DAS microseismic event automatic identification method based on image segmentation

Abstract

The invention provides a DAS microseismic event automatic identification method based on image segmentation, and the method comprises the steps: inputting two-dimensional image data of original DAS data into a pre-trained semantic segmentation model, enabling the model to carry out the classification of each pixel, and outputting a probability distribution diagram which is the same as the two-dimensional image data in size, the probability display module is used for displaying the probability that each pixel belongs to a microseismic event; by performing event connected domain analysis on the probability distribution diagram, the position and range of the micro-seismic event can be determined, and on the basis of event connected domain analysis, the position of a seismic source can be further determined by combining relevant knowledge of seismology. The recognition precision of the micro-seismic event in an extremely low signal-to-noise ratio environment can be realized; the manual marking cost can be reduced by utilizing the semantic segmentation model, so that the manual intervention is greatly reduced; for large-scale original DAS data, through event connected domain analysis on the basis of a probability distribution diagram, event detection can be rapidly completed, and real-time performance and robustness can be considered.

Description

Technical Field

[0001] The present application relates to the technical field of geophysical information processing, and particularly to an automatic recognition method for DAS microseismic events based on image segmentation. Background Art

[0002] Microseismic monitoring has become increasingly important in hydraulic fracturing operations of unconventional reservoirs such as shale gas and shale oil. Through precise monitoring and in-depth analysis of microseismic events (hereinafter referred to as microseismic events), core data such as the morphological characteristics, azimuth directions, and expansion boundaries of fractures can be accurately captured, providing strong support for optimizing fracturing strategies and improving oil and gas recovery efficiency. Distributed fiber acoustic sensing technology (DAS) can continuously and densely collect high-frequency seismic waves, enriching the dimension and depth of monitoring data. However, at the same time, the data volume generated by DAS is extremely large, and the signal-to-noise ratio is low, resulting in severe challenges in real-time data processing and accurate identification.

[0003] It should be noted that the above introduction to the technical background is only for the convenience of clearly and completely explaining the technical solutions of the present application and facilitating the understanding of those skilled in the art. It cannot be considered that the above technical solutions are well-known to those skilled in the art merely because these solutions are described in the background art part of the present application. Summary of the Invention

[0004] The purpose of the present application is to solve at least one of the technical problems in the related art to a certain extent.

[0005] To this end, the first object of the present application is to propose an automatic recognition method for DAS microseismic events based on image segmentation to perform high-precision microseismic event recognition.

[0006] The second object of the present application is to propose an automatic recognition device for DAS microseismic events based on image segmentation.

[0007] The third object of the present application is to propose an electronic device.

[0008] The fourth object of the present application is to propose a computer-readable storage medium.

[0009] The fifth object of the present application is to propose a computer program product.

[0010] To achieve the above object, the first aspect embodiment of the present application proposes an automatic recognition method for DAS microseismic events based on image segmentation, including:

[0011] Determine the two-dimensional image data corresponding to the original DAS data, where the two-dimensional image data is related to the number of acquisition channels and the number of time sampling points of the original DAS data;

[0012] Input the two-dimensional image data into a pre-trained semantic segmentation model to determine the probability distribution map of microseismic events;

[0013] Perform event connected component analysis on the probability distribution map to determine the source location of the microseismic event.

[0014] To achieve the above object, an embodiment of the second aspect of the present application proposes an automatic recognition device for DAS microseismic events based on image segmentation, including:

[0015] A first acquisition module, which is used to determine the two-dimensional image data corresponding to the original DAS data, where the two-dimensional image data is related to the number of acquisition channels and the number of time sampling points of the original DAS data;

[0016] A second acquisition module, which is used to input the two-dimensional image data into a pre-trained semantic segmentation model to determine the probability distribution map of microseismic events;

[0017] An identification module, which is used to perform event connected component analysis on the probability distribution map to determine the source location of the microseismic event.

[0018] To achieve the above object, an embodiment of the third aspect of the present application proposes an electronic device, including: a processor; a memory for storing instructions executable by the processor; wherein, the processor is configured to execute the instructions to implement the automatic recognition method for DAS microseismic events based on image segmentation proposed in the first aspect embodiment of the present application.

[0019] To achieve the above object, an embodiment of the fourth aspect of the present application proposes a non-transitory computer-readable storage medium, when the instructions in the storage medium are executed by the processor of the electronic device, enabling the electronic device to execute the method proposed in the first aspect embodiment of the present application.

[0020] To achieve the above object, an embodiment of the fifth aspect of the present application proposes a computer program product, including a computer program, which implements the method proposed in the first aspect embodiment of the present application when executed by the processor in the communication device.

[0021] The automatic recognition method for DAS microseismic events based on image segmentation provided by this application can intuitively display the distribution of seismic waves in time and space by converting the original DAS data into two-dimensional image data; input the two-dimensional image data into a pre-trained semantic segmentation model, and the model will classify each pixel and output a probability distribution map with the same size as the two-dimensional image data, which is used to show the probability that each pixel belongs to a microseismic event; by performing event connected component analysis on the probability distribution map, the position and range of the microseismic event can be determined, and on the basis of event connected component analysis, the relevant knowledge of seismology can be combined to further determine the position of the earthquake source. It can achieve the recognition accuracy of microseismic events in an extremely low signal-to-noise ratio environment; using the semantic segmentation model can reduce the manual labeling cost, thus greatly reducing manual intervention; for large-scale original DAS data, through event connected component analysis based on the probability distribution map, event detection can be quickly completed, and both real-time performance and robustness can be taken into account.

[0022] Additional aspects and advantages of this application will be given in part in the following description, will become apparent in part from the following description, or will be understood through the practice of this application. Brief Description of the Drawings

[0023] The above-mentioned and / or additional aspects and advantages of this application will become apparent and easy to understand from the following description of the embodiments in conjunction with the drawings, where:

[0024] Figure 1 It is a schematic flow chart of an automatic recognition method for DAS microseismic events based on image segmentation provided by an embodiment of this application;

[0025] Figure 2 It is a schematic flow chart of another automatic recognition method for DAS microseismic events based on image segmentation provided by an embodiment of this application;

[0026] Figure 3 It is a schematic structural diagram of an automatic recognition device for DAS microseismic events based on image segmentation provided by an embodiment of this application;

[0027] Figure 4 It is a schematic structural diagram of an electronic device provided according to an embodiment of this application. Detailed Description of the Embodiments

[0028] Here, the exemplary embodiments will be described in detail, and the examples are shown in the drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the embodiments of this application. On the contrary, they are only examples of devices and methods consistent with some aspects of the embodiments of this application as detailed in the appended claims.

[0029] The terms used in the embodiments of the present application are for the purpose of describing specific embodiments only and are not intended to limit the embodiments of the present application. The singular forms "a" and "the" used in the embodiments of the present application and the appended claims are also intended to include the plural forms unless the context clearly dictates otherwise. It should also be understood that the term "and / or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.

[0030] It should be understood that although the terms first, second, third, etc. may be used in the embodiments of the present application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the embodiments of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the words "if" and "when" as used herein may be interpreted as "when" or "when" or "in response to determining".

[0031] The embodiments of the present application are described in detail below. Examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to explain the present application and should not be construed as limiting the present application.

[0032] With the development of the exploration and development of unconventional oil and gas resources, microseismic monitoring plays an important role in the hydraulic fracturing operations of unconventional reservoirs such as shale gas and shale oil. Microseismic monitoring technology is a method for dynamically monitoring underground reservoirs using seismic waves. During oil and gas production processes such as hydraulic fracturing, due to the fracture of rocks and the injection of fluids, tiny seismic wave signals will be generated, and these signals can be captured and recorded by sensors on the ground or downhole. Through data analysis, the formation, propagation, and distribution of underground fractures can be revealed. By detecting and analyzing microseismic events, key information such as fracture morphology, orientation, and propagation range can be obtained, which helps to optimize the fracturing plan and improve the recovery rate, where:

[0033] Through the spatio-temporal distribution characteristics of microseismic events, the morphology of fractures can be inferred. For example, parameters such as the length, width, and strike of fractures can be estimated by locating and analyzing microseismic events; the azimuth of microseismic events can be determined by parameters such as the arrival time difference between the P-wave (the P-wave represents the primary wave, which is the fastest seismic wave generated during an earthquake and is also the first to be detected; the P-wave is a compressional wave that propagates by compressing and stretching the material along the wave propagation direction and is used to determine the earthquake epicenter location and propagation path) and the S-wave (the S-wave represents the secondary seismic wave, which is the second-fastest seismic wave and is the second to be detected after an earthquake occurs; the S-wave is a shear wave that can only propagate through solid materials because they require the material to resist lateral bending; the vibration direction of the S-wave is always perpendicular to the wave surface, forming a plane wave and is used to provide more detailed seismic structure information), and these parameters can provide accurate azimuth information of the fractures in the underground space. As hydraulic fracturing progresses, the fractures will gradually expand. By continuously monitoring microseismic events, the fracture expansion process can be tracked in real time, thereby understanding the fracture distribution in the reservoir.

[0034] In some embodiments, according to the microseismic event monitoring results, parameters such as the injection rate and injection pressure of the fracturing fluid can be adjusted to optimize the fracture formation and expansion process. For example, by increasing the injection pressure or changing the injection rate, the fractures can be guided to expand in a more favorable direction.

[0035] In some embodiments, the microseismic event monitoring results can also be used to determine the subsequent fracturing well positions. By analyzing the fracture morphology and azimuth, more favorable well positions can be selected for fracturing operations to improve the oil and gas recovery rate.

[0036] In some embodiments, by comparing the microseismic event distribution characteristics before and after fracturing, the effect of the fracturing operation can be evaluated. If the fracture morphology, azimuth, and expansion range meet the expected goals, it indicates that the fracturing plan is effective; otherwise, the plan needs to be adjusted.

[0037] Traditional microseismic event monitoring schemes are geophone and surface array monitoring. Commonly used methods are the short-time average / long-time average method, cross-correlation detection, etc. Among them, the basic principle of the short-time average / long-time average method is to identify the occurrence of microseismic events by comparing the short-time average energy / long-time average energy of the signals; cross-correlation detection is to use the similarity between two signals to detect microseismic events and is used to identify repeated earthquake events or analyze the propagation characteristics of seismic waves.

[0038] With the emergence of DAS technology, more abundant data has been obtained for microseismic event monitoring. For example, DAS technology can provide high-density measurement points along the optical fiber and continuously collect high-frequency seismic wave fields. However, the signal-to-noise ratio of DAS data is generally low, resulting in severe challenges for traditional methods in terms of real-time processing and recognition accuracy. In a low signal-to-noise ratio environment, the detection effect of traditional time-domain or frequency-domain filtering means is limited, and it is difficult to accurately distinguish microseismic events from background noise. Moreover, DAS data has a large number of channels, a high sampling frequency, and a long acquisition time, causing pressure in processing massive data and increasing the difficulty of real-time detection and positioning. In microseismic event monitoring, it is necessary to balance real-time performance and accuracy. However, the fault tolerance space of traditional solutions is relatively small, and missed detections or false detections are likely to occur. However, the solutions relying on manual marking or manual intervention are too costly and difficult to meet large-scale applications in terms of accuracy and efficiency.

[0039] The following describes the automatic recognition method for DAS microseismic events based on image segmentation according to the embodiments of the present application with reference to the accompanying drawings.

[0040] Figure 1 It is a schematic flowchart of an automatic recognition method for DAS microseismic events based on image segmentation provided by the embodiments of the present application.

[0041] As Figure 1 shown, the method includes but is not limited to the following steps:

[0042] S101, determine the two-dimensional image data corresponding to the original DAS data, where the two-dimensional image data is related to the number of acquisition channels and the number of time sampling points of the original DAS data.

[0043] In some embodiments, the original DAS data usually consists of data obtained by multiple acquisition channels at different time sampling points. These data can be digital values after analog signals are converted by A / D. Each value corresponds to a specific channel and time point. In the acquisition system of the original DAS data, the number of acquisition channels and the number of time sampling points are two important parameters. Among them, the number of acquisition channels refers to the number of independent channels that can simultaneously acquire the original DAS data. Each acquisition channel represents an independent data acquisition point and can monitor different physical quantities (such as vibration, temperature, pressure, etc.) or different spatial positions. The number of acquisition channels directly determines the number of data points that the acquisition system can monitor simultaneously; in practical applications, the selection of the number of acquisition channels depends on multiple factors, including the size of the monitored target area, the types and quantities of physical quantities to be monitored, the cost budget of the system, etc.; although more acquisition channels mean that more abundant data can be obtained, it will also increase the complexity and cost of the system. Therefore, a trade-off should be made according to specific requirements.

[0044] The time sampling point refers to the number of times the acquisition system collects data for each acquisition channel within a specific time interval (e.g., 5 kHz). This time interval is usually referred to as the sampling interval or sampling rate, and the number of time sampling points is the number of data points collected within this interval. The more time sampling points, the more finely the system can capture the changes in the data, thus providing more accurate information. The choice of sampling rate also depends on multiple factors, including the change speed of the physical process to be monitored, the performance limitations of the system, etc. A higher sampling rate can provide higher-resolution data but also increases the storage and processing requirements of the data. Therefore, a trade-off should be made according to specific requirements.

[0045] In some embodiments, the dimension of the two-dimensional image data to be generated can be determined based on the number of acquisition channels and the number of time sampling points in the original DAS data. For example, the horizontal axis can be characterized as the number of time sampling points, that is, the time sequence of each sampling point is used as the abscissa of the image; the vertical axis can be characterized as the number of acquisition channels, that is, different acquisition channels are used as the ordinate of the image. According to the dimension, the original DAS data is reorganized according to the number of acquisition channels and the number of time sampling points to form a two-dimensional matrix. The rows of this two-dimensional matrix represent different acquisition channels, and the columns represent different time sampling points. Among them, each pixel value corresponds to a data point in the two-dimensional matrix.

[0046] S102, input the two-dimensional image data into a pre-trained semantic segmentation model to determine the probability distribution map of microseismic events.

[0047] In some embodiments, before inputting the two-dimensional image data into the semantic segmentation model, the two-dimensional image data is preprocessed according to the input requirements of the semantic segmentation model. Preprocessing is an important link in image processing, machine learning, or microseismic event analysis, and usually involves a series of operations to ensure the quality, consistency, and applicability of the two-dimensional image data, so as to meet the requirements of subsequent analysis or model training.

[0048] In some embodiments, the preprocessing of two-dimensional image data includes: data cleaning, data normalization, data augmentation, feature extraction, other requirements, etc. The two-dimensional image data may contain various noises, such as Gaussian noise, salt-and-pepper noise, etc. These noises will affect the quality of the two-dimensional image data and the results of subsequent analysis. Therefore, filters or other methods can be used to remove the noises; if there are missing pixel values (such as black or white areas) in the two-dimensional image data, filling is required to ensure the integrity of the two-dimensional image data. The pixel values of the two-dimensional image data can be scaled to a specific range (such as 0-1 or 0-255), which helps to eliminate the brightness differences between different images for subsequent processing and analysis; the histogram of the two-dimensional image data can also be adjusted to make its distribution more uniform, thereby enhancing the contrast of the two-dimensional image. Operations such as rotation or flipping can be performed on the two-dimensional image data to increase the diversity of the data and improve the generalization ability of the semantic segmentation model; the two-dimensional image data can also be cropped or scaled according to requirements to adapt to different analysis or training needs. Algorithms such as Canny and Sobel can be used to detect the edge features in the two-dimensional image data; the texture features of the two-dimensional image data can be extracted by using methods such as gray-level co-occurrence matrix. If the two-dimensional image data contains information irrelevant to subsequent analysis (such as background, labels, etc.), it can be removed. It should be noted that suitable algorithms should be selected according to specific requirements to preprocess the two-dimensional image data, which will not be elaborated here.

[0049] In some embodiments, a suitable semantic segmentation model can be selected, such as U-Net, DeepLab, FCN, BASNet, etc. The semantic segmentation model can be trained according to the idea of event-background binary segmentation to ensure that the model outputs a two-dimensional probability map with the same size as the input. This two-dimensional probability map is characterized as the probability distribution map of microseismic events. In the probability distribution map, the value of each pixel point is between 0 and 1, and this value represents the probability of belonging to the microseismic event.

[0050] In some embodiments, to improve the processing efficiency of the semantic segmentation model, the two-dimensional image data can be divided into multiple batches for processing, and each batch contains a certain number of images. Then the semantic segmentation model classifies each pixel and outputs a segmentation result with the same size as the input image. It should be noted that the BASNet network architecture can improve the segmentation accuracy of the semantic segmentation model in edge monitoring and low signal-to-noise ratio regions through indicators such as hybrid loss function, binary cross-entropy, structural similarity, and intersection over union.

[0051] It should be added that the output of the semantic segmentation model is usually a multi-dimensional array, which contains the probabilities of each pixel belonging to different categories. The probability distribution belonging to the microseismic event can be extracted from the model output. To more intuitively understand the results, the probability distribution can be visualized. For example, the probability values can be mapped to different color entropies to generate a probability distribution map in the form of a color heat map or a grayscale form of the probability distribution map.

[0052] S103. Perform event connected component analysis on the probability distribution map to determine the source location of the microseismic event.

[0053] In some embodiments, the probability distribution map is usually used to display the probability distribution of a certain event or variable at different values. In seismic analysis, the probability distribution map may be used to represent the amplitude, frequency, or other related parameter distributions of seismic waveforms. For microseismic events, the probability distribution map may show the distribution characteristics of microseismic signals at different times, spaces, or frequencies. In order to locate the specific location of the microseismic event, a probability threshold can be set. Only when the probability of a pixel point exceeds this threshold is it considered that the pixel point belongs to the microseismic event. On the probability distribution map, the microseismic event may appear as some connected regions. By analyzing these connected regions, information such as the location, size, and shape of the microseismic event can be determined.

[0054] In some embodiments, schemes such as label algorithms, morphological operations, and graph theory methods can be used to identify the connected components in the probability distribution map. Then, feature extraction is performed on the identified connected components, including parameters such as amplitude, frequency, and duration, for further analysis. By comparing the feature parameters of different connected components and using attributes such as similarity features or correlations, such as similarity or distance metrics, signals that may belong to the same microseismic event are identified.

[0055] In some embodiments, for signals that may belong to the same microseismic event, methods such as partitioned source location methods and non-iterative source location methods can be used to estimate the source location using the features of the signals. For the source location, methods such as the least squares method can also be used to optimize the source location to further improve the location accuracy. Finally, methods such as waveform inversion and seismic phase identification can also be used to cross-validate the source location to improve the accuracy.

[0056] In a feasible implementation, a clustering algorithm can also be used to analyze the probability distribution map. A region of adjacent pixel points with a probability greater than the probability threshold is regarded as a candidate "microseismic event", and then the source location of the microseismic event is determined.

[0057] In summary, the automatic DAS microseismic event recognition method based on image segmentation proposed in this application can intuitively display the distribution of seismic waves in time and space by converting the original DAS data into two-dimensional image data. The two-dimensional image data is input into a pre-trained semantic segmentation model, which classifies each pixel and outputs a probability distribution map with the same size as the two-dimensional image data to show the probability of each pixel belonging to a microseismic event. By performing event connected component analysis on the probability distribution map, the location and range of microseismic events can be determined, and based on the event connected component analysis, relevant seismological knowledge can be combined to further determine the location of the seismic source. It can achieve the recognition accuracy of microseismic events in an extremely low signal-to-noise ratio environment; using the semantic segmentation model can reduce the manual labeling cost, thus greatly reducing manual intervention. For large-scale original DAS data, through event connected component analysis based on the probability distribution map, event detection can be quickly completed, taking into account both real-time performance and robustness.

[0058] Figure 2 It is a schematic flowchart of another automatic DAS microseismic event recognition method provided by an embodiment of this application.

[0059] As Figure 2 shown, this method includes but is not limited to the following steps:

[0060] S201, determine the overlapping region according to the number of acquisition channels and the number of time sampling points of the original DAS data.

[0061] In the embodiment of this application, in the acquisition of the original DAS data, if multiple acquisition channels acquire the same signal source, the data of these acquisition channels may overlap in time. By comparing the data of different acquisition channels in the same time period, the overlapping region can be determined.

[0062] Optionally, since the data acquired by the same acquisition channel in different time periods may overlap due to the periodic change of the signal source, the overlapping time period can be determined by analyzing the periodicity of the signal, and then the overlapping region can be determined.

[0063] As an example, the actually acquired original DAS data can be first segmented in the time dimension into several small tiles of 400×400, and 50% overlap is reserved in time and space, and then the overlapping region can be determined.

[0064] S202, perform sliding window slicing on the original DAS data according to the overlapping region to obtain two-dimensional image data.

[0065] In the embodiments of the present application, the time window size of each slice (i.e., the length of the slice on the time axis) can be determined according to the size of the overlapping region and the required image resolution. Next, the step size between slices (i.e., the overlapping part of adjacent slices on the time axis) is determined, where the step size can be smaller than the slice to achieve a higher degree of overlap.

[0066] Optionally, according to the acquisition channels involved in the overlapping region, a combination of acquisition channels for generating two-dimensional image data is selected. If the data of multiple acquisition channels need to be merged, fusion strategies such as weighted average, maximum value, and minimum value can be considered.

[0067] In the embodiments of the present application, an empty two-dimensional array can be set first to store the image data after slicing; the starting position and index of the sliding window are initialized. As an example, a Hanning window can be used for the sliding window operation. For each selected acquisition channel, a data segment corresponding to the slice window size is extracted from the original DAS data. If multiple acquisition channels are involved, the data segments of these acquisition channels are merged according to the selected fusion strategy, and the merged data segment is stored in the corresponding position of the two-dimensional data. Next, according to the step size, the starting position of the sliding window is updated, and the sliding window slicing process is repeated until the entire original DAS data is traversed to obtain two-dimensional image data.

[0068] Optionally, after generating the two-dimensional image data, operations such as image enhancement, edge detection, and feature extraction can also be performed. These operations help to extract useful information and improve the visual effect of the two-dimensional image data.

[0069] Optionally, the generated two-dimensional image data can be quantitatively evaluated according to specific evaluation metrics (such as signal-to-noise ratio, contrast, clarity, etc.). By comparing the evaluation results under different slice parameters and fusion strategies, the optimal two-dimensional image data is selected.

[0070] S203, input the two-dimensional image data into a pre-trained semantic segmentation model to determine the probability distribution map of microseismic events.

[0071] In the embodiments of the present application, before determining the two-dimensional image data corresponding to the original DAS data, a semantic segmentation model is first determined and trained. Semantic segmentation aims to classify each pixel of an image into a specific category. For the original DAS data, a fully convolutional network, DeepLab, or BASNet network architecture can be selected as the semantic segmentation model. Taking the BASNet network architecture as an example, it includes a prediction module and a residual refinement module, and its hybrid loss function combines binary cross-entropy (BCE), structural similarity (SSIM), and intersection over union (IoU) losses to supervise the training process at the pixel level, block level, and map level.

[0072] Optionally, the seismic wave propagation velocity parameters and anisotropy parameters can be determined according to the geological structure of the target area. Among them, the shear wave velocity measurement and the compressional wave velocity measurement can be carried out on the geological structure of the target area to determine the seismic wave propagation velocity parameters; the empirical relationship equation of velocity and density proposed by Castagna et al. can also be used to calculate the seismic wave propagation velocity parameters according to the rock type and empirical coefficient; if there is an obvious anisotropic geological structure (such as layered rocks, fault zones, etc.) in the target area, the propagation velocity of seismic waves may change with the change of the propagation direction and the particle polarization direction. Therefore, the seismic wave propagation velocity parameters can be calculated by the wave equation in an anisotropic elastic medium. The Thomsen parameter modeling (Tsvankin parameter modeling can also be used) can be carried out on the geological structure of the target area to determine the anisotropy parameters; the wide-azimuth seismic data, cross-well seismic data, etc. can also be used to obtain the anisotropy parameters through inversion methods.

[0073] Optionally, full-waveform forward modeling can be carried out on the seismic wave propagation velocity parameters and anisotropy parameters to obtain the simulated data of the seismic wave field. Among them, a three-dimensional elastic wave equation can be constructed according to the seismic wave propagation velocity parameters and anisotropy parameters. Then, the finite difference method is used to numerically solve the three-dimensional elastic wave equation for full-waveform forward modeling to obtain the simulated data of the seismic wave field.

[0074] Optionally, the background noise in the original DAS data is superimposed on the simulated data to obtain a composite data set. Among them, according to the preset sampling frequency and channel spacing, the convolution method is used to superimpose the background noise in the original DAS data and the simulated data to simulate the composite data under different signal-to-noise ratio conditions, and the composite data is collected to obtain a composite data set.

[0075] Optionally, a semantic segmentation model is determined according to the composite data set. Among them, the label information for simulating microseismic events is extracted from the composite data set; according to the label information, the training data of the semantic segmentation model is determined; according to the BASnet network architecture and the training data, the semantic segmentation model is obtained.

[0076] As an example, for a microseismic source, assuming a double-couple mechanism (Double-Couple Source), which is consistent with the type of fault slip induced by hydraulic fracturing, the influence of anisotropy on stress-strain is considered (using the correct elastic constant matrix) during the numerical iteration. When calculating the strain corresponding to the fiber axis, when the medium is anisotropic, the following formula is obtained through the stress-strain tensor relationship:

[0077]

[0078] Among them, σ xx 、σ yy, σ zz are the main diagonal components of the stress tensor, i.e., the normal stress components in the x, y, and z directions respectively; ε xx , ε yy , ε zz are the main diagonal components of the strain tensor, i.e., the normal strain components in the x, y, and z directions; c ij is the elastic constant of the acoustic wave propagation medium, which is deduced from the velocity and density or other anisotropic parameters. Among them, the seismic wave propagation velocity parameters (including the shear wave velocity parameter and the longitudinal wave velocity parameter) and the anisotropic parameters can be converted into the elastic constant c ij through finite element analysis or finite difference method, and then the velocity and stress are iteratively updated in the staggered network.

[0079] It should be noted that due to the existence of orthorhombic anisotropy (ORT) or vertical transverse isotropy (VTI), some elements in c ij are equal or zero.

[0080] The "axial strain along the fiber optic wind direction" obtained by DAS observation. For example, when the fiber optic direction is close to the z-axis, σ zz / c 33 can be approximately taken as the main observable; if the fiber optic is tilted in three-dimensional space, the stress-strain tensor needs to be projected onto the fiber optic direction vector to obtain the strain signal ε t in the fiber optic direction; according to the fiber optic distribution in the target area, all the strain signals ε t are converged into the simulation data of the seismic wave field.

[0081] Then the simulation data is embedded into the randomly selected measured background noise to simulate the actual low signal-to-noise ratio environment on site, and a composite data set is obtained. Among them, a threshold is defined for the simulation data, and the area with an absolute amplitude greater than 1 standard deviation can be regarded as the microseismic event area; after the simulation data is superimposed with the background noise, the label is adjusted according to the signal-to-noise ratio of the actual waveform to mark the appropriate microseismic event boundary area.

[0082] In the embodiments of the present application, based on the semantic segmentation model, the two-dimensional image data is classified pixel by pixel to obtain the probability value of each pixel belonging to the microseismic event. Among them, according to the preset flipping angle (as an example, it can be rotated 0° / 90° / 180° / 270°, horizontally or vertically flipped, etc.), each pixel in the two-dimensional image data is geometrically transformed to obtain the target two-dimensional image data. Then, according to the pixel-by-pixel classification operation of the semantic segmentation model on the target two-dimensional image data, the probability value of each pixel belonging to the microseismic event is obtained. Next, the probability values of adjacent pixels are spliced to determine the probability distribution map of the microseismic event.

[0083] S204. Perform event connected component analysis on the probability distribution map to determine the source location of the microseismic event.

[0084] In the embodiment of the present application, the probability distribution map can be binarized according to a preset threshold to obtain a binarized matrix. Among them, the pixel points with a probability greater than the preset threshold are regarded as microseismic event pixel points, and the microseismic event pixel points are assigned a value of 1; the pixel points with a probability less than or equal to the preset threshold are assigned a value of 0. Then, perform connected component analysis on the binarized matrix to identify all connected regions formed by the microseismic event pixel points, and determine each connected region as the source location of the microseismic event.

[0085] In summary, the automatic identification method for DAS microseismic events based on image segmentation proposed in the present application can intuitively display the distribution of seismic waves in time and space by converting the original DAS data into two-dimensional image data; input the two-dimensional image data into a pre-trained semantic segmentation model, and the model will classify each pixel and output a probability distribution map with the same size as the two-dimensional image data to show the probability of each pixel belonging to a microseismic event; by performing event connected component analysis on the probability distribution map, the location and range of the microseismic event can be determined, and based on the event connected component analysis, the relevant knowledge of seismology can be combined to further determine the source location. It can achieve the recognition accuracy of microseismic events in an extremely low signal-to-noise ratio environment; using the semantic segmentation model can reduce the manual labeling cost, thus greatly reducing manual intervention; for large-scale original DAS data, through event connected component analysis based on the probability distribution map, event detection can be quickly completed, and both real-time performance and robustness can be taken into account.

[0086] Figure 3 It is a schematic structural diagram of an automatic identification device for DAS microseismic events based on image segmentation provided by an embodiment of the present application. As Figure 3 shown, the automatic identification device 300 for DAS microseismic events based on image segmentation includes:

[0087] The first acquisition module 301 is used to determine the two-dimensional image data corresponding to the original DAS data, where the two-dimensional image data is related to the acquisition channel number and time sampling points of the original DAS data;

[0088] The second acquisition module 302 is used to input the two-dimensional image data into a pre-trained semantic segmentation model to determine the probability distribution map of microseismic events;

[0089] The recognition module 303 is used to perform event connected component analysis on the probability distribution map to determine the source location of the microseismic event.

[0090] Figure 4A schematic structural diagram of an electronic device according to an embodiment of the present application. Figure 4 The electronic device shown is merely an example and should not impose any limitations on the functions and usage scope of the embodiments of the present application.

[0091] As Figure 4 shown, the electronic device 400 includes a processor 401, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM, Read Only Memory) 402 or a program loaded from a memory 406 into a random access memory (RAM, Random Access Memory) 403. In the RAM 403, various programs and data required for the operation of the electronic device 400 are also stored. The processor 401, the ROM 402, and the RAM 403 are connected to each other via a bus 404. An input / output (I / O, Input / Output) interface 405 is also connected to the bus 404.

[0092] The following components are connected to the I / O interface 405: a memory 406 including a hard disk, etc.; and a communication part 407 including a network interface card such as a LAN (Local Area Network) card, a modem, etc., and the communication part 407 performs communication processing via a network such as the Internet; a drive 408 is also connected to the I / O interface 405 as required.

[0093] Specifically, according to an embodiment of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program carried on a computer-readable medium, and the computer program includes program codes for performing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 407. When the computer program is executed by the processor 401, the above functions defined in the method of the present application are executed.

[0094] In an exemplary embodiment, a storage medium including instructions is also provided, such as a memory including instructions, and the above instructions can be executed by the processor 401 of the electronic device 400 to complete the above method. Optionally, the storage medium may be a non-transitory computer-readable storage medium. For example, the non-transitory computer-readable storage medium may be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.

[0095] In the present application, a computer-readable storage medium may be any tangible medium that contains or stores a program, which can be used by or in conjunction with an instruction execution system, apparatus, or device. In the present application, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries computer-readable program code. Such a propagated data signal may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on a computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0096] Other embodiments of the present invention will be readily apparent to those skilled in the art after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present invention that follow the general principles of the present invention and include known common knowledge or conventional technical means in the technical field not disclosed in this application. The specification and examples are only to be considered as exemplary, and the true scope and spirit of this application are pointed out by the following claims.

[0097] It should be understood that this application is not limited to the exact structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is only limited by the appended claims.

Claims

1. A DAS microseismic event automatic identification method based on image segmentation, characterized in that: include: Determining two-dimensional image data corresponding to the original DAS data, wherein the two-dimensional image data is related to the number of acquisition channels and the number of time sampling points of the original DAS data; Inputting the two-dimensional image data into a pre-trained semantic segmentation model to determine a probability distribution map of microseismic events; An event connected domain analysis is performed on the probability distribution diagram to determine the source location of the microseismic event.

2. The method according to claim 1, characterized in that Before determining the two-dimensional image data corresponding to the original DAS data, the method further includes: Determine the seismic wave propagation velocity parameters and anisotropy parameters according to the geological structure of the target area; Performing full waveform forward modeling on the seismic wave propagation velocity parameters and the anisotropy parameters to obtain simulation data of the seismic wave field; superimposing the background noise in the original DAS data with the simulated data to obtain a composite data set; A semantic segmentation model is determined according to the composite data set.

3. The method according to claim 2, characterized in that Determining the seismic wave propagation velocity parameters and anisotropy parameters according to the geological structure of the target area includes: Performing shear wave velocity measurement and longitudinal wave velocity measurement on the geological structure of the target area to determine the seismic wave propagation velocity parameters; The geological structure of the target area is modeled using Thomsen parameters to determine the anisotropy parameters.

4. The method according to claim 2, characterized in that: The full waveform forward modeling of the seismic wave propagation velocity parameter and the anisotropy parameter to obtain simulation data of the seismic wave field includes: Constructing a three-dimensional elastic wave equation according to the seismic wave propagation velocity parameter and the anisotropy parameter; The three-dimensional elastic wave equation is numerically solved by using a finite difference method to perform full waveform forward modeling and obtain simulation data of the seismic wave field.

5. The method according to claim 2, characterized in that: The method of superimposing the background noise in the original DAS data with the simulated data to obtain a composite data set includes: According to the preset sampling frequency and channel spacing, the convolution method is used to superimpose the background noise in the original DAS data with the simulated data to simulate the composite data under different signal-to-noise ratio conditions, and the composite data is collected to obtain the composite data set.

6. The method according to claim 2, characterized in that Determining a semantic segmentation model according to the composite data set includes: Extracting label information for simulating microseismic events from the composite data set; Determining training data for the semantic segmentation model according to the label information; According to the BASnet network architecture and the training data, the semantic segmentation model is obtained.

7. The method according to claim 1, characterized in that The determining of the two-dimensional image data corresponding to the original DAS data includes: Determining the overlapping area according to the number of acquisition channels and the number of time sampling points of the original DAS data; According to the overlapping area, the original DAS data is sliced ​​by sliding window to obtain the two-dimensional image data.

8. The method according to claim 1, characterized in that The step of inputting the two-dimensional image data into a pre-trained semantic segmentation model to determine a probability distribution map of microseismic events includes: Based on the semantic segmentation model, the two-dimensional image data is classified pixel by pixel to obtain a probability value of each pixel belonging to a microseismic event; The probability values ​​of adjacent pixel points are spliced ​​to determine the probability distribution map of the microseismic event.

9. The method according to claim 8, characterized in that The pixel-by-pixel classification of the two-dimensional image data based on the semantic segmentation model to obtain a probability value of each pixel belonging to a microseismic event includes: According to a preset flip angle, geometrically transform each pixel in the two-dimensional image data to obtain target two-dimensional image data; The semantic segmentation model is used to perform pixel-by-pixel classification on the target two-dimensional image data to obtain a probability value of each pixel point belonging to a microseismic event.

10. The method according to any one of claims 1 to 9, wherein the performing event connected domain analysis on the probability distribution graph to determine the source location of the microseismic event comprises: According to a preset threshold, the probability distribution map is binarized to obtain a binarized matrix, wherein pixel points with a probability greater than the preset threshold are regarded as microseismic event pixel points, and the microseismic event pixel points are assigned a value of 1; pixel points with a probability less than or equal to the preset threshold are assigned a value of 0; Connected domain analysis is performed on the binary matrix to identify all connected regions formed by connecting the microseismic event pixel points, and each of the connected regions is determined as the epicenter position of the microseismic event.

Images