A natural fracture intelligent identification correction method, system, device and storage medium

Abstract

The application discloses a natural fracture intelligent identification correction method, system, device and storage medium. The method comprises the following steps: acquiring microseismic monitoring data in a hydraulic fracturing process, an initial natural fracture model, fracturing construction data and fault calibration data, and performing data quality control to obtain quality-controlled microseismic event data; determining a direct disturbance area by using a multi-source constraint method according to the microseismic event data; identifying an abnormal area according to the direct disturbance area; performing feature extraction on the microseismic event data to obtain a time-space aggregation feature; constructing a training sample set; training by using machine learning according to the training sample set and the time-space aggregation feature to obtain a natural fracture intelligent identification correction model; and predicting the abnormal area according to the natural fracture intelligent identification correction model to obtain a natural fracture existence probability prediction result. The application can improve the accuracy of natural fracture existence probability prediction.

Description

Technical Field

[0001] This invention relates to the field of unconventional oil and gas reservoir development technology, and in particular to a method, system, device and storage medium for intelligent identification and post-hoc correction of natural fractures. Background Technology

[0002] In the development of unconventional oil and gas reservoirs (shale gas, tight gas, coalbed methane, etc.), hydraulic fracturing is a key technology for improving reservoir productivity. Natural fractures, as widely distributed discontinuous structures in reservoirs, have a significant impact on the effectiveness of hydraulic fracturing: on the one hand, natural fractures can serve as advantageous seepage channels for fluids, facilitating the formation of complex fracture networks and increasing the stimulation volume; on the other hand, natural fractures can also become channels for long-distance fluid conduction, leading to engineering geological hazards such as inter-well channeling and casing deformation.

[0003] Currently, the identification and modeling of natural cracks mainly rely on the following methods: (1) Geological modeling method: Based on geological data such as outcrop observation and core description, a statistical distribution model of natural fractures is established; (2) Well logging interpretation methods: using data such as imaging logging and conventional logging curves to identify natural fractures around the wellbore; (3) Seismic attribute analysis method: using seismic attributes such as ant body tracking, coherence body, and fracture likelihood body to predict the spatial distribution of natural cracks; (4) Commercial software integrated modeling: such as Petrel software, which integrates multiple data sources to build a three-dimensional model of natural cracks.

[0004] However, the above methods are all "a priori" prediction methods before fracturing operations, and have the following shortcomings: (1) High uncertainty: Due to limitations in data resolution and interpretation methods, the natural crack model has high uncertainty and it is difficult to accurately depict the spatial distribution of cracks; (2) Lack of verification methods: After fracturing, there is a lack of effective methods to use actual response data to verify or correct the natural fracture model; (3) High subjectivity: When engineering anomalies occur on site (such as inter-well pressure channeling, casing deformation, etc.), engineers usually suspect that there are omissions in the natural fracture model. However, the correction of the model mainly relies on expert experience, which is highly subjective and lacks objective basis. (4) Unable to form a closed loop: The lack of a reverse analysis method of "fracturing response → natural fracture identification" makes it impossible to form a data-driven closed-loop optimization.

[0005] In the prior art, patent CN118691561A discloses a method for analyzing the fracturing effect based on integrated shale gas geological engineering. This method superimposes microseismic data and natural fractures into a hydraulic fracture fracturing test model, collects geometric image information of the hydraulic fracture after fracturing, and determines whether the fracturing is qualified based on the image overlap. However, this method has the following limitations: (1) Differences in technical approaches: The core technical solution of CN118691561A is image processing, threshold judgment and parameter adjustment. It mainly judges the fracturing effect by comparing the overlap of hydraulic fracture images with a preset threshold, which is a deterministic rule judgment method. In contrast, this invention uses machine learning technology to establish a nonlinear mapping relationship between the spatiotemporal aggregation characteristics of microseismic earthquakes and the existence of natural fractures through training models. It can handle complex multidimensional feature spaces and has stronger generalization ability and adaptability.

[0006] (2) Analysis of differences in objectives: CN118691561A focuses on evaluating whether the fracturing effect is qualified, and focuses on whether the geometry of the hydraulic fracture (fracture length, fracture width, cluster spacing, etc.) meets expectations; while this invention focuses on identifying the existence of natural fractures, and focuses on finding natural fractures that the initial model failed to identify in the fracturing response data, so as to achieve the a posteriori correction of the model.

[0007] (3) Differences in region division methods: CN118691561A does not involve spatial region division based on physical mechanisms; while this invention innovatively proposes an elliptical (ellipsoidal) direct disturbance region division method based on fluid pressure increment, which combines multi-source constraints (pressure diffusion theory, empirical radius, event rate dynamic correction) to determine the boundary, and has clear physical meaning and engineering interpretability.

[0008] (4) Difference in output results: CN118691561A outputs a binary judgment result of fracturing qualification / failure; while the present invention outputs a continuous value distribution of the probability of the existence of natural fractures, which can provide richer quantitative information and uncertainty assessment for engineering decision-making.

[0009] Therefore, there is an urgent need for a method that can use actual response data from fracturing operations (such as microseismic monitoring data) to reverse-identify and correct natural fracture models, providing objective and quantifiable judgment basis for engineering decisions. Summary of the Invention

[0010] This invention provides a method, system, device, and storage medium for intelligent identification and correction of natural cracks. By analyzing the microseismic response patterns induced by hydraulic fracturing, machine learning technology is used to establish a mapping relationship between the spatiotemporal clustering characteristics of microseisms and the existence of natural cracks, thereby obtaining highly accurate prediction results of the probability of the existence of natural cracks. This improves the accuracy of the prediction of the probability of the existence of natural cracks and provides an objective and quantifiable basis for engineering decision-making.

[0011] To achieve the above objectives, the present invention provides the following solution: A method for intelligent identification and correction of natural cracks, the method comprising: Acquire microseismic monitoring data, initial natural fracture model, fracturing construction data, and fault calibration data during the hydraulic fracturing process, and perform data quality control to obtain quality-controlled microseismic event data; The direct disturbance zone was determined using a multi-source constraint method based on the microseismic event data. Based on the direct disturbance area, identify the abnormal region; Feature extraction is performed on the microseismic event data to obtain spatiotemporal clustering features; Construct a training sample set; A natural crack intelligent identification and correction model is obtained by training the training sample set and the spatiotemporal clustering features using machine learning. The abnormal area is predicted based on the intelligent identification and correction model for natural cracks, and the probability prediction result of the existence of natural cracks is obtained.

[0012] To achieve the above objectives, the present invention also provides the following solution: A natural crack intelligent identification and correction system, the system comprising: The microseismic event data acquisition module is used to acquire microseismic monitoring data, initial natural fracture models, fracturing construction data, and fault calibration data during the hydraulic fracturing process, and to perform data quality control to obtain quality-controlled microseismic event data. The direct disturbance zone determination module is used to determine the direct disturbance zone based on the microseismic event data using a multi-source constraint method; An abnormal region identification module is used to identify abnormal regions based on the direct disturbance region; The spatiotemporal clustering feature extraction module is used to extract features from the microseismic event data to obtain spatiotemporal clustering features; The training sample set construction module is used to construct the training sample set; The intelligent identification and correction model determination module for natural cracks is used to train the intelligent identification and correction model for natural cracks by using machine learning based on the training sample set and the spatiotemporal clustering features. The natural crack existence probability prediction result determination module is used to predict the abnormal area based on the natural crack intelligent identification and correction model to obtain the natural crack existence probability prediction result.

[0013] To achieve the above objectives, the present invention also provides the following solution: A computer device includes a memory and a processor, the memory storing a computer program and the processor running the computer program to enable the computer device to perform a natural crack intelligent identification and correction method.

[0014] To achieve the above objectives, the present invention also provides the following solution: A computer-readable storage medium stores a computer program that is executed by a processor to implement a method for intelligent identification and correction of natural cracks.

[0015] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects: This invention provides a method for intelligent identification and correction of natural fractures. The method includes: acquiring microseismic monitoring data, an initial natural fracture model, fracturing construction data, and fault calibration data during hydraulic fracturing, and performing data quality control to obtain quality-controlled microseismic event data; determining the direct disturbance zone based on the microseismic event data using a multi-source constraint method; identifying anomalous regions based on the direct disturbance zone; extracting features from the microseismic event data to obtain spatiotemporal clustering features; constructing a training sample set; training a natural fracture intelligent identification and correction model using machine learning based on the training sample set and the spatiotemporal clustering features; and predicting the anomalous regions based on the natural fracture intelligent identification and correction model to obtain a natural fracture existence probability prediction result. This invention can improve the accuracy of predicting the probability of natural fracture existence. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A schematic diagram of the process for intelligent identification and correction of natural cracks; Figure 2 This is a schematic diagram of the spatial region division proposed in Embodiment 1; Figure 2 Figure (a) shows a schematic diagram of staged fracturing in a horizontal well and the spatial distribution of natural fractures; Figure 2 Figure (b) shows the spatial distribution of hydraulic fractures and the spatial distribution of microseismic sources induced by water injection; And the delineation of the direct disturbance region (elliptical); Figure 2 (c) shows a schematic diagram of the identification of the direct disturbance anomaly region; Figure 3This is a schematic diagram of the microseismic time-series response feature extraction proposed in Example 1; Figure 4 This is a schematic diagram of the machine learning model structure proposed in Example 1; Figure 5 The results are the verification results of the model proposed in Example 1; Figure 5 Figure (a) shows the confusion matrix for cross-validation; Figure 5 Figure (b) shows the ROC curve and AUC value; Figure 6 This is the probability prediction result of the existence of natural cracks proposed in Example 1; Figure 7 This is a schematic diagram of the structure of a natural crack intelligent identification and correction system.

[0018] Explanation of the labels in the attached diagram: 1-Natural fracture; 2-Perforation point; 3-Hydraulic fracture; 4-Direct disturbance zone; 5-Microseismic source (Richter magnitude M>0); 6-Microseismic source (Richter magnitude M<0); 7-Anomalous area. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0021] Example 1: This embodiment provides a method for intelligent identification and correction of natural cracks. This method analyzes the microseismic response patterns induced by hydraulic fracturing and uses machine learning technology to achieve intelligent evaluation of the existence of natural cracks. Figure 1 This is a schematic diagram of the process for intelligent identification and correction of natural cracks, as shown below. Figure 1 As shown, the method specifically includes the following steps 101-107, which are described in detail below: Step 101: Acquire microseismic monitoring data, initial natural fracture model, fracturing construction data, and fault calibration data during the hydraulic fracturing process, and perform data quality control to obtain quality-controlled microseismic event data.

[0022] Step 101 involves obtaining the following data for the target shale gas reservoir area (taking the southern Sichuan shale gas plate as an example in this embodiment): (1) Acquisition of microseismic monitoring data Acquire microseismic monitoring data during the hydraulic fracturing process, including: Location of the epicenter: X, Y, Z three-dimensional coordinates, in meters (m); Earthquake occurrence time: timestamp accurate to the second; Magnitude: M on the Richter scale; Positioning residual: an indicator used to evaluate positioning accuracy.

[0023] In this embodiment, microseismic monitoring data were acquired from 8 wells, with the number of microseismic events per well ranging from 500 to 2000, totaling approximately 12,000 valid events.

[0024] (2) Acquisition of the initial natural crack model To obtain the initial natural fracture model established before fracturing, in this embodiment, the natural fracture distribution model generated by Petrel software based on seismic ant body attribute tracking is used, which includes parameters such as the spatial location, orientation, and dip angle of the fractures.

[0025] (3) Data acquisition for fracturing operations Obtain relevant data for fracturing operations, including: Horizontal well trajectory: a sequence of three-dimensional coordinates of the wellbore; Segment information: the start and end depths and length of each fracturing segment; Fracturing time window: the start and end times of each fracturing segment; Injection parameters: injection volume (m³), discharge rate (m³ / min), etc.; Reservoir parameters: permeability, porosity, overall compressibility, reservoir thickness, etc. (4) Fault calibration data acquisition Obtain the location and extent of faults determined by seismic interpretation and well logging interpretation, and use this information to exclude areas affected by faults.

[0026] (5) Data quality control Quality screening of microseismic data: Positioning residual filtering: retain events with positioning residuals less than 50m; Fault exclusion: Excluding events within 100m of the fault; Earthquake Magnitude Classification and Processing: For event rate calculation: use microseismic events of all magnitudes (including events with M<0); For anomaly area identification: filter events with magnitude M ≥ 0.

[0027] After quality control, approximately 10,000 valid microseismic event data points were retained for subsequent analysis.

[0028] Step 102: Determine the direct disturbance zone using the multi-source constraint method based on the microseismic event data.

[0029] The multi-source constraint method includes a union of one or more of the following: a calculation method based on pressure diffusion theory, an estimation method based on empirical radius, and a dynamic correction method based on microseismic event rate. In three-dimensional space, the direct disturbance zone is ellipsoidal in shape; on a two-dimensional projection plane, the direct disturbance zone is elliptical, with the major axis of the ellipse parallel to the direction of hydraulic fracture propagation and the minor axis parallel to the wellbore direction. Unlike the prior art CN118691561A, this embodiment uses a multi-source constraint method based on physical mechanisms for spatial region division, rather than a simple image processing method. CN118691561A determines the region based on image overlap, lacking consideration of the physical process of fluid pressure propagation; while this embodiment establishes an elliptical direct disturbance zone model based on the pressure diffusion equation, which has clear physical meaning.

[0030] Figure 2 This is a schematic diagram of the spatial region division proposed in Embodiment 1; Figure 2 Figure (a) shows a schematic diagram of staged fracturing in a horizontal well and the spatial distribution of natural fractures; Figure 2 Figure (b) shows the spatial distribution of hydraulic fractures and the spatial distribution of water injection-induced microseismic sources, as well as the division of the direct disturbance zone (elliptical). Figure 2 Figure (c) shows a schematic diagram of the identification of the direct disturbance anomaly region.

[0031] (1) Physical definition of the direct disturbance region The direct disturbance zone is defined as the matrix fluid pressure increment being greater than a preset threshold. The region. In this embodiment, the region is taken. = 1 MPa.

[0032] Due to the high permeability of hydraulic fractures, the diffusion rate of fluid pressure along the fracture direction is much greater than the diffusion rate perpendicular to the fracture direction. Therefore: In three-dimensional space, the directly disturbed region is ellipsoidal in shape; On the two-dimensional projection plane, the directly disturbed zone is elliptical, with the major axis of the ellipse parallel to the direction of hydraulic fracture propagation and the minor axis parallel to the wellbore direction.

[0033] (2) Multi-source constraint method To improve the reliability of determining the boundary of the directly disturbed region, a multi-source constraint method is adopted, which involves taking the union of the regions determined by the following three methods: Method A: Theoretical calculation based on the pressure diffusion equation According to the pressure diffusion equation, the spatial distribution of fluid pressure increment can be expressed as: in:r The distance from the injection point. t For injection time, η The pressure diffusion coefficient is... erfc This is a complementary error function.

[0034] Considering anisotropic permeability, the pressure diffusion coefficients along the fracture direction (major axis) and perpendicular to the fracture direction (minor axis) are respectively: in: Permeability along the fracture direction, For matrix permeability, φ Porosity μ For fluid viscosity, This is the overall compression coefficient.

[0035] make The major semi-axis a and minor semi-axis b of the ellipse are obtained by solving: Incremental pressure at injection point The calculation formula is: in: Q To inject displacement, h For reservoir thickness, r w Where is the radius of the wellbore.

[0036] In this embodiment, the following parameter values ​​are used: Permeability along the fracture direction k f = 100 mD Matrix permeability k m = 0.1 mD Porosity φ = 0.05 fluid viscosity μ = 1 mPa·s Overall compression ratio c t = 1×10⁻ 4 MPa⁻¹ Injection displacement Q = 12 m³ / min reservoir thickness h = 40 m Wellbore radius r w = 0.1 m Injection time t = 120 min Pressure increment threshold ΔP th = 1 MPa The major semi-axis of the ellipse was calculated. a ≈ 280 m, short half-axis b ≈ 9 m.

[0037] Method B: Estimation method based on empirical radius Based on the injected fluid volume V and reservoir parameters during fracturing, estimate the half-length of hydraulic fracture propagation: in: V The volume of liquid injected in a single segment is (m³). h The reservoir thickness is in meters (m). φ frac The crack aperture is in meters (m).

[0038] In this embodiment, the single-stage injection volume is approximately 2000 m³, the reservoir thickness is 40 m, the estimated fracture aperture is 0.005 m, and the calculation yields... L ≈ 180 m, rounded to the nearest integer L = 200 m.

[0039] The distance from the fracturing section is less than L The region is used as the empirically estimated boundary of the direct disturbance zone.

[0040] Method C: Dynamic Correction Method Based on Microseismic Event Rate Considering that the theoretically calculated elliptical boundary may not be accurate in actual strata (e.g., the fluid path changes after hydraulic fractures communicate with natural fractures), the microseismic event rate is used for dynamic correction.

[0041] Calculate the microseismic event rate for each spatial grid cell: in: For grid cells i The number of microseismic events of all magnitudes within the region. For the volume of the grid cell, T For monitoring time windows.

[0042] Set dynamic threshold: Event rate > The area is included in the direct disturbance zone.

[0043] (3) Union of multi-source constraints The final boundary of the direct disturbance region is the union of the regions determined by the three methods mentioned above: in: For the elliptical region determined by method A, The circular region is defined by method B. This refers to the high event rate region determined by method C.

[0044] The reason for using the union method is that any area identified as a direct disturbance by any method should be excluded from the scope of anomaly area identification in order to avoid misjudging microseisms caused by the propagation of hydraulic fractures as activation of natural fractures.

[0045] Step 103: Identify abnormal regions based on the directly disturbed region. This step specifically includes: Based on the direct disturbance zone, identify anomaly areas located outside the direct disturbance zone that exhibit microseismic responses and whose initial natural crack models show no crack distribution; the presence of microseismic responses means that the number of microseismic events with magnitudes greater than or equal to a preset threshold within the area is greater than or equal to a preset threshold.

[0046] like Figure 2 As shown in (c), outside the direct disturbance zone, regions that meet the following conditions are identified as "abnormal regions": (1) Definition of abnormal regions Abnormal regions must simultaneously meet the following conditions: Located outside the direct disturbance zone; Microseismic response exists: the number of microseismic events in the region is ≥5; The initial model of natural cracks showed that there were no natural cracks in the area.

[0047] (2) Engineering significance of abnormal areas The presence of the anomalous region indicates that: There may be natural cracks in this area that were not identified by the initial model; Fluids are conducted over long distances through unidentified natural fractures, inducing microseismic responses; It may be related to abnormalities in the on-site engineering (pressure leakage, casing deformation, etc.).

[0048] In this embodiment, a total of 15 abnormal areas were identified, which coincided with the locations of engineering abnormal events (inter-well pressure cross-flow, casing deformation) recorded on site.

[0049] Step 104: Extract features from the microseismic event data to obtain spatiotemporal clustering features.

[0050] The spatiotemporal clustering characteristics include spatial and temporal characteristics. The spatial characteristics include the number of microseismic events, the spatial density of events, the degree of spatial clustering, and the distance to the nearest fracturing segment. The temporal characteristics include the time delay between the time of event occurrence and the construction time of the corresponding fracturing segment, as well as the distribution characteristics of the event time series. Figure 3 This is a schematic diagram of the microseismic time-series response feature extraction proposed in Example 1.

[0051] Extracting feature vectors reflecting "spatiotemporal clustering characteristics" from microseismic data specifically includes: (1) Spatial feature extraction For each analysis region, the following spatial features are extracted: (2) Temporal feature extraction For each analysis region, the following temporal features are extracted: (3) Optional feature extraction Depending on data availability, the following features may be selected for extraction: b Value characteristics: calculated based on the Gutenberg-Richter relationship in: N For magnitude greater than M The number of events, a and b These are the fitting parameters. b The value can be used to distinguish the type of breakage: b Higher values ​​(>1.5): Indicate a tendency towards tensile fracturing, possibly related to hydraulic fracture propagation. b Lower values ​​(<1.0): Predispose to shear fracture, possibly associated with activation of natural fractures. Focal mechanism characteristics: If focal mechanism solution data is available, the following can be extracted: DC component ratio: The proportion of the dual-couple component, reflecting the degree of shear fracture. CLVD component ratio: The proportion of the compensating linear vector dipole component, reflecting the degree of tensile fracturing.

[0052] (3) Feature normalization The extracted features are then subjected to Min-Max normalization: After normalization, all eigenvalues ​​are in the range [0,1].

[0053] Step 105: Construct the training sample set. This step includes: (1) Construction of positive samples Regions that meet the following conditions are selected as positive samples: Located outside the direct disturbance zone; The initial model of natural cracks showed the presence of natural cracks; Microseismic responses exist (number of events ≥ 5); Tag: Cracked = 1.

[0054] In this embodiment, a total of 156 positive samples were constructed.

[0055] (2) Negative sample construction Regions that meet the following conditions are selected as negative samples: Located within the direct disturbance zone; Microseismic response was observed (confirmed to be caused by the propagation of hydraulic fractures). Tag: No natural cracks = 0.

[0056] In this embodiment, a total of 312 negative samples were constructed.

[0057] (3) Sample balancing Since the number of positive and negative samples is unbalanced (1:2), the SMOTE oversampling method is used to expand the positive samples so that the ratio of positive to negative samples reaches 1:1.

[0058] Step 106: Train the model using machine learning based on the training sample set and the spatiotemporal clustering features to obtain the intelligent identification and correction model for natural cracks.

[0059] Figure 4 This is a schematic diagram of the machine learning model structure proposed in Example 1, as shown below. Figure 4 As shown, a machine learning model is constructed to establish a mapping relationship between the spatiotemporal clustering characteristics of microseismic events and the probability of the existence of natural fractures. The specific process of training the machine learning model is as follows: (1) Model selection This embodiment uses a fully connected neural network (ANN) as the classification model, and the network structure is as follows: (2) Model training Training parameter settings: Loss function: Binary Cross-Entropy Optimizer: Adam, learning rate 0.001 Batch size: 32 Number of training rounds: 100 Validation set ratio: 20% During training, an early stopping strategy is adopted, and training is stopped when the validation set loss does not decrease for 10 consecutive rounds.

[0060] (3) Model output The model output is the probability P of the existence of natural cracks, with a value range of [0,1]: A value close to 1 indicates that the area is likely to contain natural fissures. A P value close to 0 indicates that the microseismic response in the region is more likely to be caused by the propagation of hydraulic fractures rather than activation by natural fractures. P being in the middle indicates uncertainty and requires comprehensive judgment in conjunction with other information.

[0061] Step 107: Based on the intelligent identification and correction model for natural cracks, predict the abnormal area to obtain the probability prediction result of the existence of natural cracks.

[0062] Figure 5 The results are the verification results of the model proposed in Example 1; Figure 5 Figure (a) shows the confusion matrix for cross-validation; Figure 5 Figure (b) shows the ROC curve and AUC value; as shown Figure 5 As shown, the trained model is validated to evaluate its predictive performance. Figure 6 This is the probability prediction result of the existence of natural cracks proposed in Example 1. This step specifically employs the following method: (1) K-fold cross-validation The generalization ability of the model is evaluated using a 5-fold cross-validation method: Verification steps: The sample set is randomly divided into 5 subsets of equal size; Four subsets are used alternately as the training set and one subset is used as the validation set. Repeat 5 times, each time using a different validation set; Calculate the average of the 5 verification results.

[0063] Wherein: TP is a true positive (predicted to have cracks and actually has cracks), TN is a true negative, FP is a false positive, and FN is a false negative.

[0064] (2) Engineering verification The high-probability areas predicted by the model were compared and verified with the locations of anomalies in the actual engineering projects. Verification method: Collect and record engineering anomalies on-site, including inter-well pressure flow and casing deformation; Determine the spatial location of abnormal events in the project; The coincidence rate between high-probability prediction areas (P≥0.7) and abnormal engineering locations was statistically analyzed.

[0065] Verification results: In this embodiment, a total of 18 engineering abnormal events were recorded, among which: 15 cases were located within the high-probability region predicted by the model, with a consistency rate of 83.3%. Three cases are located in the medium probability region (0.4≤P<0.7). 0 is located in the low probability region (P<0.4).

[0066] Engineering verification results show that the high-probability areas predicted by the model have good consistency with the actual anomaly locations in the engineering, verifying the effectiveness of the method.

[0067] Compared with the prior art document CN118691561A, the present invention has the following essential differences: (1) Different technical approaches: CN118691561A uses image processing and threshold judgment methods to make binary judgments by comparing the overlap of hydraulic crack images with preset thresholds; this invention uses machine learning technology to establish a nonlinear mapping relationship between the spatiotemporal clustering characteristics of microseismic events and the probability of the existence of natural cracks, and outputs continuous probability values.

[0068] (2) Different regional division methods: CN118691561A does not involve spatial regional division based on physical mechanisms; this invention innovatively proposes a multi-source constraint method (pressure diffusion theory + empirical radius + union of event rate dynamic correction) to determine the boundary of the elliptical direct disturbance zone, which has clear physical significance.

[0069] (3) Different analysis objectives: CN118691561A focuses on evaluating whether the fracturing effect is qualified; this invention focuses on identifying the existence of natural fractures, realizing a closed loop from "prior prediction" to "posterior correction".

[0070] The intelligent identification and correction method for natural cracks in this embodiment has the following beneficial effects: (1) Achieving a posteriori correction: For the first time, it is proposed to use the microseismic response data induced by hydraulic fracturing to identify the existence of natural fractures in reverse, realizing a closed loop from "a priori prediction" to "post-hoc correction", filling the gap in existing technology; (2) Provide objective evidence: The probability of the existence of natural cracks is output by the machine learning model, which provides a data-driven objective basis for the correction of the natural crack model and avoids the uncertainty of human subjective judgment; (3) Scientific regional division: The boundary of the direct disturbance zone is calculated based on the spatial distribution of fluid pressure increment. The use of elliptical boundary is more in line with the actual physical laws, and the reliability of boundary determination is improved by multi-source constraint method; (4) Focus on abnormal areas: Analyze abnormal areas that have microseismic response but the crack model shows no cracks, directly serving the interpretation and risk assessment of engineering anomalies (pressure cross, casing deformation, etc.); (5) It has universality: the method does not depend on specific reservoir types or fracturing methods and can be widely applied to the development of unconventional oil and gas reservoirs such as shale gas, tight gas, and coalbed methane.

[0071] (6) Advantages compared with the prior art: Compared with CN118691561A, the present invention uses machine learning methods to replace simple threshold judgment, which can automatically learn complex patterns in microseismic data; it uses probability output to replace binary judgment, providing richer decision information; and it adopts a multi-source constraint area division method based on physical mechanism, which has stronger engineering interpretability and reliability.

[0072] Example 2: This embodiment provides a method for intelligent identification and correction of natural cracks. In addition to steps 101-107 described in the embodiment, this method includes: Step 108: Classify the predicted probability of the existence of natural cracks according to a probability threshold to obtain a spatial distribution map of the probability of the existence of natural cracks and correction suggestions. This step specifically includes: (1) Determining the probability threshold Multiple methods for determining probability thresholds are provided: Method 1: Determine the optimal probability threshold through ROC curve analysis: The threshold that maximizes the Youden Index (Youden Index = Sensitivity + Specificity - 1) is selected as the optimal threshold. In this embodiment, the optimal threshold is P = 0.65.

[0073] Method 2: Precision-Recall Curve Analysis The threshold that maximizes the F1 score is selected as the optimal threshold.

[0074] Method 3: Business Requirements Method The threshold should be set according to the actual needs of the project. For example, if the risk of missed detection is more important, the threshold can be lowered; if the risk of false alarm is more important, the threshold can be raised.

[0075] (2) Probability grading Based on the probability threshold, the prediction results are divided into three levels: (3) Output of spatial distribution map Generate a three-dimensional spatial distribution map of the probability of the existence of natural cracks, including: Spatial interpolation results of probability values; Delineating the boundaries of high-probability regions; Overlay display with the initial natural crack model.

[0076] (4) Correction suggestions output For each abnormal region, output correction suggestions: Note: The location coordinates here have been converted from geodetic coordinates to Cartesian coordinates.

[0077] (5) Output file format The output results are saved in a standard format for easy import into geological modeling software: Probability distribution data: CSV format, containing coordinates and probability values; Correction suggestion report: PDF format, including charts and text descriptions; 3D model file: GRDEC L The format can be directly imported into software such as Petrel.

[0078] Example 3: This embodiment provides a natural crack intelligent identification and correction system, which is used to implement the method described in Embodiment 1. Figure 7 This is a schematic diagram of a smart identification and correction system for natural cracks. Figure 7 The system shown includes the following modules: The microseismic event data acquisition module 201 is used to acquire microseismic monitoring data, initial natural fracture model, fracturing construction data and fault calibration data during the hydraulic fracturing process, and to perform data quality control to obtain quality-controlled microseismic event data. The direct disturbance zone determination module 202 is used to determine the direct disturbance zone based on the microseismic event data using a multi-source constraint method; The abnormal region identification module 203 is used to identify abnormal regions based on the direct disturbance region; The spatiotemporal clustering feature extraction module 204 is used to extract features from the microseismic event data to obtain spatiotemporal clustering features; Training sample set construction module 205 is used to construct the training sample set; The intelligent identification and correction model determination module 206 for natural cracks is used to train the intelligent identification and correction model for natural cracks by using machine learning based on the training sample set and the spatiotemporal clustering features. The natural crack existence probability prediction result determination module 207 is used to predict the abnormal area based on the natural crack intelligent identification and correction model to obtain the natural crack existence probability prediction result.

[0079] Example 4: This embodiment provides a natural crack intelligent identification and correction system. Unlike embodiment three, this system includes modules 201-207, and also includes the following modules: The module for determining the spatial distribution map and correction suggestions for the existence of natural cracks is used to classify the prediction results of the existence probability of natural cracks according to the probability threshold, and obtain the spatial distribution map and correction suggestions for the existence probability of natural cracks.

[0080] Example 5: This embodiment provides a system for intelligent identification and post-hoc correction of natural fractures based on the spatiotemporal clustering characteristics of microseismic events. This system is used to implement the method described in Embodiment 1. The system includes the following modules: (1) Data acquisition module The data acquisition module is used to acquire the following data: Microseismic monitoring data: Data such as focal location, time of occurrence, and magnitude are obtained from the microseismic monitoring system; Initial natural fracture model: Import the natural fracture distribution model from geological modeling software (such as Petrel); Fracturing operation data: Well trajectory, segment information, injection parameters, reservoir parameters, etc. are obtained from fracturing operation records; Fault calibration data: Obtain the location and extent of faults from seismic interpretation results.

[0081] The data acquisition module supports importing various data formats, including CSV, LAS, and GRDECL.

[0082] (2) Data preprocessing module The data preprocessing module is used to perform quality control on the acquired data, including: Positioning residual filtering: Filter high-precision positioning events based on user-defined thresholds; Fault exclusion: Automatically identifies and excludes events within the fault-affected area; Magnitude classification: distinguishes between events used for event rate calculation (all magnitudes) and anomaly area identification (M≥0); Data format conversion: Convert data of different formats into the system's internal format.

[0083] (3) Regional division module The region partitioning module is used to determine the boundary of the direct disturbance zone based on the spatial distribution of fluid pressure increments, including: Theoretical ellipse calculation: Calculate the elliptical boundary based on the pressure diffusion equation; Empirical radius estimation: Estimating the propagation range of the hydraulic fracture based on the injected fluid volume; Event rate correction: Boundaries are dynamically corrected based on the microseismic event rate; Multi-source constraint fusion: taking the union of the results from the three methods; Region visualization: Displays the region division results in a 3D view.

[0084] (4) Abnormal area identification module The anomaly region identification module is used to identify anomaly regions outside the direct disturbance zone, including: Microseismic response assessment: Count the number of microseismic events with magnitude M≥0 in each region; Crack Model Query: Query the crack distribution in each region of the initial natural crack model; Abnormal region marking: Automatically marks regions that meet abnormal conditions; List of Abnormal Regions: Generates a detailed list of abnormal regions, including information such as location and number of events.

[0085] (5) Feature extraction module The feature extraction module is used to extract spatiotemporal clustering features from microseismic data, including: Spatial feature calculation: Calculate spatial features such as the number of events, density, clustering, and distance; Time feature calculation: Calculate time features such as latency, time span, and event frequency; Optional feature calculation: Calculate optional features such as b-value and focal mechanism; Feature normalization: Normalize the extracted features; Feature vector generation: Combining various features into a feature vector, which is used as input to the model.

[0086] (6) Model training module The model training module is used to train machine learning models, including: Sample construction: Automatically constructs positive and negative samples based on user-defined conditions; Sample balancing: Provides sample balancing methods such as SMOTE and undersampling; Model selection: Supports a variety of machine learning models, including ANN, SVM, RF, XGBoost, etc. Parameter tuning: Provides parameter tuning methods such as grid search and Bayesian optimization; Model saving: Save the trained model as a file for later use.

[0087] (7) Probability Prediction Module The probability prediction module is used to make predictions using a trained model, including: Model loading: Loads a saved trained model; Feature input: Input the feature vectors of the abnormal regions into the model; Probability calculation: Calculate the probability of the existence of natural cracks in each anomalous area; Batch forecasting: Supports batch forecasting for multiple regions.

[0088] (8) Result Output Module The results output module is used to output prediction results and correction suggestions, including: Probability distribution map generation: Generates a three-dimensional spatial distribution map of the probability of the existence of natural cracks; Probability grading: Grading the prediction results according to the threshold set by the user; Correction suggestion generation: Automatically generates correction suggestions for each abnormal area; Report Export: Export the results as a PDF report or a CSV data file; Model Export: Export the probability distribution results in GRDECL format for easy import into geological modeling software.

[0089] Example 6: This embodiment provides a computer device, including a memory and a processor. The memory stores a computer program, which, when executed by the processor, implements the intelligent identification and a posteriori correction method for natural fractures based on the spatiotemporal clustering characteristics of microseismic events described in Embodiment 1.

[0090] Computer equipment can be a server, workstation, or personal computer. The processor can be a central processing unit (CPU), graphics processing unit (GPU), or other programmable logic device. Memory can be random access memory (RAM), read-only memory (ROM), hard disk drive (HDD), or solid-state drive (SSD).

[0091] Example 7: Computer-readable storage medium This embodiment provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the intelligent identification and a posteriori correction method for natural cracks based on the spatiotemporal clustering characteristics of microseismic events described in Embodiment 1.

[0092] Computer-readable storage media can be non-volatile, such as hard drives, USB flash drives, optical discs, and flash memory, or volatile, such as RAM. Computer programs can be stored on storage media in the form of source code or compiled executable files.

[0093] Embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0094] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0095] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0096] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0097] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0098] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for intelligent identification and correction of natural cracks, characterized in that, The method includes: Acquire microseismic monitoring data, initial natural fracture model, fracturing construction data, and fault calibration data during the hydraulic fracturing process, and perform data quality control to obtain quality-controlled microseismic event data; The direct disturbance zone was determined using a multi-source constraint method based on the microseismic event data. Based on the direct disturbance area, identify the abnormal region; Feature extraction is performed on the microseismic event data to obtain spatiotemporal clustering features; Construct a training sample set; A natural crack intelligent identification and correction model is obtained by training the training sample set and the spatiotemporal clustering features using machine learning. The abnormal area is predicted based on the intelligent identification and correction model for natural cracks, and the probability prediction result of the existence of natural cracks is obtained.

2. The intelligent identification and correction method for natural cracks according to claim 1, characterized in that, The method further includes: The probability prediction results of the existence of natural cracks are classified according to the probability threshold to obtain a spatial distribution map of the probability of the existence of natural cracks and correction suggestions.

3. The intelligent identification and correction method for natural cracks according to claim 1, characterized in that, The multi-source constraint method includes a union of one or more of the following: a calculation method based on pressure diffusion theory, an estimation method based on empirical radius, and a dynamic correction method based on microseismic event rate. In three-dimensional space, the direct disturbance zone is ellipsoidal in shape; on a two-dimensional projection plane, the direct disturbance zone is elliptical in shape, with the major axis of the ellipse parallel to the direction of hydraulic fracture propagation and the minor axis of the ellipse parallel to the wellbore direction.

4. The intelligent identification and correction method for natural cracks according to claim 1, characterized in that, The step of identifying abnormal regions based on the direct disturbance region specifically includes: Based on the direct disturbance zone, identify anomaly areas located outside the direct disturbance zone that exhibit microseismic responses and whose initial natural crack models show no crack distribution; the presence of microseismic responses means that the number of microseismic events with magnitudes greater than or equal to a preset threshold within the area is greater than or equal to a preset threshold.

5. The intelligent identification and correction method for natural cracks according to claim 1, characterized in that, The spatiotemporal clustering characteristics include spatial and temporal characteristics. The spatial characteristics include the number of microseismic events, the spatial density of events, the degree of spatial clustering, and the distance to the nearest fracturing segment. The temporal characteristics include the time delay between the time of event occurrence and the construction time of the corresponding fracturing segment, as well as the distribution characteristics of the event time series.

6. The intelligent identification and correction method for natural cracks according to claim 1, characterized in that, The input of the intelligent identification and correction model for natural cracks is the spatiotemporal clustering feature vector of microseismic events, and the output is the probability value of the existence of natural cracks.

7. A natural crack intelligent identification and correction system based on any one of claims 1-6, characterized in that, The system includes: The microseismic event data acquisition module is used to acquire microseismic monitoring data, initial natural fracture models, fracturing construction data, and fault calibration data during the hydraulic fracturing process, and to perform data quality control to obtain quality-controlled microseismic event data. The direct disturbance zone determination module is used to determine the direct disturbance zone based on the microseismic event data using a multi-source constraint method; An abnormal region identification module is used to identify abnormal regions based on the direct disturbance region; The spatiotemporal clustering feature extraction module is used to extract features from the microseismic event data to obtain spatiotemporal clustering features; The training sample set construction module is used to construct the training sample set; The intelligent identification and correction model determination module for natural cracks is used to train the intelligent identification and correction model for natural cracks by using machine learning based on the training sample set and the spatiotemporal clustering features. The module for determining the probability prediction result of the existence of natural cracks is used to predict the abnormal area based on the intelligent identification and correction model of natural cracks, and obtain the probability prediction result of the existence of natural cracks.

8. The intelligent identification and correction system for natural cracks according to claim 7, characterized in that, The system also includes: The module for determining the spatial distribution map and correction suggestions for the existence of natural cracks is used to classify the prediction results of the existence probability of natural cracks according to the probability threshold, and obtain the spatial distribution map and correction suggestions for the existence probability of natural cracks.

9. A computer device, characterized in that, The computer device includes a memory and a processor. The memory stores a computer program, and the processor runs the computer program to enable the computer device to perform the intelligent identification and correction method for natural cracks as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which is executed by a processor to implement the intelligent identification and correction method for natural cracks as described in any one of claims 1-7.

Images