A method and device for identifying a faulted karst

Abstract

The application provides a recognition method and device for a faulted dissolution body. The method comprises the following steps: obtaining single-channel seismic data of a target area; extracting an amplitude feature from the single-channel seismic data; the single-channel seismic data comprises a plurality of data points; identifying the data points according to the amplitude feature to obtain an initial faulted dissolution body recognition result; performing dimensionality upgrading processing on the data points to obtain upgraded data; determining a sensitive seismic attribute corresponding to the single-channel seismic data according to the upgraded data; calculating the similarity between the initial faulted dissolution body recognition result and the sensitive seismic attribute, and detecting whether the similarity is less than a first threshold value; in the case that the similarity is determined to be less than the first threshold value, inputting the single-channel seismic data and the sensitive seismic attribute into a target recognition model to obtain a target faulted dissolution body recognition result. Based on the above method, the sensitive seismic attribute which has a greater influence on the recognition of the faulted dissolution body can be mined from the single-channel seismic data, and the sensitive seismic attribute is used for the recognition of the faulted dissolution body, so that a more accurate recognition result can be obtained.

Description

Technical Field

[0001] This specification belongs to the field of geophysical exploration technology, and in particular relates to a method and apparatus for identifying fractured solutions. Background Technology

[0002] Fault-knot bodies are a type of columnar dissolution pore and cavern reservoir formed by dissolution and alteration around deep, large fault zones. The identification results of fault-knot bodies can be applied to reservoir prediction, well location deployment, and safe drilling in carbonate karst oil and gas reservoirs. Existing fault-knot body identification technologies involve inputting raw seismic data without processing into a neural network identification model, outputting the identification results. This approach fails to fully extract seismic response characteristics from the raw seismic data, nor does it fully utilize the extracted seismic response characteristics in the neural network identification model. Therefore, it suffers from insufficient input information representation capabilities and cannot obtain accurate fault-knot body identification results.

[0003] There is currently no effective solution to the aforementioned technical problems. Summary of the Invention

[0004] The purpose of the embodiments in this specification is to provide a method for identifying broken solutions, including:

[0005] Acquire single-channel seismic data for the target area; and extract amplitude features from the single-channel seismic data; wherein the single-channel seismic data includes multiple data points;

[0006] Based on the amplitude characteristics, the data points are identified to obtain the initial fractured body identification result;

[0007] The data points are subjected to dimensionality upscaling to obtain dimensionality upscaling data; and the sensitive seismic attributes corresponding to the single-channel seismic data are determined based on the dimensionality upscaling data.

[0008] Calculate the similarity between the initial fault-dissolved body identification result and the sensitive seismic attribute, and detect whether the similarity between the initial fault-dissolved body identification result and the sensitive seismic attribute is less than a first threshold;

[0009] If the similarity between the initial fault-dissolved body identification result and the sensitive seismic attribute is less than a first threshold, the single-channel seismic data and the sensitive seismic attribute are input into the target identification model to obtain the target fault-dissolved body identification result.

[0010] Furthermore, in another embodiment of the method, the amplitude characteristics include root mean square amplitude and spectral decomposition information;

[0011] The step of identifying the data points based on the amplitude characteristics to obtain the initial fractured body identification result includes:

[0012] The first label value of the data point is generated based on the difference between the root mean square amplitude and the second threshold.

[0013] The second label value of the data point is generated based on the difference between the spectral decomposition information and the third threshold.

[0014] Detect whether the first tag value and the second tag value meet preset conditions;

[0015] If the first tag value and the second tag value meet the preset conditions, the initial fracture body identification result of the data point is determined to be a fracture body.

[0016] Furthermore, in another embodiment of the method, determining the sensitive seismic attributes corresponding to the single-channel seismic data based on the upgraded data includes:

[0017] Calculate the embedding distance between the up-dimensional data corresponding to multiple data points and the seed point;

[0018] The embedding distance is normalized to obtain the normalized embedding distance results for multiple data points;

[0019] By splicing the embedding distance normalization results corresponding to multiple data points, the sensitive seismic attributes corresponding to a single seismic data channel are obtained.

[0020] Furthermore, in another embodiment of the method, calculating the embedding distance between the up-dimensional data corresponding to multiple data points and the seed point includes:

[0021] The embedding distance is determined using the following formula:

[0022]

[0023] Where i represents the data point number, x i Let d represent the data point, c represent the seed point, and d represent the seed point. c (x i ) represents x i The corresponding embedding distance, I(x) i ) represents x i The corresponding upgraded data, I(c) represents the upgraded data corresponding to c, x max I(x) represents the data point that differs most from c. max ) represents x max The corresponding upgraded data.

[0024] Furthermore, in another embodiment of the method, the target recognition model is trained in the following manner:

[0025] Obtain the training dataset and the validation dataset;

[0026] An initial recognition model is trained using the training dataset to obtain an intermediate model;

[0027] The validation dataset is input into the intermediate model to obtain intermediate results;

[0028] The accuracy of the intermediate results is detected using a loss function to obtain a first detection result;

[0029] Based on the first detection result, if the first detection result is determined to be passed, the intermediate model is used as the target recognition model.

[0030] Furthermore, in another embodiment of the method, after inputting the single-channel seismic data and the sensitive seismic attributes into the target identification model to obtain the target fault-collapse identification result, the method further includes:

[0031] Identify the drilling fluid loss area within the target region;

[0032] The accuracy of the target fracture body identification result is tested to obtain a second detection result;

[0033] Based on the second detection result, if the second detection result is passed, the target fracture body identification result is corrected using the drilling fluid leakage area to obtain the corrected target fracture body identification result.

[0034] Furthermore, in another embodiment of the method, the step of performing accuracy detection on the target fragment identification result to obtain a second detection result includes:

[0035] The target fractured body identification results are divided into target profile results and target planar results;

[0036] The first profile result is extracted from the initial fractured body identification result;

[0037] The second profile result is extracted from the aforementioned sensitive seismic attributes;

[0038] Determine the similarity between the target profile result and the first profile result; and detect whether the similarity between the target profile result and the first profile result is less than a fifth threshold;

[0039] If the similarity between the target profile result and the first profile result is less than a fifth threshold, the similarity between the target profile result and the second profile result is determined; and it is detected whether the similarity between the target profile result and the second profile result is less than a sixth threshold.

[0040] If the similarity between the target profile result and the second profile result is less than a sixth threshold, the similarity between the target plane result and the coherent technology detection plane result is determined; and it is detected whether the similarity between the target plane result and the coherent technology detection plane result is less than a seventh threshold.

[0041] If the similarity between the target plane result and the coherent technology detection plane result is less than the seventh threshold, the second detection result is determined to be passed.

[0042] Furthermore, in another embodiment of the method, the step of correcting the target fracture-dissolved body identification result using the drilling fluid leakage area to obtain a corrected target fracture-dissolved body identification result includes:

[0043] The first broken solution is extracted from the target broken solution identification results;

[0044] The first fractured solution and the drilling fluid leakage area are intersected to obtain the target fractured solution.

[0045] The target non-fractured solution is obtained by performing a difference set processing on the target fractured solution identification result and the target fractured solution;

[0046] By combining the target broken solution and the target non-broken solution, a corrected target broken solution identification result is obtained.

[0047] The purpose of the embodiments in this specification is to provide a device for identifying broken solutions, including:

[0048] An acquisition module is used to acquire single-channel seismic data of a target area and extract amplitude features from the single-channel seismic data; wherein the single-channel seismic data includes multiple data points;

[0049] The first identification module is used to identify the data points based on the amplitude characteristics to obtain the initial fracture body identification result;

[0050] An extraction module is used to perform dimensionality upscaling on the data points to obtain upscaled data; and to determine the sensitive seismic attributes corresponding to the single-channel seismic data based on the upscaled data.

[0051] The calculation module is used to calculate the similarity between the initial fault-dissolved body identification result and the sensitive seismic attribute, and to detect whether the similarity between the initial fault-dissolved body identification result and the sensitive seismic attribute is less than a first threshold.

[0052] The second identification module is used to input the single-channel seismic data and the sensitive seismic attribute into the target identification model to obtain the target seismic identification result when the similarity between the initial fault-dissolved body identification result and the sensitive seismic attribute is less than a first threshold.

[0053] This specification also provides a computer-readable storage medium storing computer instructions thereon, which, when executed, implements the above-described method for identifying broken solutions. Attached Figure Description

[0054] To more clearly illustrate the embodiments of this specification, the accompanying drawings used in the embodiments will be briefly introduced below. The drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 A schematic diagram of one embodiment of a method for identifying broken solutions provided in this specification;

[0056] Figure 2 This is a schematic flowchart of a method for identifying broken solutions, provided as an example of a specific scenario in the embodiments of this specification.

[0057] Figure 3a A schematic diagram of a three-dimensional physical simulation model of the first target region provided in the embodiments of this specification;

[0058] Figure 3b A planar representation of the fractured solution in the first target region provided for the embodiments of this specification;

[0059] Figure 4a A schematic diagram of the root mean square amplitude property of the first target region provided in the embodiments of this specification;

[0060] Figure 4b This is a schematic diagram of the spectral decomposition information attribute body of the first target region provided in the embodiments of this specification;

[0061] Figure 5a A schematic diagram of the first label value of the first target region provided in the embodiments of this specification;

[0062] Figure 5b A schematic diagram of the second label value of the first target area provided in the embodiments of this specification;

[0063] Figure 5c A schematic diagram of the dissolution label for the first target region provided in the embodiments of this specification;

[0064] Figure 6aDistribution map of sensitive seismic attributes at a 680ms time slice provided in the embodiments of this specification;

[0065] Figure 6b The distribution map of seismic amplitude values ​​at a 680ms time slice provided in the embodiments of this specification;

[0066] Figure 7a This is a diagram showing the target fractured body identification result at a 680ms time slice location obtained using the method described in this application;

[0067] Figure 7b Image showing the results of fault-dissolution identification at the 680ms time slice location obtained using seismic data and the Unet neural network model;

[0068] Figure 7c The image shows the results of fault-dissolution identification at the 680ms time slice location obtained using the sensitive seismic properties proposed by VQ-VAE.

[0069] Figure 8a A three-dimensional sculpted image of the fractured body obtained using the method described in this application;

[0070] Figure 8b A three-dimensional sculpted image of a fractured body obtained using seismic data and the Unet neural network model;

[0071] Figure 8c This is a 3D sculpting result of a fault-collapsed body obtained using the sensitive seismic properties proposed by VQ-VAE.

[0072] Figure 9 A schematic diagram of the geological model of the second target area provided in the embodiments of this specification;

[0073] Figure 10a This is a schematic diagram of the first label value of the second target region provided in the embodiments of this specification;

[0074] Figure 10b This is a schematic diagram of the second label value of the second target region provided in the embodiments of this specification;

[0075] Figure 10c A schematic diagram of the dissolution label for the second target region provided in the embodiments of this specification;

[0076] Figure 11a Distribution map of sensitive seismic attributes at a 300ms time slice provided in the embodiments of this specification;

[0077] Figure 11b This is a distribution map of seismic amplitude values ​​at a 300ms time slice provided in the embodiments of this specification;

[0078] Figure 12aThis is a diagram showing the results of dissolution identification at a 300ms time slice provided in the embodiments of this specification;

[0079] Figure 12b The three-dimensional sculpting result image of the fracture surface based on three-dimensional measured data is provided in the embodiments of this specification.

[0080] Figure 13a This is a schematic diagram of a seismic profile through a well, provided in an embodiment of this specification.

[0081] Figure 13b A diagram showing the identification results of fault-fused bodies in a well-crossing seismic profile obtained using the method described in this application;

[0082] Figure 14a The image shows the identification results of a fractured solution near a well obtained using the method described in this application.

[0083] Figure 14b This is a diagram showing the identification results of broken solutions obtained using coherent detection techniques.

[0084] Figure 15 This is a schematic diagram of the structure of a solution identification device provided in the embodiments of this specification;

[0085] Figure 16 This is a schematic diagram of one embodiment of the server structure provided in the embodiments of this specification. Detailed Implementation

[0086] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0087] Fault-induced solutions are a type of columnar dissolution pore and cavern reservoir formed by dissolution and alteration of deep and large fault zones. The identification results of fault-induced solutions can be applied to reservoir prediction, well location deployment, safe drilling, and other exploration and development work in carbonate karst oil and gas reservoirs.

[0088] Existing fault-knot body identification technologies include neural network / deep learning algorithms, which involve inputting raw seismic data without processing into a neural network identification model and outputting the identification result of the fault-knot body. These existing algorithms directly analyze raw seismic data, which is a comprehensive response of different types of geological structures and stratigraphic features, failing to highlight the seismic response characteristics of any particular geological body (especially deep geological bodies). This results in existing algorithms requiring long-term iterative computation or complex network configurations to complete the identification and detection of fault-knot bodies. Therefore, existing neural network / deep learning algorithms suffer from insufficient input information representation capabilities; they neither fully extract seismic response features from the raw seismic data nor fully utilize the extracted seismic response features when inputting them into the neural network identification model, thus failing to obtain accurate fault-knot body identification results.

[0089] Existing fault-fusion body identification technologies also include multi-attribute fusion methods. These methods extract multiple seismic attributes from raw seismic data, combine them with annotation labels, and then use the fusion results to identify fault-fusion bodies. This approach relies on manual interpretation of seismic profiles to obtain annotation labels, which depends heavily on the experience of the annotators and is highly subjective. Insufficient annotator experience can lead to low-accuracy annotation labels, and the uncertainty in label interpretation ultimately affects the accuracy of fault-fusion body identification.

[0090] In view of the above-mentioned problems of existing methods and the specific reasons for these problems, this application proposes a method and device for identifying fault-fusion bodies based on sensitive seismic attributes, which can improve the identification accuracy of fault-fusion bodies and reduce reliance on human experience.

[0091] Based on the above approach, this specification proposes a method for identifying fault-collapse bodies. The method includes: acquiring single-channel seismic data of a target area; extracting amplitude features from the single-channel seismic data; wherein the single-channel seismic data includes multiple data points; identifying the data points based on the amplitude features to obtain an initial fault-collapse body identification result; performing dimensionality upscaling on the data points to obtain upscaled data; determining the sensitive seismic attribute corresponding to the single-channel seismic data based on the upscaled data; calculating the similarity between the initial fault-collapse body identification result and the sensitive seismic attribute, and detecting whether the similarity between the initial fault-collapse body identification result and the sensitive seismic attribute is less than a first threshold; if the similarity between the initial fault-collapse body identification result and the sensitive seismic attribute is determined to be less than the first threshold, inputting the single-channel seismic data and the sensitive seismic attribute into a target identification model to obtain a target fault-collapse body identification result.

[0092] See Figure 1As shown, this specification proposes a method for identifying broken solutions. In specific implementation, this method may include the following:

[0093] S101: Acquire single-channel seismic data of the target area; and extract amplitude features from the single-channel seismic data; wherein the single-channel seismic data includes multiple data points.

[0094] In some embodiments, the target area refers to the area where a fractured solution exists. A fractured solution is a type of columnar solution-cavitary reservoir formed by dissolution and modification around a deep and large fault zone. Its reservoir types include vault-type reservoirs, porous reservoirs, fracture-pore reservoirs, and fracture reservoirs.

[0095] In some embodiments, it is necessary to acquire multiple single-channel seismic data for the target area. The single-channel seismic data is denoted as X = {x} i}={x1,x2,…,x n}(i=1,2,,n), each single-channel seismic data includes multiple data points x i , where i represents the data point number and n represents the number of data points in a single seismic track. Data point x i Specifically, this includes: the three-axis coordinate information and amplitude information of the point.

[0096] In some embodiments, the amplitude features include root-mean-square amplitude and spectral decomposition information. Root-mean-square amplitude and spectral decomposition information have a better correlation with "beaded" seismic responses ("beaded" seismic responses refer to strong reflections with a small lateral distribution but significant longitudinal differences; "beaded" seismic responses are closely related to fault-soluble bodies, and in existing technologies, areas with "beaded" seismic responses are manually identified as fault-soluble bodies). Therefore, amplitude features can be used for preliminary, non-refined identification of fault-soluble bodies, obtaining initial identification results.

[0097] In some embodiments, amplitude information is extracted from data points, and the root mean square amplitude is calculated using the amplitude information. The root mean square amplitude is calculated for each data point.

[0098] In some embodiments, the method for obtaining spectral decomposition information includes two steps: spectral decomposition and frequency fusion. First, spectral decomposition converts single-channel seismic data X into a two-dimensional complex signal X′ using short-time Fourier transform:

[0099] X′=∫X(τ)h(f,τ-t)·e -j2πfτ )dτ (1)

[0100] Where τ represents the window size of the window function, h represents the window function, f represents the frequency, t represents time, and j represents the imaginary unit (j 2 =―1).

[0101] The window function used in this application is a Gaussian window, and its expression is as follows:

[0102]

[0103] Different column vectors in the real part of X′ represent single-frequency amplitudes of different frequencies. By averaging the three single-frequency amplitudes of the low-frequency, mid-frequency, and high-frequency components (i.e., frequency fusion), we can obtain spectral decomposition information SDC of the same size as X.

[0104]

[0105] in, The operation represents taking the real part, where f1 represents the low-frequency component of a single-frequency amplitude, f2 represents the mid-frequency component of a single-frequency amplitude, and f3 represents the high-frequency component of a single-frequency amplitude; in this application, f1 = 56, f2 = 60, and f3 = 64.

[0106] Based on the above embodiments, the spectral decomposition information can be calculated for each data point in a single-channel seismic data X.

[0107] S102: Based on the amplitude characteristics, the data points are identified to obtain the initial fractured body identification result.

[0108] In some embodiments, the data points are identified based on the amplitude characteristics to obtain a first identification result, specifically including:

[0109] S1: Generate the first label value of the data point based on the difference between the root mean square amplitude and the second threshold;

[0110] S2: Generate a second label value for the data point based on the difference between the spectral decomposition information and the third threshold;

[0111] S3: Detect whether the first tag value and the second tag value meet the preset conditions;

[0112] S4: If the first tag value and the second tag value meet the preset conditions, the initial fracture body identification result of the data point is determined to be a fracture body.

[0113] In some embodiments, the data points are identified based on the amplitude characteristics to obtain a first identification result, which further includes: if the first label value and the second label value do not meet the preset conditions, the initial fractured body identification result of the data point is determined to be a non-fractured body.

[0114] In some embodiments, the second threshold is set to 25, and the third threshold is set to 25. If the root mean square amplitude is greater than or equal to the second threshold, the first label value of the data point is assigned to 1 (a first label value of 1 indicates that the point is a broken solution based on the root mean square amplitude); if the root mean square amplitude is less than the second threshold, the first label value of the data point is assigned to 0 (a first label value of 0 indicates that the point is a non-broken solution based on the root mean square amplitude). If the spectral decomposition information is greater than or equal to the third threshold, the second label value of the data point is assigned to 1 (a second label value of 1 indicates that the point is a broken solution based on the spectral decomposition information); if the spectral decomposition information is less than the third threshold, the second label value of the data point is assigned to 0 (a second label value of 0 indicates that the point is a non-broken solution based on the spectral decomposition information).

[0115] In some embodiments, detecting whether the first tag value and the second tag value meet a preset condition specifically includes:

[0116] S1: Check if the first label value and the second label value are both equal to 1;

[0117] S2: If the first label value and the second label value are both equal to 1, then the first label value and the second label value satisfy the preset condition.

[0118] S3: If the first label value and the second label value are different and equal to 1, then the first label value and the second label value do not meet the preset conditions.

[0119] In some embodiments, the data points of the broken solution in the initial broken solution identification result are assigned a broken solution label with a value of 1, and the data points of the non-broken solution in the initial broken solution identification result are assigned a broken solution label with a value of 0.

[0120] In some embodiments, the above-mentioned preset conditions are determined based on the "hard voting" rule of ensemble learning (the idea of ​​majority rule). By combining the common representation of each individual learner, the data point is finally labeled as a broken solution only when both the root mean square amplitude and spectral decomposition information consider it to be a broken solution; otherwise, it is labeled as 0.

[0121] As can be seen from the above embodiments, in the initial solution fragment identification results, each data point is assigned a solution fragment label. A solution fragment label of 1 indicates that the data point is a solution fragment, and a solution fragment label of 0 indicates that the data point is not a solution fragment. This application comprehensively uses root mean square amplitude and spectral decomposition information to interpret solution fragments, reducing the ambiguity caused by single information and improving the reliability and determinism of solution fragment labels.

[0122] S103: Perform dimensionality upscaling on the data points to obtain upscaled data; and determine the sensitive seismic attributes corresponding to the single-channel seismic data based on the upscaled data.

[0123] In some embodiments, determining the sensitive seismic attributes corresponding to the single-channel seismic data based on the upgraded data specifically includes:

[0124] S1: Calculate the embedding distance between the up-dimensional data corresponding to multiple data points and the seed point;

[0125] S2: Normalize the embedding distance to obtain the normalized embedding distance results for multiple data points;

[0126] S3: Combine the embedding distance normalization results corresponding to multiple data points to obtain the sensitive seismic attributes corresponding to a single-channel seismic data.

[0127] In some embodiments, the data points can be upgraded using the following formula to obtain upgraded data:

[0128]

[0129] Where I(x) i ) represents x i The corresponding upgraded data, e(x) i ) represents x i The corresponding continuous hidden layer variable, I u (x) represents the u-th upgraded data in the embedded dictionary, where u represents the index of the upgraded data.

[0130] In some embodiments, updimensional data are also referred to as high-dimensional embedded vectors or vectorized advanced seismic properties.

[0131] In some embodiments, the embedding distance between the up-dimensional data corresponding to multiple data points and the seed point can be calculated using the following formula:

[0132]

[0133] Where i represents the data point number, x i Let d represent the data point, c represent the seed point, and d represent the seed point. c (x i ) represents x i The corresponding embedding distance, I(x) i ) represents x i The corresponding upgraded data, I(c) represents the upgraded data corresponding to c, x max I(x) represents the data point that differs most from c. max ) represents x max The corresponding upgraded data.

[0134] In some embodiments, a small number of fractured solution data points are randomly selected from the initial fractured solution identification results as seed points.

[0135] In some embodiments, d c (x i The smaller the value, the better the data point x. i The greater the similarity between data point x and seed point c, the better. i The greater the likelihood that it is a broken solution.

[0136] In some embodiments, the embedding distance can be normalized using radial basis functions to obtain normalized embedding distance results for multiple data points:

[0137]

[0138] Among them, P c (x i ) represents x i The corresponding normalized embedding distance result is also known as the dissociation dominance property.

[0139] In some embodiments, P c (x i The range of ) is [0,1], d c (x i The smaller P is, c (x i The closer the data point x is to 1, the better. i The more likely it is to be a broken solution; d c (x i The larger ) is, the more P c (x i The closer the data point x is to 0, the better. i The more likely it is to be a non-broken solution, the more likely P is to be a non-broken solution. c (x i This can characterize the probability that a data point belongs to a broken solution. Using radial basis functions, nonlinear dimensionality reduction can be achieved for the embedding distance, reducing memory consumption and computational costs in subsequent target recognition model outputs.

[0140] In some embodiments, the embedding distance normalization results corresponding to data points belonging to the same single-channel seismic data are concatenated to obtain the sensitive seismic attribute (i.e., the dominant attribute corresponding to the single-channel seismic data) corresponding to that single-channel seismic data:

[0141]

[0142] Among them, P c (X) represents a sensitive seismic attribute.

[0143] In some embodiments, sensitive seismic attributes can highlight the differences between fault-fused and non-fault-fused bodies in terms of spatial morphology, geometric structure, and numerical distribution, reducing the difficulty for deep learning neural networks to identify fault-fused bodies, thereby improving the training speed and application capabilities of deep learning neural networks.

[0144] In some embodiments, upscaling data can be obtained using an unsupervised vector quantized variational autoencoder (VQ-VAE). The VQ-VAE network structure is similar to that of a variational autoencoder (VAE), including an encoder and a decoder. The difference between VQ-VAE and VAE is that VQ-VAE includes an embedding dictionary between the encoder and decoder. Specifically, the encoder in VQ-VAE is not restricted by a specific data distribution; under unlabeled training conditions, the encoder iterates over data points x. i Perform data encoding, extract features related to the broken solution, and input x i Mapped to a latent space, characterized as a continuous hidden layer variable e(x) that amplifies the features of fractured and non-fractured solutions. i The embedding dictionary is responsible for storing the continuous hidden layer variables e(x) output by the encoder. i Discretize and re-encode based on the continuous hidden layer variable e(x) i ) and each high-dimensional embedding vector I(x) in the embedding dictionary i Quantizing the distance between (x, y) can realize the quantization of continuous hidden layer variables e(x, y) i Within the VQ-VAE, it is transformed into the nearest up-dimensional data I(x). i The decoder then processes the upgraded data I(x) i Decode and reconstruct, output x i Approximate data x' i Furthermore, x can be calculated. i and x' i The degree of similarity, if x i and x' i The high degree of similarity indicates that x i and x' i They are quite similar, therefore I(x) i This can be used for subsequent calculations. Based on the above description, VQ-VAE is an unsupervised learner that does not require broken-solution labels; it can apply this to each data point x. i The data undergoes dimensionality upscaling to transform it into data of appropriate length. It should be noted that this application only applies to x. i The amplitude information in the data is upgraded, but the three-axis coordinate information does not need to be upgraded.

[0145] Based on the above embodiments, the upgraded data I(x) i This approach can amplify the differences in seismic response characteristics between fractured and non-fractured solutions within a high-dimensional hidden space, fully extracting effective information from single-channel seismic data. Then, nonlinear transformations such as embedding distance and radial basis functions are used to transform the upgraded data I(x) into a higher-dimensional hidden space. i This transforms the seismic properties into sensitive seismic attributes that highlight the characteristics of fault-collapsed bodies, further amplifying the differences in response characteristics and numerical distributions between fault-collapsed and non-fault-collapsed bodies. The amplitude differences between fault-collapsed and non-fault-collapsed bodies are amplified and highlighted through embedding distance, providing richer input data for target recognition models built based on deep learning algorithms (P). c (X)) overcomes the problems of strong subjectivity and poor accuracy of existing manual annotation, which helps to improve the accuracy of subsequent target recognition models in identifying fractured bodies and improves the ability of target recognition models to characterize fractured body features.

[0146] S104: Calculate the similarity between the initial fault-dissolved body identification result and the sensitive seismic attribute, and detect whether the similarity between the initial fault-dissolved body identification result and the sensitive seismic attribute is less than a first threshold.

[0147] In some embodiments, single-channel seismic data X = {x} can be extracted from the initial fault-body identification results. i}={x1,x2,…,x n The labels corresponding to each data point in the array are concatenated to form a label vector [label(x1), label(x2), ..., label(x...]. n ], label(x1) represents the label of the broken body corresponding to x1, label(x2) represents the label of the broken body corresponding to x2, label(x n ) represents x n The corresponding dissolution label; then calculate [P] c (x1),P c (x2),…,P c (x n )] and [label(x1),label(x2),…,label(x n The Euclidean distance between [P] and [P] is used as the similarity score; if the Euclidean distance is less than a first threshold, it indicates that [P] is not similar. c (x1),P c (x2),…,P c (x n )] and [label(x1),label(x2),…,label(x nThe initial fault-dissolved body identification result is considered to be similar to the sensitive seismic attributes. The sensitive seismic attributes can effectively characterize the distribution of fault-dissolved bodies and are accurate, thus serving as subsequent input data. Referring to the above steps, the similarity between the sensitive seismic attributes corresponding to multiple single-channel seismic data and the initial fault-dissolved body identification result is checked to see if all are less than the first threshold. If the similarity between the sensitive seismic attributes corresponding to any single-channel seismic data and the initial fault-dissolved body identification result is greater than or equal to the first threshold, then seed points are reselected, and the sensitive seismic attributes are recalculated.

[0148] S105: If the similarity between the initial fault-dissolved body identification result and the sensitive seismic attribute is less than a first threshold, the single-channel seismic data and the sensitive seismic attribute are input into the target identification model to obtain the target fault-dissolved body identification result.

[0149] In some embodiments, the target recognition model is trained based on a deep learning algorithm, specifically the Unet neural network model.

[0150] In some embodiments, the input and output data of the target recognition model can be represented by the following formula:

[0151] Y pre =Unet((X,P c (X)),m) (8)

[0152] Among them, single-channel seismic data X and sensitive seismic attributes P c (X) represents the input data; m represents the network parameters, including the weights w and biases b of the target recognition model; Y pre This represents the target fragment identification result, i.e., the output data, where each data point corresponds to a Y-axis. pre Y pre The range is [0,1], Y pre The closer the value is to 1, the more likely the data point is to be a broken solution. pre The closer the value is to 0, the more likely the data point is to be a non-dissolved solution.

[0153] In some embodiments, Y can be pre Data points greater than or equal to the threshold for identifying broken solutions are considered broken solutions, and Y is... pre Data points below the threshold for identifying broken solutions are considered as non-broken solutions.

[0154] In some embodiments, the target recognition model is trained in the following manner:

[0155] S1: Obtain the training dataset and validation dataset;

[0156] S2: Train the initial recognition model using the training dataset to obtain the intermediate model;

[0157] S3: Input the validation dataset into the intermediate model to obtain intermediate results;

[0158] S4: Use a loss function to detect the accuracy of the intermediate results and obtain the first detection result;

[0159] S5: Based on the first detection result, if the first detection result is determined to be passed, the intermediate model is used as the target recognition model.

[0160] In some embodiments, the training dataset includes: single-channel seismic data corresponding to the first sample region, fault-collapse labels, sensitive seismic attributes, and target fault-collapse identification results. The validation dataset includes: single-channel seismic data corresponding to the second sample region, fault-collapse labels, sensitive seismic attributes, and target fault-collapse identification results.

[0161] In existing technologies, cross-entropy loss is typically used as the loss function. Models using cross-entropy loss generally achieve high recognition accuracy, but their accuracy decreases when the number of sample classes is imbalanced. To alleviate the problem of imbalanced positive and negative sample ratios in the identification of broken bodies, this application uses the Focal Loss function:

[0162]

[0163] Where FocalLoss represents the loss function value, α represents the shared weight, and Unet((X0,P) c (X0)),m) represent intermediate results, X0 represents single-channel seismic data in the validation dataset, P c (X0) represents the sensitive seismic properties in the validation dataset, γ represents the modulation coefficient, and Y0 represents the fault-dissolution label in the validation dataset.

[0164] In some embodiments, the shared weight α specifically represents "the shared weight of positive and negative sample pairs in the total loss". Generally, there are far fewer broken samples (broken sample label = 1) than non-broken samples (broken sample label = 0), so α needs to be set to be larger. Typically, α is between 0.5 and 1. In this application, α is 0.9 and γ is 2.

[0165] The target recognition model provided in this application can fully learn the difference information between fractured and non-fractured bodies without losing the original seismic amplitude information, thereby achieving the goal of obtaining reasonable recognition results with a small amount of input data.

[0166] In some embodiments, the accuracy of the intermediate results is detected using a loss function to obtain a first detection result, specifically including:

[0167] S1: Detect whether the loss function value is less than the fourth threshold;

[0168] S2: If the loss function value is determined to be less than the fourth threshold, the first detection result is determined to be passed;

[0169] S3: If the loss function is determined to be greater than or equal to the fourth threshold, the first detection result is determined to be a failure.

[0170] In some embodiments, if the loss function value is less than the fourth threshold, it indicates that the accuracy of the intermediate result is high, and the intermediate model can be used as the target recognition model; if the loss function value is greater than or equal to the fourth threshold, it indicates that the accuracy of the intermediate result is low, and the intermediate model needs to be trained and the network parameters adjusted until the loss function value is less than the fourth threshold, indicating that the accuracy requirement is met.

[0171] The method provided in this application can achieve a two-way improvement between deep learning methods and multi-attribute fusion methods, and obtain more accurate results for identifying broken bodies.

[0172] In some embodiments, after inputting the single-channel seismic data and the sensitive seismic attributes into the target identification model to obtain the target fault-knot identification result, the method further includes:

[0173] S1: Obtain the drilling fluid loss area in the target region;

[0174] S2: Perform accuracy detection on the target broken body identification result to obtain a second detection result;

[0175] S3: Based on the second detection result, if the second detection result is passed, the target fracture body identification result is corrected using the drilling fluid leakage area to obtain the corrected target fracture body identification result.

[0176] In some embodiments, the accuracy of the target fracture body identification result is measured to obtain a second detection result, specifically including:

[0177] S1: Divide the target fracture body identification results into target profile results and target planar results;

[0178] S2: Extract the first profile result from the initial fractured body identification result;

[0179] S3: Extract the second profile results from the aforementioned sensitive seismic attributes;

[0180] S4: Determine the similarity between the target profile result and the first profile result; and detect whether the similarity between the target profile result and the first profile result is less than a fifth threshold;

[0181] S5: If the similarity between the target profile result and the first profile result is less than a fifth threshold, determine the similarity between the target profile result and the second profile result; and detect whether the similarity between the target profile result and the second profile result is less than a sixth threshold;

[0182] S6: If the similarity between the target profile result and the second profile result is less than the sixth threshold, determine the similarity between the target plane result and the coherent technology detection plane result; and detect whether the similarity between the target plane result and the coherent technology detection plane result is less than the seventh threshold;

[0183] S7: If the similarity between the target plane result and the coherent technology detection plane result is less than the seventh threshold, the second detection result is determined to be passed.

[0184] In some embodiments, based on the triaxial coordinate information in the data points, the target area can be divided into a profile area and a planar area; the target fracture solution identification results of each data point in the profile area are used as the target profile results; and the target fracture solution identification results of each data point in the planar area are used as the target planar results.

[0185] In some embodiments, the fracture labels of each data point in the profile region are used as the first profile result.

[0186] In some embodiments, the sensitive seismic properties of each data point in the profile region are used as the second profile result.

[0187] In some embodiments, Euclidean distance can be selected as the similarity between the target profile result and the first profile result, the target profile result and the second profile result, and the target plane result and the coherent detection plane result. Of course, it should be noted that other similarity measurement parameters can also be selected, such as the squared Euclidean distance, Manhattan distance, etc.

[0188] Based on the above embodiments, if the target profile result is relatively close to the first profile result and also relatively close to the second profile result, the accuracy of the target profile result is considered to be relatively high. If the similarity between the target profile result and the first profile result is greater than or equal to a fifth threshold, or if the similarity between the target profile result and the second profile result is greater than or equal to a sixth threshold, the accuracy of the target profile result is considered to be relatively low.

[0189] In some embodiments, the coherent detection plane result is obtained as follows: coherent detection is performed on the target region to obtain the coherent detection result of the target region; from the coherent detection result, the coherent detection result corresponding to the plane region is extracted as the coherent detection plane result. Coherent detection is a method for obtaining fracture results in the prior art, and it has relatively good fracture recognition accuracy for plane regions. If the target plane result is close to the coherent detection plane result (i.e., the similarity between the target plane result and the coherent detection plane result is less than the seventh threshold), the accuracy of the target plane result is considered to be relatively high.

[0190] In some embodiments, the target fracture-dissolved body identification result is corrected using the drilling fluid leakage area to obtain a corrected target fracture-dissolved body identification result, specifically including:

[0191] S1: Extract the first broken solution from the target broken solution identification results;

[0192] S2: Intersect the first broken solution body and the drilling fluid leakage area to obtain the target broken solution body;

[0193] S3: Perform a difference set processing on the target broken solution identification result and the target broken solution to obtain the target non-broken solution;

[0194] S4: Combine the target broken solution and the target non-broken solution to obtain the corrected target broken solution identification result.

[0195] In some embodiments, the first broken solution is all data points identified as broken solutions in the target broken solution identification result.

[0196] In some embodiments, the drilling fluid loss area must be a broken solution. Therefore, by intersecting the first broken solution and the drilling fluid loss area, a more accurate target broken solution can be obtained.

[0197] In some embodiments, the corrected target fracture solution identification results can be combined with the triaxial coordinate information of the data points to create a 3D sculpted result of the fracture solution. During subsequent well location deployment and drilling operations, the 3D sculpted result of the fracture solution can be used to design well locations that avoid the fracture solution, thereby preventing drilling accidents such as drilling fluid loss and ensuring safe drilling.

[0198] Based on the above embodiments, the target fracture-solution identification results are checked or corrected from two aspects: profile-combined analysis and drilling loss risk quality control, further improving the accuracy of the target fracture-solution identification results. This application can not only improve the characterization accuracy of multi-scale fracture-solution contours, but also identify the internal structure in different fracture-vuggy combination patterns.

[0199] In a specific scenario example, it can be followed Figure 2 The method shown achieves fault-dissolved body identification. First, seismic data is acquired, including different attribute values ​​such as root-mean-square amplitude, relative acoustic impedance, and spectral decomposition properties (i.e., spectral decomposition information). Attribute optimization is performed to extract root-mean-square amplitude and spectral decomposition properties that have a better correlation with the "beaded" seismic response. Using a hard-voting strategy in ensemble learning, fault-dissolved body labels are generated: if both the root-mean-square amplitude and spectral decomposition properties identify a fault-dissolved body, the fault-dissolved body label value is equal to 1; otherwise, the fault-dissolved body label value is equal to 0. A vector quantization variational autoencoder is then used to output vectorized high-level seismic properties (I(x...). i According to I(x) i The process involves generating sensitive seismic attributes and performing feature analysis on fault-dissolved and non-fault-dissolved bodies. This involves determining the similarity between the fault-dissolved body label and the sensitive seismic attributes. If the similarity is low (the fault-dissolved body label and the sensitive seismic attributes are relatively similar), the requirement is met; if the similarity is low (the fault-dissolved body label and the sensitive seismic attributes are dissimilar), the requirement is not met, and the vector quantization variational autoencoder needs to be corrected before outputting the vectorized high-level seismic attributes again. If the requirements are met, the dimensionality reduction advantage attribute (P) is obtained. c (X)). The dimensionality reduction advantage attribute and seismic data are used as input data and input into the fault-dissolved body intelligent identification model (Unet model). The target fault-dissolved body identification result is output and input into the quality controller. The identification accuracy of the profile area in the target fault-dissolved body identification result is detected by using the first profile result and the second profile result. The planar area in the target fault-dissolved body identification result is corrected by using the drilling fluid leakage area. The three-dimensional sculpting result of the fault-dissolved body is drawn using the corrected result.

[0200] In some specific scenario examples, Figure 3a This represents a three-dimensional physical simulation model of the first target region. Figure 3b This is a plan view of the fractured solution in the simulated design of the first target area. Zones 1 to 18 contain several holes. Each zone represents a region with holes. The 3D physical simulation model of the first target area includes 780 main survey lines (Inlines) and 640 connecting survey lines (Crosslines), covering an area of ​​20 × 17 km. 2 The number of time sampling points is 800, the sampling interval is 2ms, and the time recording range is 0–1600ms. According to... Figure 3b It can be seen that the three-dimensional physical simulation model includes single or multiple holes of different sizes, shapes and spatial distributions, and there is an obvious "X-shaped" fracture in the center of the three-dimensional physical simulation model.

[0201] In a specific scenario example Figure 4aThis represents the root mean square amplitude attribute volume of the first target region. Figure 4b This represents the spectral decomposition information attribute body of the first target region. Figure 4a and Figure 4b compared to Figure 3a This allows for a more distinct depiction of the characteristics of the broken-solution.

[0202] In a specific scenario example Figure 5a The diagram shows the first label value of the first target region. Data points corresponding to the root mean square amplitude attribute volume with a value greater than or equal to 25 are assigned a first label value of 1 (i.e., the black part), and data points corresponding to the root mean square amplitude attribute volume with a value less than 25 are assigned a first label value of 0 (i.e., the white part). Figure 5b The diagram illustrates the second label value for the first target region. Data points corresponding to spectral decomposition information attribute volumes greater than or equal to 25 are assigned a second label value of 1 (i.e., the black part), while data points corresponding to spectral decomposition information attribute volumes less than 25 are assigned a second label value of 0 (i.e., the white part). Figure 5c A schematic diagram of the fractured body label representing the first target region. Data points with a first label value of 1 and a second label value of 1 are assigned a fractured body label of 1 (i.e., the black part), and data points with a first label value of 0 and / or a second label value of 0 are assigned a fractured body label of 0 (i.e., the white part).

[0203] In a specific scenario example Figure 6a This is a map showing the distribution of sensitive seismic attributes corresponding to fault-collapsed (light gray) and non-fault-collapsed (dark gray) bodies at the 680ms time slice. Figure 6b This is a map showing the distribution of seismic amplitude values ​​corresponding to fractured solutions (light gray) and non-fractured solutions (dark gray) at a time slice of 680ms. Figure 6a The horizontal axis represents the normalized values ​​of the sensitive earthquake attributes. Figure 6a The horizontal axis represents the normalized value of the earthquake amplitude. According to... Figure 6a It can be seen that the normalized values ​​of the sensitive seismic attributes corresponding to fault-knot bodies are concentrated in the range of 0.9 to 1, while the normalized values ​​of the sensitive seismic attributes corresponding to non-fault-knot bodies are concentrated in the range of 0.5 to 0.8. Figure 6b The distribution of these organisms is quite mixed and difficult to distinguish. Figure 6a and Figure 6b The comparison shows that the sensitive seismic properties proposed by VQ-VAE and radial basis functions are better able to distinguish between fractured and non-fractured solutions than single-channel seismic data.

[0204] In a specific scenario example Figure 7a The image shows the target fractured solution identification results obtained using the method described in this application at a time slice position of 680ms. The light-colored area represents the fractured solution, and the dark-colored area represents the non-fractured solution. Figure 7b This represents the results of fault-collapse identification at the 680ms time slice location obtained using only seismic data and the Unet neural network model. Light-colored areas represent fault-collapse, and dark-colored areas represent non-fault-collapse. Figure 7c This indicates the results of fault-collapse identification at the 680ms time slice location obtained using the sensitive seismic attributes proposed by VQ-VAE. Light-colored areas represent fault-collapse, and dark-colored areas represent non-fault-collapse. Figure 7a , Figure 7b , Figure 7c This indicates that the identification results of the fracture-dissolved bodies using the three methods largely coincide with the locations of the fracture-dissolved bodies on the seismic slices. Compared to the other two methods, the method described in this application achieves more precise identification of small-scale cavities (rectangular areas) and even smaller-scale fissures (circular areas) with similar spatial locations.

[0205] In a specific scenario example Figure 8a This indicates the three-dimensional sculpted result of the fractured body obtained using the method described in this application. Figure 8b This indicates the 3D sculpting results of the fracture surface obtained using only seismic data and the Unet neural network model. Figure 8c This represents the 3D sculpting results of the fractured solution obtained using the sensitive seismic properties proposed by VQ-VAE. Figure 8a , Figure 8b , Figure 8c and Figure 3b The comparison shows that the method described in this application is less affected by noise, the sculpted three-dimensional fracture solution results are clear and complete, more consistent with the real situation, and fewer small-scale fracture solutions are missed.

[0206] In a specific scenario example Figure 9 This geological model represents the second target area, constructed based on three-dimensional measured data. It contains 800 main survey lines (Inline) and 800 connecting survey lines (Crossline), with 451 time sampling points at a 2ms interval, covering an area of ​​approximately 150 km². 2 The second target area exhibits shear fractures, with sedimentary facies transitioning from continental to marine from shallow to deep. The shallow target reservoir is a continental clastic reservoir, while the deep reservoir is a marine carbonate reservoir. Acidic fluids flowing through the fracture zones erode and alter the carbonate rocks, forming fractured solutions with good porosity and permeability, and developing high-angle fractures.

[0207] In a specific scenario example Figure 10a The diagram shows the first label value of the second target region. Data points corresponding to the root mean square amplitude attribute volume with a value greater than or equal to 25 are assigned a first label value of 1 (i.e., the black part), and data points corresponding to the root mean square amplitude attribute volume with a value less than 25 are assigned a first label value of 0 (i.e., the white part). Figure 10bThe diagram illustrates the second label value for the second target region. Data points corresponding to spectral decomposition information attribute volumes greater than or equal to 25 are assigned a second label value of 1 (i.e., the black part), while data points corresponding to spectral decomposition information attribute volumes less than 25 are assigned a second label value of 0 (i.e., the white part). Figure 10c The diagram shows the broken-solution label for the second target region. Data points with a first label value of 1 and a second label value of 1 are assigned a broken-solution label of 1 (i.e., the black part), while data points with a first label value of 0 and / or a second label value of 0 are assigned a broken-solution label of 0 (i.e., the white part).

[0208] In a specific scenario example Figure 11a This is a map showing the distribution of sensitive seismic attributes corresponding to fault-collapsed (light gray) and non-fault-collapsed (dark gray) bodies at a 300ms time slice. Figure 11b This is a map showing the distribution of seismic amplitude values ​​at a 300ms time slice, corresponding to ruptured solutions (light gray) and non-ruptured solutions (dark gray). Figure 11a and Figure 11b The comparison shows that sensitive seismic attributes are better able to distinguish between fractured and non-fractured solutions than seismic data.

[0209] In a specific scenario example Figure 12a This indicates the identification results of the broken solution at the 300ms time slice. Figure 12b This represents the 3D sculpting results of a fractured body based on 3D measured data. According to... Figure 12a It can be seen that the fractured body identification results obtained by the method provided in this application basically coincide with the "beaded" seismic response on the profile. Figure 12b This indicates that the method provided in this application accurately depicts the location and size of the hole, which is basically consistent with the actual hole.

[0210] In a specific scenario example Figure 13a This represents a seismic profile through a well, drawn based on seismic data. Figure 13b express Figure 13a The corresponding identification results of the broken-solution obtained using the method described in this application. Figure 13a and Figure 13b In the diagram, the horizontal curve represents the layer logging curve, and the vertical curve represents the overall logging curve. Figure 13a In the diagram, white areas represent amplitude values ​​less than 0, and black areas represent amplitude values ​​greater than 0. Figure 13b In the diagram, white areas represent broken solutions, and black areas represent non-broken solutions. The location of the broken solution obtained by the method described in this application (i.e., Figure 13b The white area in the middle shows obvious strong amplitude anomaly "beaded" reflection characteristics, and the location of drilling fluid loss also predicts the broken solution.

[0211] In a specific scenario example Figure 14a This indicates the identification result of a fractured solution near a well obtained using the method described in this application. TP represents the well name, and the circle represents the location of the TP well. Figure 14b This indicates the identification result of fractured solution near a certain well obtained by coherent technical detection. TP represents the well name, and the circle indicates the location of well TP. The location of the fractured solution identified by coherent technical detection is basically consistent with the method described in this application, further confirming the reliability of the identification results of this application. The interpreted fractured solution can guide subsequent drilling trajectory design and avoid accidents such as stuck pipe and lost drilling.

[0212] Based on the above-described method for identifying broken-solution bodies, this specification also provides an embodiment of a device for identifying broken-solution bodies, see reference. Figure 15 As shown, the identification device for broken solutions specifically includes the following modules: acquisition module 1501, first identification module 1502, extraction module 1503, calculation module 1504, and second identification module 1505.

[0213] The acquisition module 1501 is used to acquire single-channel seismic data of the target area and extract amplitude features from the single-channel seismic data; wherein the single-channel seismic data includes multiple data points;

[0214] The first identification module 1502 is used to identify the data points based on the amplitude characteristics to obtain the initial fracture body identification result;

[0215] The extraction module 1503 is used to perform dimensionality upscaling on the data points to obtain dimensionality upscaling data; and to determine the sensitive seismic attributes corresponding to the single-channel seismic data based on the dimensionality upscaling data.

[0216] The calculation module 1504 is used to calculate the similarity between the initial fault-dissolved body identification result and the sensitive seismic attribute, and to detect whether the similarity between the initial fault-dissolved body identification result and the sensitive seismic attribute is less than a first threshold.

[0217] The second identification module 1505 is used to input the single-channel seismic data and the sensitive seismic attribute into the target identification model to obtain the target seismic identification result when the similarity between the initial fault-dissolved body identification result and the sensitive seismic attribute is less than a first threshold.

[0218] In some embodiments, the first identification module 1502 is specifically configured to generate a first label value for the data point based on the difference between the root mean square amplitude and the second threshold; generate a second label value for the data point based on the difference between the spectral decomposition information and the third threshold; detect whether the first label value and the second label value meet preset conditions; and determine that the initial fractured body identification result of the data point is a fractured body if the first label value and the second label value meet the preset conditions.

[0219] In some embodiments, the extraction module 1503 is specifically used to calculate the embedding distance between the upgraded data corresponding to multiple data points and the seed point; normalize the embedding distance to obtain the normalized embedding distance result corresponding to multiple data points; and concatenate the normalized embedding distance result corresponding to multiple data points to obtain the sensitive seismic attribute corresponding to the single-channel seismic data.

[0220] In some embodiments, the device for identifying broken bodies is further configured to train a target recognition model by: acquiring a training dataset and a verification dataset; training an initial recognition model using the training dataset to obtain an intermediate model; inputting the verification dataset into the intermediate model to obtain an intermediate result; detecting the accuracy of the intermediate result using a loss function to obtain a first detection result; and, based on the first detection result, using the intermediate model as the target recognition model if the first detection result is determined to be passed.

[0221] In some embodiments, the device for identifying broken solutions is further configured to: acquire a drilling fluid leakage area in a target area; perform accuracy detection on the identification result of the target broken solution to obtain a second detection result; and, based on the second detection result, if the second detection result is passed, correct the identification result of the target broken solution using the drilling fluid leakage area to obtain a corrected identification result of the target broken solution.

[0222] It should be noted that the units, devices, or modules described in the above embodiments can be implemented by computer chips or physical entities, or by products with certain functions. For ease of description, the above devices are described by dividing them into various modules according to their functions. Of course, in implementing this specification, the functions of each module can be implemented in one or more software and / or hardware, or the module that implements the same function can be implemented by a combination of multiple sub-modules or sub-units, etc. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional 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. Furthermore, the coupling or direct coupling or communication connection between the devices or units shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0223] This specification also provides a computer storage medium for a method of identifying fault-dissolved bodies. The computer storage medium stores computer program instructions, which, when executed by a processor, perform the following: acquiring single-channel seismic data of a target area; extracting amplitude features from the single-channel seismic data; wherein the single-channel seismic data includes multiple data points; identifying the data points based on the amplitude features to obtain an initial fault-dissolved body identification result; performing dimensionality upscaling on the data points to obtain upscaled data; determining the sensitive seismic attribute corresponding to the single-channel seismic data based on the upscaled data; calculating the similarity between the initial fault-dissolved body identification result and the sensitive seismic attribute, and detecting whether the similarity between the initial fault-dissolved body identification result and the sensitive seismic attribute is less than a first threshold; if the similarity between the initial fault-dissolved body identification result and the sensitive seismic attribute is determined to be less than the first threshold, inputting the single-channel seismic data and the sensitive seismic attribute into a target identification model to obtain a target fault-dissolved body identification result.

[0224] In this embodiment, the storage medium includes, but is not limited to, Random Access Memory (RAM), Read-Only Memory (ROM), cache, hard disk drive (HDD), or memory card. The memory can be used to store computer program instructions. The network communication unit can be an interface configured according to standards specified in the communication protocol for network connection communication.

[0225] In this embodiment, the specific functions and effects implemented by the program instructions stored in the computer storage medium can be explained in comparison with other implementation methods, and will not be repeated here.

[0226] This specification also provides a server, including a processor and a memory for storing processor-executable instructions. In specific implementations, the processor can perform the following steps according to the instructions: acquiring single-channel seismic data of a target area; and extracting amplitude features from the single-channel seismic data; wherein the single-channel seismic data includes multiple data points; identifying the data points based on the amplitude features to obtain an initial fault-dissolved body identification result; performing dimensionality upscaling on the data points to obtain upscaled data; and determining the sensitive seismic attribute corresponding to the single-channel seismic data based on the upscaled data; calculating the similarity between the initial fault-dissolved body identification result and the sensitive seismic attribute, and detecting whether the similarity between the initial fault-dissolved body identification result and the sensitive seismic attribute is less than a first threshold; if it is determined that the similarity between the initial fault-dissolved body identification result and the sensitive seismic attribute is less than the first threshold, inputting the single-channel seismic data and the sensitive seismic attribute into a target identification model to obtain a target fault-dissolved body identification result.

[0227] To execute the above instructions more accurately, please refer to... Figure 16 As shown in the embodiments of this specification, another specific server is also provided, wherein the server includes a network communication port 1601, a processor 1602 and a memory 1603, and the above structures are connected by internal cables so that the various structures can perform specific data interaction.

[0228] Specifically, the network communication port 1601 can be used to acquire single-channel seismic data of the target area.

[0229] The processor 1602 is specifically used to extract amplitude features from the single-channel seismic data; wherein the single-channel seismic data includes multiple data points; based on the amplitude features, the data points are identified to obtain an initial fault-dissolved body identification result; the data points are subjected to dimensionality upscaling to obtain upscaled data; and the sensitive seismic attribute corresponding to the single-channel seismic data is determined based on the upscaled data; the similarity between the initial fault-dissolved body identification result and the sensitive seismic attribute is calculated, and it is detected whether the similarity between the initial fault-dissolved body identification result and the sensitive seismic attribute is less than a first threshold; if it is determined that the similarity between the initial fault-dissolved body identification result and the sensitive seismic attribute is less than the first threshold, the single-channel seismic data and the sensitive seismic attribute are input into the target identification model to obtain the target fault-dissolved body identification result.

[0230] The memory 1603 can be used to store the corresponding instruction program.

[0231] In this embodiment, the network communication port 1601 can be a virtual port bound to different communication protocols, thereby enabling the sending or receiving of different data. For example, the network communication port can be a port responsible for web data communication, a port responsible for FTP data communication, or a port responsible for email data communication. Furthermore, the network communication port can also be a physical communication interface or communication chip. For example, it can be a wireless mobile network communication chip, such as GSM or CDMA; it can also be a Wi-Fi chip; or it can be a Bluetooth chip.

[0232] In this embodiment, the processor 1602 can be implemented in any suitable manner. For example, the processor can take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers, and embedded microcontrollers, etc. This specification is not limiting.

[0233] In this embodiment, the memory 1603 may include multiple layers. In a digital system, anything that can store binary data can be a memory. In an integrated circuit, a circuit with storage function but no physical form is also called a memory, such as RAM, FIFO, etc. In a system, a storage device with a physical form is also called a memory, such as a memory stick, TF card, etc.

[0234] While this specification provides the steps of operation for the methods described in the embodiments or flowcharts, more or fewer steps may be included based on conventional or non-inventive means. The order of steps listed in the embodiments is merely one possible order of execution among many steps and does not represent the only possible order. In actual device or client product execution, the methods shown in the embodiments or drawings may be executed sequentially or in parallel (e.g., in a parallel processor or multi-threaded processing environment, or even a distributed data processing environment). The terms "comprising," "including," or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, product, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, product, or apparatus. Without further limitations, the presence of other identical or equivalent elements in a process, method, product, or apparatus that includes said elements is not excluded. The terms "first," "second," etc., are used to denote names and do not indicate any particular order.

[0235] Those skilled in the art will also know that, besides implementing the controller using purely computer-readable program code, the same functions can be achieved by logically programming the method steps, making the controller function as logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers (PLCs), and embedded microcontrollers. Therefore, such a controller can be considered a hardware component, and the devices within it used to implement various functions can also be considered structures within that hardware component. Alternatively, the devices used to implement various functions can be considered as both software modules implementing the method and structures within a hardware component.

[0236] This specification can be described in the general context of computer-executable instructions that are executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, classes, etc., that perform a specific task or implement a specific abstract data type. This specification can also be practiced in distributed computing environments, where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0237] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that this specification can be implemented by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solutions of this specification can essentially be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, mobile terminal, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments of this specification.

[0238] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on its differences from other embodiments. This specification can be used in numerous general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable electronic devices, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices, etc.

[0239] Although this specification has been described by way of examples, those skilled in the art will recognize that many variations and modifications are possible without departing from the spirit of this specification, and it is intended that the appended claims cover such variations and modifications without departing from the spirit of this specification.

Claims

1. A method for identifying broken solutions, characterized in that, include: Acquire single-channel seismic data for the target area; Amplitude features are extracted from the single-channel seismic data; wherein the single-channel seismic data includes multiple data points; Based on the amplitude characteristics, the data points are identified to obtain the initial fractured body identification result; The data points are subjected to dimensionality upscaling to obtain dimensionality upscaling data; and the sensitive seismic attributes corresponding to the single-channel seismic data are determined based on the dimensionality upscaling data. Calculate the similarity between the initial fault-dissolved body identification result and the sensitive seismic attribute, and detect whether the similarity between the initial fault-dissolved body identification result and the sensitive seismic attribute is less than a first threshold; If the similarity between the initial fault-dissolved body identification result and the sensitive seismic attribute is less than a first threshold, the single-channel seismic data and the sensitive seismic attribute are input into the target identification model to obtain the target fault-dissolved body identification result.

2. The method according to claim 1, characterized in that, The amplitude characteristics include root mean square amplitude and spectral decomposition information; The step of identifying the data points based on the amplitude characteristics to obtain the initial fractured body identification result includes: The first label value of the data point is generated based on the difference between the root mean square amplitude and the second threshold. The second label value of the data point is generated based on the difference between the spectral decomposition information and the third threshold. Detect whether the first tag value and the second tag value meet preset conditions; If the first tag value and the second tag value meet the preset conditions, the initial fracture body identification result of the data point is determined to be a fracture body.

3. The method according to claim 1, characterized in that, Based on the upgraded data, the sensitive seismic attributes corresponding to the single-channel seismic data are determined, including: Calculate the embedding distance between the up-dimensional data corresponding to multiple data points and the seed point; The embedding distance is normalized to obtain the normalized embedding distance results for multiple data points; By splicing the embedding distance normalization results corresponding to multiple data points, the sensitive seismic attributes corresponding to a single seismic data channel are obtained.

4. The method according to claim 3, characterized in that, Calculate the embedding distance between the upgraded data corresponding to multiple data points and the seed point, including: The embedding distance is determined using the following formula: Where i represents the data point number, x i Let d represent the data point, c represent the seed point, and d represent the seed point. c (x i ) represents x i The corresponding embedding distance, I(x) i ) represents x i The corresponding upgraded data, I(c) represents the upgraded data corresponding to c, x max I(x) represents the data point that differs most from c. max ) represents x max The corresponding upgraded data.

5. The method according to claim 1, characterized in that, The target recognition model is trained in the following manner: Obtain the training dataset and the validation dataset; An initial recognition model is trained using the training dataset to obtain an intermediate model; The validation dataset is input into the intermediate model to obtain intermediate results; The accuracy of the intermediate results is detected using a loss function to obtain a first detection result; Based on the first detection result, if the first detection result is determined to be passed, the intermediate model is used as the target recognition model.

6. The method according to claim 1, characterized in that, After inputting the single-channel seismic data and the sensitive seismic attributes into the target identification model to obtain the target fault-collapse identification result, the method further includes: Identify the drilling fluid loss area within the target region; The accuracy of the target fracture body identification result is tested to obtain a second detection result; Based on the second detection result, if the second detection result is passed, the target fracture body identification result is corrected using the drilling fluid leakage area to obtain the corrected target fracture body identification result.

7. The method according to claim 6, characterized in that, The accuracy of the target fracture body identification result is measured to obtain a second detection result, including: The target fractured body identification results are divided into target profile results and target planar results; The first profile result is extracted from the initial fractured body identification result; The second profile result is extracted from the aforementioned sensitive seismic attributes; Determine the similarity between the target profile result and the first profile result; and detect whether the similarity between the target profile result and the first profile result is less than a fifth threshold; If the similarity between the target profile result and the first profile result is less than a fifth threshold, the similarity between the target profile result and the second profile result is determined; and it is detected whether the similarity between the target profile result and the second profile result is less than a sixth threshold. If the similarity between the target profile result and the second profile result is less than a sixth threshold, the similarity between the target plane result and the coherent technology detection plane result is determined; and it is detected whether the similarity between the target plane result and the coherent technology detection plane result is less than a seventh threshold. If the similarity between the target plane result and the coherent technology detection plane result is less than the seventh threshold, the second detection result is determined to be passed.

8. The method according to claim 6, characterized in that, The target fractured body identification result is corrected using the drilling fluid leakage area to obtain a corrected target fractured body identification result, including: The first broken solution is extracted from the target broken solution identification results; The first fractured solution and the drilling fluid leakage area are intersected to obtain the target fractured solution. The target non-fractured solution is obtained by performing a difference set processing on the target fractured solution identification result and the target fractured solution; By combining the target broken solution and the target non-broken solution, a corrected target broken solution identification result is obtained.

9. A device for identifying broken solutions, characterized in that, include: The acquisition module is used to acquire single-channel seismic data for the target area; Amplitude features are extracted from the single-channel seismic data; wherein the single-channel seismic data includes multiple data points; The first identification module is used to identify the data points based on the amplitude characteristics to obtain the initial fracture body identification result; An extraction module is used to perform dimensionality upscaling on the data points to obtain upscaled data; and to determine the sensitive seismic attributes corresponding to the single-channel seismic data based on the upscaled data. The calculation module is used to calculate the similarity between the initial fault-dissolved body identification result and the sensitive seismic attribute, and to detect whether the similarity between the initial fault-dissolved body identification result and the sensitive seismic attribute is less than a first threshold. The second identification module is used to input the single-channel seismic data and the sensitive seismic attribute into the target identification model to obtain the target seismic identification result when the similarity between the initial fault-dissolved body identification result and the sensitive seismic attribute is less than a first threshold.

10. A computer-readable storage medium, characterized in that, It stores computer instructions that, when executed by a processor, implement the steps of the method according to any one of claims 1 to 8.

Images