A method for predicting fracture based on morphological analysis and signal processing

Abstract

The present application belongs to the field of geophysics and petroleum engineering technology, and specifically discloses a fracture prediction method based on morphological analysis and signal processing, comprising: S1, performing USM sharpening processing on post-stack seismic data to obtain sharpened data; S2, performing pearson-based similarity calculation on the USM sharpened data as input to obtain similarity data; S3, performing histogram equalization operation on the similarity data as input to obtain equalization data; S4, performing regional fusion operation processing on the equalization data to obtain regional fusion data; and S5, performing skeleton extraction processing on the regional fusion data to obtain fracture information. The present application has no limitations in application range, and fracture information with high resolution, large number of recognitions, strong contrast, good continuity and high fidelity can be obtained by processing post-stack seismic data, the characteristics of small fractures are accurately described, necessary basis is provided for well location design, horizontal well steering, fracturing design and the like, and the present application is suitable for prediction and interpretation of various fractures.

Description

Technical Field

[0001] This invention belongs to the fields of geophysics and petroleum engineering technology, and relates to a fracture prediction method, specifically a fracture prediction method based on morphological analysis and signal processing. Background Technology

[0002] A fracture is a rupture surface formed in underground rocks under geological stress. Based on its fracture morphology, fractures are classified into different scales and orientations. Fracture interpretation is fundamental to oil and gas exploration and is widely used in structural geology, petroleum geology, engineering geology, active earthquake prediction, and other Earth science-related research and production tasks. Because the rationality and accuracy of fracture interpretation results directly affect subsequent structural modeling, oil and gas reservoir prediction, and oil and gas resource evaluation, the seismic exploration industry has long been committed to improving the accuracy and efficiency of fracture interpretation. In recent years, the industry has successively developed fault identification technologies based on seismic attributes, image analysis, and deep learning. By analyzing seismic parameters sensitive to fractures, such as seismic waveforms, amplitudes, and frequencies, these technologies highlight the discontinuities in seismic reflection interfaces caused by faults, thereby continuously improving the efficiency and accuracy of fracture interpretation.

[0003] Currently, commonly used fault prediction methods include the Likelihood algorithm and the Ant Colony algorithm. The Likelihood algorithm processes extracted Fault Likelihood (FL) attributes for fracture prediction and 3D geological modeling, suitable for the exploration and development of tight sandstone gas reservoirs, and can provide spatial classification of fault development zones. The Ant Colony algorithm is a method for identifying geological fault layers using ant colony algorithms, mainly used for the identification and location of seismic fault layers. However, using these methods to predict faults, especially micro-fractures, in seismic data has many limitations in terms of prediction technology and application. For example, the Likelihood algorithm's prediction results show indistinct local micro-fracture features, while the Ant Colony algorithm's prediction results suffer from low resolution and a limited number of faults. Furthermore, the applicability of these existing prediction techniques varies, and they may not be suitable for predicting local data. Therefore, relying on only one fault prediction method yields very limited fault information. In summary, existing fault prediction techniques using post-stack seismic data generally suffer from a small number of faults, low resolution, and poor continuity and fidelity, making it difficult to meet the needs of high-precision fault prediction and limiting the efficiency and accuracy of subsequent tectonic interpretation. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention aims to provide a fracture prediction method based on morphological analysis and signal processing. This method identifies fracture information with high resolution, large number of fractures, strong contrast, good continuity, and high fidelity, thereby accurately depicting the characteristics of minute fractures and providing necessary basis for well location design, horizontal well guidance, fracturing design, and other related tasks.

[0005] To achieve the above objectives, the technical method adopted by this invention is as follows: A fracture prediction method based on morphological analysis and signal processing, comprising the following steps:

[0006] S1. The post-stack seismic data is subjected to USM sharpening to enhance the high-frequency content of the seismic data and highlight small faults, resulting in sharpened data.

[0007] S2. Using the USM sharpened data as input, perform Pearson-based similarity calculations to obtain similarity data;

[0008] S3. Using similarity data as input, perform histogram equalization. Through histogram equalization, weak signal enhancement is performed on the fracture information in the fracture data to obtain equalized data.

[0009] S4. Perform regional fusion processing on the balanced data to obtain regional fused data;

[0010] S5. Perform skeleton extraction processing on the regional fusion data to obtain fracture information.

[0011] As a limitation, the implementation steps of USM sharpening in S1 are as follows:

[0012] S11. Low-pass filtering: The original seismic data is filtered using a Gaussian filter to obtain low-frequency component seismic data.

[0013] S12. Calculate the high-frequency component, subtract the low-frequency component seismic data from the original seismic data to obtain the high-frequency component seismic data, which contains detailed information of the original seismic data.

[0014] S13. Amplify high-frequency components by multiplying the high-frequency seismic data by a coefficient greater than 1 to amplify detailed information.

[0015] S14. Superimpose high-frequency components. Finally, superimpose the amplified high-frequency component seismic data onto the original seismic data to obtain the sharpened seismic data, also known as the sharpened data.

[0016] As a limitation, in S2, the Pearson similarity calculation measures the degree of linear correlation between two sets of continuous variables. Similarity calculations are performed on each sample point and its surrounding data. Based on the obtained waveform similarity features, corresponding fault attribute data is generated. During the calculation process, the size of the unit calculation window and the size of the migration window are determined according to the characteristics of the actual post-stack seismic data, generating similarity data of faults covering the range of the post-stack seismic data. The Pearson similarity calculation formula is as follows:

[0017]

[0018] In the formula: R xy Let be the Pearson correlation coefficient between variables x and y, where n is the number of seismic data samples, and x is the number of samples. i Let y be the value of the i-th sample point of x. i Let be the value of the i-th sample point of y.

[0019] As a limitation, the implementation process of histogram equalization in S3 includes:

[0020] S31. Count the number of pixels n at each gray level of the original input image. i Where i = 0, 1, ..., L-1; L is the total number of gray levels;

[0021] S32. Calculate the histogram of the original image, i.e., the probability density of each gray level, P. i (r i ) = n i / n, where n is the total number of pixels in the original image;

[0022] S33. Calculate the cumulative distribution function s k (r k ), Where k = 0, 1, ..., L-1;

[0023] S34. Calculate the final output gray level g. k ,

[0024] g k =INT[(g max -g min )s k (r k )+g min +0.5] / (L-1), where k = 0, 1, ..., L-1;

[0025] In this expression, INT[] is the integer operator, let g min =0, g max =L-1, then this calculation formula simplifies to g k =INT[(L-1)sk (r k )+0.5] / (L-1);

[0026] S35. Use the gray level function f of the original image. k and the g obtained above k By modifying the gray levels of the original image according to the mapping relationship, the histogram is obtained as an approximately uniformly distributed output image, thus obtaining balanced data.

[0027] As a limitation, the steps of the region fusion operation in S4 include:

[0028] S41. Determine the uniformity merging criterion and divide the image into a set of arbitrary non-overlapping initial regions, typically 4 parts;

[0029] S42. Division: For any region, if it does not meet the criteria, then divide it into 4 equal parts.

[0030] S43. Repeat step S42 until all segmented regions meet the criteria.

[0031] S44. Merge: For two adjacent regions, if the new region obtained by merging these two regions meets the criteria, then the two regions are merged.

[0032] S45. Repeat S44 until all merges are completed, generating regional merged data.

[0033] As a limitation, the skeleton extraction algorithm in S5 adopts the Khalid Sheed algorithm, and the specific steps are as follows:

[0034] S51. Mark the boundaries of the image;

[0035] S52. If a point has three non-zero neighboring points, delete the point.

[0036] S53. If a point has 3 or 4 non-zero neighboring points, delete the point.

[0037] S54. If a point has 3, 4, or 5 non-zero neighbors in its neighborhood, delete the point.

[0038] S55. If there are 3, 4, 5, or 6 non-zero points adjacent to this point in its neighborhood, delete this point.

[0039] S56. If there are 3, 4, 5, 6, or 7 non-zero points adjacent to this point in its neighborhood, delete this point.

[0040] S57. Unmark the remaining boundary points. If no points have been modified in S56, stop the iteration; otherwise, return to S51.

[0041] Due to the adoption of the above technical solutions, the beneficial effects achieved by this invention compared with the prior art are as follows: This invention processes post-stack seismic data through steps such as USM sharpening, Pearson-based similarity calculation, histogram equalization, region fusion, and skeleton extraction. Through morphological analysis and signal processing, it ultimately obtains fracture information with high resolution, a large number of identified fractures, strong contrast, good continuity, and high fidelity, and can accurately characterize the features of minute fractures. Compared with the ant body algorithm, the fracture prediction results of this invention can obtain clearer fault morphology, and compared with the Likelihood algorithm, the features of local minute fractures are more obvious. This invention provides strong support for improving the accuracy and efficiency of minute fracture interpretation, ensuring the efficient completion of three-dimensional structural interpretation work, providing necessary basis for well location design, horizontal well guidance, fracturing design, etc., and is applicable to the prediction and interpretation of various types of fractures. Attached Figure Description

[0042] Figure 1 This is a flowchart of the process in Embodiment 1 of the present invention;

[0043] Figure 2 This is the initial post-stack seismic data map in Embodiment 2 of the present invention;

[0044] Figure 3 This is a data image after USM sharpening processing in step S1 of Embodiment 2 of the present invention;

[0045] Figure 4 This refers to the similarity data graph obtained in step S2 of Embodiment 2 of the present invention;

[0046] Figure 5 This is the histogram equalization data obtained in step S3 of Embodiment 2 of the present invention;

[0047] Figure 6 This is the region fusion data map obtained in step S4 of Embodiment 2 of the present invention;

[0048] Figure 7 This is a diagram of the skeleton extraction data obtained in step S5 of embodiment 2 of the present invention;

[0049] Figure 8 This is a data graph obtained using the ant colony algorithm in Embodiment 2 of the present invention;

[0050] Figure 9 This is a data graph obtained using the Likelihood algorithm in Embodiment 2 of the present invention. Detailed Implementation

[0051] The preferred embodiments of the present invention will now be described with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustrative and explanatory purposes only and do not constitute a limitation thereof.

[0052] Example 1

[0053] This embodiment discloses a fracture prediction method based on morphological analysis and signal processing, such as Figure 1 As shown, the specific steps include:

[0054] S1. Perform Unsharp Mask (USM) sharpening on the post-stack seismic data to enhance the high-frequency components and highlight minute faults, resulting in sharpened data. This sharpening method involves first applying a Gaussian blur to the seismic data, then multiplying the blurred data by a specific coefficient. USM-based sharpening can remove some minor noise, producing more realistic and reliable results than simply using convolutional sharpening operators. Specifically, this step includes:

[0055] S11. Low-pass filtering: The original seismic data is filtered using a Gaussian filter to obtain low-frequency component seismic data.

[0056] S12. Calculate the high-frequency component by subtracting the low-frequency component seismic data from the original seismic data to obtain the high-frequency component seismic data. This high-frequency component seismic data contains detailed information from the original seismic data.

[0057] S13. Amplify high-frequency components by multiplying the high-frequency seismic data by a coefficient greater than 1 to amplify detailed information.

[0058] S14. Superimpose high-frequency components. Finally, superimpose the amplified high-frequency component seismic data onto the original seismic data to obtain the sharpened seismic data, also known as the sharpened data.

[0059] S2. Using USM sharpened data as input, perform Pearson-based similarity calculations to obtain similarity data. The Pearson-based similarity calculation method measures the degree of linear correlation between two sets of continuous variables. It calculates the similarity between each sample point and its surrounding data, obtaining waveform similarity features and forming corresponding fault attribute data. During the calculation, the size of the unit calculation window and the size of the migration window need to be determined based on the characteristics of the actual post-stack seismic data. The optimized parameters are used to generate fault data covering the range of the post-stack seismic data. The Pearson calculation formula is as follows:

[0060]

[0061] In the formula: R xy Let be the Pearson correlation coefficient between variables x and y, where n is the number of seismic data samples, and x is the number of samples. i Let y be the value of the i-th sample point of x. i Let be the value of the i-th sample point of y.

[0062] S3. Using similarity data as input, perform histogram equalization. Through histogram equalization, weak signal enhancement is applied to the fracture information in the fracture data to obtain equalized data. To further improve the resolution and contrast of fracture information, fracture data is used as input for histogram equalization. Histogram equalization is a method of adjusting contrast using the statistical histogram of seismic data. This method allows amplitudes to be better distributed on the histogram. This can be used to enhance local contrast without affecting overall contrast, especially when the contrast of the data is quite similar. In other words, through histogram equalization, weak signal enhancement of fracture information in the fracture data volume can be applied to highlight fracture information. Specifically, the implementation process of histogram equalization in this embodiment includes:

[0063] S31. Count the number of pixels n at each gray level of the original input image. i Where i = 0, 1, ..., L-1, and L is the total number of gray levels;

[0064] S32. Calculate the histogram of the original image, i.e., the probability density of each gray level, P. i (r i ) = n i / n, where n is the total number of pixels in the original image;

[0065] S33. Calculate the cumulative distribution function s k (r k ), Where k = 0, 1, ..., L-1;

[0066] S34. Calculate the final output gray level g. k ,

[0067] g k =INT[(g max -g min )s k (r k )+g min +0.5] / (L-1), where k = 0, 1, ..., L-1;

[0068] In this expression, INT[] is the integer operator, let g min =0, g max =L-1, then this calculation formula simplifies to g k =INT[(L-1)s k (r k )+0.5] / (L-1);

[0069] S35. Use the gray level function f of the original image. k and the g obtained above k By modifying the gray levels of the original image according to the mapping relationship, the histogram is obtained as an approximately uniformly distributed output image, thus obtaining balanced data.

[0070] S4. Perform region merging on the balanced data to obtain region-merged data. To further improve the continuity of fracture information and reduce invalid non-fracture information, the "histogram equalization" data is used as input for region merging. Region merging utilizes the hierarchical concept of the pyramid or quadtree data structure of image data to divide the image into a set of arbitrarily disjoint initial regions. That is, it can start from this pyramid or quadtree data structure of the image, splitting and merging these regions according to a given uniformity detection criterion, gradually improving the performance of region division until the image is finally divided into the fewest uniform regions. The newly generated region-merged data can effectively identify the true extent of continuous fractures, fill in the missing content at both ends of the fracture information, and reduce invalid non-fracture information. The specific algorithm steps are as follows:

[0071] S41. Determine the uniformity merging criterion and divide the image into a set of four arbitrary non-overlapping initial regions;

[0072] S42. Division: For any region, if it does not meet the criteria, then divide it into 4 equal parts.

[0073] S43. Repeat step S42 until all segmented regions meet the criteria.

[0074] S44. Merge: For two adjacent regions, if the new region obtained by merging these two regions meets the criteria, then the two regions are merged.

[0075] S45. Repeat S44 until all merges are completed, generating regional merged data.

[0076] S5. The region fusion data is processed for skeleton extraction to obtain fracture information, making the fracture recognition results clearer. The skeleton extraction algorithm uses the Khalid Sheed algorithm, also known as the K3M algorithm, which is a thinning algorithm for image skeleton extraction. This algorithm belongs to the class of iterative boundary erosion algorithms. Its basic idea is to iteratively erode the boundaries of objects in the binary image, gradually thinning the image while ensuring that points that meet certain conditions are retained or deleted, ultimately obtaining the image skeleton. The steps are as follows:

[0077] S51. Mark the boundaries of the image;

[0078] S52. If a point has three non-zero neighboring points, delete the point.

[0079] S53. If there are 3 or 4 non-zero points adjacent to the point in its neighborhood, delete the point.

[0080] S54. If a point has 3, 4, or 5 non-zero neighbors in its neighborhood, delete the point.

[0081] S55. If a point has 3, 4, 5, or 6 non-zero neighboring points, delete the point.

[0082] S56. If there are 3, 4, 5, 6, or 7 non-zero points adjacent to a point in its neighborhood, delete the point.

[0083] S57. Unmark the remaining boundary points. If no points have been modified in S56, stop the iteration; otherwise, return to S51.

[0084] This embodiment provides a fracture prediction method based on morphological analysis and signal processing. The initial post-stack seismic data is processed sequentially through USM sharpening, Pearson similarity calculation, histogram equalization, region fusion, and skeleton extraction. The final result is fracture information with high resolution, large number of identified fractures, strong contrast, good continuity, and high fidelity. It can accurately characterize the features of small fractures and provide necessary basis for well location design, horizontal well guidance, and fracturing design.

[0085] Example 2

[0086] Based on Example 1, this example illustrates a further application of a fracture prediction method based on morphological analysis and signal processing. Applying this invention to the Southwest Oil and Gas Field provides strong support for oil companies to improve the accuracy and efficiency of interpreting micro-fractures, ensures the efficient completion of three-dimensional structural interpretation work, and provides necessary basis for well location design, horizontal well guidance, fracturing design, etc.

[0087] like Figure 2 The image shows the initial state of the post-stack seismic data, which is then processed according to the workflow of this invention:

[0088] S1. Perform USM sharpening to enhance the high-frequency components of the seismic data, highlight minor fractures, and generate new sharpened data. Figure 3 );

[0089] S2. Using the USM sharpened data as input, perform Pearson-based similarity calculations to obtain similarity data. Figure 4 );

[0090] S3. Using similarity data as input, perform histogram equalization. Through histogram equalization, weak signal enhancement is applied to the fracture information in the fracture data to highlight the fracture information, resulting in balanced data. Figure 5 );

[0091] S4. Perform region merging on the balanced data to generate newly generated region merging data. Figure 6 It can effectively identify the true extent of continuous fractures, fill in the missing information at both ends of the fracture information, and reduce invalid non-fracture information.

[0092] S5. Perform skeleton extraction processing on the region fusion data to make the fracture recognition results clearer. Figure 7 Finally, the fracture information was obtained.

[0093] like Figure 8 , Figure 9 The images shown are fracture information data obtained using the ant body algorithm and the Likelihood algorithm, respectively. Compared with the ant body algorithm, the present invention can obtain a clearer fault morphology, and compared with the Likelihood algorithm, the characteristics of local micro-fractures are more obvious.

[0094] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this disclosure, and are not intended to limit them. Although this disclosure has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this disclosure.

Claims

1. A fracture prediction method based on morphological analysis and signal processing, characterized in that, Includes the following steps: S1. The post-stack seismic data is subjected to USM sharpening to enhance the high-frequency content of the seismic data and highlight small faults, resulting in sharpened data. S2. Using the USM sharpened data as input, perform Pearson-based similarity calculations to obtain similarity data; S3. Using similarity data as input, perform histogram equalization. Through histogram equalization, weak signal enhancement is performed on the fracture information in the fracture data to obtain equalized data. S4. Perform regional fusion processing on the balanced data to obtain regional fused data; S5. Perform skeleton extraction processing on the regional fusion data to obtain fracture information.

2. The fracture prediction method based on morphological analysis and signal processing according to claim 1, characterized in that, The steps for implementing USM sharpening in S1 are as follows: S11. Low-pass filtering: The original seismic data is filtered using a Gaussian filter to obtain low-frequency component seismic data. S12. Calculate the high-frequency component, subtract the low-frequency component seismic data from the original seismic data to obtain the high-frequency component seismic data, which contains detailed information of the original seismic data. S13. Amplify high-frequency components by multiplying the high-frequency seismic data by a coefficient greater than 1 to amplify detailed information. S14. Superimpose high-frequency components. Finally, superimpose the amplified high-frequency component seismic data onto the original seismic data to obtain the sharpened seismic data, also known as the sharpened data.

3. The fracture prediction method based on morphological analysis and signal processing according to claim 1, characterized in that, In S2, the Pearson similarity calculation measures the linear correlation between two sets of continuous variables. Similarity calculations are performed on each sample point and its surrounding data. Based on the obtained waveform similarity features, corresponding fault attribute data is generated. During the calculation process, the size of the unit calculation window and the size of the migration window are determined according to the characteristics of the actual post-stack seismic data, generating similarity data of faults covering the range of the post-stack seismic data. The Pearson similarity calculation formula is as follows: In the formula: R xy Let be the Pearson correlation coefficient between variables x and y, where n is the number of seismic data samples, and x is the number of samples. i Let y be the value of the i-th sample point of x. i Let be the value of the i-th sample point of y.

4. The fracture prediction method based on morphological analysis and signal processing according to claim 1, characterized in that, The implementation process of histogram equalization in S3 includes: S31. Count the number of pixels n at each gray level of the original input image. i Where i = 0, 1, ..., L-1; L is the total number of gray levels; S32. Calculate the histogram of the original image, i.e., the probability density of each gray level, P. i (r i ) = n i / n, where n is the total number of pixels in the original image; S33. Calculate the cumulative distribution function s k (r k ), Where k = 0, 1, ..., L-1; S34. Calculate the final output gray level g. k , g k =INT[(g max -g min )s k (r k )+g min +0.5] / (L-1), where k = 0, 1, ..., L-1; In this expression, INT[] is the integer operator, let g min =0, g max =L-1, then this calculation formula simplifies to g k =INT[(L-1)s k (r k )+0.5] / (L-1); S35. Use the gray level function f of the original image. k and the g obtained above k By modifying the gray levels of the original image according to the mapping relationship, the histogram is obtained as an approximately uniformly distributed output image, thus obtaining balanced data.

5. The fracture prediction method based on morphological analysis and signal processing according to claim 1, characterized in that, The steps of the region fusion operation in S4 include: S41. Determine the uniformity merging criterion and divide the image into a set of arbitrary non-overlapping initial regions, typically 4 parts; S42. Division: For any region, if it does not meet the criteria, then divide it into 4 equal parts. S43. Repeat step S42 until all segmented regions meet the criteria. S44. Merge: For two adjacent regions, if the new region obtained by merging these two regions meets the criteria, then the two regions are merged. S45. Repeat S44 until all merges are completed, generating regional merged data.

6. The fracture prediction method based on morphological analysis and signal processing according to claim 1, characterized in that, The skeleton extraction algorithm in S5 adopts the Khalid Sheed algorithm, and the specific steps are as follows: S51. Mark the boundaries of the image; S52. If a point has three non-zero neighboring points, delete the point. S53. If a point has 3 or 4 non-zero neighboring points, delete the point. S54. If a point has 3, 4, or 5 non-zero neighbors in its neighborhood, delete the point. S55. If there are 3, 4, 5, or 6 non-zero points adjacent to this point in its neighborhood, delete the point. S56. If there are 3, 4, 5, 6, or 7 non-zero points adjacent to this point in its neighborhood, delete this point. S57. Unmark the remaining boundary points. If no points have been modified in S56, stop the iteration; otherwise, return to S51.

Images