A method for fine interpretation of fractured-cavity reservoirs near wells using acoustic long-range detection

Abstract

This invention discloses a method for fine interpretation of well-side fractured-vuggy reservoirs using acoustic long-range detection, relating to the field of oil and gas exploration. Specifically, the method includes: extracting three amplitude attributes—root mean square amplitude, instantaneous amplitude, and mean absolute amplitude—from acoustic long-range detection data to complete a preliminary characterization of the fractured-vuggy body; extracting coherent bodies, ant colony algorithms, and mathematical morphology attributes based on coherence analysis, ant colony algorithms, and morphological methods to enhance the continuity of the fractured-vuggy body and suppress imaging noise; employing principal component analysis to standardize multiple attributes and then selecting core principal components through cumulative variance thresholding to complete multi-attribute fusion and improve the accuracy of fractured-vuggy reservoir characterization; utilizing Canny and Sobel operators to achieve edge detection and refinement of the fractured-vuggy body, automatically dividing pores and fracture zones based on morphological features, and calculating geological parameters of the fracture zones through Hough transform.

Description

Technical Field

[0001] This invention relates to the field of oil and gas exploration, specifically to a method for fine interpretation of well-side fractured-cavity reservoirs using acoustic remote detection. Background Technology

[0002] Detailed interpretation of fractured-vuggy reservoirs near wells is a crucial aspect of carbonate reservoir evaluation. Acoustic long-range imaging technology is a powerful tool for detecting fractured-vuggy reservoirs near wells using reflected waves. Currently, there are two main types of interpretation methods:

[0003] (1) Direct interpretation by interpreters. Based on the experience of interpreters, continuous in-phase axes in the acoustic remote detection imaging are interpreted as crack zones, and irregular discontinuity distribution of in-phase axes is interpreted as holes.

[0004] (2) Acoustic remote detection attribute analysis method. Commonly used amplitude attributes (root mean square amplitude, instantaneous amplitude, mean absolute amplitude) describe crack characteristics. These attributes can not only filter the acoustic remote detection profile and improve the signal-to-noise ratio of the profile, but also highlight the crack morphology in the profile, making it easier for interpreters to identify cracks.

[0005] In technology (1), direct human interpretation is highly dependent on human experience, has strong subjectivity, poor consistency of results, and is prone to misjudgment and omission, making it impossible to achieve quantitative characterization of reservoirs.

[0006] In the application of technology (2), although traditional amplitude properties can effectively characterize the spatial morphological features of the well-side fracture cavity, it is difficult to achieve effective connection of the fracture body and enhance the reflection continuity. Summary of the Invention

[0007] The purpose of this invention is to address the above-mentioned shortcomings by proposing a method for fine interpretation of well-side fractured-cavity reservoirs using acoustic long-range detection.

[0008] The present invention specifically adopts the following technical solution:

[0009] A method for detailed interpretation of fractured-cavitary reservoirs near wells using acoustic long-range detection includes the following steps:

[0010] (1) Based on the acoustic remote detection data, extract three types of amplitude attributes: root mean square amplitude, instantaneous amplitude, and average absolute amplitude, and complete the preliminary characterization of the cavity.

[0011] (2) Based on coherence method, ant colony algorithm and mathematical morphology method, coherent body, ant body and morphological attributes are extracted to enhance the continuity of the slit body and suppress imaging noise.

[0012] (3) Principal component analysis is used to standardize multiple attributes and then screen core principal components by cumulative variance threshold to complete multi-attribute fusion and improve the accuracy of fissure reservoir characterization.

[0013] (4) The edge detection and refinement of the fracture zone are realized by using the Canny and Sobel operators, the pores and fracture zones are automatically divided based on morphological features, and the geological parameters of the fracture zone are calculated by Hough transform.

[0014] Preferably, in step (1), the root mean square amplitude is extracted by calculating the root mean square value of the amplitude at each sampling point within a fixed time window of the acoustic far-field detection signal. The calculation formula is as follows:

[0015] (1)

[0016] In the formula, A RMS Let a be the root mean square amplitude. i The amplitude values ​​of each sampling point within a fixed time window are defined, and n is the number of sampling points within the time window.

[0017] The formula for extracting instantaneous amplitude is:

[0018] (2)

[0019] In the formula, A IAH Let u(t) be the instantaneous amplitude based on the Hilbert transform, u(t) be the seismic signal, and q(t) be the result of the Hilbert transform of u(t), representing the orthogonal components of the seismic signal.

[0020] The formula for extracting the mean absolute amplitude is:

[0021] (3)

[0022] In the formula, A AA Let N be the average absolute amplitude, N be the total number of sampling points, and a be the amplitude of the sampling point.

[0023] Preferred,

[0024] The specific expression for the coherent volume in step (2) is:

[0025] (4)

[0026] In the formula, For coherent volume, n is the sampling point number, j is the depth channel number, N represents the number of sampling points within a specified time window, and J represents the number of depth points within the time window;

[0027] Ant body attributes effectively solve the discreteness problem of phase tracking through swarm intelligence optimization. Its calculation results show three major characteristics: the bifurcated network distribution of fault zones, fault location identification, and multi-scale synchronous identification of main faults and micro-cracks.

[0028] First, a high-value region in the data is selected using a thresholding method to determine the starting search position of the ant colony, as shown in the following formula:

[0029] (5)

[0030] In the formula, T init The initial threshold is usually chosen as 0.5;

[0031] Then, the ant explores its eight neighboring directions from its current position. When choosing a specific location to move to, it considers both pheromone concentration and heuristic information, and determines the final direction of movement through probability calculation, as shown in the following formula:

[0032] (6)

[0033] In the formula, Let be the probability that the Kth ant is at position (i,j). Let (i,j) be the pheromone concentration at position (i,j). This is heuristic information, usually the inverse of the path, where α represents the relative importance of the pheromone and β represents the relative importance of the heuristic factor. Let be the set of possible paths for the Kth ant to choose in its next step when it is at point i.

[0034] Secondly, after calculating the transition probabilities for all target points in the vicinity, the final direction of travel is determined by a roulette wheel method, as shown in the following formula:

[0035] (7)

[0036] In the formula, The target point chosen by the k-th ant for the next step is then randomly generated in the range [0,1]. By comparing the random number with q, the q that is greater than the random number and is the smallest in the set is selected as the final direction of movement.

[0037] Finally, ants release pheromones proportionally during their movement, which decay over time to suppress noise. Simultaneously, ants release pheromones along their path, and the amount of pheromones deposited is positively correlated with data values, continuously reinforcing key structures such as faults or cracks, as shown in the following formula:

[0038] (8)

[0039] In the formula, ρ is the pheromone evaporation rate, and N is the number of ants. This refers to the amount of pheromone deposited.

[0040] By utilizing morphological closing operations, noise elimination in acoustic long-range detection imaging is achieved, providing a clear direction for the interpretation of crack volumes. Specifically, this includes:

[0041] ① Erosion operation

[0042] Let A and B be two sets, where A represents background information and B represents the cavities / holes. If set A is eroded by set B, then the erosion operation is defined as follows:

[0043] (9)

[0044] In the formula, It is the erosion operator, where x is any element in the set and Φ is the empty set;

[0045] ② Dilation operation

[0046] The dilation operation is the opposite of the erosion operation. It can be used to find local maxima. Let A and B be two sets. If set A is dilated by set B, the dilation operation is defined as follows:

[0047] (10)

[0048] In the formula, ⊕ is the dilation operator;

[0049] ③ Closing operation

[0050] Closing operations, which combine dilation followed by erosion, effectively fill internal holes, smooth external boundaries, and connect minute gaps in an image while preserving the target's shape and position. This can be represented as:

[0051] (11).

[0052] Preferably, in step (3), it is assumed that the original data consists of N attribute bodies, which are combined into a new data body Array of size m*n, where m is the number of elements in the attribute body and n is the number of attribute bodies; each element in the matrix can be represented as x. ij Where i is the group number and j is the channel number, first, the combined data volume Array is Z-score normalized:

[0053] (12)

[0054] In the formula, μ j Let σ be the mean of the j-th attribute body. j Let z be the standard deviation of the j-th attribute. ij The standardized attribute data, where the mean of each attribute data is 0 and the variance is 1;

[0055] Then, using the standardized attribute volume, the covariance matrix R is calculated.

[0056] (13)

[0057] The calculated eigenvalues ​​are then processed according to... The order is rearranged, and the corresponding feature vector is ,in n new attribute volumes are formed by eigenvectors, i.e., principal components. :

[0058] (14)

[0059] In the formula, ,

[0060] In principal component Y j In, λ j Let λ represent the variance of the j-th attribute. A larger variance indicates a greater contribution of that component to the total variance, and more information is contained within the attribute body. Therefore, the eigenvalue λ can be used to determine the variance. j Calculate the importance c of the j-th principal component in the data volume. j :

[0061] (15)

[0062] In the formula, λ j The principal components have been arranged in descending order. The contribution rate satisfies c1≥c2≥···≥c n ,

[0063] By setting a cumulative variance threshold, the top k principal components are automatically selected for data fusion.

[0064] (16)

[0065] In the formula, Att pca This is the result of multi-attribute fusion.

[0066] Preferably, in step (4), assuming the seismic profile is 𝑃(𝑥,𝑦), Gaussian filtering is used to suppress high-frequency noise, expressed as:

[0067] (17)

[0068] In the formula, 𝐺aussian(𝑥,𝑦) is the Gaussian kernel function. For product operations, the σ value controls the smoothness.

[0069] Based on the edge detection results, the Sobel operator is used to refine the edges of the reflector. The principle is to calculate the gradients in both the horizontal and vertical directions. The larger the gradient magnitude, the more likely it is to be an edge on the same phase axis. The gradient value can be calculated as follows:

[0070] (18)

[0071] In the formula, It is the gradient value of the cave body. It is a horizontal gradient. It is the vertical gradient;

[0072] (19)

[0073] In the formula, S x For the horizontal gradient kernel, S y Using a vertical gradient kernel and a dual-threshold detection method, strong edges are directly preserved by setting high and low thresholds, while weak edges are only preserved if they are connected to strong edges; the rest are discarded.

[0074] Based on the different acoustic long-range detection imaging characteristics of the cavity, and using edge detection results, the acoustic long-range detection imaging results are divided into holes and cracks; holes appear as discrete points in the acoustic long-range detection imaging; cracks appear as strips; the specific division process is as follows:

[0075] ① Calculate the horizontal continuous length; the horizontal continuous length refers to the total number H of pixels (i,j) in the image that can be continuously connected to the left and right in the same row (horizontal direction) with a value of 0.

[0076] ② Calculate the vertical continuous length. The vertical continuous length refers to the total number V of pixels (i,j) in the image that can be continuously connected upwards and downwards in the same column (vertical direction) with a value of 0.

[0077] ③ If the ratio of the total vertical continuous length (V) to the total horizontal continuous length (H) of the reflector or the ratio of the total horizontal continuous length to the total vertical continuous length is greater than 1, the reflector is a crack zone and is marked as 1; otherwise, it is a hole and is marked as 0. For the crack zone model, the major and minor axes of the crack are obtained through Hough transform. The crack opening can be obtained through the ratio of the major and minor axes. The angle between the major axis and the positive y-axis is the tilt angle.

[0078] The present invention has the following beneficial effects:

[0079] This method, based on coherence theory and optimization theory, constructs coherent volume and ant-like properties for long-range acoustic detection, enabling fracture continuity detection and description for evaluating fracture effectiveness. Principal component analysis is used to achieve multi-attribute fusion of special properties from long-range acoustic detection, enabling fine characterization and identification of fractured-vuggy reservoirs near the well. Multi-step optimized edge detection using Canny and Sobel operators achieves high-precision boundary extraction of fractured-vuggy bodies, accurate classification of pores and fractures is achieved based on pixel distribution, and automated quantitative calculation of fracture opening and dip angle is realized using Hough transform. This provides a new method and workflow for fine interpretation of long-range acoustic detection and quantitative characterization of fractured-vuggy reservoirs, offering strong support for improving the efficiency of oil and gas reservoir exploration and development. Attached Figure Description

[0080] Figure 1 A flowchart for a detailed interpretation method of well-side fractured-cavity reservoirs using acoustic remote detection;

[0081] Figure 2 A schematic diagram illustrating the results of well acoustic long-range detection imaging interpretation. Detailed Implementation

[0082] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings and specific examples:

[0083] Combination Figure 1 A method for detailed interpretation of fractured-cavitary reservoirs near wells using acoustic long-range detection includes the following steps:

[0084] (1) Based on the acoustic remote detection data, extract three types of amplitude attributes: root mean square amplitude, instantaneous amplitude, and average absolute amplitude, and complete the preliminary characterization of the cavity.

[0085] (2) Based on coherence method, ant colony algorithm and mathematical morphology method, coherent body, ant body and morphological attributes are extracted to enhance the continuity of the slit body and suppress imaging noise.

[0086] (3) Principal component analysis is used to standardize multiple attributes and then screen core principal components by cumulative variance threshold to complete multi-attribute fusion and improve the accuracy of fissure reservoir characterization.

[0087] (4) The edge detection and refinement of the fracture zone are realized by using the Canny and Sobel operators, the pores and fracture zones are automatically divided based on morphological features, and the geological parameters of the fracture zone are calculated by Hough transform.

[0088] Preferably, in step (1), the root mean square amplitude is extracted by calculating the root mean square value of the amplitude at each sampling point within a fixed time window of the acoustic far-field detection signal. The calculation formula is as follows:

[0089] (1)

[0090] In the formula, A RMS Let a be the root mean square amplitude. i The amplitude values ​​of each sampling point within a fixed time window are defined, and n is the number of sampling points within the time window.

[0091] The formula for extracting instantaneous amplitude is:

[0092] (2)

[0093] In the formula, A IAH Let u(t) be the instantaneous amplitude based on the Hilbert transform, u(t) be the seismic signal, and q(t) be the result of the Hilbert transform of u(t), representing the orthogonal components of the seismic signal.

[0094] The formula for extracting the mean absolute amplitude is:

[0095] (3)

[0096] In the formula, A AA The average absolute amplitude is N, where N is the total number of sampling points, and a is the average absolute amplitude. i The amplitude of the sampling point.

[0097] Preferred,

[0098] The specific expression for the coherent volume in step (2) is:

[0099] (4)

[0100] In the formula, For coherent volume, n is the sampling point number, j is the depth channel number, N represents the number of sampling points within a specified time window, and J represents the number of depth points within the time window;

[0101] Ant body attributes effectively solve the discreteness problem of phase tracking through swarm intelligence optimization. Its calculation results show three major characteristics: the bifurcated network distribution of fault zones, fault location identification, and multi-scale synchronous identification of main faults and micro-cracks.

[0102] First, a high-value region in the data is selected using a thresholding method to determine the starting search position of the ant colony, as shown in the following formula:

[0103] (5)

[0104] In the formula, T init The initial threshold is usually chosen as 0.5;

[0105] Then, the ant explores its eight neighboring directions from its current position. When choosing a specific location to move to, it considers both pheromone concentration and heuristic information, and determines the final direction of movement through probability calculation, as shown in the following formula:

[0106] (6)

[0107] In the formula, Let be the probability that the Kth ant is at position (i,j). Let (i,j) be the pheromone concentration at position (i,j). This is heuristic information, usually the inverse of the path, where α represents the relative importance of the pheromone and β represents the relative importance of the heuristic factor. Let be the set of possible paths for the Kth ant to choose in its next step when it is at point i.

[0108] Secondly, after calculating the transition probabilities for all target points in the vicinity, the final direction of travel is determined by a roulette wheel method, as shown in the following formula:

[0109] (7)

[0110] In the formula, The target point chosen by the k-th ant for the next step is then randomly generated in the range [0,1]. By comparing the random number with q, the q that is greater than the random number and is the smallest in the set is selected as the final direction of movement.

[0111] Finally, ants release pheromones proportionally during their movement, which decay over time to suppress noise. Simultaneously, ants release pheromones along their path, and the amount of pheromones deposited is positively correlated with data values, continuously reinforcing key structures such as faults or cracks, as shown in the following formula:

[0112] (8)

[0113] In the formula, ρ is the pheromone evaporation rate, and N is the number of ants. This refers to the amount of pheromone deposited.

[0114] By utilizing morphological closing operations, noise elimination in acoustic long-range detection imaging is achieved, providing a clear direction for the interpretation of crack volumes. Specifically, this includes:

[0115] ① Erosion operation

[0116] Let A and B be two sets, where A represents background information and B represents the cavities / holes. If set A is eroded by set B, then the erosion operation is defined as follows:

[0117] (9)

[0118] In the formula, It is the erosion operator, where x is any element in the set and Φ is the empty set;

[0119] ② Dilation operation

[0120] The dilation operation is the opposite of the erosion operation. It can be used to find local maxima. Let A and B be two sets. If set A is dilated by set B, the dilation operation is defined as follows:

[0121] (10)

[0122] In the formula, ⊕ is the dilation operator;

[0123] ③ Closing operation

[0124] Closing operations, which combine dilation followed by erosion, effectively fill internal holes, smooth external boundaries, and connect minute gaps in an image while preserving the target's shape and position. This can be represented as:

[0125] (11).

[0126] In step (3), it is assumed that the original data consists of N attribute bodies, which are combined into a new data body Array of size m*n, where m is the number of elements in the attribute body and n is the number of attribute bodies; each element in the matrix can be represented as x. ij Where i is the group number and j is the channel number, first, the combined data volume Array is Z-score normalized:

[0127] (12)

[0128] In the formula, μ j Let σ be the mean of the j-th attribute body. j Let z be the standard deviation of the j-th attribute. ij The standardized attribute data, where the mean of each attribute data is 0 and the variance is 1;

[0129] Then, using the standardized attribute volume, the covariance matrix R is calculated.

[0130] (13)

[0131] The calculated eigenvalues ​​are then processed according to... The order is rearranged, and the corresponding feature vector is ,in n new attribute volumes are formed by eigenvectors, i.e., principal components. :

[0132] (14)

[0133] In the formula, ,

[0134] In principal component Y j In, λ j Let λ represent the variance of the j-th attribute. A larger variance indicates a greater contribution of that component to the total variance, and more information is contained within the attribute body. Therefore, the eigenvalue λ can be used to determine the variance. j Calculate the importance c of the j-th principal component in the data volume. j :

[0135] (15)

[0136] In the formula, λ j The principal components have been arranged in descending order. The contribution rate satisfies c1≥c2≥···≥c n ,

[0137] By setting a cumulative variance threshold, the top k principal components are automatically selected for data fusion.

[0138] (16)

[0139] In the formula, Att pca This is the result of multi-attribute fusion.

[0140] In step (4), assuming the seismic profile is 𝑃(𝑥,𝑦), Gaussian filtering is used to suppress high-frequency noise, expressed as:

[0141] (17)

[0142] In the formula, 𝐺aussian(𝑥,𝑦) is the Gaussian kernel function. For product operations, the σ value controls the smoothness.

[0143] Based on the edge detection results, the Sobel operator is used to refine the edges of the reflector. The principle is to calculate the gradients in both the horizontal and vertical directions. The larger the gradient magnitude, the more likely it is to be an edge on the same phase axis. The gradient value can be calculated as follows:

[0144] (18)

[0145] In the formula, 𝐶𝑎𝑣𝑒 𝐺 It is the gradient value of the cave body. It is a horizontal gradient. It is the vertical gradient;

[0146] (19)

[0147] In the formula, S xFor the horizontal gradient kernel, S y Using a vertical gradient kernel and a dual-threshold detection method, strong edges are directly preserved by setting high and low thresholds, while weak edges are only preserved if they are connected to strong edges; the rest are discarded.

[0148] Based on the different acoustic long-range detection imaging characteristics of the cavity, and using edge detection results, the acoustic long-range detection imaging results are divided into holes and cracks; holes appear as discrete points in the acoustic long-range detection imaging; cracks appear as strips; the specific division process is as follows:

[0149] ① Calculate the horizontal continuous length; the horizontal continuous length refers to the total number H of “graphic pixels (with a value of 0)” that can be continuously connected to the left and right in the same row (horizontal direction) of the image when pixels (i,j) are in the same row;

[0150] ② Calculate the vertical continuous length. The vertical continuous length refers to the total number V of "graphic pixels (with a value of 0)" that can be continuously connected upwards and downwards in the same column (vertical direction) of the image when the pixel points (i,j) are in the same column.

[0151] ③ If the ratio of the total vertical continuous length (V) to the total horizontal continuous length (H) of the reflector or the ratio of the total horizontal continuous length to the total vertical continuous length is greater than 1, the reflector is a crack zone and is marked as 1; otherwise, it is a hole and is marked as 0. For the crack zone model, the major and minor axes of the crack are obtained through Hough transform. The crack opening can be obtained through the ratio of the major and minor axes. The angle between the major axis and the positive y-axis is the tilt angle.

[0152] The Ordovician carbonate strata in the Tarim Basin are the primary targets for oil and gas exploration. Outcrop observations indicate that the Ordovician carbonate reservoirs are predominantly characterized by fractures and cavities. Beneath the outcrops, the reservoir exhibits fractures and pores, providing space for oil and gas storage. To effectively evaluate and characterize this reservoir, in-well acoustic remote sensing imaging technology was designed for this area.

[0153] Well A is an important exploration well in the area, and acoustic remote sensing data was acquired using the XMAC-F1 instrument. Using a series of techniques and steps proposed in this invention, the data was processed to obtain a well-side acoustic remote sensing imaging profile. Figure 2 The third channel shows the acoustic long-range imaging profile of well A at 0°, and the fourth channel shows the fracture identification results after multi-attribute fusion. It can be seen that there is a phase axis approximately 20m long in the 6160-6185m interval, and the fracture dip angle is calculated to be approximately 15° using Hough transform. Numerous blocky phase axes exist in the 6190-6205m depth range, indicating that pores also develop within the reservoir.

[0154] The application of this technology in the actual work of identifying and finely characterizing fractures and cavities in Well A has achieved good results, providing strong technical support for oil and gas exploration and reservoir location, which is of great significance.

[0155] Of course, the above description is not intended to limit the present invention, and the present invention is not limited to the examples given above. Any changes, modifications, additions or substitutions made by those skilled in the art within the scope of the present invention should also fall within the protection scope of the present invention.

Claims

1. A method for detailed interpretation of fractured-cavitary reservoirs near wells using acoustic long-range detection, characterized in that, Includes the following steps: (1) Based on the acoustic wave remote detection data, extract three types of amplitude attributes: root mean square amplitude, instantaneous amplitude, and mean absolute amplitude, and complete the preliminary characterization of the crevice body; (2) Based on coherence method, ant colony algorithm and mathematical morphology method, extract coherent body, ant body and morphological attributes to enhance the continuity of the slit body and suppress imaging noise; (3) Principal component analysis is used to standardize multiple attributes and then the core principal components are screened by the cumulative variance threshold to complete the fusion of multiple attributes and improve the accuracy of the characterization of the fracture-cavity reservoir. (4) The edge detection and refinement of the fracture zone are realized by using the Canny and Sobel operators, the pores and fracture zones are automatically divided based on morphological features, and the geological parameters of the fracture zone are calculated by Hough transform.

2. The method for detailed interpretation of well-side fractured-cavity reservoirs using acoustic long-range detection as described in claim 1, characterized in that, In step (1), the root mean square amplitude is extracted by calculating the root mean square value of the amplitude at each sampling point within a fixed time window of the acoustic far-field detection signal. The calculation formula is as follows: （1） In the formula, A RMS Let a be the root mean square amplitude. i The amplitude values ​​of each sampling point within a fixed time window are defined, and n is the number of sampling points within the time window. The formula for extracting instantaneous amplitude is: （2） In the formula, A IAH Let u(t) be the instantaneous amplitude based on the Hilbert transform, u(t) be the seismic signal, and q(t) be the result of the Hilbert transform of u(t), representing the orthogonal components of the seismic signal. The formula for extracting the mean absolute amplitude is: （3） In the formula, A AA The average absolute amplitude is N, where N is the total number of sampling points, and a is the average absolute amplitude. i This represents the amplitude values ​​at each sampling point within a fixed time window.

3. The method for detailed interpretation of well-side fractured-cavity reservoirs using acoustic long-range detection as described in claim 1, characterized in that, The specific expression for the coherent volume in step (2) is: （4） In the formula, For coherent volume, n is the sampling point number, j is the depth channel number, N represents the number of sampling points within a specified time window, and J represents the number of depth points within the time window; Ant body attributes effectively solve the discreteness problem of phase tracking through swarm intelligence optimization. Its calculation results show three major characteristics: the bifurcated network distribution of fault zones, fault location identification, and multi-scale synchronous identification of main faults and micro-cracks. First, a high-value region in the data is selected using a thresholding method to determine the starting search position of the ant colony, as shown in the following formula: （5） In the formula, T init The initial threshold is usually chosen as 0.5; Then, the ant explores its eight neighboring directions from its current position. When choosing a specific location to move to, it considers both pheromone concentration and heuristic information, and determines the final direction of movement through probability calculation, as shown in the following formula: （6） In the formula, Let be the probability that the Kth ant is at position (i,j). Let (i,j) be the pheromone concentration at position (i,j). This is heuristic information, usually the inverse of the path, where α represents the relative importance of the pheromone and β represents the relative importance of the heuristic factor. Let be the set of possible paths for the Kth ant to choose in its next step when it is at point i. Secondly, after calculating the transition probabilities for all target points in the vicinity, the final direction of travel is determined by a roulette wheel method, as shown in the following formula: （7） In the formula, The target point chosen by the k-th ant for the next step is then randomly generated in the range [0,1]. By comparing the random number with q, the q that is greater than the random number and is the smallest in the set is selected as the final direction of movement. Finally, ants release pheromones proportionally during their movement, which decay over time to suppress noise. Simultaneously, ants release pheromones along their path, and the amount of pheromones deposited is positively correlated with data values, continuously reinforcing key structures such as faults or cracks, as shown in the following formula: （8） In the formula, ρ is the pheromone evaporation rate, and N is the number of ants. This refers to the amount of pheromone deposited. By utilizing morphological closing operations, noise elimination in acoustic long-range detection imaging is achieved, providing a clear direction for the interpretation of crack volumes. Specifically, this includes: ① Erosion operation Let A and B be two sets, where A represents background information and B represents the cavities / holes. If set A is eroded by set B, then the erosion operation is defined as follows: （9） In the formula, It is the erosion operator, where x is any element in the set and Φ is the empty set; ② Dilation operation The dilation operation is the opposite of the erosion operation. It can be used to find local maxima. Let A and B be two sets. If set A is dilated by set B, the dilation operation is defined as follows: （10） In the formula, ⊕ is the dilation operator; ③ Closing operation Closing operations, which combine dilation followed by erosion, effectively fill internal holes, smooth external boundaries, and connect minute gaps in an image while preserving the target's shape and position. This can be represented as: （11）。 4. The method for detailed interpretation of well-side fractured-cavity reservoirs using acoustic long-range detection as described in claim 1, characterized in that, In step (3), it is assumed that the original data consists of N attribute bodies, which are combined into a new data body Array of size m*n, where m is the number of elements in the attribute body and n is the number of attribute bodies; each element in the matrix can be represented as x. ij Where i is the group number and j is the channel number, first, the combined data volume Array is Z-score normalized: （12） In the formula, μ j Let σ be the mean of the j-th attribute body. j Let z be the standard deviation of the j-th attribute. ij The standardized attribute data, where the mean of each attribute data is 0 and the variance is 1; Then, using the standardized attribute volume, the covariance matrix R is calculated. （13） The calculated eigenvalues ​​are then processed according to... The order is rearranged, and the corresponding feature vector is ,in n new attribute volumes are formed by eigenvectors, i.e., principal components. : （14） In the formula, , In principal component Y j In, λ j Let λ represent the variance of the j-th attribute. A larger variance indicates a greater contribution of that component to the total variance, and more information is contained within the attribute body. Therefore, the eigenvalue λ can be used to determine the variance. j Calculate the importance c of the j-th principal component in the data volume. j : （15） In the formula, λ j The principal components have been arranged in descending order. The contribution rate satisfies c1≥c2≥···≥c n , By setting a cumulative variance threshold, the top k principal components are automatically selected for data fusion. （16） In the formula, Att pca This is the result of multi-attribute fusion for long-range acoustic detection.

5. The method for detailed interpretation of well-side fractured-cavity reservoirs using acoustic long-range detection as described in claim 1, characterized in that, In step (4), assuming the seismic profile is 𝑃(𝑥,𝑦), Gaussian filtering is used to suppress high-frequency noise, expressed as: （17） In the formula, 𝐺aussian(𝑥,𝑦) is the Gaussian kernel function. For product operations, the σ value controls the smoothness. Based on the edge detection results, the Sobel operator is used to refine the edges of the reflector. The principle is to calculate the gradients in both the horizontal and vertical directions. The larger the gradient magnitude, the more likely it is to be an edge on the same phase axis. The gradient value can be calculated as follows: （18） In the formula, It is the gradient value of the cave body. It is a horizontal gradient. It is the vertical gradient; （19） In the formula, S x For the horizontal gradient kernel, S y Using a vertical gradient kernel and a dual-threshold detection method, strong edges are directly preserved by setting high and low thresholds, while weak edges are only preserved if they are connected to strong edges; the rest are discarded. Based on the different acoustic long-range detection imaging characteristics of the cavity, and using edge detection results, the acoustic long-range detection imaging results are divided into holes and cracks; holes appear as discrete points in the acoustic long-range detection imaging; cracks appear as strips; the specific division process is as follows: ① Calculate the horizontal continuous length; the horizontal continuous length refers to the total number H of pixels (i,j) in the image that can be continuously connected to the left and right in the same row (horizontal direction) with a value of 0. ② Calculate the vertical continuous length. The vertical continuous length refers to the total number V of pixels (i,j) in the image that can be continuously connected upwards and downwards in the same column (vertical direction) with a value of 0. ③ If the ratio of the total vertical continuous length (V) to the total horizontal continuous length (H) of the reflector or the ratio of the total horizontal continuous length to the total vertical continuous length is greater than 1, it indicates that the reflector is a crack zone and is marked as 1; Otherwise, it is a hole and marked as 0; for the crack zone model, the major and minor axes of the crack are obtained through Hough transform, and the crack opening can be obtained by the ratio of the major and minor axes. The angle between the major axis and the positive y-axis is the dip angle.

Images