Method for intelligently identifying macroscopic fracture extension track in fracturing focus cluster

Abstract

The invention discloses a method for intelligently identifying a macroscopic fracture extension track in a fracturing focus cluster, and the method comprises the steps: collecting an elastic wave waveform released by rock fracture or fracture, and carrying out the filtering of the elastic wave waveform; acquiring a head oscillation starting point of the high-quality waveform by using a deep learning method; dividing a dynamic time step length, and dividing the seismic source cluster according to the dynamic time step length; carrying out multi-stage space unit body division on the seismic source data body in a single time step length; integrating the macroscopic crack identified by each parameter to obtain a macroscopic crack extension track in the unit body; and the fractures identified in the space unit bodies are reconstructed on the space level and the time level, and the macroscopic fracture extension track in the whole fracturing extension process is obtained. According to the method, the macroscopic crack and the secondary small-scale crack area are distinguished, so that high-precision identification of the macroscopic crack expansion track can be realized.

Description

Technical Field

[0001] The present invention relates to a rock material structure detection technology, and in particular to a method for intelligently identifying macro crack extension trajectories in a fracturing source cluster. Background Art

[0002] Fracturing, which uses high-pressure fluid to fracture rocks and drive the expansion of fractures, is the core technology for permeability enhancement of unconventional oil and gas reservoirs, permeability enhancement of hot dry rock geothermal heat storage to improve heat extraction efficiency, permeability enhancement of coal seam gas in underground mines and surrounding rock shear stress transfer, and permeability enhancement of uranium reservoirs to improve in-situ leaching efficiency. High-precision monitoring of fracturing fractures is the basis for controlling the expansion of fracturing fractures and an important support point for increasing the production of energy and strategic minerals.

[0003] Earthquakes, microseisms, ground sounds and acoustic emissions are elastic waves with frequencies ranging from low to high, which are released in fractures of different scales. The main means of monitoring hydraulic fractures is to receive elastic waves through sensors to invert the fracture position, which is collectively referred to as the source inversion of fractures. However, the physical mechanism of the initiation and fusion of a large number of small-scale fractures into macro fractures determines that the fracture morphology obtained by source inversion is clustered; by improving the accuracy of source inversion to obtain higher-precision source clusters, this method of identifying cracks is currently the main method and development trend of hydraulic fracture monitoring, but this method will cause macro fractures to be hidden in the source cluster, making it impossible to distinguish between macro fractures and secondary small-scale fracture areas, which seriously limits the accuracy of hydraulic fracture monitoring. Summary of the invention

[0004] Purpose of the invention: In view of the above problems, the purpose of the present invention is to provide a method for intelligently identifying the extension trajectory of macroscopic fractures in a fracturing source cluster, thereby realizing the intelligent identification of the extension trajectory of macroscopic fractures and secondary small-scale fracture areas in the source cluster.

[0005] Technical solution: A method for intelligently identifying macro crack extension trajectories in a fracturing source cluster of the present invention comprises the following steps:

[0006] Step 1, collecting elastic wave waveforms released by rock fracture or rupture, filtering the elastic wave waveforms using an intelligent filtering model, and recording the filtered elastic wave waveforms as high-quality waveforms;

[0007] Step 2: Use deep learning methods to obtain the head starting point of the high-quality waveform, and then obtain the arrival time of the high-quality waveform, and obtain the source point according to the arrival time;

[0008] Step 3, obtaining the earthquake source clusters of the earthquake source points, dividing the dynamic time steps, dividing the earthquake source clusters according to the dynamic time steps, and obtaining the earthquake source data volume within a single time step;

[0009] Step 4, dividing the earthquake source data volume within a single time step into multi-level spatial units, and analyzing the distribution characteristics of multiple parameters of earthquake source characteristics within the spatial units;

[0010] Step 5, identifying the macro cracks corresponding to each parameter according to the distribution characteristics, and obtaining the extension trajectory of the macro cracks in the unit body by integrating the macro cracks identified by each parameter;

[0011] Step 6: Reconstruct the identified fractures in each spatial unit at both spatial and temporal levels to obtain the macroscopic fracture extension trajectory during the entire fracturing expansion process.

[0012] Further, in step 3, dividing the dynamic time steps includes:

[0013] Dynamically identify the aggregation form of the point cluster in the earthquake source space, and determine whether the aggregation form conforms to the characteristics of non-planar turning expansion or bifurcation of the crack. If it conforms, continue to search for time points earlier than the current time until the aggregation form presents a single crack, and the corresponding time is taken as the end point of the time step;

[0014] If the source distribution shows a single form throughout the entire process of crack expansion, it is necessary to determine whether the amount of spatial source data within a single time step is sufficient for the analysis of the multi-parameter spatial distribution law of source characteristics. If not, the time step needs to be increased until the minimum threshold of spatial source data is reached.

[0015] Furthermore, in step 4, dividing the earthquake source data volume within a single time step into multi-level spatial units includes:

[0016] The earthquake source data volume in a single time step is divided into multiple first-level earthquake source unit volumes;

[0017] The primary source unit body is divided into a plurality of secondary source unit bodies;

[0018] The secondary seismic source unit volume is divided into multiple tertiary seismic source unit volumes.

[0019] Furthermore, dividing the earthquake source data volume within a single time step into multiple first-level earthquake source unit volumes includes:

[0020] For the approximately elliptical fracturing crack with the injection point as the centroid, a cylindrical coordinate system is constructed with the injection point as the origin O and the normal line of the fracturing crack surface identified by the earthquake source cluster as the axis A. The cylindrical coordinate system parameters r and θ are located on the projection surface of the fracturing crack. When the crack does not expand along the trajectory with the injection point as the centroid, the centroid of the stage expansion of the irregular fracture expansion surface is selected as the origin O, and the normal line of the fracturing crack surface identified by the earthquake source cluster is used as the axis A to establish a cylindrical coordinate system. The cylindrical coordinate system parameters r and θ are located on the projection surface of the irregular fracturing crack.

[0021] The infinite plane where the axis A and the polar coordinate r are located is the feature plane S0, and the two infinite planes S1 and S2 that are symmetric planes with the feature plane S0 and have the same normal are the data screening intervals, and the vertical distance between plane S1 and plane S2 is w;

[0022] When the characteristic surface S0 rotates, plane S1 and plane S2 rotate with the characteristic surface; let the characteristic surface S0 pass through the injection point, and rotate the characteristic surface in increments of α degrees until it reaches 180°. During each rotation, the source data between plane S1 and plane S2 are used as the first-level source unit body, and the total number of first-level source unit bodies is 180 / α.

[0023] Further, the first-level seismic source unit body is divided into a plurality of second-level seismic source unit bodies including:

[0024] A local spatial rectangular coordinate system is established for each first-level seismic source unit body obtained, the Y axis is in the same direction as the polar coordinate r of the cylindrical coordinate system, the X, Y and Z axes conform to the directions specified by the right-hand system, the Z axis is the thickness direction of the first-level seismic source unit body, and the X axis is the width direction of the first-level seismic source unit body. The seismic source in the first-level seismic source unit body is projected onto the YZ plane to obtain a two-dimensional seismic source data volume;

[0025] The first-level seismic source unit body is cut along the Y direction to form multiple parallel lines with a spacing of D and parallel to the X-axis, among which the seismic source point cluster covered between two adjacent parallel lines is regarded as the second-level seismic source unit body.

[0026] Furthermore, the secondary seismic source unit body is divided into a plurality of tertiary seismic source unit bodies including:

[0027] Each secondary source unit body is split along the X direction to form multiple parallel lines with a spacing of d and parallel to the Y axis, where the source point cluster covered between two adjacent parallel lines is the tertiary source unit body; the size of the tertiary source unit body is D×d.

[0028] Furthermore, in step 4, analyzing the distribution characteristics of multiple parameters of earthquake source characteristics within the spatial unit body includes:

[0029] The third-level source unit is taken as the smallest statistical unit, and the multi-parameters of the waveform corresponding to the source are statistically analyzed. Then, the distribution characteristics of the multi-parameters of the source characteristics are determined in the second-level source unit using a mathematical statistical model.

[0030] Further, step 5 includes:

[0031] According to the distribution law of each parameter, the coordinate value of the macro crack formation position in the secondary source unit body in the local space rectangular coordinate system of the primary source unit body is obtained;

[0032] Set the picking threshold of the microcrack area on both sides of the macrocrack, and obtain the coordinate value of the boundary of the microcrack area in the secondary source unit in the local space rectangular coordinate system of the primary source unit according to the picking threshold;

[0033] When the macro-crack formation positions and micro-crack zone boundary positions of all secondary units within the primary source unit are determined, the macro-cracks of all secondary units within the primary source unit and the micro-crack zone boundary positions are connected in sequence to obtain the macro-crack extension trajectory and micro-crack zone formation positions within the primary source strip of the seismic source unit.

[0034] Further, step 6 comprises:

[0035] Reconstruct the third-level seismic source unit body that has completed the statistical analysis of the parameters into the second-level seismic source unit body, reconstruct all the second-level seismic source unit bodies into the first-level seismic source unit bodies, and reconstruct all the first-level seismic source unit bodies into the segment macro cracks corresponding to each time step;

[0036] The macroscopic fractures of the segments are connected according to the time nodes within each time step to obtain the trajectory of the macroscopic fractures.

[0037] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages:

[0038] 1. The present invention can separate the secondary fracture area from the macro fracture surface and enhance the monitoring and identification accuracy of three-dimensional fractures;

[0039] 2. The present invention can identify the three-dimensional fracturing crack network and its dynamic evolution through multi-stage and multi-unit division in time and space;

[0040] 3. The present invention achieves quantitative identification of macroscopic three-dimensional cracks and secondary small-scale crack areas by describing the multi-parameter distribution of the cross-sectional crack extension direction;

[0041] 4. The present invention improves the accuracy and efficiency of three-dimensional macro crack identification through waveform denoising, source cluster distribution morphology recognition, time step division and source characteristic parameter distribution pattern characterization. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A flow chart of a method for intelligently identifying macroscopic fracture extension trajectories in a fracturing source cluster;

[0043] Figure 2 Schematic diagram of earthquake source cluster;

[0044] Figure 3 It is a schematic diagram of the cylindrical coordinate system and the SO-S1-S2 section plane;

[0045] Figure 4Schematic diagram of introducing cylindrical coordinate system and SO-S1-S2 section plane into the source cluster;

[0046] Figure 5 It is a structural schematic diagram of the first-level source unit;

[0047] Figure 6 It is a schematic diagram of dividing the secondary seismic source unit into three-level seismic source unit bodies;

[0048] Figure 7 Schematic diagram of the boundary position of the crack area;

[0049] Figure 8 This is a schematic diagram of the actual crack formation location;

[0050] Fig. 9 It is a schematic diagram of reconstruction into a first-level source unit;

[0051] Fig.10 Schematic diagram of the fracture trajectory after reconstruction at the time level. DETAILED DESCRIPTION

[0052] In order to make the objectives, technical solutions and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments.

[0053] The method for intelligently identifying the extension trajectory of macroscopic fractures in the fracturing source cluster described in this embodiment is as follows: Figure 1 As shown, the method at least includes the following steps 1 to 6.

[0054] Among them, step 1 is to collect the elastic wave waveform released by rock fracture or rupture, and use the intelligent filtering model to filter the elastic wave waveform, and record the filtered elastic wave waveform as a high-quality waveform.

[0055] The acquisition of high-quality rock fracture or rupture waveforms is the basis for high-precision source inversion and waveform feature analysis. In actual engineering and experiments, interference signals, namely background noise, are inevitable. Background noise is random and difficult to eliminate by the unified rules specified by traditional filtering methods. In this example, before formally acquiring signals, rock fracture tests in the laboratory and engineering are first carried out under the condition of avoiding noise interference to obtain elastic wave waveforms released by rock fracture or rupture, and obtain waveform characteristic parameters of high-quality waveforms. Among them, waveforms without many messy interference signals and whose various parameter values ​​of the collected waveforms reach the minimum threshold value are considered to be high-quality waveforms.

[0056] Furthermore, high-quality waveforms of rock fractures or cracks are used as samples to compare with the fracture or crack waveforms collected under background noise conditions, and deep learning is carried out to filter out the background noise and obtain high-quality waveforms. Through artificial intelligence training, an intelligent filtering model in the waveform acquisition process is obtained.

[0057] In the above-mentioned deep learning process, neural networks can be used for training, such as time series transformer, diffusion model, conditional generative adversarial network, variational autoencoder, deep convolutional generative adversarial network, self-supervised learning and contrastive learning and other neural network models. Through any of the above neural network models, artificial intelligence training is carried out to conduct rock fracture tests in laboratories and engineering projects under the condition of avoiding noise interference, and the elastic wave waveform released by rock fracture or rupture is obtained as the waveform characteristic parameters of the standard waveform. At this time, the waveform is a standard high-quality waveform without noise. The input of the neural network is a large amount of elastic wave waveform data obtained at the test or engineering site, and the output is a standard high-quality waveform. Then, the neural network model is used for training and learning, and the waveform characteristic parameters of the obtained standard waveform are compared to filter out the noise interference signal except the standard waveform. The trained neural network model is used as an intelligent filtering model, and the elastic wave waveform can be intelligently filtered through the intelligent filtering model to obtain a high-quality waveform.

[0058] Step 2: Use deep learning methods to obtain the head starting point of the high-quality waveform, and then obtain the arrival time of the high-quality waveform, and obtain the source point based on the arrival time.

[0059] Picking up the starting points of all waveforms released by the same source, that is, the significant sudden starting points of the waveforms that fluctuate smoothly before the source is generated, can obtain the waveform arrival time, which is a key parameter for source inversion. However, the sudden onset time of the waveform is short and is presented by multiple data points. It is difficult to pick up the first onset point, especially when the background noise cannot be filtered out. It is more difficult to identify. Therefore, in this example, the deep learning method is preferred to label the onset points of the waveforms in different background noises as training set data, and then train the neural network model to obtain an intelligent high-precision recognition model for waveform onset points suitable for different noise levels, and obtain the onset time more accurately to improve the accuracy of source inversion. The neural network model can be a variety of neural network models such as time series Transformer, diffusion model, conditional generative adversarial network, variational autoencoder, deep convolutional generative adversarial network, self-supervised learning and contrastive learning. The input item is a high-quality waveform after filtering, and the output item is the onset point of all waveforms of the same source. The neural network model is used to perform deep training on the identification of the onset point, and the head onset point of the high-quality waveform can be obtained by using the trained neural network model.

[0060] After completing steps 1 and 2 above, not only the accuracy of earthquake source identification is improved, but also high-quality initial earthquake source data is obtained.

[0061] Step 3, obtain the source clusters of the source points, divide the dynamic time steps, divide the source clusters according to the dynamic time steps, and obtain the source data volume within a single time step.

[0062] After the earthquake source point is determined, the earthquake source cluster of the earthquake source point can be obtained through the crack monitoring related instruments, but at this time, the earthquake source cluster not only contains the real crack trajectory, but also includes the secondary small-scale crack area, which seriously limits the monitoring accuracy of the hydraulic fracture. In this embodiment, the macro cracks in the earthquake source cluster are separated from the secondary small-scale crack area, in order to achieve high-precision identification of the macro crack extension trajectory.

[0063] After obtaining the source cluster of the source point, it is necessary to identify the preliminary crack morphology depicted by the spatial source cluster, and divide the source data into multiple characteristic time stages based on this. Usually, the time analysis step is of equal length, but in view of the limitation that the spatial distribution of the source is too complex due to crack bifurcation or networking during the whole process of crack expansion, which restricts the identification of macro cracks, the setting of the time analysis step needs to be dynamically adjusted. In this embodiment, the single crack expansion mode presented by the source in each time analysis step is used as the basis for the dynamic adjustment and selection of the time analysis step, and the whole process of complex crack expansion is divided into multiple analysis steps formed in sequence by multiple single cracks, and the whole is divided into parts. The macroscopic real crack surface hidden in the source cluster in each analysis step is identified, and finally the whole process crack expansion morphology is reconstructed at the time level. If the source distribution is mainly single-shaped during the whole process of crack expansion, it is necessary to judge whether the amount of spatial source data in a single time step is sufficient to support the determination of the macro crack criterion, and the adjustment of the analysis time step needs to reach the minimum threshold of the data volume.

[0064] Further, in step 3, dividing the dynamic time steps includes:

[0065] Dynamically identify the aggregation form of the point cluster in the earthquake source space, and determine whether the aggregation form conforms to the characteristics of non-planar turning expansion or bifurcation of the crack. If it conforms, continue to search for time points earlier than the current time until the aggregation form presents a single crack, and the corresponding time is taken as the end point of the time step;

[0066] If the source distribution shows a single form throughout the entire process of crack expansion, it is necessary to determine whether the amount of spatial source data within a single time step is sufficient for the analysis of the multi-parameter spatial distribution law of source characteristics. If not, the time step needs to be increased until the minimum threshold of spatial source data is reached.

[0067] The minimum threshold source data volume is set to determine whether the spatial source data volume within a single time step exceeds this threshold value. If it is greater than or equal to the minimum threshold value, it is considered that the spatial source data volume within a single time step is sufficient for the analysis of the multi-parameter spatial distribution law of source characteristics, otherwise it is not.

[0068] Step 4: divide the earthquake source data volume within a single time step into multi-level spatial units, and analyze the distribution characteristics of multiple parameters of earthquake source characteristics within the spatial units.

[0069] In this step 4, the selected source point cluster will be preliminarily judged for its distribution morphology, and the spatial unit division method of the source data will be determined, which is the basis for statistics, identification and analysis of the multi-parameter distribution characteristics of the source characteristics in each spatial unit. In general, the source data body within a single time step is spatially divided into multi-regional and multi-level units, and a mathematical statistical model is constructed and analyzed for each level of the divided source unit body. The mathematical statistical model can be the more commonly used normal distribution, binomial distribution, Poisson distribution and exponential distribution, etc., but is not limited to the above models.

[0070] Furthermore, in step 4, dividing the earthquake source data volume within a single time step into multi-level spatial units includes:

[0071] The earthquake source data volume in a single time step is divided into multiple first-level earthquake source unit volumes;

[0072] The primary source unit body is divided into a plurality of secondary source unit bodies;

[0073] The secondary seismic source unit volume is divided into multiple tertiary seismic source unit volumes.

[0074] Specifically, dividing the source data volume within a single time step into multiple first-level source unit volumes includes:

[0075] like Figure 2 As shown in the figure, for an approximately elliptical fracturing crack with the injection point as the centroid, perspective ① and perspective ② are schematic diagrams of the source cluster at different perspectives. The injection point is taken as the origin O, and the normal line of the fracturing crack surface identified by the source cluster is taken as the axis A to construct a cylindrical coordinate system, as shown in Figure 3 As shown in the middle figure (a), the cylindrical coordinate system parameters r and θ are located on the projection surface of the hydraulic fracture, and the coordinates of any point M in the source cluster are expressed as (r, θ, z), where z is the ordinate in the spatial coordinate system O-XYZ. When the fracture does not extend along the trajectory with the injection point as the centroid, the centroid of the stage expansion of the irregular fracture extension surface is selected as the origin O, and the normal line of the hydraulic fracture surface identified by the source cluster is used as the axis A to establish a cylindrical coordinate system. The cylindrical coordinate system parameters r and θ are located on the projection surface of the irregular hydraulic fracture.

[0076] The infinite plane where the axis A and the polar coordinate r are located is the feature plane S0, and the two infinite planes S1 and S2 that are symmetric planes with the feature plane S0 and have the same normal are the data screening intervals. The vertical distance between plane S1 and plane S2 is w. Figure 3The middle figure (b) shows a schematic diagram of the SO-S1-S2 wireless large cross-section space, and figure (c) is a schematic diagram of introducing the SO-S1-S2 cross-section into the cylindrical coordinate system.

[0077] When the characteristic surface S0 rotates, planes S1 and S2 rotate with the characteristic surface; let the characteristic surface S0 pass through the injection point, and rotate the characteristic surface by increments of α degrees until it reaches 180°. During each rotation, the source data between planes S1 and S2 are used as the first-level source unit body, and the first-level source unit bodies total 180 / α. Figure 4 The middle figure (a) shows a schematic diagram of introducing a cylindrical coordinate system into a source cluster, and the figure (b) shows a schematic diagram of introducing the SO-S1-S2 section and cylindrical coordinate system into a source cluster.

[0078] Further, the first-level seismic source unit body is divided into a plurality of second-level seismic source unit bodies including:

[0079] like Figure 5 As shown, a local spatial rectangular coordinate system O-XYZ is established for each first-level seismic source unit body. The Y axis is in the same direction as the polar coordinate r of the cylindrical coordinate system, and is also the length direction of the first-level seismic source unit body. The X, Y and Z axes conform to the directions specified by the right-hand system. The Z axis is the thickness direction of the first-level seismic source unit body and is in the same direction as the axis A. The X axis is the width direction of the first-level seismic source unit body. The seismic source in the first-level seismic source unit body is projected onto the YZ plane to obtain a two-dimensional seismic source data body. Figure 5 In the middle, the ① perspective is the projection view of the earthquake source under the YZ projection plane, and the ② perspective is the projection view of the earthquake source under the ZY projection plane.

[0080] The first-level seismic source unit body is cut along the Y direction to form multiple parallel lines with a spacing of D and parallel to the X-axis, among which the seismic source point cluster covered between two adjacent parallel lines is regarded as the second-level seismic source unit body.

[0081] Furthermore, if Figure 6 As shown, the secondary source unit body is divided into a plurality of tertiary source unit bodies including:

[0082] Each secondary source unit body is split along the X direction to form multiple parallel lines with a spacing of d and parallel to the Y axis, where the source point cluster covered between two adjacent parallel lines is the tertiary source unit body; the size of the tertiary source unit body is D×d.

[0083] It should be noted that the determination of the size of the third-level seismic source unit depends on the number of seismic sources. The number of seismic sources in the third-level seismic source unit must be higher than the pre-set minimum threshold. If it is not met, the unit size needs to be increased. At the same time, the number of seismic sources in the third-level seismic source unit must not exceed the pre-set maximum threshold. If it exceeds, the unit size needs to be increased according to actual needs.

[0084] Furthermore, in step 4, analyzing the distribution characteristics of multiple parameters of earthquake source characteristics within the spatial unit body includes:

[0085] The third-level source unit is taken as the smallest statistical unit, and the multi-parameters of the waveform corresponding to the source are statistically analyzed. Then, the distribution characteristics of the multi-parameters of the source characteristics are determined in the second-level source unit using a mathematical statistical model.

[0086] In one example, if the statistical multiple parameters are accumulative scalars, the sum of the statistics is calculated, such as energy, number of earthquake sources, ringing count, etc.; if the sum of the statistical parameters has no physical meaning, the average value of the parameters is calculated, such as wave velocity, amplitude, frequency, earthquake source mechanism, duration, rise time, etc.

[0087] On the basis of the determination of multiple parameters of earthquake source characteristics in the third-level unit, the distribution of multiple parameters of earthquake source characteristics along the second-level unit is determined, and the distribution of multiple parameters of earthquake source characteristics in the second-level earthquake source unit is determined by using a mathematical statistical model. Typical multiple parameters of earthquake source characteristics here include energy, number of earthquake sources, ring count, wave velocity, amplitude, frequency, earthquake source mechanism, duration, rise time, etc., but are not limited to the above parameters. The mathematical statistical models used here include normal distribution, binomial distribution, Poisson distribution or exponential distribution, etc., but are not limited to the above models.

[0088] When analyzing the source parameter distribution of the secondary unit, the source characteristic parameters of some tertiary units will be abnormally high or low, which will reduce the accuracy of the mathematical model in describing the distribution of source characteristic parameters in the secondary unit. Therefore, the artificial intelligence method can be used for intelligent identification, such as pre-setting a reasonable reference range of parameter characteristic values, and then pre-setting a reasonable error range, inputting the characteristic parameter data of the three-level source unit body obtained by analysis into the artificial intelligence method, and then the artificial intelligence method is used for screening and identification, and the data within the set parameter reference range is output, and the remaining characteristic parameters that are too high or too low are identified and removed, and the input is a reasonable characteristic parameter value. For each three-level source unit body with abnormal source characteristic parameters in the secondary source unit body, the data in the unit body is eliminated, and then the mathematical statistics of the source characteristic parameter distribution of the secondary source unit body are performed.

[0089] Step 5, identifying the macro cracks corresponding to each parameter based on the distribution characteristics, and obtaining the extension trajectory of the macro cracks in the unit body by integrating the macro cracks identified by various parameters.

[0090] Further, step 5 includes:

[0091] According to the distribution law of each parameter, the coordinate value of the macro crack formation position in the secondary source unit body in the local space rectangular coordinate system of the primary source unit body is obtained;

[0092] Set the picking threshold of the microcrack area on both sides of the macrocrack, and obtain the coordinate value of the boundary of the microcrack area in the secondary source unit in the local space rectangular coordinate system of the primary source unit according to the picking threshold;

[0093] When the macro crack formation positions and micro crack zone boundary positions of all secondary units in the primary source unit are determined, the macro cracks of all secondary units in the primary source unit are connected in sequence, and the micro crack zone boundary positions are connected in sequence to obtain the macro crack extension trajectory and micro crack zone formation position in the primary source unit body source strip. Figure 7 The dashed outline in the middle shows the boundary position of the crack area, and the figure shows the corresponding distribution models at positions ① and ② respectively.

[0094] For each secondary source unit, the mathematical statistical model type and model parameters obtained by different parameters are different due to differences in parameter distribution laws, but the coordinates corresponding to the peak or valley value of the model are the formation positions of macro cracks. For example, if the characteristic parameters in the secondary source unit obey the normal distribution model, the coordinates corresponding to the peak point of the normal distribution model are the formation positions of the macro cracks in the secondary source unit, such as Figure 8 As shown in the figure; if the frequency or wave velocity in the secondary source unit body obeys the distribution model with high on both sides and low in the middle, then the position corresponding to the valley point is the position where the macro crack is formed, and the coordinate value of the macro crack formation position in the local coordinate system of the primary source unit body can be obtained. Figure 8 The distribution models at positions ① and ② are also shown in the figure.

[0095] According to the test results in the laboratory and mechanical theory, the picking threshold of the microcrack zone on both sides of the macro crack is determined. For example, if the acoustic emission energy obeys the normal distribution function in the secondary source unit, the cumulative acoustic emission energy in the selected space range can be obtained by integrating on both sides of the model symmetry axis. If the energy obtained by integration reaches a certain threshold p, such as 90%, compared with the total energy in the secondary source unit, the integral boundary is the boundary of the microcrack zone, so that the coordinate value at the boundary of the microcrack zone can be determined. For example, if the acoustic emission wave velocity has a high distribution characteristic of high on both sides and low in the middle in the secondary source unit, according to the test results and theoretical analysis, the wave velocity v0 is used as the threshold, and the area with a wave velocity lower than v0 is selected as the microcrack zone. Further, the coordinate value of the local coordinate system of the primary source unit where the boundary of the microcrack zone in the secondary source unit is located can be obtained.

[0096] After the formation positions of the macro-cracks and the boundary positions of the micro-crack zones of all the secondary units within the primary source unit are determined, the macro-cracks and the boundary positions of the micro-crack zones of all the secondary units within the primary source unit are connected in sequence, and the macro-cracks and micro-crack zones within the source strip of the primary source unit can be obtained.

[0097] In addition, for multiple characteristic parameters such as source energy, source number, ring count, wave velocity, amplitude, frequency, source mechanism, duration, and rise time, the above method can be used to obtain the macro crack extension trajectory and the micro crack zone formation position. In order to obtain the final identification result by integrating the crack trajectories identified by multiple characteristic quantities, the identification results of the source characteristic parameters with identification errors exceeding the threshold are firstly eliminated, and the crack trajectories identified by different source characteristic parameters are used as samples. Then, the deep learning method is used for training to obtain the weights of the crack trajectories identified by different source characteristic parameters, and the deep learning model is obtained to determine the final macro crack extension trajectory and micro crack zone in the level unit. The structure of the deep learning model is not limited.

[0098] Step 6: Reconstruct the identified fractures in each spatial unit at both spatial and temporal levels to obtain the macroscopic fracture extension trajectory during the entire fracturing expansion process.

[0099] After the characteristic parameter data analysis of the cracks is completed, the divided areas and units at all levels are reconstructed as a whole. First, the smallest unit body is reconstructed at the time and space levels, and then the interlayer reconstruction is carried out from the smallest to the largest unit body step by step. The multi-level and multi-directional analysis of the target source body data can realize the accurate identification of the real crack trajectory. After the mathematical statistical model analysis of the three-level source unit body is completed, all the three-level source unit bodies are reconstructed and synthesized into the second-level source unit body in a splitting manner; similarly, after the analysis of the second-level source unit body is completed, it is reconstructed into the first-level source unit body according to the splitting path, such as Fig. 9 As shown in the figure, perspective ① and perspective ② are projection images of the first-level seismic source unit after reconstruction at different perspectives; after the first-level seismic source unit is reconstructed, it is reconstructed on the time level according to the division of time steps, and finally a complete fracture trajectory is formed.

[0100] Specifically, step 6 includes:

[0101] The third-level seismic source unit body that has completed the statistical analysis of the parameters is reconstructed into a second-level seismic source unit body, all the second-level seismic source unit bodies are reconstructed into a first-level seismic source unit body, and all the first-level seismic source unit bodies are reconstructed into the segment macro-cracks corresponding to each time step. At this time, the macro-crack area corresponding to the divided single time step is reconstructed, and the reconstruction of the units of each level divided in each time period on the spatial level is realized. Then, the segment macro-cracks are connected according to the time nodes in each time step, and the crack reconstruction on the time level is completed to obtain the trajectory of the macro-cracks, such as Fig.10 As shown, perspective ① and perspective ② are projection images of the cracks reconstructed at the time level under different perspectives.

[0102] This method proposes a method for intelligently identifying the extension trajectory of macroscopic fractures in the cluster of seismic source of fracturing. The characterization of fracture extension trajectory is upgraded from "only optimizing the location of the seismic source" to "identifying the macroscopic fractures hidden in the cluster of seismic source", and artificial intelligence is introduced to improve the recognition accuracy and efficiency, distinguishing macroscopic fractures from secondary small-scale fracture areas, and realizing high-precision identification of the extension trajectory of macroscopic fractures. The seismic source that characterizes the fracturing fracture is divided into multi-stage multi-region (unit) in time and space, the spatial distribution of multiple parameters of the seismic source characteristics in the sub-region is analyzed to find the macroscopic fractures, and finally multi-region reconstruction and multi-stage reconstruction are performed to obtain the macroscopic fractures hidden in the cluster of seismic source; artificial intelligence algorithms are introduced in multiple stages of the identification process.

[0103] The crack identification method proposed in the present invention is not only applicable to the identification of hydraulic fractures, but also to the crack monitoring of any rock material, and is applicable to the trajectory identification of two-dimensional and three-dimensional cracks. Earthquakes, microseisms, ground sounds and acoustic emissions are all elastic waves with frequencies from low to high, and they are essentially the same, with only frequency differences. Therefore, this method is also applicable to the monitoring of earthquakes, microseisms, ground sounds and acoustic emissions.

Claims

1. A method for intelligently identifying macroscopic fracture extension trajectories in a fracturing source cluster, characterized in that: The steps include: Step 1, collecting elastic wave waveforms released by rock fracture or rupture, filtering the elastic wave waveforms using an intelligent filtering model, and recording the filtered elastic wave waveforms as high-quality waveforms; Step 2: Use deep learning methods to obtain the head starting point of the high-quality waveform, and then obtain the arrival time of the high-quality waveform, and obtain the source point according to the arrival time; Step 3, obtaining the earthquake source clusters of the earthquake source points, dividing the dynamic time steps, dividing the earthquake source clusters according to the dynamic time steps, and obtaining the earthquake source data volume within a single time step; Step 4, dividing the earthquake source data volume within a single time step into multi-level spatial units, and analyzing the distribution characteristics of multiple parameters of earthquake source characteristics within the spatial units; Step 5, identifying the macro cracks corresponding to each parameter according to the distribution characteristics, and obtaining the extension trajectory of the macro cracks in the unit body by integrating the macro cracks identified by each parameter; Step 6: Reconstruct the identified fractures in each spatial unit at both spatial and temporal levels to obtain the macroscopic fracture extension trajectory during the entire fracturing expansion process.

2. The method for intelligently identifying macroscopic fracture extension trajectories in a fracturing source cluster according to claim 1, characterized in that: In step 3, dividing the dynamic time steps includes: Dynamically identify the aggregation form of the point cluster in the earthquake source space, and determine whether the aggregation form conforms to the characteristics of non-planar turning expansion or bifurcation of the crack. If it conforms, continue to search for a time point earlier than the current time until the aggregation form presents a single crack, and the corresponding time is taken as the end point of the time step; If the source distribution shows a single form throughout the entire process of crack expansion, it is necessary to determine whether the amount of spatial source data within a single time step is sufficient for the analysis of the multi-parameter spatial distribution law of source characteristics. If not, the time step needs to be increased until the minimum threshold of spatial source data is reached.

3. The method for intelligently identifying macroscopic fracture extension trajectories in a fracturing source cluster according to claim 1, characterized in that: In step 4, dividing the earthquake source data volume within a single time step into multi-level spatial units includes: The earthquake source data volume in a single time step is divided into multiple first-level earthquake source unit volumes; The primary source unit body is divided into a plurality of secondary source unit bodies; The secondary seismic source unit volume is divided into multiple tertiary seismic source unit volumes.

4. The method for intelligently identifying macroscopic fracture extension trajectories in a fracturing source cluster according to claim 3, characterized in that: The source data volume in a single time step is divided into multiple first-level source units, including: For the approximately elliptical fracturing crack with the injection point as the centroid, a cylindrical coordinate system is constructed with the injection point as the origin O and the normal line of the fracturing crack surface identified by the earthquake source cluster as the axis A. The cylindrical coordinate system parameters r and θ are located on the projection surface of the fracturing crack. When the crack does not expand along the trajectory with the injection point as the centroid, the centroid of the stage expansion of the irregular fracture expansion surface is selected as the origin O, and the normal line of the fracturing crack surface identified by the earthquake source cluster is used as the axis A to establish a cylindrical coordinate system. The cylindrical coordinate system parameters r and θ are located on the projection surface of the irregular fracturing crack. The infinite plane where the axis A and the polar coordinate r are located is the feature plane S0, and the two infinite planes S1 and S2 that are symmetric planes with the feature plane S0 and have the same normal are the data screening intervals, and the vertical distance between plane S1 and plane S2 is w; When the characteristic surface S0 rotates, plane S1 and plane S2 rotate with the characteristic surface; let the characteristic surface S0 pass through the injection point, and rotate the characteristic surface in increments of α degrees until it reaches 180°. During each rotation, the source data between plane S1 and plane S2 are used as the first-level source unit body, and the total number of first-level source unit bodies is 180 / α.

5. The method for intelligently identifying macroscopic fracture extension trajectories in a fracturing source cluster according to claim 4, characterized in that: The first-level seismic source unit body is divided into multiple second-level seismic source unit bodies including: A local spatial rectangular coordinate system is established for each first-level seismic source unit body obtained, the Y axis is in the same direction as the polar coordinate r of the cylindrical coordinate system, the X, Y and Z axes conform to the directions specified by the right-hand system, the Z axis is the thickness direction of the first-level seismic source unit body, and the X axis is the width direction of the first-level seismic source unit body. The seismic source in the first-level seismic source unit body is projected onto the YZ plane to obtain a two-dimensional seismic source data volume; The first-level seismic source unit body is cut along the Y direction to form multiple parallel lines with a spacing of D and parallel to the X-axis, among which the seismic source point cluster covered between two adjacent parallel lines is regarded as the second-level seismic source unit body.

6. The method for intelligently identifying macroscopic fracture extension trajectories in a fracturing source cluster according to claim 5, characterized in that: The secondary source unit body is divided into multiple tertiary source unit bodies including: Each secondary source unit body is split along the X direction to form multiple parallel lines with a spacing of d and parallel to the Y axis, where the source point cluster covered between two adjacent parallel lines is the tertiary source unit body; the size of the tertiary source unit body is D×d.

7. The method for intelligently identifying macroscopic fracture extension trajectories in a fracturing source cluster according to claim 6, characterized in that: In step 4, the distribution characteristics of multiple parameters of earthquake source characteristics within the spatial unit are analyzed, including: The third-level source unit is taken as the smallest statistical unit, and the multi-parameters of the waveform corresponding to the source are statistically analyzed. Then, the distribution characteristics of the multi-parameters of the source characteristics are determined in the second-level source unit using a mathematical statistical model.

8. The method for intelligently identifying macroscopic fracture extension trajectories in a fracturing source cluster according to claim 7, characterized in that: Step 5 includes: According to the distribution law of each parameter, the coordinate value of the macro crack formation position in the secondary source unit body in the local space rectangular coordinate system of the primary source unit body is obtained; Set the picking threshold of the microcrack area on both sides of the macrocrack, and obtain the coordinate value of the boundary of the microcrack area in the secondary source unit in the local space rectangular coordinate system of the primary source unit according to the picking threshold; When the macro-crack formation positions and micro-crack zone boundary positions of all secondary units within the primary source unit are determined, the macro-cracks of all secondary units within the primary source unit and the micro-crack zone boundary positions are connected in sequence to obtain the macro-crack extension trajectory and micro-crack zone formation positions within the primary source strip of the seismic source unit.

9. The method for intelligently identifying macroscopic fracture extension trajectories in a fracturing source cluster according to claim 8, characterized in that: Step 6 includes: Reconstruct the third-level seismic source unit body that has completed the statistical analysis of the parameters into the second-level seismic source unit body, reconstruct all the second-level seismic source unit bodies into the first-level seismic source unit bodies, and reconstruct all the first-level seismic source unit bodies into the segment macro cracks corresponding to each time step; The macroscopic fractures of the segments are connected according to the time nodes within each time step to obtain the trajectory of the macroscopic fractures.

Images