for interpreting sand bodies by fusing multi-frequency attributes based on t-distributed stochastic neighbor embedding are implemented.

[0049] The method for interpreting sand bodies by fusing multi-frequency attributes based on t-distributed stochastic neighbor embedding provided by the present invention uses the t-distributed stochastic neighbor embedding algorithm to fuse multi-attributes to interpret sand bodies. The t-distributed stochastic neighbor embedding algorithm is an unsupervised non-linear dimensionality reduction algorithm based on manifold learning. Based on the unsupervised and non-linear characteristics of the t-distributed stochastic neighbor embedding algorithm, it can better fit the complex non-linear relationship between sand body thickness and multi-frequency attributes in the absence of prior knowledge, and can accurately delimit the sand body boundary and interpret the sand body thickness. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] 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 use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention, and for those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on the structures shown in these drawings.

[0051] Figure 1 It is a schematic diagram of a three-dimensional seismic data volume to be interpreted containing channel sand bodies in an embodiment of the present invention.

[0052] Figure 2 It is a schematic diagram of a horizon slice of three-dimensional seismic data in an embodiment of the present invention.

[0053] Figure 3 It is the normalized Fourier spectrum applied to actual seismic data in an embodiment of the present invention.

[0054] Figure 4 It is a horizon slice of the coherence attribute applied to actual seismic data in an embodiment of the present invention.

[0055] Figure 5 It is a horizon slice of the RGB fusion interpretation result applied to actual seismic data in an embodiment of the present invention, and the fused frequency attributes are 20Hz, 50Hz, and 70Hz.

[0056] Figure 6 It is a horizon slice of the sand body interpretation result applied to actual seismic data in an embodiment of the present invention.

[0057] The realization, functional features, and advantages of the object of the present invention will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0058] 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. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without making creative efforts fall within the protection scope of the present invention.

[0059] It should be noted that all directional indications (such as up, down, left, right, front, back...) in the embodiments of the present invention are only used to explain the relative position relationship and movement conditions between components in a specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indications will also change accordingly.

[0060] In addition, the descriptions involving "first", "second", etc. in the present invention are only for descriptive purposes, and cannot be understood as indicating or implying their relative importance or implicitly indicating the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one such feature. In addition, "and / or" throughout the text includes three scenarios. Taking A and / or B as an example, it includes the technical solution of A, the technical solution of B, and the technical solution that A and B are satisfied simultaneously. In addition, the technical solutions between the various embodiments can be combined with each other, but it must be based on the fact that those of ordinary skill in the art can implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the protection scope required by the present invention.

[0061] In some embodiments, a multi-frequency attribute fusion sand body interpretation method based on t-distributed stochastic neighbor embedding includes the following steps:

[0062] S100. Calculate the Fourier transform spectrum of the seismic data to be analyzed, and determine the number of frequency components to be extracted;

[0063] S200. Calculate the time-frequency spectrum of the seismic data to be analyzed;

[0064] S300. Extract the multi-frequency attributes of the seismic data according to the number of frequency components and the time-frequency spectrum;

[0065] S400. Determine the parameters for fusing the multi-frequency attributes according to the size of the seismic data. For large-scale seismic data sets with a large number of seismic traces and a large number of time sampling points, a larger perplexity value can be selected to accurately retain the global structure of the analyzed seismic data.

[0066] S500 combines the parameters of multi - frequency attributes and the t - Distributed Stochastic Neighbor Embedding algorithm to fuse multi - frequency attributes for interpreting the sand body thickness and spatial distribution. According to the fused result map and the color scale depth, the thickness and spatial distribution information of the sand body can be obtained (where the larger the color scale value, the thicker the sand body thickness, and vice versa).

[0067] A method for interpreting sand bodies by fusing multi - frequency attributes based on t - Distributed Stochastic Neighbor Embedding provided by the present invention uses the t - Distributed Stochastic Neighbor Embedding (t - SNE) algorithm to fuse multi - attributes to interpret sand bodies. The t - Distributed Stochastic Neighbor Embedding algorithm is an unsupervised non - linear dimensionality reduction algorithm based on manifold learning. Based on the unsupervised and non - linear characteristics of the t - Distributed Stochastic Neighbor Embedding algorithm, it can better fit the complex non - linear relationship between the sand body thickness and multi - frequency attributes in the absence of prior knowledge, and can accurately delimit the sand body boundary and interpret the sand body thickness. In addition, the t - Distributed Stochastic Neighbor Embedding algorithm provides adjustable parameters, making the task of delineating the sand body distribution more flexible.

[0068] In related technologies, the RGB mixing technology is difficult to be used to solve complex and non - linear sand body interpretation problems. Principal Component Analysis (PCA) can transform the original data into a set of linearly independent representations in each dimension through linear transformation, and can be used to extract the main feature components of the data and reduce the dimension of high - dimensional data. However, PCA cannot obtain good results for non - linear dependence relationships. Machine - learning - based methods, especially unsupervised machine - learning methods, can effectively solve the problem of interpreting underground geological structures. Some machine - learning methods, such as K - means, Self - Organizing Map (SOM), Random Forest (RF), Multiple Linear Regression, etc., are widely used.

[0069] Currently, multi - attribute analysis methods have been widely used in sand body interpretation, but selecting appropriate seismic attributes is still a challenge. Seismic signals are a typical non - stationary wide - band signal, which contains rich geological information. The frequency attributes and spectral attributes of seismic signals can fully describe the geological structure. Therefore, using time - frequency transformation to extract the multi - frequency attributes of seismic signals and then making full use of the geological information contained in these multi - frequency attributes is helpful for interpreting sand bodies. A method for interpreting sand bodies by fusing multi - frequency attributes based on t - Distributed Stochastic Neighbor Embedding (t - SNE) solves the defective technical problem that related technologies are only applicable to linear models.

[0070] Specifically, for step S100 of calculating the Fourier transform spectrum of the seismic data to be analyzed and determining the number of frequency components to be extracted, it includes:

[0071] S110 reads three - dimensional seismic data u(t, inline, xline);

[0072] S120. Calculate the Fourier spectrum of three-dimensional seismic data;

[0073] S130. Determine the number M of frequency components to be extracted according to the bandwidth range of the Fourier spectrum.

[0074] More specifically, the steps of S200. calculating the time-frequency spectrum of the seismic data to be analyzed include:

[0075] S210. Calculate three-dimensional seismic data through the S transform;

[0076] S220. For the given three-dimensional seismic data u(t, inline, xline), the time-frequency spectrum result obtained by its S transform is denoted as ST(τ, f). The definition of the S transform is

[0077]

[0078] wherein, inline and xline represent the single-channel seismic signals to be processed of the three-dimensional seismic data, t represents the time variable, f represents the frequency variable, and τ reveals the position information of the Gaussian window function w(t).

[0079] By using the S transform (ST) as the time-frequency transformation tool to extract multi-frequency attributes as the attributes to be fused. Since seismic signals are typical wide-band signals and contain rich geological information. Fusing frequency attributes of different scales can make full use of the geological information in the frequency band of seismic signals, which helps to interpret the sand body thickness. The standard deviation of the Gaussian window function of the S transform is inversely proportional to the frequency variable f. The S transform provides a higher time resolution at high frequencies and a higher frequency resolution at low frequencies.

[0080] Furthermore, the steps of S300. extracting the multi-frequency attributes of seismic data according to the number of frequency components and the time-frequency spectrum include:

[0081] S310. Extract M multi-frequency attributes ST(τ, f m , inline, xline), m = 1, 2... M, where f m represents the frequencies of the M multi-frequency attributes to be extracted.

[0082] Furthermore, in S400, when determining the parameters for fusing multi-frequency attributes based on the magnitude of seismic data, the parameter σ for fusing multi-frequency attributes is determined according to the data scale of the seismic data to be analyzed. The perplexity σ can be regarded as a smoothed measure of the number of effective neighbors. A larger perplexity σ refers to a larger number of neighbors and can accurately retain the global structure of the analyzed data set, while a smaller perplexity σ refers to a smaller number of neighbors and tends to retain the local features in the analyzed data set. For large-scale seismic data sets with a large number of seismic traces and time sampling points, a larger perplexity σ is usually more effective.

[0083] S500. The steps for fusing multi-frequency attributes by combining the parameters for fusing multi-frequency attributes and the t-distributed stochastic neighbor embedding algorithm include:

[0084] S510. Use the t-distributed stochastic neighbor embedding algorithm to fuse the multi-frequency attributes ST(τ, f m , inline, xline), where m = 1, 2... M.

[0085] S520. Calculate the similarity p i of the data point x j mapped to x j|i and the similarity p j of the data point x i mapped to x i|j .

[0086] S530. Calculate the joint probability distribution p j|i from p i|j and p ij . Calculate the joint probability q i between the points y j and y ij in the low-dimensional space distribution.

[0087] S540. To iteratively optimize the probability distribution in the low-dimensional space, t-SNE uses the KL divergence to evaluate the accuracy of the mapping of q ij to p ij , and minimizes the KL divergence through the gradient descent method.

[0088] Among them, p i|j and p j|i represent the conditional probability of the similarity between data points in the high-dimensional space distribution, p ij represents the joint probability of the similarity between data points in the high-dimensional space distribution, and q ij represents the joint probability of the similarity between data points in the low-dimensional space distribution. The KL divergence, also known as relative entropy, is a metric used to measure the similarity between two probability distributions.

[0089] Specifically, S520. Calculate xi The similarity p mapped to the data point x j and x j|i and x j The similarity p mapped to the data point x i The steps of mapping to the similarity p of the data point x i|j also include:

[0090] S521. Based on a set of three-dimensional seismic data {x1, x2, x3,..., x n}}, x i represents a feature vector at a time point of an earthquake signal. The number of elements of the feature vector is M. The similarity p of mapping the data point x i to the data point x j is expressed as j|i is expressed as

[0091]

[0092]

[0093] S522. The similarity p of mapping the data point x j to the data point x j is expressed as j|i is expressed as

[0094]

[0095] wherein, is the variance of the Gaussian function centered on the data point x i .

[0096] S530. Calculating the joint probability distribution p j|i from p i|j and p ij , and calculating the joint probability q i between the points y j and y ij in the low-dimensional space distribution also includes:

[0097] S531. The joint probability distribution p ij is expressed as

[0098]

[0099] S532. Given that the data set in the low-dimensional space is {y1, y2, y3,..., y n}}, the joint probability q i between the data points y i and y ij can be calculated as

[0100]

[0101] where n is the number of sampling points of the 3D seismic data.

[0102] Appendix Figure 1 showing the position of the 3D seismic data volume and the interpreted horizons.

[0103] Appendix Figure 2 Inline slice of the interpreted horizons, where 6 points represent the positions of 6 wells, used to verify the effectiveness of the multi-frequency attribute fusion sand body interpretation method based on t-distribution stochastic neighbor embedding subsequently.

[0104] Appendix Figure 3 is the Fourier spectrum, showing that the frequency band range of the 3D seismic data volume is about 0 - 100 Hz, and the dominant frequency is relatively high.

[0105] Appendix Figure 4 indicating that the coherence attribute can describe the lateral discontinuity of the geological structure, that is, it can obviously describe the lateral discontinuous structures such as seismic faults and channel edges, but it cannot characterize the longitudinal thickness of the channel sand body.

[0106] Appendix Figure 5 indicating that the RGB mixing result can characterize the channel features more clearly than the coherence attribute, but it cannot accurately describe the thickness of the river sand body and does not match the well logging data interpretation result.

[0107] Appendix Figure 6 verifying the effectiveness and accuracy of the sand body thickness interpretation of the multi-frequency attribute fusion sand body interpretation method based on t-distribution stochastic neighbor embedding of the present invention.

[0108] The above are only the preferred embodiments of the present invention, and do not limit the patent scope of the present invention accordingly. Any equivalent structural transformation made under the inventive concept of the present invention, or direct / indirect application in other related technical fields, is included in the patent protection scope of the present invention.

Claims

1. A multi-frequency attribute fusion sand body interpretation method based on t-distribution stochastic neighbor embedding, characterized in that, Including the following steps: Calculating the Fourier transform spectrum of the seismic data to be analyzed and determining the number of frequency components to be extracted; Calculating the time-frequency spectrum of the seismic data to be analyzed; Extracting the multi-frequency attributes of the seismic data according to the number of frequency components and the time-frequency spectrum; Determining the parameters for fusing the multi-frequency attributes according to the size of the seismic data; Fusing the multi-frequency attributes by combining the parameters for fusing the multi-frequency attributes and the t-distributed stochastic neighbor embedding algorithm to interpret the sand body thickness and spatial distribution.

2. The interpretation method according to claim 1, characterized in that, The step of calculating the Fourier transform spectrum of the seismic data to be analyzed and determining the number of frequency components to be extracted includes: Reading the three-dimensional seismic data u(t, imline, xline); Calculating the Fourier spectrum of the three-dimensional seismic data; Determining the number of frequency components M to be extracted according to the bandwidth range of the Fourier spectrum.

3. The interpretation method according to claim 2, characterized in that, The step of calculating the time-frequency spectrum of the seismic data to be analyzed includes: Calculating the three-dimensional seismic data through the S transform; For the given three-dimensional seismic data u(t, inline, xline), the time-frequency spectrum result obtained by its S transform is denoted as ST(τ, f), and the definition of the S transform is wherein, inline and xline represent the single-channel seismic signals to be processed of the three-dimensional seismic data, t represents the time variable, f represents the frequency variable, and τ reveals the position information of the Gaussian window function w(t).

4. The interpretation method according to claim 3, characterized in that, The step of extracting the multi-frequency attributes of the seismic data according to the number of frequency components and the time-frequency spectrum includes: Extract M multi-frequency attributes ST(τ, f, inline, xline) from the time-frequency spectrum ST(τ, f, inline, xline) of the seismic data, where ST(τ, f m , inline, xline), m = 1, 2... M, and f m represents the frequencies of the M multi-frequency attributes to be extracted.

5. The interpretation method according to claim 4, characterized in that, The step of fusing the multi-frequency attributes by combining the parameters for fusing the multi-frequency attributes and the t-distributed stochastic neighbor embedding algorithm includes: Using the t-distributed stochastic neighbor embedding algorithm that fuses multi-frequency attributes ST(τ, f according to the parameter σ that fuses multi-frequency attributes m , inline, xline), m = 1, 2…M; Calculate x i Map to data point x j Similarity p j|i and x j Map to data point x i Similarity p i|j ; From p j|i and p i|j Calculate the joint probability distribution p ij , calculate the joint probability q i between points y j and y ij ; Evaluate q using the KL divergence ij Map to p ij for accuracy, and minimize the KL divergence by gradient descent; where, p i|j and p j|i represent the conditional probability of similarity between data points in the high-dimensional space distribution, p ij represents the joint probability of similarity between data points in the high-dimensional space distribution, q ij represents the joint probability of similarity between data points in the low-dimensional space distribution.

6. The interpretation method according to claim 5, characterized in that, The calculation of x i is mapped to the data point x j with a similarity p j|i and x is mapped to the data point x i with a similarity p i|j The steps also include: Based on a set of three-dimensional seismic data {x1, x2, x3, ..., x n}, x i represents a feature vector at a time point of a seismic signal, the number of elements of the feature vector is M, and the data point x i is mapped to the similarity p j of the data point x j|i which is expressed as Map the data point x j to the data point x j with similarity p j|i is denoted as Among them, is the variance of the Gaussian function centered on the data point x i .

7. The interpretation method according to claim 6, characterized in that, The p j|i and p i|j to calculate the joint probability distribution p ij , calculating the joint probability q i between points y j and y ij in the low-dimensional space distribution further includes: Joint probability distribution p ij Denoted as Given that the dataset in the low-dimensional space is {y1, y2, y3, …, y n}, the joint probability q i between data points y j and y ij can be calculated as where n is the number of sampling points of the three-dimensional seismic data.

8. The interpretation method according to claim 7, wherein, The use of KL divergence to evaluate q ij mapped to p ij The steps to minimize the KL divergence by the gradient descent method also include: The t-distributed stochastic neighbor embedding algorithm uses the KL divergence to evaluate q ij , mapping p ij for accuracy: Realizing the minimization of the KL divergence in the above formula through the gradient descent method: where P represents the joint probability distribution in the high-dimensional distribution, and Q represents the joint probability distribution in the low-dimensional distribution.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, wherein, When the processor executes the computer program, it realizes the steps of a sand body interpretation method based on t-distributed stochastic neighbor embedding multi-frequency attribute fusion according to any one of claims 1 to 8.

10. A computer-readable storage medium storing a computer program, wherein, When the computer program is executed by the processor, it realizes the steps of a sand body interpretation method based on t-distributed stochastic neighbor embedding multi-frequency attribute fusion according to any one of claims 1 to 8.

Images