Fracture detection method based on gridding and DBSCAN clustering algorithm

Abstract

The invention provides a fracture detection method based on gridding and a DBSCAN clustering algorithm. The fracture detection method comprises the steps of obtaining post-stack seismic data as input data of a subsequent calculation step; carrying out fracture sensitive attribute calculation on the post-stack seismic data; clustering analysis is carried out on fracture detection attributes through a DBSCAN algorithm so as to realize automatic tracking of fracture detection results, on this basis, through gridding processing, a reasonable threshold value is set for tracked fracture data according to geometric topology information such as an inclination angle, an inclination angle and a length in combination with prior geological knowledge, cleaning of the fracture detection results is realized, and the fracture detection results are cleared. And finally, a fracture detection data body with high precision and high signal-to-noise ratio is obtained. The method is suitable for the seismic data processing field with a complex fracture system and a low seismic data signal-to-noise ratio. Compared with the prior art, the method has the advantages that the finally obtained fracture detection result is high in signal-to-noise ratio, high in resolution, obvious in linear feature, good in continuity, high in interpretability and better in conformity with geological knowledge of a work area.

Description

Technical Field

[0001] The present invention relates to the field of oil and gas exploration and development, and more specifically to a fracture detection method based on gridding and DBSCAN clustering algorithm. Background Art

[0002] Fault system is an important controlling factor for oil and gas accumulation and distribution, and fault detection is an important basic work in oil and gas exploration. Existing fault detection technology solutions calculate seismic fault sensitive attribute data bodies, characterize abnormal values ​​in seismic attributes as faults, and then realize fault detection. Existing fault detection technology solutions mainly have the following problems:

[0003] 1. The existing technical solutions are sensitive to the noise in the seismic data. When the signal-to-noise ratio of the seismic data is low, many false anomalies will appear in the fault detection data body, resulting in poor continuity and poor interpretability of the fault detection results, which is inconsistent with the actual geological understanding.

[0004] 2. The existing technical solutions do not fully consider the application of prior information. In actual work areas, the main tendency, dip, scale and topological relationship of the faults can often be obtained through other channels such as imaging logging or geostress analysis. The existing technical solutions lack systematic design and consideration for the reasonable application of this prior information. Summary of the invention

[0005] The present invention overcomes the deficiencies in the prior art and provides a fracture detection method based on gridding and DBSCAN clustering algorithm.

[0006] The purpose of the present invention is achieved through the following technical solutions.

[0007] A fracture detection method based on gridding and DBSCAN clustering algorithm, the specific steps include:

[0008] S1, obtaining post-stack seismic data as input data for subsequent calculation steps;

[0009] S2, calculate the fracture sensitivity attributes of post-stack seismic data;

[0010] S3, using the DBSCAN density clustering algorithm to perform cluster analysis on the sensitive seismic attributes of each fault calculated in step S2;

[0011] S4, gridding the clustering results calculated in step S3;

[0012] S5. Perform spatial analysis on the clustering results after grid division, calculate the extension length, inclination and dip angle of each class in the clustering results by using the coordinate information of the grid and the value information of the grid unit, calculate the topological structure of all classes by using the surface information of the grid and the value information of the grid unit, and divide the clustering results into two topological structures of isolated fracture and intersecting fracture according to the topological structure;

[0013] S6. Obtain the structural geomechanics research results and imaging logging data results of the study area to determine the prior information of fault tendency, fault dip and fault extension length;

[0014] S7, using the prior information obtained in step S6 and the topological structure of the fault in step S5 to set a corresponding threshold value, filtering the clustering results after meshing in step S4, removing isolated clustering results with an extension length less than the threshold value, and removing clustering results with fault dip and azimuth that are inconsistent with the prior information, so as to clean the fault clustering results;

[0015] S8. Output the cleaned fracture clustering result as the result of fracture detection.

[0016] The fault-sensitive seismic attributes of step S2 include seismic dip attributes, azimuth attributes, coherence attributes and curvature attributes.

[0017] The specific steps of S3 include:

[0018] S31, setting the valid value range of the fracture sensitive attribute, when the value of the sampling point is within the valid value range, marking the location of the point as a fracture point, when the value of the sampling point is outside the valid value range, marking the location of the point as a background point;

[0019] S32, setting the initial value of the domain distance parameter and the initial value of the minimum sample number parameter in the DBSCAN algorithm;

[0020] S33, marking all sampling points of the fracture-sensitive data as unvisited;

[0021] S34, randomly selecting a sampling point P of the fracture attribute data and marking the point as visited, and counting the number of all the marked fracture points within the neighborhood distance range of the sampling point;

[0022] S35, when the number of statistics is greater than the minimum sample value, all sampling points marked as breakpoints within the neighborhood distance range of the sampling point are marked as object geometry N, and all points within the object geometry N are divided into a cluster C; when the number of statistics is less than the minimum sample value, the point is marked as noise, the next sampling point is randomly selected, and step S33 is re-executed;

[0023] S36, randomly select a sampling point in the object set N, mark the point as visited, count the number of all the points marked as broken within the neighborhood distance range of the sampling point, and when the counted number is greater than the minimum number of samples, add all the sampling points marked as broken within the neighborhood distance range of the sampling point to N, and when the point is not a member of other clusters, add the point to cluster C;

[0024] S37, repeat step S36 until all points in N are traversed;

[0025] S38. Traverse all sampling points in the fracture sensitive data until all sampling points are marked as visited.

[0026] In step S4, the grid division adopts a structured hexahedral grid, the length, width and height of the grid correspond to the line spacing, track spacing and sampling rate of the original fracture sensitive attributes respectively, the center point of the grid is located at the location of the original sampling point, and the value of the grid unit is the clustering result obtained in step S3.

[0027] In step S4, the eight vertex coordinates of the hexahedral mesh are obtained by converting the center point coordinates and the length, width and height of the mesh, and the six faces of the hexahedral mesh are represented by the four vertex coordinates of each face.

[0028] The beneficial effects of the present invention are:

[0029] A fracture detection method based on gridding and DBSCAN clustering algorithm is provided, and the invention is applicable to the field of seismic data processing with complex fracture system and low signal-to-noise ratio of seismic data.

[0030] This scheme uses the DBSCAN clustering algorithm to implement cluster analysis of fracture attribute data feature values ​​and realize automatic tracking of fracture data. According to the geometric topological information such as dip, inclination and length of the tracked fracture data, a reasonable threshold value is set in combination with prior geological knowledge to clean the fracture detection results, and finally obtain a fracture detection data body with high precision and high signal-to-noise ratio. The fracture detection results finally obtained by the scheme have high signal-to-noise ratio, high resolution, obvious linear features, good continuity, strong interpretability, and better conformity with the geological knowledge of the work area. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is a schematic diagram of the process of the present invention;

[0032] Figure 2 is the fracture detection attribute profile feature map;

[0033] Figure 3 yes Figure 2 The plane feature map extracted from the fracture detection attribute profile feature when the sampling time is 1 second;

[0034] Figure 4 Yes Figure 3 Plane feature map after cluster analysis of fracture detection attributes;

[0035] Figure 5 It is the result map of imaging logging interpretation;

[0036] Figure 6 It is a rose diagram of the fault orientation obtained based on the statistical interpretation results of imaging logging;

[0037] Figure 7 It is a fracture length distribution bar chart obtained based on the research results of structural mechanics;

[0038] Figure 8 Yes Figure 4 The plane feature map after fracture cleaning is shown in the middle result;

[0039] Fig. 9 is a profile characteristic diagram of the fracture detection property obtained in this embodiment. DETAILED DESCRIPTION

[0040] The technical solution of the present invention is further described below through specific embodiments.

[0041] Example

[0042] like Figure 1 As shown, a fracture detection method based on gridding and DBSCAN clustering algorithm includes the following specific steps:

[0043] S1. Obtain post-stack seismic data as input data for subsequent calculation steps.

[0044] S2, such as Figure 2 and Figure 3 As shown in Figure 2, the fracture sensitivity attributes of post-stack seismic data are calculated.

[0045] Fault-sensitive seismic attributes include seismic dip attributes, azimuth attributes, coherence attributes, and curvature attributes. The geological meanings of the above attributes are as follows:

[0046] Seismic dip: describes the local seismic reflection dip. This attribute indicates the structural characteristics of the seismic reflection and can identify large-scale fault systems. Faults often appear as black linear features on the dip attribute profile. On the other hand, large depressions can also show the same characteristics.

[0047] Seismic azimuth attribute: describes the tendency of seismic reflection, often used in conjunction with seismic dip.

[0048] Coherence attributes: describe the discontinuity of seismic reflections and are often used to describe large-scale fracture features and discontinuities of geological bodies.

[0049] Curvature attributes: can better describe vertical discontinuities and fracture channels associated with faults. These fracture channels are often nearly vertical fracture surfaces that cut through the reservoir horizontally, and their lengths are generally from several meters to tens of meters.

[0050] S3, such as Figure 4 As shown, cluster analysis is performed on each fault sensitive seismic attribute calculated in step S2.

[0051] This embodiment adopts the DBSCAN density clustering algorithm. The specific steps of applying the DBSCAN algorithm to the fault sensitive seismic attributes include:

[0052] S31, setting the valid value range of the fracture sensitive attribute, when the value of the sampling point is within the valid value range, marking the location of the point as a fracture point, when the value of the sampling point is outside the valid value range, marking the location of the point as a background point;

[0053] S32, setting the initial value of the domain distance parameter and the initial value of the minimum sample number parameter in the DBSCAN algorithm;

[0054] S33, marking all sampling points of the fracture-sensitive data as unvisited;

[0055] S34, randomly selecting a sampling point P of the fracture attribute data and marking the point as visited, and counting the number of all the marked fracture points within the neighborhood distance range of the sampling point;

[0056] S35, when the number of statistics is greater than the minimum sample value, all sampling points marked as breakpoints within the neighborhood distance range of the sampling point are marked as object geometry N, and all points within the object geometry N are divided into a cluster C; when the number of statistics is less than the minimum sample value, the point is marked as noise, the next sampling point is randomly selected, and step S33 is re-executed;

[0057] S36, randomly select a sampling point in the object set N, mark the point as visited, count the number of all the points marked as broken within the neighborhood distance range of the sampling point, and when the counted number is greater than the minimum number of samples, add all the sampling points marked as broken within the neighborhood distance range of the sampling point to N, and when the point is not a member of other clusters, add the point to cluster C;

[0058] S37, repeat step S36 until all points in N are traversed;

[0059] S38. Traverse all sampling points in the fracture sensitive data until all sampling points are marked as visited.

[0060] S4. Gridding the clustering results calculated in step S3.

[0061] This embodiment adopts a structured hexahedral grid, the length, width and height of the grid respectively correspond to the line spacing, track spacing and sampling rate of the original fracture sensitive attributes, the center point of the grid is located at the position of the original sampling point, the value of the grid unit is the clustering result obtained in step S3, the eight vertex coordinates of the hexahedral grid are obtained by converting the center point coordinates and the length, width and height of the grid, and the six faces of the hexahedral grid are characterized by the four vertex coordinates of each face.

[0062] S5. Perform spatial analysis on the clustering results after grid division. Use the grid coordinate information and grid cell value information to calculate the extension length, inclination, and dip angle of each class in the clustering results. Use the grid surface information and grid cell value information to calculate the topological structure of all classes. Then, divide the clustering results into two topological structures: isolated fractures and intersecting fractures according to the topological structure.

[0063] An isolated break indicates that the class does not geometrically intersect with other classes in three-dimensional space, and an intersecting break indicates that the class geometrically intersects with other classes in three-dimensional space.

[0064] S6. Obtain the structural geomechanics research results and imaging logging data results of the study area to determine the prior information on fault tendency, fault dip and fault extension length.

[0065] like Figure 5 As shown, the fracture inclination in the embodiment is a low-angle fracture.

[0066] like Figure 6 As shown, the fault orientation in the embodiment is mainly near east-west and southwest-northeast.

[0067] like Figure 7 As shown, the length of the fracture in the embodiment is concentrated between 80-500 meters.

[0068] S7, such as Figure 8 As shown, the prior information obtained in step S6 and the fractured topological structure in step S5 are used to set a corresponding threshold value, and the clustering result after the grid division in step S4 is filtered.

[0069] According to the fracture dip, fracture azimuth and fracture length data obtained in step S6 of this embodiment, the clustering results with an extension length less than 80 meters and isolated are eliminated, and the clustering results with large dip and northwest-southeast orientation that are inconsistent with the prior information are eliminated, so as to achieve the cleaning of the fracture clustering results.

[0070] S8, such as Fig. 9 As shown, the cleaned fracture clustering result is output as the result of fracture detection.

[0071] The embodiments of the present invention are described in detail above, but the contents described are only preferred embodiments of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A fracture detection method based on gridding and DBSCAN clustering algorithm, characterized in that: The specific steps include: S1, obtaining post-stack seismic data as input data for subsequent calculation steps; S2, calculate the fracture sensitivity attributes of post-stack seismic data; S3, using the DBSCAN density clustering algorithm to perform cluster analysis on the sensitive seismic attributes of each fault calculated in step S2; S4, gridding the clustering results calculated in step S3; S5. Perform spatial analysis on the clustering results after grid division, calculate the extension length, inclination and dip angle of each class in the clustering results by using the coordinate information of the grid and the value information of the grid unit, calculate the topological structure of all classes by using the surface information of the grid and the value information of the grid unit, and divide the clustering results into two topological structures of isolated fracture and intersecting fracture according to the topological structure; S6. Obtain the structural geomechanics research results and imaging logging data results of the study area to determine the prior information of fault tendency, fault dip and fault extension length; S7, using the prior information obtained in step S6 and the topological structure of the fracture in step S5 to set a corresponding threshold value, filtering the clustering results after meshing in step S4, removing isolated clustering results with an extension length less than the threshold value, and removing clustering results with fracture inclination angles inconsistent with the prior information, so as to clean the fracture clustering results; S8. Output the cleaned fracture clustering result as the result of fracture detection.

2. A fracture detection method based on gridding and DBSCAN clustering algorithm according to claim 1, characterized in that: The fault-sensitive seismic attributes of step S2 include seismic dip attributes, azimuth attributes, coherence attributes and curvature attributes.

3. The fracture detection method based on gridding and DBSCAN clustering algorithm according to claim 1, characterized in that: The specific steps of S3 include: S31, setting the valid value range of the fracture sensitive attribute, when the value of the sampling point is within the valid value range, marking the location of the point as a fracture point, when the value of the sampling point is outside the valid value range, marking the location of the point as a background point; S32, setting the initial value of the domain distance parameter and the initial value of the minimum sample number parameter in the DBSCAN algorithm; S33, marking all sampling points of the fracture-sensitive data as unvisited; S34, randomly selecting a sampling point P of the fracture attribute data and marking the point as visited, and counting the number of all the marked fracture points within the neighborhood distance range of the sampling point; S35, when the number of statistics is greater than the minimum sample value, all sampling points marked as breakpoints within the neighborhood distance range of the sampling point are marked as object geometry N, and all points within the object geometry N are divided into a cluster C; when the number of statistics is less than the minimum sample value, the point is marked as noise, the next sampling point is randomly selected, and step S33 is re-executed; S36, randomly select a sampling point in the object set N, mark the point as visited, count the number of all the points marked as broken within the neighborhood distance range of the sampling point, and when the counted number is greater than the minimum number of samples, add all the sampling points marked as broken within the neighborhood distance range of the sampling point to N, and when the point is not a member of other clusters, add the point to cluster C; S37, repeat step S36 until all points in N are traversed; S38. Traverse all sampling points in the fracture sensitive data until all sampling points are marked as visited.

4. The fracture detection method based on gridding and DBSCAN clustering algorithm according to claim 1, characterized in that: In step S4, the grid division adopts a structured hexahedral grid, the length, width and height of the grid correspond to the line spacing, track spacing and sampling rate of the original fracture sensitive attributes respectively, the center point of the grid is located at the location of the original sampling point, and the value of the grid unit is the clustering result obtained in step S3.

5. A fracture detection method based on gridding and DBSCAN clustering algorithm according to claim 4, characterized in that: In step S4, the eight vertex coordinates of the hexahedral mesh are obtained by converting the center point coordinates and the length, width and height of the mesh, and the six faces of the hexahedral mesh are represented by the four vertex coordinates of each face.

Images