A fracture prediction method and system

Abstract

The present invention discloses a fracture prediction method and system, including: performing fracture enhancement and interpretive processing on the three-dimensional seismic data of a target work area; preferably key attributes for fracture prediction, calculating attributes for the data volume obtained from the fracture enhancement and interpretive processing to obtain various attribute calculation results containing attribute value information of points at different positions; extracting various attribute values at fractures of different levels from the various attribute calculation results to form a sample matrix; determining determination rules for cross-sections of different levels according to the geological characteristics of the target work area and in combination with the various attribute calculation results, thereby forming a cross-section feature recognition library; constructing a fracture prediction model according to the sample matrix and the cross-section feature recognition library; obtaining the sample matrix of the area to be predicted, and using the fracture prediction model to perform fracture level prediction based on scale. The present invention solves the problem of multi-solution in manual interpretation of strike-slip cross-sections and the problem of uniqueness in fracture characterization of a single discontinuous attribute.

Description

Technical Field

[0001] The present invention relates to the technical field of seismic geological interpretation in oil and gas exploration, and particularly to a fracture prediction method and system. Background Art

[0002] Fracture prediction is one of the core tasks in seismic exploration interpretation, which is directly related to the efficiency and benefit of oil and gas field exploration and development. The main method of fracture prediction is to utilize pre-stack and post-stack seismic data information, extract spatial attributes sensitive to fracture characterization for analysis, and at the same time provide important reference data for fault interpretation.

[0003] Generally speaking, fracture prediction techniques can be divided into post-stack prediction and pre-stack prediction. Post-stack prediction mainly uses 3D seismic data to calculate attributes sensitive to fracture characterization such as discontinuity geometry and image recognition for fracture prediction. Different fracture attribute volumes have different geological significances. For fracture characterization, different attributes reflect different response characteristics and different fracture levels. Pre-stack prediction mainly studies fractures within the target layer. Generally, based on the anisotropic characteristics of underground strata, the differences in attribute values such as time difference, velocity, amplitude, frequency, and wave impedance of pre-stack CMP or OVT gather azimuth angles are used for fracture and crack prediction. The main methods include pre-stack P-wave anisotropic ellipse fitting, anisotropic AVO inversion, anisotropic parameter inversion, etc., for semi-quantitatively predicting the development direction and intensity of fractures or quantitatively predicting anisotropic parameters.

[0004] In addition, in addition to single-attribute prediction, there is also a type of fracture prediction that comprehensively applies machine learning methods with multiple attributes. At present, machine learning methods have been applied to a certain extent in the field of seismic exploration. The more suitable ones are mainly two categories: pattern recognition and deep learning. From practice, strictly speaking, deep learning is based on a multi-layer neural network (more than 5 layers) with a large number of sample points (usually tens of thousands of sample points), and the required pre-input conditions are high. For prediction problems, when the number of sample points is insufficient, the calculation effect cannot be guaranteed. At this time, the pattern recognition method still has great use. Pattern recognition requires a small number of sample points and can achieve the prediction effect after supervised training. Therefore, for fracture prediction, there has been a neural network clustering method in the industry to obtain the fault body result, but the actual effect is not ideal and the high-precision fracture surface characteristics cannot be effectively obtained.

[0005] Therefore, the existing technology needs a new solution for predicting strike-slip fracture surfaces and classifying fractures, which can solve the problem of multiple interpretations of strike-slip fracture zone interpretation styles and / or the problem of uniqueness of single discontinuous fracture characterization. Summary of the Invention

[0006] To solve the above technical problems, the present invention provides a fracture prediction method, including: performing fracture enhancement and interpretive processing on 3D seismic data of a target work area; selecting key attributes for fracture prediction, performing attribute calculation on the fracture enhancement and interpretive processing data volume, and obtaining various attribute calculation results containing attribute value information of points at different positions; extracting various attribute values at fractures of different levels from the various attribute calculation results to form a sample matrix; determining determination rules for cross-sections of different levels according to the geological characteristics of the target work area and in combination with the various attribute calculation results, thereby forming a cross-section feature recognition library; constructing a fracture prediction model according to the sample matrix and the cross-section feature recognition library; obtaining the sample matrix corresponding to the area to be predicted, and using the fracture prediction model to perform scale-based level prediction on the fractures in the area to be predicted to obtain a fracture level prediction data volume.

[0007] Preferably, the key attributes include dip angle attribute, texture attribute, coherence attribute, fault enhancement attribute, ant body attribute, and maximum likelihood body attribute.

[0008] Preferably, in the step of performing fracture enhancement and interpretive processing on 3D seismic data of a target work area, according to the 3D seismic data, dip angle control enhancement processing and lateral enhancement processing are sequentially performed to obtain the fracture enhancement and interpretive processing data volume.

[0009] Preferably, in the step of extracting various attribute values at fractures of different levels from the various attribute calculation results to form a sample matrix, it includes: referring to well data of the target area, calibrating the attribute values of points at different positions in the various attribute calculation results in terms of well-seismic time and depth, and determining the positions of breakpoints, fractures, and non-breakpoints in the various attribute calculation results; according to the positions of breakpoints, fractures, and non-breakpoints in the various attribute calculation results, respectively extracting seismic attribute values at the wellbore trajectory from each calculation result diagram to form cross-well fracture attribute curves for different key attributes; extracting various attribute values at key positions including fracture positions and non-fracture positions of different levels from the cross-well fracture attribute curves of different key attributes, and based on this, obtaining the sample matrix.

[0010] Preferably, in the process of determining determination rules for cross-sections of different levels, referring to the statistical situation of breakpoints of different levels at the wellbore trajectory, imaging logging data, the offset of seismic profile in-phase axes, and the various attribute calculation results, the determination rules for cross-sections of different levels are determined by selecting main and secondary discriminant factors, qualitative display, or determining quantitative value ranges.

[0011] Preferably, when all the conditions involved in the main discriminant factors and the secondary discriminant factors meet three conditions, and at least one of these three conditions is a condition of the main discriminant factor, it is determined that the current prediction target is a first-class section; when all the conditions involved in the main discriminant factors and the secondary discriminant factors meet one or two conditions, and at least one of the one or two conditions is a main discriminant factor condition, it is determined that the current prediction target is a second-class section.

[0012] Preferably, the main discriminant factors include: the case where the leakage volume is greater than a preset value, the case where a section is shown in the imaging logging results, the case where a significant discontinuity of the in-phase axis is shown in the seismic profile, and the case where the attribute value in the calculation result of the fault enhancement type attribute is greater than the threshold value of this type of attribute; the secondary discriminant factors include: the case where the attribute value in the calculation result of the dip type attribute is greater than the threshold value of this type of attribute, the case where the attribute value in the calculation result of the texture type attribute is greater than the threshold value of this type of attribute, the case where the attribute value in the calculation result of the coherence type attribute is greater than the threshold value of this type of attribute, the case where the attribute value in the calculation result of the ant body type attribute is greater than the threshold value of this type of attribute, and the case where the attribute value in the calculation result of the maximum likelihood body type attribute is greater than the threshold value of this type of attribute.

[0013] Preferably, in the step of constructing a fracture prediction model according to the sample matrix and the section feature recognition library, it includes: preferably selecting a classifier model adapted to the current fracture prediction method, training the selected classifier model according to the sample matrix and the determination rules of different-level sections, and outputting model parameters after training to form the fracture prediction model, where the best classifier model is preferably the K-nearest neighbor classifier.

[0014] On the other hand, the present invention also provides a fracture prediction system, including: a seismic data interpretation module configured to perform fracture enhancement interpretive processing on the three-dimensional seismic data of the target work area; an attribute calculation module configured to preferably select key attributes for fracture prediction and calculate attributes for the fracture enhancement interpretive processing data volume to obtain various attribute calculation results containing attribute value information of different position points; a sample selection module configured to extract various attribute values at different-level fractures from the various attribute calculation results to form a sample matrix; a determination rule generation module configured to determine the determination rules of different-level sections according to the geological characteristics of the target work area and in combination with the various attribute calculation results, so as to form a section feature recognition library; a model construction module configured to construct a fracture prediction model according to the sample matrix and the section feature recognition library; and a fracture prediction module configured to obtain the sample matrix corresponding to the area to be predicted, and use the fracture prediction model to perform scale-based level prediction on the fractures in the area to be predicted to obtain a fracture level prediction data volume.

[0015] Preferably, the key attributes include dip angle attribute, texture attribute, coherence attribute, fault enhancement attribute, ant body attribute, and maximum likelihood body attribute.

[0016] Compared with the prior art, one or more embodiments of the above solution may have the following advantages or beneficial effects:

[0017] The present invention discloses a fracture prediction method and system. The method and system are a method for predicting fracture surfaces and classifying fractures using machine learning methods, which is applicable to fracture interpretation in strike-slip fault zones and also has certain reference significance for fields such as clastic rock reservoirs and volcanic rock reservoirs in complex fault-controlled basins. By comparing the effects of conventional fracture prediction, this method has effectiveness and practicality. On the one hand, due to the small fault displacement, complex longitudinal stacking, and planar segmentation in strike-slip fault zones, it is difficult to combine the cross-section space, and there are multiple sets of fault interpretation results in manual interpretation. Using the present invention, automatic tracking of fractures can be achieved, effectively solving the problems of cross-section space combination and the multi-solution problem of manual interpretation of strike-slip cross-sections. In addition, since a single fracture attribute often only reflects one aspect of the fracture, and strike-slip fault zones have various seismic responses such as in-phase axis offset, bending, and abnormal reflection, a single attribute cannot be used to comprehensively predict fractures and classify fractures. Therefore, the present invention uses multiple fracture attributes and their attribute data value ranges to comprehensively identify fracture types at different scale levels, solving the problem of the uniqueness of fracture characterization by a single discontinuous attribute.

[0018] Other features and advantages of the present invention will be described in the following specification, and in part, will be obvious from the specification, or will be understood by implementing the present invention. The objectives and other advantages of the present invention can be achieved and obtained through the structures specifically pointed out in the specification, claims, and drawings. Description of the Drawings

[0019] The drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the present invention and do not constitute a limitation to the present invention. In the drawings:

[0020] Figure 1 is a step diagram of the fracture prediction method according to an embodiment of the present application.

[0021] Figure 2 is a specific flowchart of the fracture prediction method according to an embodiment of the present application.

[0022] Figure 3 is a schematic diagram of the effect of fracture enhancement interpretive processing in the fracture prediction method according to an embodiment of the present application.

[0023] Figure 4 is a schematic diagram of the cross-section of the calculation results of various attributes in the fracture prediction method according to an embodiment of the present application.

[0024] Figure 5 It is a schematic diagram of the plane of the calculation results of various attributes in the fracture prediction method of the embodiment of the present application.

[0025] Figure 6 It is an example diagram of the sample optimization result in the fracture prediction method of the embodiment of the present application.

[0026] Figure 7 It is a cross-sectional view of the fracture prediction result in the fracture prediction method of the embodiment of the present application.

[0027] Figure 8 It is a plan view of the target layer of the fracture prediction result in the fracture prediction method of the embodiment of the present application.

[0028] Figure 9 It is a block diagram of the modules of the fracture prediction system of the embodiment of the present application. Detailed implementation manners

[0029] The following will combine the accompanying drawings and embodiments to detail the implementation manners of the present invention, so as to fully understand how the present invention uses technical means to solve technical problems and achieve the implementation process of technical effects and implement accordingly. It should be noted that as long as there is no conflict, the various embodiments in the present invention and the various features in each embodiment can be combined with each other, and the formed technical solutions are all within the protection scope of the present invention.

[0030] In addition, the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions. And although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0031] The present invention relates to fracture prediction technology. Fracture prediction is one of the core tasks in seismic exploration interpretation, which is directly related to the efficiency and benefit of oil and gas field exploration and development. The main method of fracture prediction is to use pre-stack and post-stack seismic data information, extract spatial attributes sensitive to fracture characterization for analysis, and at the same time provide important reference data for fault interpretation.

[0032] Generally speaking, fracture prediction techniques can be divided into post-stack prediction and pre-stack prediction. Post-stack prediction mainly uses 3D seismic data to calculate attributes sensitive to fracture characterization, such as discontinuity geometry and image recognition, for fracture prediction. Different fracture attribute volumes have different geological meanings. For fracture characterization, different attributes reflect different response characteristics and different fracture levels. Pre-stack prediction mainly studies the fractures within the target layer. Generally, based on the anisotropic characteristics of underground strata, the differences in attribute values such as time difference, velocity, amplitude, frequency, and wave impedance in the pre-stack CMP or OVT gather azimuth are used for fracture and crack prediction. The main methods include pre-stack P-wave anisotropic ellipse fitting, anisotropic AVO inversion, anisotropic parameter inversion, etc., for semi-quantitatively predicting the development direction and intensity of fractures or quantitatively predicting anisotropic parameters.

[0033] In addition, in addition to single-attribute prediction, there is also a type of fracture prediction that comprehensively applies machine learning methods with multiple attributes. At present, machine learning methods have been applied to a certain extent in the field of seismic exploration. The two main types that are more suitable are pattern recognition and deep learning. From practice, strictly speaking, deep learning is based on a multi-layer neural network (more than 5 layers) with a large number of sample points (usually tens of thousands of sample points), and the required pre-input conditions are high. For prediction problems, when the number of sample points is insufficient, the calculation effect cannot be guaranteed. At this time, the pattern recognition method still has great use. Pattern recognition requires a small number of sample points and can achieve the prediction effect after supervised training. Therefore, for fracture prediction, the neural network clustering method has been used in the industry to obtain the fault body results, but the actual effect is not ideal and the high-precision fracture surface characteristics cannot be effectively obtained.

[0034] To solve the above technical problems, the present invention proposes a fracture prediction method and system based on machine learning. The method and system include: performing fracture enhancement and interpretive processing on the three-dimensional seismic data of the target work area; selecting six key attributes for fracture prediction, calculating the attributes of the fracture enhancement and interpretive processing data volume, and obtaining various attribute calculation results (profiles or plane of the target layer) containing attribute value information of different position points; extracting various key attribute values of different-level fracture positions from various attribute calculation results to form a sample matrix, and completing the accurate reading of fracture attribute sample values; determining the judgment rules for different-level fault surfaces according to the geological characteristics of the target work area and combining various attribute calculation results, thereby establishing a fracture characteristic recognition library for the target work area; through multi-map joint analysis and quality control means, optimizing the initial prediction model, and constructing a fracture prediction model according to the sample matrix and the fracture characteristic recognition library; obtaining the sample matrix corresponding to the area to be predicted, and using the fracture prediction model to perform scale-based level prediction on the fractures in the new area to be predicted, and obtaining a fracture level prediction data volume including fracture prediction and fracture level division. Thus, the fracture prediction method provided by the present invention can perform fracture prediction and fracture grading on strike-slip fault surfaces, and solves the problems of multi-solution in the interpretation style of the existing strike-slip fault zone and / or the uniqueness of the representation of a single discontinuous fracture.

[0035] Figure 1 It is a step diagram of the fracture prediction method of the embodiment of the present application. Figure 2 It is a specific flow chart of the fracture prediction method of the embodiment of the present application. The following combines Figure 1 and Figure 2 to illustrate the fracture prediction method of the present invention.

[0036] Step S110: Perform fracture enhancement and interpretive processing on the three-dimensional seismic data of the target work area. Specifically, first, it is necessary to obtain the three-dimensional seismic data of the target work area to be predicted, and draw the seismic profile of the target work area and / or the plane map at the target layer to be predicted in the target work area. Then, perform fracture enhancement and interpretive processing on the currently obtained three-dimensional seismic data.

[0037] In step S110, it is necessary to perform dip control enhancement processing and lateral enhancement processing on the three-dimensional seismic data of the target work area in sequence to obtain a fracture enhancement and interpretive processing data volume. After collecting the seismic data of the target work area, first perform dip control enhancement processing.

[0038] Such as Figure 2As shown, in step S1, it is necessary to first calculate the similarity of adjacent traces according to the characteristics of formation dip angle and azimuth angle changes shown in 3D seismic data to complete the dip control enhancement process, thereby improving the seismic lateral signal-to-noise ratio and at the same time improving the fault identification accuracy. During the dip control enhancement process, it is necessary to fully consider the dip angle changes of each sampling point (in the seismic profile of the target work area or in the plan view at the target horizon to be predicted in the target work area), and select a comparison window centered on the calculation sample point. Each sampling point involved in the calculation is defined with a corresponding calculation weight. Further, in order to ensure that the calculation results of all sampling points are positive values, the relevant initial value of each sampling point is defined as 1, so that the correlation of the sampling point will generate a weight of 2 after the dip enhancement calculation is completed. Among them, for the weight calculation in the calculation sample point window, a weighted average method needs to be used. Specifically, first calculate the amplitude and weight of each dip control sampling point in the window, and then perform weighted average processing on the sampling points in all comparison time windows to obtain the weight corresponding to the currently selected window. Since the method of calculating the dip angle and azimuth angle is adopted, the ability of the dip enhancement processing result to depict faults is significantly enhanced, which is beneficial to the subsequent research on faults and fractures.

[0039] Further, after the dip enhancement processing is completed for the 3D seismic data of the target work area, on this basis, continue to perform lateral enhancement processing on the dip enhancement processing result. By stacking the signals along the seismic horizon and at the same time retaining the discontinuity and seismic wave impedance characteristics, the seismic data is smoothed, and the discontinuity (such as faults) is more easily identified. In this way, the dip control enhancement can further improve the quality of seismic data and is more conducive to better implementing horizon and fault interpretation.

[0040] Figure 3 It is a schematic diagram of the effect of fracture enhancement and interpretive processing in the fracture prediction method of the embodiment of the present application. As Figure 3 shown, Figure 3 The left figure clearly shows the fracture surface characteristics of the strike-slip fault zone. After the dip control enhancement processing, referring to Figure 3 the middle figure, while the seismic signal-to-noise ratio is significantly improved, the negative flower structure trace on layer A is significantly highlighted; below layer A, the main fault plane of the strike-slip zone is clearly visible and penetrates to the deep part. In addition, the associated fault planes existing on the periphery of the main fault plane can also be clearly shown. Continuing to refer to Figure 3 the right figure, on the basis of the dip control enhancement processing, perform horizon and fault enhancement technology to further denoise, highlight the fault plane, and exclude the fracture response characteristics outside the strike-slip zone range, so as to achieve clear fault plane depiction only effective within the main strike-slip zone range.

[0041] After the fracture enhancement interpretation processing is completed, the process proceeds to step S120. Step S120 preferably performs attribute calculation on the fracture enhancement interpretation processing data body based on the key attributes of fracture prediction, and obtains various attribute calculation results containing attribute value information of different position points.

[0042] Specifically, in step S120, it is necessary to select key attribute features related to the scale of section characterization based on the principle of the scale of section characterization. Each attribute feature is most sensitive to the fault type under a certain characterization scale. In an embodiment of the present invention, the section characterization scale is preferably three categories, specifically including strike-slip fault zones, large-scale faults and small-scale faults. In the preferred key attribute features, it is necessary to make each preferred key attribute feature reflect one of the fault types. Therefore, the six key attributes selected in the embodiment of the present invention include: dip attribute, texture attribute, coherence attribute, fault enhancement attribute, ant body attribute and maximum likelihood body attribute. Among them, dip attribute and texture attribute can reflect strike-slip fault zones; coherence attribute and fault enhancement attribute can reflect large-scale faults; and ant body and maximum likelihood body can reflect small-scale faults.

[0043] Furthermore, the dip attribute is to calculate the relative changes in dip and azimuth between various sample points of seismic data through complex channels, discrete scanning robust algorithms, etc., reflecting the fault and bend characteristics of the stratum, and describing the larger scale faults, folds, and structural styles. The texture attribute is based on the image texture characteristics of the grayscale co-occurrence matrix to analyze the contrast image changes between pixels and their speed of change. It is effective for the internal structure of the buried hill, large-scale fault system detection, lithological sedimentary body identification, and sedimentary phase characterization. The coherence body is to compare the similarity between channels, multiple channels, etc. in the Z direction or along the dip direction to realize the identification of faults, rivers, and other geological anomalies. The fault enhancement attribute is to improve the imaging of the fault interface by inputting the coherent data body after the dip and strike noise is controlled, and the direction-constrained ant colony algorithm is used to achieve high-precision continuous imaging of the fault and improve the recognition ability of the medium-scale fault. The ant body attribute is based on the positive feedback mechanism of the ant algorithm, and a model for optimizing the search using swarm intelligence is established to complete the tracking and identification of faults, which is very effective for the identification of small faults. The maximum likelihood volume attribute is to find the similarity between the voxel unit and the overall sampling point, retain the minimum similarity, and connect the voxels with similarity, so as to highlight the small differences of the voxels and detect small faults. These six attributes can well characterize faults from strike-slip fault zones to faults of different scales.

[0044] Further, after the above six attributes are optimized, step S120 will calculate the attributes according to the principle of the scale of section characterization, based on the principles that the fault attributes are well matched with the seismic phase axis misalignment, the strike-slip fault morphology characterized by the fault attributes is stable, the fault attributes outside the fault zone are weak, the fault surface characterized by the fault attributes is crisp, and each attribute has a certain degree of distinction, and the six attributes of dip attribute, texture attribute, coherence attribute, fault enhancement attribute, ant body attribute, and maximum likelihood body attribute are optimized; then, on the fault enhancement explanatory processing data volume generated in step S110, the dip angle calculation, texture attribute calculation, coherence attribute calculation, fault enhancement attribute calculation, ant body attribute calculation, and maximum likelihood body attribute calculation are performed on the fault enhancement explanatory processing data volume, respectively, to obtain various attribute calculation results (dip angle calculation results, texture attribute calculation results, coherence attribute calculation results, fault enhancement attribute calculation results, ant body attribute calculation results, and maximum likelihood body attribute calculation results). Among them, for each type of attribute calculation result, the calculation of each type of attribute is completed after the fracture enhancement explanatory processing is performed on the seismic profile of the target work area or the plan view of the target layer to be predicted in the target work area. Therefore, the calculation result of each type of attribute is a profile or plan view for each type of attribute, and in the profile or plan view of each type of attribute, there is attribute value (domain) information for the current attribute at different position points.

[0045] Figure 4 It is a schematic diagram of the cross-section of the calculation results of various attributes in the fracture prediction method of the embodiment of the present application. Figure 5 Schematic diagram of the plane of calculation results of various attributes in the fracture prediction method of the embodiment of the present application. Figure 4 and Figure 5 The dip angle attribute and texture attribute are more beneficial to the identification of strike-slip fault zones, so as to clarify the distribution range of faults; the coherence attribute and fault enhancement attribute can highlight the characteristics of larger-scale faults; while the ant body attribute and maximum likelihood body attribute are more sensitive to small faults and cracks.

[0046] After completing the calculation of various attributes, the process proceeds to step S130 to perform intensive processing on the sample values ​​of various fracture attributes. Step S130 extracts various key attribute values ​​at fractures of different scale levels from the calculation results of various fracture attributes obtained in step S120 to form a sample matrix.

[0047] In step S130, first, the well data of the target area needs to be obtained. Then, referring to the well data information, the attribute values of different position points in each type of attribute calculation result (profile or plan view) obtained in step S120 are calibrated for well-seismic time and depth, so as to determine the positions of breakpoints, fractures, and non-breakpoints in each type of attribute calculation result. Then, according to the positions of breakpoints, fractures, and non-breakpoints in each type of fracture attribute calculation result, the seismic attribute values at the wellbore trajectory are respectively extracted from each type of fracture attribute calculation result diagram to generate a cross-well fracture attribute curve for different key fracture attributes. That is to say, for each type of fracture attribute calculation result diagram, the attribute values at the same wellbore trajectory in the diagram are extracted, so as to construct a cross-well fracture attribute curve for six fracture attributes at the current wellbore trajectory. It should be noted that if there are multiple wells in each type of attribute calculation result, cross-well fracture attribute curves for six fracture attributes can be obtained for each well. Finally, for the same well, from the cross-well fracture attribute curves of six fracture attributes, the fracture attribute values of various key positions including different-level fracture positions and non-fracture positions are respectively extracted. Based on this, a sample matrix including all fracture attributes is obtained, thus completing the accurate reading of fracture attribute sample values for the current wellbore trajectory.

[0048] Specifically, after obtaining the attribute calculation result diagram of each type of fracture attribute, first perform well-seismic time-depth calibration processing, and on the basis of this processing, clarify the positions of breakpoints, fractures, and non-breakpoints; then, continuously extract six seismic attribute values at the well trajectory in each type of attribute calculation result diagram (which has completed time and depth calibration) to construct a cross-well fracture attribute curve, so as to accurately extract several specific fracture and non-fracture depth points (n) and their attribute characteristic values reflecting different levels of fractures on the well, and form a 6×n sample matrix. In one embodiment, if 26 different-level fracture points and non-fracture points are identified from the fracture attribute calculation result diagram for the same well, and the fracture attribute characteristic values of these 26 key positions are obtained, a 6×26 sample matrix is formed.

[0049] Figure 6 It is an example diagram of the sample optimization result in the fracture prediction method of the embodiment of the present application. The six attribute values at the position of 8139m of a certain well in the target work area are respectively 0, 255, 0.62, 0.79, 3.59, and 6.53.

[0050] After completing the construction of the sample matrix for each well in the target work area, step S140 is entered. In step S140, according to the geological characteristics of the target work area and combined with the various attribute calculation results obtained in step S120, the determination rules for cross-sections of different scale levels are determined, so as to form a cross-section feature recognition library (for the target work area).

[0051] Specifically, first, according to actual requirements, it is necessary to determine the levels of fractures of different scale levels included in the prediction result of the fracture prediction method described in the embodiments of the present invention, and the fracture types of each level. Specifically, the determined prediction types include three fracture types (the first type of fracture, the second type of fracture, and non-fracture). Among them, the first type of fracture refers to a fracture at a scale where the vertical throw is greater than 10 m or the fracture width is greater than 10 m. Such fractures can cause characteristics such as the dislocation and bending response of seismic profile isophase axes, and the general geometric attributes are recognizable fault surfaces. The second type of fracture refers to a fracture at a scale where the vertical throw is less than 10 m and the fracture width is less than 10 m. Such fractures cannot cause the characteristics of dislocation and bending response of seismic profile isophase axes, and individual high-precision geometric attributes are recognizable fault surfaces. Non-fracture refers to a fracture without vertical throw and fracture width. The seismic profile of such fractures has no characteristics of dislocation and bending response of isophase axes, and all geometric attributes cannot be recognized. After successively determining the fracture types of different scale levels appearing in the current prediction result (for example: specifying the three numerical values of 2, 1, and 0 to represent the three classifications of the first type of fault surface, the second type of fault surface, and non-fault surface respectively), thus, the establishment of the fracture prediction type is completed.

[0052] Furthermore, when determining the judgment rules for each prediction result type, specifically, in the actual application process, it is necessary to refer to the statistical situation of breakpoints of different scale levels at the same wellbore trajectory, imaging logging data, the dislocation situation of seismic profile isophase axes, and the attribute calculation results of various fracture attributes, and determine the judgment rules for identifying fault surfaces of different scale levels by selecting main discriminant factors, secondary discriminant factors, qualitative display, or determining quantitative value ranges.

[0053] In the embodiments of the present invention, the main discriminant factors have multiple first-type conditions, including: the situation where the lost circulation volume is greater than a preset value, the situation where a fault surface is displayed in the imaging logging result, the situation where an obvious dislocation of isophase axes is displayed in the seismic profile, and the situation where the attribute value in the calculation result of fault enhancement type attributes is greater than the threshold of this type of attribute. And the secondary discriminant factors have multiple second-type conditions, including: the situation where the attribute value in the calculation result of dip type attributes is greater than the threshold of this type of attribute, the situation where the attribute value in the calculation result of texture type attributes is greater than the threshold of this type of attribute, the situation where the attribute value in the calculation result of coherence type attributes is greater than the threshold of this type of attribute, the situation where the attribute value in the calculation result of ant body type attributes is greater than the threshold of this type of attribute, and the situation where the attribute value in the calculation result of maximum likelihood body type attributes is greater than the threshold of this type of attribute.

[0054] In the first embodiment, if among all the conditions involved in the primary discrimination factor and the secondary discrimination factor (i.e., the sum of all the first-class conditions and all the second-class conditions), three conditions are satisfied, and among these three conditions, there is at least one first-class condition of the primary discrimination factor, at this time, the current fracture prediction result is determined to be the first-class fracture surface (i.e., the large fracture surface).

[0055] In the second embodiment, if among all the conditions involved in the primary discrimination factor and the secondary discrimination factor (i.e., the sum of all the first-class conditions and all the second-class conditions), 1 or 2 conditions are satisfied, and among the 1 or 2 satisfied conditions, there is at least 1 first-class condition of the primary discrimination factor, at this time, the current fracture prediction result is determined to be the second-class fracture surface (i.e., the small fracture surface).

[0056] In the third embodiment, if neither the determination rule for the first-class fracture surface nor the determination rule for the second-class fracture surface is satisfied, at this time, the current fracture prediction result is determined to be a non-fracture surface.

[0057] For example, according to the actual situation of this work area, the determination principle for the fracture surface level is determined. The four factors of the lost circulation volume being greater than 30m 3 , the imaging logging showing a fracture surface, the profile showing an obvious dislocation of the in-phase axis, and the fault enhancement attribute value being greater than 100 are used as the primary discrimination factors, and the fracture response value ranges of other attributes are used as the secondary discrimination factors. When more than 3 discrimination factors are satisfied and at least 1 primary factor is included, it is defined as a large fracture surface, and the given value range is 2; when 1 to 2 discrimination factors are satisfied and at least 1 primary factor is included, it is defined as a small fracture surface, and the given value range is 1; in other cases, it is defined as a non-fracture surface, and the given value range is 0, thus establishing a fracture surface feature recognition library for the target work area.

[0058] In this way, after determining the type of the prediction result and the determination rules for each type of prediction result, the establishment of the fracture surface feature recognition library is achieved, and then it enters step S150.

[0059] Step S150: According to the sample matrix constructed in step S130 and the fracture surface feature recognition library constructed in step S140, construct a fracture prediction model to predict the fracture type at the un-drilled location in the target work area.

[0060] Specifically, first in step S150, a classifier model suitable for the current fracture prediction method needs to be selected. Among them, the best classifier model is preferably the K-nearest neighbor classifier.

[0061] Further, in the process of optimizing the classifier model, it is first necessary to select a variety of alternative classifiers, including: decision tree, discriminant analysis, support vector machine (SVM), K-nearest neighbor (KNN), ensemble classifier, naive Bayes, neural network, etc. Since the algorithms corresponding to each classifier have their corresponding applicable conditions, in the actual application process, the classifier is mainly optimized according to the accuracy of the actual model establishment. Then, seismic data near multiple historical wells in the target work area are collected and a corresponding sample matrix is obtained for each historical well. All sample matrices are divided into a test data group and a training data group. Using the training data group and the determined decision rules, each alternative classifier is trained, and the trained alternative classifiers are tested using the test data group. During the testing process of various alternative classifiers, multi-graph joint analysis quality control means can be used to display and compare the model test accuracy results of each alternative classifier. Among them, the multi-graph joint analysis quality control means include, but are not limited to: scatter plot, error matrix plot, ROC curve, and parallel coordinate plot. Thus, according to the comparison results, the best classifier model is selected.

[0062] After selecting the best classifier, step S150 will also train the selected classifier model according to the sample matrix generated in step S130 and the decision rules for cross-section at different scale levels obtained in step S140. After the training is completed, the model parameters are output to form a fracture prediction model. In this way, after obtaining the fracture prediction model, it enters step S160.

[0063] Step S160 obtains the sample matrix corresponding to the new area to be predicted, and uses the fracture prediction model to perform scale-based level prediction on the fractures in the new area to be predicted, obtaining a fracture level prediction data volume. In step S160, it enters the usage stage of the fracture prediction model. In this stage, it is necessary to obtain the 3D seismic data at other areas in the target work area that need to be fracture-predicted (for example: areas without drilled wells or without wells), and then process the 3D seismic data of the new area to be predicted according to the methods described in steps S110 to S130 to obtain the sample matrix for the current new area. Then, according to the sample matrix corresponding to the new area, the current sample matrix is input into the constructed fracture prediction model to obtain the fracture prediction result (fracture level prediction data volume) for the current new area, and identify the fracture types (large cross-section, small cross-section, or non-cross-section) at different positions in this new area. In this way, the division of the fracture levels in the new area to be predicted is completed.

[0064] Figure 7 It is a cross-sectional view of the fracture prediction result in the fracture prediction method of the embodiment of the present application. Figure 8 It is a plan view of the target layer of the fracture prediction result in the fracture prediction method of the embodiment of the present application. Figure 7The displayed is the cross-section prediction effect. It can be seen that the black part is the obvious fault plane, the gray position is the predicted crack position, and the corresponding relationship at the well point position has a good effect and a high degree of coincidence. Figure 8 What is displayed is Figure 3 the planar attribute of the fault level at the position of Layer A in Figure 8 As can be seen from

[0065] In this way, the embodiment of the present invention uses the above steps S110 to S160, starting from the fracture enhancement and optimization of seismic data, and through steps such as fracture attribute optimization, accurate selection of fracture attribute sample values, establishment of a fracture feature recognition library, establishment of a prediction model, and model output, to achieve fine prediction of the fault plane within the strike-slip fault zone and realize the division of fault plane levels, which is applicable to the fracture interpretation of the strike-slip fault zone and also has certain reference significance for fields such as clastic rock reservoirs and volcanic rock reservoirs in complex fault-controlled basins.

[0066] On the other hand, based on the above fracture prediction method, the present invention also proposes a fracture prediction system. Figure 9 It is the module block diagram of the fracture prediction system of the embodiment of the present application. As Figure 9 shown, the fracture prediction system described in the present invention includes: a seismic data interpretation module 101, an attribute calculation module 102, a sample optimization module 103, a decision rule generation module 104, a model construction module 105, and a fracture prediction module 106.

[0067] Further, the seismic data interpretation module 101 is implemented according to the method described in step S110 above, and is configured to perform fracture enhancement interpretive processing on the 3D seismic data of the target work area. The attribute calculation module 102 is implemented according to the method described in step S120 above, and is configured to preferably select the key attributes for fracture prediction, perform attribute calculation on the fracture enhancement interpretive processing data volume, and obtain various attribute calculation results containing the attribute value information of different position points. The sample optimization module 103 is implemented according to the method described in step S130 above, and is configured to extract the key attribute values of different levels of fractures from various attribute calculation results to form a sample matrix. The decision rule generation module 104 is implemented according to the method described in step S140 above, and is configured to determine the decision rules for different levels of fault planes according to the geological characteristics of the target work area and in combination with the above various attribute calculation results, thereby forming a fracture feature recognition library. The model construction module 105 is implemented according to the method described in step S150 above, and is configured to construct a fracture prediction model according to the above sample matrix and fracture feature recognition library. The fracture prediction module 106 is implemented according to the method described in step S160 above, and is configured to obtain the sample matrix corresponding to the area to be predicted, and use the constructed fracture prediction model to predict the fractures in the area to be predicted, and obtain the fracture level prediction data volume.

[0068] Among them, the above key attributes include: dip angle attribute, texture attribute, coherence attribute, fault enhancement attribute, ant body attribute, and maximum likelihood body attribute.

[0069] The present invention discloses a fracture prediction method and system. The method and system are a method for predicting fracture planes and classifying fractures using machine learning methods, which is applicable to fracture interpretation in strike-slip fault zones and also has certain reference significance for fields such as clastic rock reservoirs and volcanic rock reservoirs in complex fault-controlled basins. By comparing the conventional fracture prediction effects, this method has effectiveness and practicability. On the one hand, due to the small fault throw, complex vertical stacking, and planar segmentation in strike-slip fault zones, it is difficult to combine the fault planes in space, and there are multiple sets of fault interpretation results in manual interpretation. Using the present invention can achieve automatic fracture tracking, effectively solving the problem of fault plane space combination and the problem of multiple solutions in manual interpretation of strike-slip fault planes. In addition, since a single fracture attribute often only reflects one aspect of the fracture, and there are various seismic responses such as in-phase axis offset, bending, and abnormal reflection in strike-slip fault zones, a single attribute cannot be used to comprehensively predict fractures and classify fractures. Therefore, the present invention uses multiple fracture attributes and their attribute data value ranges to comprehensively identify fracture types at different scale levels, solving the problem of the uniqueness of the representation of single discontinuous attribute fractures.

[0070] As described above, only the preferred specific embodiments of the present invention are provided, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

[0071] It should be understood that the embodiments disclosed in the present invention are not limited to the specific structures, processing steps or materials disclosed herein, but should extend to equivalent alternatives of these features understood by those of ordinary skill in the relevant art. It should also be understood that the terms used herein are for the purpose of describing specific embodiments only and do not imply limitation.

[0072] The phrase "an embodiment" or "embodiments" mentioned in the specification means that the specific features, structures or characteristics described in connection with the embodiments are included in at least one embodiment of the present invention. Therefore, the phrases "an embodiment" or "embodiments" that appear throughout the specification do not necessarily all refer to the same embodiment.

[0073] Although the embodiments disclosed in the present invention are as above, the content described is only the embodiments adopted for the convenience of understanding the present invention and is not used to limit the present invention. Any person skilled in the technical field to which the present invention pertains may make any modifications and changes in the form of implementation and details without departing from the spirit and scope disclosed by the present invention. However, the patent protection scope of the present invention shall still be subject to the scope defined by the appended claims.

Claims

1. A fracture prediction method, comprising: Performing fracture enhancement interpretive processing on 3D seismic data of a target work area; Optimizing key attributes for fracture prediction, performing attribute calculations on the fracture enhancement interpretive processing data volume, and obtaining various attribute calculation results containing attribute value information of points at different positions; Extracting various attribute values at fractures of different levels from the various attribute calculation results to form a sample matrix; Determining determination rules for cross-sections of different levels according to the geological characteristics of the target work area and in combination with the various attribute calculation results, thereby forming a cross-section feature recognition library; Constructing a fracture prediction model according to the sample matrix and the cross-section feature recognition library; Obtaining the sample matrix corresponding to the area to be predicted, and using the fracture prediction model to perform scale-based level prediction on the fractures in the area to be predicted to obtain a fracture level prediction data volume, wherein In the step of determining the determination rules for cross-sections of different levels, it includes: Referring to the statistical situation of breakpoints of different levels at the wellbore trajectory, imaging logging data, the offset of the in-phase axis of the seismic profile, and the various attribute calculation results, and determining the determination rules for cross-sections of different levels by selecting main and secondary discriminant factors, qualitatively displaying or determining quantitative value ranges, wherein The main discriminant factors include: the situation where the lost circulation volume is greater than a preset value, the situation where a cross-section is shown in the imaging logging result, the situation where an obvious offset of the in-phase axis is shown in the seismic profile, and the situation where the attribute value in the fault enhancement type attribute calculation result is greater than the threshold of this type of attribute; The secondary discriminant factors include: the situation where the attribute value in the dip angle type attribute calculation result is greater than the threshold of this type of attribute, the situation where the attribute value in the texture type attribute calculation result is greater than the threshold of this type of attribute, the situation where the attribute value in the coherence type attribute calculation result is greater than the threshold of this type of attribute, the situation where the attribute value in the ant body type attribute calculation result is greater than the threshold of this type of attribute, and the situation where the attribute value in the maximum likelihood body type attribute calculation result is greater than the threshold of this type of attribute, wherein When all the conditions involved in the main discriminant factors and the secondary discriminant factors satisfy three conditions, and at least one of these three conditions is a condition of the main discriminant factor, it is determined that the current one to be predicted is a first type of cross-section; When all the conditions involved in the main discriminant factors and the secondary discriminant factors satisfy 1 or 2 conditions, and at least 1 of the 1 or 2 conditions is a condition of the main discriminant factor, it is determined that the current one to be predicted is a second type of cross-section.

2. The fracture prediction method according to claim 1, wherein The key attributes include dip angle attribute, texture attribute, coherence attribute, fault enhancement attribute, ant body attribute, and maximum likelihood body attribute.

3. The fracture prediction method according to claim 2, wherein In the step of performing fracture enhancement interpretive processing on the 3D seismic data of the target work area, according to the 3D seismic data, performing dip angle control enhancement processing and lateral enhancement processing in sequence to obtain the fracture enhancement interpretive processing data volume.

4. The fracture prediction method according to claim 1, wherein In the step of extracting various attribute values at fractures of different levels from the various attribute calculation results to form a sample matrix, it includes: Referring to the well data of the target area, perform well-seismic time and depth calibration on the attribute values of different position points in the calculation results of the various types of attributes, and determine the positions of breaks, fractures, and non-breaks in the calculation results of the various types of attributes; According to the positions of breaks, fractures, and non-breaks in the calculation results of the various types of attributes, extract the seismic attribute values at the wellbore trajectory from each calculation result graph respectively to form cross-well fracture attribute curves for different key attributes; Extract the attribute values of various types at the key positions including different-level fracture positions and non-fracture positions from the cross-well fracture attribute curves of the different key attributes, and based on this, obtain the sample matrix; 5. The fracture prediction method according to any one of claims 1 to 4, characterized in that, In the step of constructing a fracture prediction model according to the sample matrix and the cross-section feature recognition library, it includes: Screen a classifier model suitable for the current fracture prediction method, train the selected classifier model according to the sample matrix and the determination rules of different-level cross-sections, and output the model parameters after training to form the fracture prediction model, where the selected classifier model is the K-nearest neighbor classifier; 6. A fracture prediction system, including: A seismic data interpretation module configured to perform fracture enhancement interpretive processing on the 3D seismic data of the target work area; An attribute calculation module configured to optimize the key attributes for fracture prediction and perform attribute calculation on the fracture enhancement interpretive processing data volume to obtain various attribute calculation results containing attribute value information of different position points; A sample optimization module configured to extract the attribute values of various types at different-level fractures from the various attribute calculation results to form a sample matrix; A determination rule generation module configured to determine the determination rules of different-level cross-sections according to the geological characteristics of the target work area and in combination with the various attribute calculation results, so as to form a cross-section feature recognition library; A model construction module configured to construct a fracture prediction model according to the sample matrix and the cross-section feature recognition library; A fracture prediction module configured to obtain the sample matrix corresponding to the area to be predicted, and use the fracture prediction model to perform scale-based level prediction on the fractures in the area to be predicted to obtain a fracture level prediction data volume, where In the process of determining the determination rules of different-level cross-sections, it includes: Referring to the statistical situation of different-level breaks at the wellbore trajectory, imaging logging data, the dislocation of the in-phase axis of the seismic profile, and the calculation results of the various types of attributes, determine the determination rules of different-level cross-sections by selecting main and secondary discriminant factors, qualitative display, or determining quantitative value ranges, where The main discriminant factors include: the situation where the leakage volume is greater than a preset value, the situation where a cross-section is shown in the imaging logging results, the situation where an obvious dislocation of the in-phase axis is shown in the seismic profile, and the situation where the attribute value in the calculation result of the fault enhancement type attribute is greater than the threshold value of this type of attribute; The secondary discriminant factors include: the situation where there is an attribute value greater than the class attribute threshold in the calculation result of the dip angle class attribute, the situation where there is an attribute value greater than the class attribute threshold in the calculation result of the texture class attribute, the situation where there is an attribute value greater than the class attribute threshold in the calculation result of the coherence class attribute, the situation where there is an attribute value greater than the class attribute threshold in the calculation result of the ant body class attribute, and the situation where there is an attribute value greater than the class attribute threshold in the calculation result of the maximum likelihood body class attribute. Among them, When all the conditions involved in the primary discriminant factor and the secondary discriminant factor meet three conditions, and at least one of these three conditions is a condition of the primary discriminant factor, it is determined that the current item to be predicted is the first type of section; When all the conditions involved in the primary discriminant factor and the secondary discriminant factor meet one or two conditions, and at least one of the one or two conditions is a condition of the primary discriminant factor, it is determined that the current item to be predicted is the second type of section.

7. The fracture prediction system according to claim 6, wherein The key attributes include dip angle attribute, texture attribute, coherence attribute, fault enhancement attribute, ant body attribute, and maximum likelihood body attribute.

Images