Three-dimensional seismic fault automatic interpretation method

Abstract

The invention provides a three-dimensional seismic fault automatic interpretation method. The three-dimensional seismic fault automatic interpretation method comprises the steps of 1, performing original seismic interpretation preprocessing; step 2, carrying out fine description on the fault by using sensitive seismic attributes; step 3, automatic interpretation of the fault is carried out; step 4, invalid fault filtering is carried out; step 5, selecting the filtered fault to carry out discretization processing; step 6, recombining discretized data by using a big data deep learning algorithm; and 7, repeating the steps 3 to 6 until the fracture system conforming to geological knowledge in the target work area is obtained. According to the three-dimensional seismic fault automatic interpretation method, the accuracy of subsequent work is improved, a fracture system can be better described, the fault recognition capability is improved, the working efficiency is greatly improved, the influence of a stratum interface on the fault is eliminated, and the automatic interpretation result has high credibility.

Description

Technical Field

[0001] The invention relates to the technical field of geophysical exploration seismic data interpretation, and in particular to a three-dimensional seismic fault automatic interpretation method. Background Art

[0002] The fault plane refers to the plane where the rock mass on both sides of the fault has significant sliding displacement, which is manifested as the dislocation, displacement and twisting of the seismic axes on both sides on the seismic profile. Accurate and reasonable fault interpretation is one of the most important tasks in the study of oil and gas basins, which directly affects the correctness of exploration decisions and the effect of oil and gas field development.

[0003] There are currently three methods for fault interpretation: manual, semi-automatic, and automatic. For areas with complex faults, if semi-automatic and traditional manual interpretation methods are used, the interpretation is difficult and the work efficiency is low. At the same time, it is easily affected by the interpretation experience, making the results highly multi-solution; automatic interpretation uses seismic data processing and ant body automatic tracking technology to make the fault line clear, the breakpoint crisp, and the fault continuity enhanced. The software automatically identifies small faults that cannot be identified by the naked eye, which can reduce the influence of researchers' subjective experience. At the same time, automatic interpretation technology is more convenient and quick, and can increase work efficiency by more than ten times. However, the credibility of automatic interpretation of earthquake faults in existing technologies is not high.

[0004] In the Chinese patent application with application number: CN201110380782.2, a fault automatic interpretation device for a three-dimensional seismic data body is involved. The coherence volume imaging module in the fault automatic interpretation device for a three-dimensional seismic data body is used to detect seismic faults and output a coherence volume to characterize the faults. The ant colony tracking module uses a directional ant colony algorithm to automatically track and identify the three-dimensional faults on the coherence volume and outputs the fault automatic tracking results. The fault automatic separation module is used to judge and automatically separate each fault according to the fault automatic tracking results. The fault attribute automatic calculation module measures and calculates the fault geometric elements and fault dynamic attributes of each separated fault. The fault automatic interpretation device for a three-dimensional seismic data body in the invention can automatically interpret faults for various seismic data, has the advantages of high calculation efficiency, objective and accurate fault interpretation results, etc., effectively improves the accuracy of fault interpretation and improves seismic interpretation efficiency.

[0005] The shortcomings of this invention are: in the application of Chinese patent CN201110380782.2, there is a lack of processing of original seismic data and comparison of multiple seismic attributes, and it is impossible to ensure the acquisition of high-quality ant volume data, thereby failing to ensure the authenticity and objectivity of fault data.

[0006] In the Chinese patent application with application number: CN201710891808.7, a seismic interpretation method for faults is involved, which belongs to the field of oil exploration technology. The seismic interpretation method for faults of the invention first determines the time-depth conversion relationship between the strata encountered by the well and the seismic profile of the well bypass, and then establishes the time-depth relationship conversion of all wells in the target work area. Using the established time-depth relationship conversion between all wells and earthquakes in the work area, the main seismic profile and the contact profile are first interpreted, and the trajectory position of the target fault on the main profile and the contact profile is preliminarily determined. Then, the fault trajectory line on the rising plate seismic profile and the descending plate seismic profile is determined, and the target fault trajectory line on the rising plate seismic profile and the descending seismic profile that are parallel to the target fault is used to correct and interpret the target fault trajectory line on the main seismic profile and the contact profile. The seismic interpretation method for faults of the invention greatly improves the accuracy of seismic interpretation, which is of great significance for the fine development of old oilfield areas.

[0007] The shortcomings of this invention are: in the application of Chinese patent CN201710891808.7, it is necessary to interpret the work area, which not only increases the workload, but also increases the errors caused by the different levels of interpreters. The invention also lacks the constraints of seismic data processing and seismic attributes on fault interpretation results.

[0008] In the Chinese patent application with application number: CN201410152774.6, a fast three-dimensional fault interpretation method based on plane navigation is involved, which includes: ① performing fault-enhanced interpretive processing on seismic data to obtain a seismic data body; ② extracting a time slice of the target layer segment reflecting the plane distribution characteristics of the fault on the seismic data body, and using it as a plane navigation slice for fault interpretation; ③ selecting a fault on the plane navigation slice; ④ extracting a seismic profile in a direction perpendicular to the fault strike, and completing the profile interpretation of the fault at the projection point of the selected fault section; ⑤ gridding the navigation plane interpretation of the selected fault and the interpretation results of the fault on the extracted seismic profile, and projecting them onto another newly extracted seismic profile; ⑥ under the guidance of the navigation of the fault plane interpretation, gradually rolling extracting the seismic profile in a direction perpendicular to the fault strike and completing the fault interpretation of the seismic profile, and completing the interpretation of all seismic profiles of the fault; ⑦ repeating steps ③ to ⑥ until all fault interpretations in the entire area are completed.

[0009] The shortcoming of this invention is that in the application of Chinese patent CN201410152774.6, steps ④, ⑤, and ⑥ require manual interpretation of the seismic profile, which will increase the error caused by the different levels of interpreters. This invention fully applies automatic interpretation and big data processing technology, completely avoiding the impact caused by manual interpretation.

[0010] In the Chinese patent application with application number: CN202110164203.4, a fault interpretation method is involved, which belongs to the field of oil exploration and development technology. The method includes: calibrating the geological stratification of existing wells in the target work area on the through-well seismic profile, determining the layer segment with developed faults according to the calibration results of the geological stratification of the existing wells, and using it as the target layer segment; using the ratio of the stratum thickness at the upper and lower sections of the larger fault to identify the type of the larger fault in the target layer segment, and determining whether each larger fault is a single fault with only one section or a staged multi-segment fault with multiple sections; using the maximum change rate of the seismic reflection amplitude in the target layer segment to identify the micro-faults in the target layer segment; combining the identification results of the larger faults and micro-faults in the target layer segment, the fault interpretation of the target work area is completed. This invention can solve the phenomenon of inconsistent stratum strike between complex fault blocks and unreasonable thickness relationship of the target layer segment, so that the fault interpretation results are more in line with the laws of geological tectonic movement and sedimentary characteristics.

[0011] The shortcomings of this invention are: in the application of Chinese patent CN202110164203.4, the identification of faults by layered calibration and stratum thickness is not a complete interpretation. Large faults and small faults are identified by different methods, and the technical route is cumbersome and complicated. In addition, the automatic interpretation of three-dimensional structural fault markers is not realized, and the explanatory processing of seismic data is also lacking.

[0012] The above existing technologies are significantly different from the present invention, and fail to achieve interpretative processing and sensitive seismic attribute constraints; fail to achieve automatic identification of small faults that conform to geological laws, and fail to solve the technical problems we want to solve. For this reason, we have invented a new three-dimensional seismic fault automatic interpretation method. Summary of the invention

[0013] The purpose of the present invention is to provide a three-dimensional seismic fault automatic interpretation method for improving the efficiency and accuracy of fault interpretation.

[0014] The object of the present invention can be achieved by the following technical measures: a three-dimensional seismic fault automatic interpretation method, the three-dimensional seismic fault automatic interpretation method comprising:

[0015] Step 1, preprocessing the original seismic interpretation;

[0016] Step 2: Use sensitive seismic attributes to finely characterize the fault;

[0017] Step 3, automatic interpretation of faults;

[0018] Step 4, filtering invalid faults;

[0019] Step 5, select the filtered faults for discretization;

[0020] Step 6: Recombine the discretized data using big data deep learning algorithms;

[0021] Step 7, repeat steps 3 to 6 until a fault system that is consistent with geological knowledge in the target work area is obtained.

[0022] The purpose of the present invention can also be achieved by the following technical measures:

[0023] In step 1, the original seismic data is processed to enhance the dislocation and twisting characteristics of the same-direction axes on both sides of the fault in the seismic section.

[0024] In step 1, the sampling interval L and the continuity threshold A are set by performing continuity analysis on the same-direction axes of the original seismic data.

[0025] In step 1, when the co-axial direction is within the sampling interval L and the co-axial similarity is greater than A, it is considered that there is no breakpoint. The signal-to-noise ratio of the seismic data is improved by the filtering algorithm to enhance the continuity of the co-axial direction. When the co-axial direction is within the sampling interval L and the co-axial similarity is less than A, the diffusion filtering algorithm is used to highlight the abnormal points in the co-axial direction, strengthen the dislocation and twisting characteristics of the co-axial direction on both sides of the fault of the seismic profile, and then obtain the fault enhancement data volume.

[0026] In step 2, seismic attribute calculations are performed to obtain various seismic body attributes in the target area. By evaluating and analyzing the correlation of various body attributes, the seismic body attributes that are most sensitive to subtle structural changes of low-order faults are selected. Combined with ant body tracking technology, the three-dimensional spatial distribution and combination characteristics of the faults are further clarified.

[0027] In step 2, to ensure the integrity and accuracy of the system, the fracture system should be characterized as finely as possible.

[0028] In step 3, ant bodies are used to automatically interpret complex faults in three-dimensional space.

[0029] In step 4, the credibility threshold is set by using the strike attribute, dip attribute and area attribute to filter out the low-angle invalid faults in step 3, eliminating the influence of the stratigraphic interface and structural undulation on the results.

[0030] In step 5, the fault fragments filtered in step 4 are selected for discretization processing to convert the surface data into discrete point data.

[0031] In step 6, the big data deep learning algorithm is used to recombine the discretized data: abnormal discrete points are removed and the discretized data is recombinated to obtain a preliminary fault combination.

[0032] In step 6, the big data deep learning algorithm model relationship is:

[0033] B j =f(netj ) (1)

[0034]

[0035] Where: f() is the signal function; θ j is the internal threshold of the neuron; B j is the output of the neuron; W 0j To adjust θ j The weight value of A = {a1, a2, ..., a n} is the input signal; W={W 1j , W 2j , …, W nj} is the connection strength vector between the neuron numbered j and the n other neurons that transmit information to it.

[0036] In step 7, steps 3 to 6 are looped, and ant bodies are used to perform quality control along the stratigraphic slices. If the calculation results are consistent with the geological understanding of the target area, the loop is terminated and the final fracture system is output, otherwise it returns to step 3.

[0037] The three-dimensional seismic fault automatic interpretation method of the present invention has the following advantages: 1. The present invention uses professional processing software to enhance the fault edge, making the fault line clearer and improving the accuracy of subsequent work. 2. The present invention can better characterize the fault system and improve the ability to identify faults by comprehensively utilizing sensitive attributes and ant colony tracking technology. 3. The present invention uses automatic fault extraction technology and big data deep learning algorithms to reduce the influence of researchers' subjective experience and greatly improve work efficiency. 4. The present invention uses strike attributes, dip attributes and area attributes to filter faults, eliminating the influence of stratum interfaces on faults. At the same time, the ant body is used to perform quality control on the fault system along the layer section, so that the automatic interpretation results have high credibility. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 is a schematic diagram of an original seismic profile involved in a specific embodiment of the present invention;

[0039] Figure 2 is a schematic diagram of a seismic section after fault edge enhancement processing involved in a specific embodiment of the present invention;

[0040] Figure 3 It is a schematic diagram of preliminary automatic interpretation results of three-dimensional space faults involved in a specific embodiment of the present invention;

[0041] Figure 4 It is a schematic diagram of the principle of a big data deep learning algorithm involved in a specific embodiment of the present invention;

[0042] Figure 5is a schematic diagram of a filtering, combining and post-quality control fracture system involved in a specific embodiment of the present invention;

[0043] Figure 6 The flowchart is a specific embodiment of the three-dimensional seismic fault automatic interpretation method of the present invention. DETAILED DESCRIPTION

[0044] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0045] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations and / or combinations thereof.

[0046] like Figure 6 As shown, Figure 6 The flowchart of the automatic interpretation method of three-dimensional seismic fault of the present invention is as follows. The automatic interpretation method of three-dimensional seismic fault comprises:

[0047] 1) Original seismic interpretation preprocessing: Figure 1 The original seismic data profile has a fuzzy fault line. The original seismic data is processed using commercial software. When the common axis is within the sampling interval L and the common axis similarity is greater than A, it is considered that there is no breakpoint. The filtering algorithm is used to improve the signal-to-noise ratio of the seismic data and enhance the continuity of the common axis. When the common axis is within the sampling interval L and the common axis similarity is less than A, the diffusion filtering algorithm is used to highlight the abnormal points in the common axis and enhance the dislocation and twisting characteristics of the common axis on both sides of the fault in the seismic profile, such as Figure 2 As shown, the fault enhancement data volume has clear fault lines and crisp breakpoints compared to the original seismic volume, which is helpful for the subsequent automatic interpretation of faults;

[0048] 2) Use sensitive seismic attributes to finely characterize faults: Use the seismic attribute calculation module in relevant commercial software to obtain various seismic body attributes in the target area. By evaluating and analyzing the correlation of various body attributes, select the seismic body attributes that are most sensitive to subtle structural changes such as low-order faults, and combine them with ant body tracking technology to further clarify the three-dimensional spatial distribution and combination characteristics of faults;

[0049] 3) Automatic interpretation of faults: The above-mentioned professional commercial seismic interpretation software is used to automatically interpret faults in three-dimensional space using the automatic fragment interpretation method.

[0050] 4) Further using the strike attribute, dip attribute and area attribute, the credibility threshold is set and the low-angle invalid faults in step 3) are filtered out to eliminate the influence of the stratigraphic interface and structural undulation on the results.

[0051] 5) Select the fault fragments filtered in step 4) for discretization, and convert the surface data into discrete point data.

[0052] 6) Using big data deep learning algorithms, recombine the discretized data: remove abnormal discrete points, and recombine the discretized data to obtain a preliminary fault combination. The mathematical relationship of the model is:

[0053] B j =f(net j ) (1)

[0054]

[0055] Where: f() is the signal function; θ j is the internal threshold of the neuron; B j is the output of the neuron; W 0j To adjust θ j The weight value of A = {a1, a2, ..., a n} is the input signal; W={W 1j , W 2j , …, W nj} is the connection strength vector between the neuron numbered j and the n other neurons that transmit information to it.

[0056] 7) Loop through steps 3) to 6) and use the ant body to perform quality control along the stratigraphic slices. If the calculation results are consistent with the geological knowledge of the target area, terminate the loop and output the final fracture system. Otherwise, return to step 3).

[0057] In the step 1), the sampling interval L and the continuity threshold A are set by performing continuity analysis on the iso-axis of the original seismic data. When the iso-axis is within the sampling interval L and the iso-axis similarity is greater than A, it is considered that there is no breakpoint, and the signal-to-noise ratio of the seismic data is improved by the filtering algorithm to enhance the continuity of the iso-axis; when the iso-axis is within the sampling interval L and the iso-axis similarity is less than A, the diffusion filtering algorithm is used to highlight the abnormal points in the iso-axis, strengthen the dislocation and twisting characteristics of the iso-axis on both sides of the fault of the seismic profile, and then obtain the fault enhancement data body.

[0058] In step 2), in order to ensure the integrity and accuracy of the system, the fracture system should be characterized as finely as possible.

[0059] In step 4), relevant professional software is used to further utilize strike attributes, dip attributes and area attributes to set a credibility threshold and filter out low-angle invalid faults, thereby eliminating the impact of stratum interfaces and structural undulations on the results.

[0060] In the step 5), the filtered fault data of the target work area is discretized to convert the surface data into discrete point data.

[0061] In step 6), a big data deep learning algorithm is used to remove abnormal discrete points, and the discrete point data is recombined to obtain a preliminary fault combination.

[0062] The following are several specific embodiments of the present invention.

[0063] Example 1

[0064] In a specific embodiment 1 of the present invention, the 3D seismic fault automatic interpretation method includes the following steps:

[0065] 1) Raw seismic interpretive preprocessing.

[0066] like Figure 1 As shown in the figure, the sampling interval L and continuity threshold A are set by analyzing the continuity of the co-axial of the original seismic data. When the co-axial is within the sampling interval L and the similarity of the co-axial is greater than A, it is considered that there is no breakpoint. The filtering algorithm is used to improve the signal-to-noise ratio of the seismic data and enhance the continuity of the co-axial. When the co-axial is within the sampling interval L and the similarity of the co-axial is less than A, the diffusion filtering algorithm is used to highlight the abnormal points in the co-axial, strengthen the dislocation and twisting characteristics of the co-axial on both sides of the fault of the seismic profile, and then obtain the fault enhancement data volume ( Figure 2 ).

[0067] 2) Use sensitive seismic attributes to finely characterize faults:

[0068] By using the seismic attribute calculation module in relevant commercial software, various seismic body attributes in the target area are obtained. Through evaluation and correlation analysis of various body attributes, the seismic body attributes that are most sensitive to subtle structural changes such as low-order faults are selected. Under the constraint of the fault direction in the target area, the ant body tracking technology is combined to further clarify the planar fault distribution and combination characteristics.

[0069] 3) Automatic interpretation of faults:

[0070] The above-mentioned professional commercial seismic interpretation software is used to automatically interpret complex faults in three-dimensional space using ant bodies.

[0071] 4) Invalid fault filtering:

[0072] The software is used to calculate the strike attributes, dip attributes and area attributes of all faults, and invalid faults such as low-angle faults are deleted according to the geological characteristics of the target work area.

[0073] 5) Select the fault after filtering in step 4) and perform discretization.

[0074] 6) Recombining discretized data through big data deep learning algorithms:

[0075] Using big data deep learning algorithms, the discretized data is reassembled: abnormal discrete points are removed, and the discretized data is reassembled to obtain a preliminary fault combination. The mathematical relationship of the big data deep learning model is:

[0076] B j =f(net j ) (1)

[0077]

[0078] Where: f() is the signal function; θ j is the internal threshold of the neuron; B j is the output of the neuron; W 0j To adjust θ j The weight value of A = {a1, a2, ..., a n} is the input signal; W={W 1j , W 2j , …, W nj} is the connection strength vector between the neuron numbered j and the n other neurons that transmit information to it.

[0079] 7) Repeat steps 3) to 6) until a fault system that is consistent with geological knowledge in the target work area is obtained.

[0080] Example 2

[0081] In a specific embodiment 2 of the present invention, a three-dimensional seismic fault automatic interpretation method is introduced by taking the target work area in oil field A as an example.

[0082] 1) Raw seismic interpretive preprocessing.

[0083] The sampling interval L and continuity threshold A are set by analyzing the continuity of the co-axial of the original seismic data. When the co-axial is within the sampling interval L and the similarity of the co-axial is greater than A, it is considered that there is no breakpoint. The signal-to-noise ratio of the seismic data is improved by the filtering algorithm, and the continuity of the co-axial is enhanced; when the co-axial is within the sampling interval L and the similarity of the co-axial is less than A, the diffusion filtering algorithm is used to highlight the abnormal points in the co-axial, strengthen the dislocation and twisting characteristics of the co-axial on both sides of the fault of the seismic profile, and then obtain the fault enhancement data volume.

[0084] 2) Use sensitive seismic attributes to finely characterize faults:

[0085] Using the attribute extraction module of seismic interpretation professional commercial software, the seismic data obtained in step 1) after interpretive processing is extracted to obtain a variety of different seismic attribute bodies: variance body, chaos body, amplitude contrast body. The different seismic attribute slices of the target work area are compared, and the sensitive attribute body of this work area is selected as the amplitude contrast body, and then the ant colony tracking technology is used to finely characterize the fault system.

[0086] 3) Automatic interpretation of faults:

[0087] like Figure 3 As shown, the above-mentioned professional commercial seismic interpretation software is used to set parameter values ​​such as sample plane deviation, sample size threshold and section deflection degree of freedom, and the ant attribute body is used to obtain the automatic interpretation results of faults in three-dimensional space.

[0088] 4) Invalid fault filtering:

[0089] Use the above-mentioned professional commercial seismic interpretation software to calculate all strike attributes, dip attributes and area attributes, combine the credibility values ​​provided by the software, and delete invalid faults such as low-angle faults based on the geological characteristics of the work area.

[0090] 5) Select the fault after filtering in step 4) and perform discretization.

[0091] Convert area data into point data.

[0092] 6) Recombining discretized data through big data deep learning algorithms:

[0093] like Figure 4 As shown, the model mapping relationship obtained by supervised learning of the calibration area of ​​oil field A is applied to the target work area, and the point data obtained in step 5) is used to obtain the fracture system after the discrete data of the target work area is combined. The mathematical relationship of the model is:

[0094] B j =f(net j ) (1)

[0095]

[0096] Where: f() is the signal function; θ j is the internal threshold of the neuron; B j is the output of the neuron; W 0j To adjust θ j The weight value of A = {a1, a2, ..., a n} is the input signal; W={W 1j , W 2j , …, W nj} is the connection strength vector between the neuron numbered j and the n other neurons that transmit information to it.

[0097] 7) According to the geological characteristics of the work area, use the ant body to slice along the layer to perform quality control on the output fault, and repeat steps 3) to 6) until the fault system ( Figure 5 ).

[0098] Example 3

[0099] In a specific embodiment 3 of the present invention, a three-dimensional seismic fault automatic interpretation method is introduced by taking the target work area in oil field A as an example.

[0100] 1) Raw seismic interpretive preprocessing.

[0101] like Figure 1 As shown in the figure, by analyzing the continuity of the co-axial of the original seismic data, the sampling interval L and the continuity threshold A are set, where L is 8 seismic traces and the continuity A is 90%. When the co-axial is within the sampling interval L and the similarity of the co-axial is greater than A, it is considered that there is no breakpoint. The signal-to-noise ratio of the seismic data is improved by the filtering algorithm, and the continuity of the co-axial is enhanced; when the co-axial is within the sampling interval L and the similarity of the co-axial is less than A, the diffusion filtering algorithm is used to highlight the abnormal points in the co-axial, strengthen the dislocation and twisting characteristics of the co-axial on both sides of the fault of the seismic profile, and then obtain the fault enhancement data volume. Figure 2 .

[0102] 2) Use sensitive seismic attributes to finely characterize faults:

[0103] By using the seismic attribute calculation module in relevant commercial software, various seismic body attributes in the target area are obtained. Through evaluation and correlation analysis of various body attributes, the seismic body attributes that are most sensitive to subtle structural changes such as low-order faults are selected. Under the constraint of the fault direction in the target area, the ant body tracking technology is combined to further clarify the planar fault distribution and combination characteristics.

[0104] 3) Automatic interpretation of faults:

[0105] The above-mentioned professional commercial seismic interpretation software is used to automatically interpret complex faults in three-dimensional space using ant bodies.

[0106] 4) Invalid fault filtering:

[0107] The software is used to calculate the strike attributes, dip attributes and area attributes of all faults, and invalid faults such as low-angle faults are deleted according to the geological characteristics of the target work area.

[0108] 5) Select the fault after filtering in step 4) and perform discretization.

[0109] 6) Recombining discretized data through big data deep learning algorithms:

[0110] Using big data deep learning algorithms, the discretized data is reassembled: abnormal discrete points are removed, and the discretized data is reassembled to obtain a preliminary fault combination. The mathematical relationship of the big data deep learning model is:

[0111] B j =f(net j ) (1)

[0112]

[0113] Where: f() is the signal function; θ j is the internal threshold of the neuron; B j is the output of the neuron; W 0j To adjust θ j The weight value of A = {a1, a2, ..., a n} is the input signal; W={W 1j , W 2j , …, W nj}, is the connection strength vector between the neuron numbered j and the n other neurons that transmit information to it.

[0114] 7) Repeat steps 3) to 6) until a fault system that is consistent with geological knowledge in the target work area is obtained.

[0115] Example 4

[0116] In a specific embodiment 4 of the present invention, a three-dimensional seismic fault automatic interpretation method is introduced by taking the target work area in oil field A as an example.

[0117] 1) Raw seismic interpretive preprocessing.

[0118] The sampling interval L and continuity threshold A are set by analyzing the continuity of the co-axial of the original seismic data. When the co-axial is within the sampling interval L and the similarity of the co-axial is greater than A, it is considered that there is no breakpoint. The signal-to-noise ratio of the seismic data is improved by the filtering algorithm, and the continuity of the co-axial is enhanced; when the co-axial is within the sampling interval L and the similarity of the co-axial is less than A, the diffusion filtering algorithm is used to highlight the abnormal points in the co-axial, strengthen the dislocation and twisting characteristics of the co-axial on both sides of the fault of the seismic profile, and then obtain the fault enhancement data volume.

[0119] 2) Use sensitive seismic attributes to finely characterize faults:

[0120] By using the seismic attribute calculation module in the relevant commercial software, various seismic body attributes such as chaos body, variance body, maximum curvature body, etc. in the target area are obtained. By evaluating and analyzing the correlation of various body attributes, the maximum curvature body attribute that is most sensitive to subtle structural changes such as low-order faults in this work area is selected. Under the constraint of the fault direction in the target area, the ant body tracking technology is combined to produce the ant tracking body by setting parameters such as tracking deviation, effective step size and stopping standard, which greatly improves the resolution of micro-fault identification and further clarifies the distribution and combination characteristics of planar faults.

[0121] 3) Automatic interpretation of faults:

[0122] The above-mentioned professional commercial seismic interpretation software is used to automatically interpret complex faults in three-dimensional space using ant bodies.

[0123] 4) Invalid fault filtering:

[0124] The software is used to calculate the strike attributes, dip attributes and area attributes of all faults, and invalid faults such as low-angle faults are deleted according to the geological characteristics of the target work area.

[0125] 5) Select the fault after filtering in step 4) and perform discretization.

[0126] 6) Recombining discretized data through big data deep learning algorithms:

[0127] Using big data deep learning algorithms, the discretized data is reassembled: abnormal discrete points are removed, and the discretized data is reassembled to obtain a preliminary fault combination. The mathematical relationship of the big data deep learning model is:

[0128] B j =f(net j )(1)

[0129]

[0130] Where: f() is the signal function; θ j is the internal threshold of the neuron; B j is the output of the neuron; W 0j To adjust θ j The weight value of A = {a1, a2, ..., a n} is the input signal; W={W 1j , W 2j , …, W nj}, is the connection strength vector between the neuron numbered j and the n other neurons that transmit information to it.

[0131] 7) Repeat steps 3) to 6) until a fault system that is consistent with geological knowledge in the target work area is obtained.

[0132] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions recorded in the aforementioned embodiments or replace some of the technical features therein with equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0133] Except for the technical features described in the specification, all other technical features are known technologies to those skilled in the art.

Claims

1. A three-dimensional seismic fault automatic interpretation method, characterized in that: The 3D seismic fault automatic interpretation method comprises: Step 1, preprocessing the original seismic interpretation; Step 2: Use sensitive seismic attributes to finely characterize the fault; Step 3, automatic interpretation of faults; Step 4, filtering invalid faults; Step 5, select the filtered faults for discretization; Step 6: Recombine the discretized data using big data deep learning algorithms; Step 7, repeat steps 3 to 6 until a fault system that is consistent with geological knowledge in the target work area is obtained.

2. The three-dimensional seismic fault automatic interpretation method according to claim 1, characterized in that: In step 1, the original seismic data is processed to enhance the dislocation and twisting characteristics of the same-direction axes on both sides of the fault in the seismic section.

3. The three-dimensional seismic fault automatic interpretation method according to claim 2, characterized in that: In step 1, the sampling interval L and the continuity threshold A are set by performing continuity analysis on the same-direction axes of the original seismic data.

4. The three-dimensional seismic fault automatic interpretation method according to claim 3, characterized in that: In step 1, when the co-axial direction is within the sampling interval L and the co-axial similarity is greater than A, it is considered that there is no breakpoint. The signal-to-noise ratio of the seismic data is improved by the filtering algorithm to enhance the continuity of the co-axial direction. When the co-axial direction is within the sampling interval L and the co-axial similarity is less than A, the diffusion filtering algorithm is used to highlight the abnormal points in the co-axial direction, strengthen the dislocation and twisting characteristics of the co-axial direction on both sides of the fault of the seismic profile, and then obtain the fault enhancement data volume.

5. The three-dimensional seismic fault automatic interpretation method according to claim 1, characterized in that: In step 2, seismic attribute calculations are performed to obtain various seismic body attributes in the target area. By evaluating and analyzing the correlation of various body attributes, the seismic body attributes that are most sensitive to subtle structural changes of low-order faults are selected. Combined with ant body tracking technology, the three-dimensional spatial distribution and combination characteristics of the faults are further clarified.

6. The three-dimensional seismic fault automatic interpretation method according to claim 5, characterized in that: In step 2, to ensure the integrity and accuracy of the system, the fracture system should be characterized as finely as possible.

7. The three-dimensional seismic fault automatic interpretation method according to claim 1, characterized in that: In step 3, ant bodies are used to automatically interpret complex faults in three-dimensional space.

8. The three-dimensional seismic fault automatic interpretation method according to claim 1, characterized in that: In step 4, the credibility threshold is set by using the strike attribute, dip attribute and area attribute to filter out the low-angle invalid faults in step 3, eliminating the influence of the stratigraphic interface and structural undulation on the results.

9. The three-dimensional seismic fault automatic interpretation method according to claim 1, characterized in that: In step 5, the fault fragments filtered in step 4 are selected for discretization processing to convert the surface data into discrete point data.

10. The three-dimensional seismic fault automatic interpretation method according to claim 1, characterized in that: In step 6, the big data deep learning algorithm is used to recombine the discretized data: abnormal discrete points are removed and the discretized data is recombinated to obtain a preliminary fault combination.

11. The three-dimensional seismic fault automatic interpretation method according to claim 10, characterized in that: In step 6, the big data deep learning algorithm model relationship is: B j =f(net j ) (1) Where: f() is the signal function; θ j is the internal threshold of the neuron; B j is the output of the neuron; W 0j To adjust θ j The weight value of A = {a1, a2, ..., a n } is the input signal; W={W 1j , W 2j , …, W nj } is the connection strength vector between the neuron numbered j and the n other neurons that transmit information to it.

12. The three-dimensional seismic fault automatic interpretation method according to claim 1, characterized in that: In step 7, steps 3 to 6 are looped, and ant bodies are used to perform quality control along the stratigraphic slices. If the calculation results are consistent with the geological understanding of the target area, the loop is terminated and the final fracture system is output, otherwise it returns to step 3.

Images