A Method for Extracting Weak Signals and Three-Dimensional Localization of Deep Shale Oil Surface Microseismic Reactions

Abstract

This application relates to the field of oil and gas field development technology, specifically to a method for extracting and three-dimensionally locating weak seismic signals from deep shale oil reservoirs. The method includes: during fracturing operations in deep shale oil reservoirs within the target exploration area, acquiring weak seismic signals induced by fracturing fractures through each geophone channel; calculating the characteristic difference degree representing the deviation of waveform morphology between signals, using this as a metric to classify the weak seismic signals from all geophone channels, thereby selecting a target signal set; determining the disturbance assessment value of each weak seismic signal within the target signal set, selecting a benchmark signal, extracting the time window from the benchmark signal to the complete attenuation characteristic of the overlying formation fracture, and based on this, extracting effective signals from the weak seismic signals within the target signal set for three-dimensional location of the fracturing fractures. This application improves the accuracy of three-dimensional location of fracturing fractures in deep shale oil reservoirs.

Description

Technical Field

[0001] This application relates to the field of oil and gas field development technology, specifically to a method for extracting weak signals from surface microseismic activity and three-dimensional localization in deep shale oil formations. Background Technology

[0002] In the exploration and development of deep shale oil, hydraulic fracturing is a core method for achieving commercial reservoir production. Due to the complex geological conditions of deep reservoirs, indiscriminate fracturing can easily lead to insufficient fracture propagation or fractures penetrating the reservoir, severely impacting single-well production and wasting fracturing fluid. Therefore, real-time acquisition of the three-dimensional spatial distribution of fracturing fractures using surface microseismic monitoring technology is crucial for guiding the dynamic adjustment of fracturing operation parameters.

[0003] However, due to the complex geological characteristics of deep shale oil reservoirs, including high stress and strong heterogeneity, the fracturing generates weak seismic waves. These signals undergo strong formation absorption and geometric attenuation as they propagate to the surface, and are further amplified by various background noises such as surface environmental noise and mechanical interference. This results in extremely weak signal energy and a very low signal-to-noise ratio received by ground geophones. Traditional microseismic monitoring methods struggle to effectively extract the true fracture response characteristics when processing such weak signals from deep formations. Furthermore, complex interferences easily generate pseudo-signals that resemble the actual signal waveforms, leading to large positioning errors and a high misjudgment rate in the fracturing of deep shale oil reservoirs, thus affecting the development efficiency of deep shale oil. Summary of the Invention

[0004] To address the aforementioned technical issues, a method for extracting weak signals from deep shale oil surface microseismic data and establishing three-dimensional localization is provided to resolve existing problems.

[0005] The solution to the technical problem presented in this application is to provide a method for extracting weak signals from deep shale oil surface microseismic formations and for three-dimensional localization, comprising the following steps:

[0006] When performing fracturing operations in deep shale oil reservoirs in the target exploration area, weak seismic signals induced by fracturing fractures are collected through each geophone channel;

[0007] By calculating the characteristic difference degree, which characterizes the degree of waveform morphology deviation between signals, based on the inconsistency of local fluctuation characteristics between weak seismic signals from any two geophone channels, and using this as a metric for distance, weak seismic signals from all geophone channels are classified, thereby selecting the target signal set.

[0008] For the target signal set, the degree of deviation of the distribution characteristics of different weak seismic signals in the frequency domain is evaluated to reflect the degree of deviation from the common law of the actual fracturing response in the frequency domain, and the disturbance assessment value of each weak seismic signal in the target signal set is determined.

[0009] Based on the disturbance assessment value, a reference signal is selected. The time window from the start of the fracture in the overlying strata to the complete attenuation feature is extracted from the reference signal. Based on this, the effective signal is extracted from the weak seismic signals in the target signal set, and a deep neural network model is used to perform three-dimensional localization of the hydraulic fracture.

[0010] Preferably, the process of obtaining the feature difference degree is as follows:

[0011] After waveform alignment processing of the weak seismic signals from all detector channels, all peaks in the weak seismic signals are obtained for each detector channel, and the full width at half maximum (FWHM) and kurtosis of each peak are calculated. The time corresponding to each peak is defined as the peak time. The peak value, peak time, FWHM, and kurtosis of each peak in the weak seismic signal are used to form a feature vector. The feature vectors of all peaks in the weak seismic signal are arranged in the order of the peak times to form the peak feature sequence of the detector channel.

[0012] Calculate the difference between the eigenvectors at the same position in the peak characteristic sequences of any two detector channels, and use this as the relative difference;

[0013] The characteristic difference degree is positively correlated with the relative difference.

[0014] Preferably, the process of obtaining the target signal set is as follows: using the feature difference degree as the distance metric, the weak seismic signals of all detector channels are clustered to obtain multiple clusters; the number of weak seismic signals contained in each cluster is counted, and the cluster corresponding to the largest number is selected and defined as the target signal set.

[0015] Preferably, the calculation process for the disturbance assessment value is as follows:

[0016] Frequency domain analysis was performed on each weak seismic signal within the target signal set to extract various spectral features;

[0017] For any two weak seismic signals in the target signal set, calculate the difference between them in the same spectral characteristics, and use it as a single feature difference.

[0018] Using each weak seismic signal in the target signal set as a sample, and all kinds of spectral features as evaluation indicators, the weights of various spectral features are calculated using the entropy weight method.

[0019] For the target signal set, the single feature differences between each weak seismic signal and all other weak seismic signals in various spectral features are positively fused to serve as the relative deviation of each weak seismic signal in various spectral features.

[0020] Based on the weights of various spectral features, the relative deviations of all spectral features of each weak seismic signal in the target signal set are weighted and summed to obtain the disturbance assessment value for each weak seismic signal in the target signal set.

[0021] Preferably, the various spectral characteristics include the dominant frequency, bandwidth, and energy concentration of the weak seismic signal in the frequency domain.

[0022] Preferably, the process of obtaining the dominant frequency and energy concentration is as follows: each weak seismic signal is decomposed using a wavelet packet decomposition algorithm to obtain the reconstructed signal of each frequency band, the energy of the reconstructed signal of each frequency band is calculated, the center frequency of the frequency band with the maximum energy is selected as the dominant frequency, the sum of the energy of the reconstructed signals of all frequency bands is counted as the total energy, and the proportion of the energy of the reconstructed signal of the frequency band where the dominant frequency is located in the total energy is taken as the energy concentration.

[0023] Preferably, the process of obtaining the bandwidth is as follows: select the frequency range covered by a continuous frequency band whose energy exceeds a preset threshold from all reconstructed signals under each weak seismic signal, and use this range as the bandwidth.

[0024] Preferably, the reference signal is the weak seismic signal corresponding to the minimum disturbed assessment value within the target signal set.

[0025] Preferably, the extraction process of the time window is as follows:

[0026] The first arrival time of the P-wave of the reference signal is extracted using the first arrival picking algorithm. The time when the reference signal is traced back a preset time along the time axis from the first arrival time of the P-wave is defined as the starting time.

[0027] Calculate the instantaneous energy at each moment in the reference signal, obtain the maximum peak value among all instantaneous energies after the P-wave arrival time, and search backward from the moment corresponding to the maximum peak value for the moment when the instantaneous energy drops to a preset percentage of the maximum peak value, which is defined as the end time;

[0028] The time period contained in the reference signal from the start time to the end time is defined as the time window.

[0029] Preferably, the step of using a deep neural network model to perform three-dimensional localization of the hydraulic fracturing fracture includes: establishing a three-dimensional formation velocity model of the target exploration area; and using a UNet-Transformer hybrid network model to perform three-dimensional localization of the hydraulic fracturing fracture based on the three-dimensional spatial coordinates of all effective signals and their corresponding geophones, as well as the three-dimensional formation velocity model.

[0030] This application has at least the following beneficial effects:

[0031] This application calculates the characteristic difference degree and uses it as a distance metric to classify weak seismic signals from all geophone channels, thereby selecting a target signal set. Its beneficial effect lies in assessing the deviation of waveform morphology between any two geophone channels in the time domain. Furthermore, by utilizing the spatial continuity of vibration signals generated by actual hydraulic fracturing fractures, which theoretically should be effectively received by most geophone channels and exhibit similar waveform morphology, and considering that pseudo-signals generated by local interference only appear on a few channels, cluster analysis is used to select the target signal set. This allows for the accurate removal of waveform islands caused by local random interference such as construction machinery and surface vehicles in environments with extremely low signal-to-noise ratios, reducing the impact of pseudo-signals on subsequent signals. Interference in fracture location; determining the perturbation assessment value of each weak seismic signal within the target signal set. Its beneficial effect lies in introducing a deeper frequency domain feature assessment based on preliminary time-domain screening. Since real fractured rock exhibits common response patterns in frequency domain characteristics such as dominant frequency and bandwidth, calculating the perturbation assessment value allows for precise quantification of the physical purity of each signal within the target signal set. Furthermore, a benchmark signal with the lowest physical signal-to-noise ratio and least affected by formation absorption attenuation distortion and environmental noise pollution is selected from the target signal set. This signal most closely approximates the common patterns of real fracture response in terms of frequency domain distribution characteristics, exhibits the lowest degree of interference, and possesses the highest reliability, providing a reliable basis for subsequent time window extraction. The established reference standard fundamentally eliminates the risk of distortion in effective capture due to the selection of inferior signals as time references. Extracting the time window from the reference signal to capture the complete characteristics of the fracture initiation and attenuation of the overlying strata offers the advantage of establishing a unified time window based on the most reliable reference signal. This ensures that the capture range fully encompasses the precursory characteristics of P-wave initiation, the high-energy oscillations of S-waves, and the entire attenuation process of the coda, providing complete waveform information for fracture location. Extracting effective signals from weak seismic signals within the target signal set and employing a deep neural network model for three-dimensional fracture location offers the advantage of eliminating pure noise segments before and after the time window through unified time window capture. The effective signal retains only the valid waveform containing fracture feature information, improving the signal-to-noise ratio and the purity of the positioning input. Secondly, since the waveforms of each signal have been aligned, this time window ensures that the signal segments intercepted by different detector channels physically correspond to the same fracture event, avoiding the physical mismatch problem caused by independent window selection of each detector channel in traditional methods. In addition, since the effective signal input to the deep neural network has undergone time-domain spoofing, frequency-domain purification, and absolutely guarantees the physical integrity of the wave field time difference, it completely eliminates the misleading effect of strong background noise on the learning of network weights, realizing high-precision three-dimensional mapping of the spatial coordinates of deep weak vibrations and improving the three-dimensional positioning accuracy of fracturing fractures in deep shale oil reservoirs. Attached Figure Description

[0032] The following section, in conjunction with the accompanying drawings, provides a more detailed description of a method for extracting weak signals and three-dimensionally locating microseismic signals in deep shale oil formations.

[0033] Figure 1 A flowchart illustrating the steps of a method for extracting weak signals from deep shale oil surface microseismic formations and establishing three-dimensional localization, provided in this application embodiment;

[0034] Figure 2 A flowchart illustrating the steps of the method for obtaining the disturbed evaluation value provided in the embodiments of this application. Detailed Implementation

[0035] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description, in conjunction with the accompanying drawings and implementation examples, provides a method for extracting weak signals and establishing three-dimensional localization of deep shale oil surface microseismic data proposed in this application. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0036] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0037] Please see Figure 1 The diagram illustrates a flowchart of a method for extracting weak signals and three-dimensionally locating deep shale oil surface microseismic signals according to an embodiment of this application. The method includes the following steps:

[0038] Step 1: When carrying out fracturing operations in the deep shale oil reservoir of the target exploration area, weak seismic signals induced by the fracturing fractures are collected through each geophone channel.

[0039] In the hydraulic fracturing development of deep shale oil, ground monitoring of fracturing-induced rock microfracture signals is typically required to assess the real-time propagation morphology and spatial distribution of fractures. Analysis and localization of these microfracture signals can effectively invert the spatial distribution of the fracture network, providing crucial information for evaluating fracturing effectiveness, optimizing construction parameters, and adjusting subsequent development plans. However, deep shale oil reservoirs are typically buried at depths exceeding 3000 meters, and the reservoir region possesses complex geological characteristics such as high stress and strong heterogeneity. Signals generated by fractures undergo strong geometric attenuation and medium absorption as they propagate upwards to the surface, while also being superimposed with various background noises and random interferences. This results in extremely weak signal energy and a low signal-to-noise ratio received at the surface, making it difficult to accurately extract effective features. Consequently, this limits the accuracy of fracturing fracture spatial localization and the reliability of interpretation.

[0040] Based on the above analysis, in order to improve the ability to capture weak vibration signals generated by fracturing in deep shale oil reservoirs and ensure the integrity of the signals, this embodiment uses the fracturing wellhead as the core to construct a high-density surface observation array, with the specific deployment method as follows:

[0041] Within the target exploration area, a ground observation array is constructed using a circular or cross-shaped layout, centered on the fractured wellhead. In this embodiment, a cross-shaped layout is selected, wherein the array channel spacing is set to 50~100m, and the number of layout points in a single array is greater than or equal to 96, to ensure all-round, high-resolution reception of weak signals generated by deep fractures. As for other implementation methods, the implementer can set them according to the actual situation.

[0042] High-sensitivity broadband geophones are deployed at each point in the ground observation array to synchronously acquire the weak vibrations induced by rock fracturing during the fracturing process of deep shale oil reservoirs and obtain the weak seismic signals of each geophone channel; among them, the broadband geophones are selected as three-component piezoelectric broadband geophones or moving-coil high-sensitivity geophones.

[0043] Secondly, the weak seismic signals are preprocessed, including:

[0044] Since the signal energy at depth is weak and easily affected by noise, resulting in poor signal quality, a Convolutional Denoising Autoencoder (CDAE) is used to denoise the weak seismic signal. CDAE is a well-known technology and will not be described in detail here.

[0045] Secondly, to eliminate waveform time shifts caused by differences in propagation distance between different geophone channels, a cross-correlation time delay estimation method is used to perform waveform alignment processing on the weak seismic signals of all geophone channels. This aligns the signal waveforms induced by the same rock fracture in all geophone channels on the time axis. The cross-correlation time delay estimation method is a well-known technique and will not be described in detail here. It should be noted that the waveform alignment processing refers to using a cross-correlation algorithm to eliminate the surface propagation travel time difference caused by differences in spatial distance between the geophones, thereby aligning the start-up phase of all weak seismic signals on the time axis.

[0046] Thus, weak seismic signals from each detector channel during the fracturing operation were obtained.

[0047] Step 2: By analyzing the inconsistency of local fluctuation characteristics between weak seismic signals from any two geophone channels, the characteristic difference degree, which characterizes the degree of waveform deviation between signals, is calculated. Using this as a distance metric, weak seismic signals from all geophone channels are classified, thereby selecting the target signal set.

[0048] Furthermore, during surface monitoring of deep shale oil, the waveform and spectral characteristics of rock fracture vibration signals generated by the same fracturing event should theoretically maintain a high degree of consistency when propagating to different geophones. However, due to the complex geological structure of deep reservoirs, differences in propagation paths, and the influence of non-uniform interference, some geophone channels may receive distorted or pseudo-signals that differ significantly from the true signal characteristics, resulting in differential distortion of signals received by different geophone channels. Therefore, before extracting the characteristics of weak seismic signals, it is necessary to conduct a consistency analysis of the signals from different geophone channels, specifically:

[0049] Obtain all peaks in the weak seismic signal and calculate the full width at half maximum (FWHM) and kurtosis of each peak;

[0050] In this embodiment, the Automatic Multiscale Peak Detection (AMPD) algorithm is used to obtain the peak. The calculation of the AMPD algorithm, the full width at half maximum (FWHM), and the kurtosis are all well-known techniques and will not be described in detail here.

[0051] Define the time corresponding to each wave peak as the wave peak time, and form a feature vector by combining the peak value, wave peak time, full width at half maximum (FWHM), and kurtosis of each wave peak in the weak seismic signal;

[0052] It should be noted that the full width at half maximum (FWHM) reflects the duration of the source rupture. The larger the value, the wider the peak and the longer the signal duration. The larger the kurtosis, the sharper the peak, indicating that the energy is released in a concentrated manner in a short period of time, and that the source mechanism is simple and the rupture is decisive.

[0053] For each detector channel, the feature vectors of all peaks in the weak seismic signal are arranged in the order of peak times to form the peak feature sequence of that detector channel.

[0054] Calculate the difference between the eigenvectors at the same position in the peak characteristic sequences of any two detector channels, and use this as the relative difference;

[0055] In this embodiment, the components of the feature vectors are normalized. The specific process is as follows: all feature vectors in the weak seismic signals of all detector channels are obtained to form a feature set. Each feature vector contains four components, corresponding to the peak time, peak amplitude, full width at half maximum (FWHM), and kurtosis, respectively. For any component, the maximum and minimum values ​​of that component in the feature set are selected, and the maximum-minimum normalization method is used to normalize each component. The maximum-minimum normalization method is a well-known technique and will not be described in detail here. Next, the Euclidean distance between the normalized feature vectors at the same position in the peak feature sequences of any two detector channels is calculated as the relative difference. The Euclidean distance is a well-known technique and will not be described in detail here.

[0056] The characteristic difference degree of the weak seismic signal between any two detector channels is positively correlated with the relative difference;

[0057] It should be noted that a positive correlation means that the dependent variable increases as the independent variable increases and decreases as the independent variable decreases.

[0058] In this embodiment, the mean of all relative differences in the peak feature sequences of any two detector channels is taken as the feature difference degree; wherein, when the peak feature sequences of the two detector channels are not the same length, the shortest peak feature sequence is selected for corresponding calculation.

[0059] It should be noted that the greater the relative difference, the greater the difference in the shape of the two wave peaks; the greater the characteristic difference, the more significant the overall waveform of the weak seismic signal from the two detector channels is, and the greater the possibility that the signal is affected by interference.

[0060] Secondly, considering that genuine weak seismic signals are spatially continuous, they should theoretically be effectively received by most detector channels and exhibit similar characteristics; while spurious signals generated by local interference often only appear on a few channels. Based on this characteristic, the weak seismic signals of all detector channels are classified as follows:

[0061] Using the aforementioned feature difference as a distance metric, weak seismic signals from all detector channels are clustered to obtain multiple clusters;

[0062] In this embodiment, agglomerative hierarchical clustering algorithm is used for clustering. The agglomerative hierarchical clustering algorithm is a well-known technology and will not be described in detail here.

[0063] The number of weak seismic signals contained in each cluster is counted, and the cluster with the largest number is selected and defined as the target signal set.

[0064] It should be noted that the cluster with the largest number represents the consistent response of most detectors, and its waveform characteristics are closest to the common characteristics of real weak seismic signals. However, the signals in clusters with a smaller number may be subject to more severe interference or be pseudo-signals. Including them in subsequent analysis will reduce the accuracy of weak signal extraction and crack location accuracy. If there are multiple clusters corresponding to the largest number, the cluster with the smallest mean characteristic difference of all weak seismic signals within the cluster is selected as the target signal set.

[0065] Thus, the target signal set is obtained.

[0066] Step 3: For the target signal set, evaluate the degree of deviation of the distribution characteristics of different weak seismic signals in the frequency domain to reflect the degree of deviation from the common law of the actual fracturing response in the frequency domain, and determine the disturbance assessment value of each weak seismic signal in the target signal set.

[0067] Furthermore, due to the long signal propagation path and the passage through highly heterogeneous strata, the signals captured by the ground geophones not only contain the true microseismic features reflecting rock fractures, but also are superimposed with a large amount of operational noise and strata artifacts. While relying solely on differences in time-domain waveforms can eliminate obvious anomalous signals, it cannot finely distinguish pseudo-features in spectral details. Therefore, analyzing the differences in spectral characteristics of different weak seismic signal phases within the target signal set in the frequency domain and calculating the disturbance assessment value is crucial. The flowchart of the method for obtaining the disturbance assessment value provided in this application embodiment is shown below. Figure 2 As shown, it specifically includes:

[0068] Frequency domain analysis was performed on each weak seismic signal within the target signal set to extract various spectral features;

[0069] In this embodiment, the frequency domain analysis process is as follows: Each weak seismic signal is decomposed using a wavelet packet decomposition algorithm to obtain the reconstructed signal of each frequency band. The energy of the reconstructed signal in each frequency band is calculated, and the center frequency of the frequency band with the highest energy is selected as the dominant frequency. The sum of the energy of the reconstructed signals in all frequency bands is calculated as the total energy. The proportion of the energy of the reconstructed signal in the frequency band containing the dominant frequency in the total energy is defined as the energy concentration. The frequency range covered by continuous frequency bands with energy exceeding a preset threshold is selected as the bandwidth. The preset threshold is 5% of the total energy. Alternatively, the implementer can set this threshold according to the actual situation. The dominant frequency, bandwidth, and energy concentration are defined as spectral features. The wavelet packet decomposition algorithm is a well-known technology and will not be described further here.

[0070] It should be noted that the dominant frequency reflects the scale characteristics of rock fracturing. The lower the dominant frequency, the larger the fracturing scale and the higher the degree of development of the main fracture; the higher the dominant frequency, the smaller the fracturing scale, with micro fractures being the main feature. The bandwidth reflects the complexity of the fracture network. The wider the bandwidth, the more diverse the fracturing scale, the more complex the fracture network, and the larger the reservoir stimulation volume; the narrower the bandwidth, the more uniform the fracturing scale and the simpler the fracture morphology. The energy concentration reflects the purity and reliability of the signal. The higher the energy concentration, the more concentrated the signal energy is near the dominant frequency, the less the signal is affected by noise, and the higher the data reliability; the lower the energy concentration, the more dispersed the signal energy, which may be subject to strong interference or a more complex fracturing process.

[0071] For any two weak seismic signals in the target signal set, calculate the difference between them in the same spectral characteristics, and use it as a single feature difference.

[0072] In this embodiment, for the first Spectral characteristics, calculate the first interval between any two weak seismic signals. The absolute value of the difference between the spectral features is compared with the value of the first weak seismic signal in the target signal set. The ratio between the maximum values ​​of a spectral feature is taken as the single feature difference. For ease of understanding, it is assumed that for the spectral feature of energy concentration, the absolute value of the difference in energy concentration between any two weak seismic signals is calculated, and the ratio of this difference to the maximum energy concentration of all weak seismic signals in the target signal set is taken as the single feature difference.

[0073] It should be noted that the greater the difference in the single feature, the lower the similarity between the two weak seismic signals in that spectral feature, and the more significant the difference in the signal source or propagation path.

[0074] Using each weak seismic signal in the target signal set as a sample, and all kinds of spectral features as evaluation indicators, the weights of various spectral features are calculated using the entropy weight method.

[0075] It should be noted that the entropy weight method is a well-known technique and will not be elaborated here. The larger the weight, the more significant the difference between the spectral features and different signals, and the greater the contribution to identifying pseudo signals, reflecting the ability of each spectral feature to distinguish between real and pseudo signals.

[0076] For the target signal set, the single feature differences between each weak seismic signal and all other weak seismic signals in various spectral features are positively fused to serve as the relative deviation of each weak seismic signal in various spectral features.

[0077] In this embodiment, forward fusion is specifically represented as an additive relationship, a multiplicative relationship, etc. In this embodiment, the mean of the single feature differences between each weak seismic signal and all other weak seismic signals in various spectral features is used as the relative deviation of each weak seismic signal in various spectral features.

[0078] For the target signal set, based on the weights of various spectral features, the relative deviations of all spectral features of each weak seismic signal are weighted and summed to obtain the disturbance assessment value for each weak seismic signal.

[0079] It should be noted that the relative deviation reflects the degree of deviation of a single weak seismic signal from the mainstream signal in the target signal set at a certain spectral feature. The larger the value, the more abnormal the signal is at that spectral feature, and the greater the possibility of interference. The disturbance assessment value reflects the degree of deviation of a single weak seismic signal from the mainstream signal across all spectral features. The larger the value, the more abnormal the signal is as a whole, and it is very likely a spurious signal or severely interfered with. The greater the impact on the extraction and analysis of weak signal features generated by the fracture may be. Conversely, it indicates that the signal is reliable as a whole and can truly reflect the spectral features of the hydraulic fracturing fracture.

[0080] Thus, the disturbance assessment value of each weak seismic signal within the target signal set is obtained.

[0081] Step 4: Based on the disturbed assessment value, a reference signal is selected, and the time window from the start of the fracture of the overlying strata to the complete attenuation feature is extracted from the reference signal. Based on this, the effective signal is extracted from the weak seismic signals in the target signal set, and a deep neural network model is used to perform three-dimensional localization of the hydraulic fracture.

[0082] Furthermore, in surface monitoring of fracturing fractures in deep shale oil formations, the arrival times of elastic waves (including P-waves, S-waves, and wakes) generated by the same micro-fracture event differ across different geophone channels. If the signals from each geophone channel are individually time-windowed, the captured signal segments will not correspond to the same physical event on the time axis, thus affecting the accuracy of subsequent fracture location. Therefore, a time window that completely covers the P-wave, S-wave, and wake is extracted from the signal with the least interference, and this time window is uniformly applied to all geophone channels to ensure that the signal segments captured by each geophone channel physically correspond to the same micro-fracture event. Specifically:

[0083] The weak seismic signal corresponding to the minimum disturbed assessment value within the target signal set is defined as the reference signal;

[0084] The first arrival time of the P-wave in the reference signal is extracted using the first arrival picking algorithm; the time when the reference signal is traced back a preset time along the time axis from the first arrival time of the P-wave is defined as the starting time.

[0085] In this embodiment, the preset duration ranges from 50 to 200 milliseconds, with the preset duration set to 100 milliseconds. By reserving a buffer window of 50 to 200 milliseconds in advance, the weak oscillation precursor wave characteristic information before the first arrival of the P wave is fully preserved. It should be noted that the first arrival picking algorithm is a well-known technology and will not be described in detail here.

[0086] Calculate the instantaneous energy at each moment in the reference signal, obtain the maximum peak value among all instantaneous energies after the P-wave arrival time, and search backward from the moment corresponding to the maximum peak value for the moment when the instantaneous energy drops to a preset percentage of the maximum peak value, which is defined as the end time;

[0087] In this embodiment, the AMPD (Automatic Multiscale Peak Detection) algorithm is used to obtain the peak of the instantaneous energy at all times after the first arrival of the P wave, and the maximum peak value of all peaks is selected. The AMPD algorithm is a well-known technology and will not be described in detail here. Secondly, the preset percentage is set to 5%. As another implementation method, the implementer can set it according to the actual situation.

[0088] It should be noted that the time corresponding to 1000 milliseconds after the arrival of the P wave is defined as the maximum limiting time. If the instantaneous energy searched from the time corresponding to the maximum peak value does not drop below the preset percentage of the maximum peak value before the maximum limiting time, the end time is set as the maximum limiting time.

[0089] The time period contained in the reference signal from the start time to the end time is defined as the time window;

[0090] It should be noted that the calculation of instantaneous energy is a well-known technique and will not be described in detail here. In this embodiment, the energy envelope of the reference signal is extracted using Hilbert transform, where the energy at each moment on the energy envelope is the instantaneous energy at that moment. Hilbert transform is a well-known technique and will not be described in detail here.

[0091] The signal segment corresponding to the time window is extracted from each weak seismic signal in the target signal set, defined as a valid signal, and constitutes a valid signal set;

[0092] It should be noted that the weak seismic signals in the target signal set have already undergone waveform alignment, so the signal waveforms induced by the same rock fracture are aligned on the time axis. By extracting the effective signals, the characteristics of the corresponding weak signals generated by the fracture can be accurately reflected, thereby improving the accuracy of fracture location and reducing the impact of false signals caused by complex geological features on the accuracy of the analysis of the corresponding weak signals generated by the fracture.

[0093] Furthermore, based on the effective signal set, the fractures generated during the fracturing process in deep shale oil are located, specifically as follows:

[0094] Establish a three-dimensional formation velocity model of the target exploration area;

[0095] In this embodiment, the construction of the three-dimensional formation velocity model is a well-known technique and will not be described in detail here. Specifically, well logging data and three-dimensional seismic exploration data of the target exploration area are collected, a three-dimensional geological entity model is established using geological modeling software, the three-dimensional geological entity model is used as the formation framework, the initial velocity of the rock strata is obtained through well logging data, and the wave impedance inversion technique in the three-dimensional seismic exploration data is used to perform velocity interpolation and spatial correction for the entire area to obtain the inverted velocity field of the target exploration area; according to the spatial distribution range of the ground high-density observation array and the depth span of the target reservoir, the inverted velocity field is re-divided into a three-dimensional discrete grid, and the final three-dimensional formation velocity model is output.

[0096] Based on the effective signal set and the three-dimensional spatial coordinates of the detectors corresponding to the effective signals, as well as the three-dimensional formation velocity model, a deep neural network model is used to perform three-dimensional localization of fractures in deep shale oil.

[0097] In this embodiment, the deep neural network model adopts the UNet-Transformer hybrid network model, wherein the loss function of the UNet-Transformer hybrid network model is the mean squared error loss function, and the optimizer is the adaptive moment estimation optimizer (Adam); the UNet-Transformer hybrid network model is a well-known technology and will not be described in detail here; secondly, the training process of the UNet-Transformer hybrid network model is as follows:

[0098] Data on perforation calibration guns from historical fracturing operations were collected, and their corresponding three-dimensional coordinates were used as the true location labels. Physical forward modeling was performed based on the constructed three-dimensional formation velocity model. The coordinates of underground fracture rupture points were randomly set, weak seismic signals were synthesized, and the set fracture rupture point coordinates were used as the simulated true location labels.

[0099] After injecting background noise into the perforation calibration shot data and the synthesized weak seismic signal, the effective signal set is extracted using the above process, and combined with the corresponding geophone spatial coordinates and three-dimensional formation velocity model to form a training sample set for training the UNet-Transformer hybrid network model.

[0100] Based on the three-dimensional spatial coordinates of the fractures, the real-time distribution of the fractures within the target reservoir is determined. When the vertical coordinates of the fractures approach or exceed the set boundaries of the top and bottom plates of the target reservoir, a risk of cross-layer penetration is identified, and the pumping rate of the fracturing operation is reduced or a temporary plugging agent is injected. When the horizontal coordinate extension range of the fractures does not reach the set stimulation volume threshold, the pumping rate of the fracturing operation is increased or the sand-carrying concentration of the fracturing fluid is changed to achieve effective control of the fracturing operation, ensure effective fracture propagation within the target reservoir, and improve fracturing efficiency and reservoir activation.

[0101] It should be understood that, although Figure 1 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order in which these steps are executed, and they can be performed in other orders. Figure 1 At least some of the steps in the process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.

[0102] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0103] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application. Therefore, any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of this application, without departing from the content of the technical solution of this application, shall fall within the protection scope of the technical solution of this application.

Claims

1. A method for extracting weak signals and three-dimensionally locating microseismic signals from deep shale oil formations, characterized in that, The method includes the following steps: When performing fracturing operations in deep shale oil reservoirs in the target exploration area, weak seismic signals induced by fracturing fractures are collected through each geophone channel; By calculating the characteristic difference degree, which characterizes the degree of waveform morphology deviation between signals, based on the inconsistency of local fluctuation characteristics between weak seismic signals from any two geophone channels, and using this as a metric for distance, weak seismic signals from all geophone channels are classified, thereby selecting the target signal set. For the target signal set, the degree of deviation of the distribution characteristics of different weak seismic signals in the frequency domain is evaluated to reflect the degree of deviation from the common law of the actual fracturing response in the frequency domain, and the disturbance assessment value of each weak seismic signal in the target signal set is determined. Based on the disturbance assessment value, a reference signal is selected. The time window from the start of the fracture in the overlying strata to the complete attenuation feature is extracted from the reference signal. Based on this, the effective signal is extracted from the weak seismic signals in the target signal set, and a deep neural network model is used to perform three-dimensional localization of the hydraulic fracture.

2. The method for extracting weak signals and three-dimensionally locating microseismic signals from deep shale oil formations as described in claim 1, characterized in that, The process of obtaining the feature difference degree is as follows: After waveform alignment processing of the weak seismic signals from all detector channels, all peaks in the weak seismic signals are obtained for each detector channel, and the full width at half maximum (FWHM) and kurtosis of each peak are calculated. The time corresponding to each peak is defined as the peak time, and the peak value, peak time, FWHM, and kurtosis of each peak in the weak seismic signal are combined into a feature vector. The feature vectors of all peaks in the weak seismic signal are arranged in the order of peak times to form the peak feature sequence of the detector channel; Calculate the difference between the eigenvectors at the same position in the peak characteristic sequences of any two detector channels, and use this as the relative difference; The characteristic difference degree is positively correlated with the relative difference.

3. The method for extracting weak signals and three-dimensionally locating microseismic signals from deep shale oil formations as described in claim 1, characterized in that, The process of obtaining the target signal set is as follows: using the feature difference degree as the distance metric, the weak seismic signals of all detector channels are clustered to obtain multiple clusters; the number of weak seismic signals contained in each cluster is counted, and the cluster corresponding to the largest number is selected and defined as the target signal set.

4. The method for extracting weak signals and three-dimensionally locating microseismic signals from deep shale oil formations as described in claim 1, characterized in that, The calculation process for the disturbance assessment value is as follows: Frequency domain analysis was performed on each weak seismic signal within the target signal set to extract various spectral features; For any two weak seismic signals in the target signal set, calculate the difference between them in the same spectral characteristics, and use it as a single feature difference. Using each weak seismic signal in the target signal set as a sample, and all kinds of spectral features as evaluation indicators, the weights of various spectral features are calculated using the entropy weight method. For the target signal set, the single feature differences between each weak seismic signal and all other weak seismic signals in various spectral features are positively fused to serve as the relative deviation of each weak seismic signal in various spectral features. Based on the weights of various spectral features, the relative deviations of all spectral features of each weak seismic signal in the target signal set are weighted and summed to obtain the disturbance assessment value for each weak seismic signal in the target signal set.

5. The method for extracting weak signals and three-dimensionally locating microseismic signals from deep shale oil formations as described in claim 4, characterized in that, The various spectral characteristics include the dominant frequency, bandwidth, and energy concentration of weak seismic signals in the frequency domain.

6. The method for extracting weak signals and three-dimensionally locating microseismic signals from deep shale oil formations as described in claim 5, characterized in that, The process of obtaining the dominant frequency and energy concentration is as follows: each weak seismic signal is decomposed using the wavelet packet decomposition algorithm to obtain the reconstructed signal of each frequency band, the energy of the reconstructed signal of each frequency band is calculated, the center frequency of the frequency band with the maximum energy is selected as the dominant frequency, and the sum of the energy and value of the reconstructed signals of all frequency bands is counted as the total energy. The energy concentration is defined as the proportion of the energy of the reconstructed signal in the frequency band where the main frequency is located to the total energy.

7. The method for extracting weak signals and three-dimensionally locating microseismic signals from deep shale oil formations as described in claim 6, characterized in that, The process of obtaining the bandwidth is as follows: select the frequency range covered by a continuous frequency band whose energy exceeds a preset threshold from all reconstructed signals under each weak seismic signal, and use this range as the bandwidth.

8. The method for extracting weak signals and three-dimensionally locating microseismic signals from deep shale oil formations as described in claim 1, characterized in that, The reference signal is the weak seismic signal corresponding to the minimum disturbed assessment value within the target signal set.

9. The method for extracting weak signals and three-dimensionally locating microseismic signals from deep shale oil formations as described in claim 1, characterized in that, The extraction process of the time window is as follows: The first arrival time of the P-wave of the reference signal is extracted using the first arrival picking algorithm. The time when the reference signal is traced back a preset time along the time axis from the first arrival time of the P-wave is defined as the starting time. Calculate the instantaneous energy at each moment in the reference signal, obtain the maximum peak value among all instantaneous energies after the P-wave arrival time, and search backward from the moment corresponding to the maximum peak value for the moment when the instantaneous energy drops to a preset percentage of the maximum peak value, which is defined as the end time; The time period contained in the reference signal from the start time to the end time is defined as the time window.

10. The method for extracting weak signals and three-dimensionally locating microseismic signals from deep shale oil formations as described in claim 1, characterized in that, The method of using a deep neural network model to perform three-dimensional localization of hydraulic fracturing fractures includes: establishing a three-dimensional formation velocity model of the target exploration area; and using a UNet-Transformer hybrid network model to perform three-dimensional localization of hydraulic fracturing fractures based on the three-dimensional spatial coordinates of all effective signals and their corresponding geophones, as well as the three-dimensional formation velocity model.

Images