Earthquake monitoring data denoising method and device
Through the methods of space-time segmentation and denoising network processing, the problem of low denoising efficiency of distributed fiber microseismic monitoring data is solved, and efficient denoising and real-time processing is achieved.
Patent Information
- Application Number
- CN202311694766.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-12-11
- Publication Date
- 2025-06-13
AI Technical Summary
Due to the large amount of data, the existing denoising methods are inefficient and cannot be processed in real time.
Through space-time segmentation, the seismic monitoring data of the distributed optical fiber is divided into multiple preset data slices, and these slices are input into a pre-established denoising network for processing, and finally the denoised data volume is spliced together.
It effectively improves the denoising efficiency of distributed fiber micro-seismic monitoring data, solves the problem of low denoising efficiency caused by large data volume, and realizes real-time processing.
Smart Images
Figure CN120143259A_ABST
Abstract
Description
Technical Field
[0001] This application belongs to the technical field of seismic acquisition, and particularly relates to a method and device for denoising seismic monitoring data. Background Art
[0002] Seismic data acquisition through distributed optical fiber is a new acquisition method in the field of seismic acquisition. This acquisition method is based on the principle of fiber backscattering Rayleigh scattering and uses optical fiber as a sensor to acquire seismic signals. Compared with conventional seismic geophones, distributed optical fiber has the advantages of low cost, high density, high sensitivity, high construction efficiency, and long-term monitoring of transmission.
[0003] However, in microseismic monitoring, since the optical cable needs to be laid in the well, the data quality will be affected by the coupling conditions of the optical cable. If the laying of the optical cable does not meet the specified requirements or there are external interference sources during the acquisition process, the signal-to-noise ratio in the acquired data will be very low, and the acquired data needs to be denoised to improve the effects of subsequent imaging, inversion, etc.
[0004] The existing methods for suppressing the noise of seismic data mainly utilize the differences between the effective signal and the noise signal in the time-space domain or the transform domain to achieve the purpose of enhancing the effective signal and suppressing the noise. Specifically, the denoising methods in the transform domain generally perform denoising based on different transforms such as wavelet transform, S transform, and curvelet transform. These denoising methods need to transform the seismic signal into the transform domain by means of various mathematical transform methods, and then utilize the differences between the effective signal and the noise signal in the transform domain to separate the two.
[0005] However, the data volume of distributed optical fiber microseismic monitoring data is very large. Based on the above methods for noise suppression, there will be problems of low efficiency and inability to process in real time.
[0006] In view of the above problems, no effective solution has been proposed yet. Summary of the Invention
[0007] The purpose of this application is to provide a method and device for denoising seismic monitoring data, which can effectively improve the denoising efficiency of distributed optical fiber microseismic monitoring data.
[0008] The method and device for denoising seismic monitoring data provided by this application are implemented as follows:
[0009] A method for denoising seismic monitoring data includes:
[0010] Obtain the seismic monitoring data of the distributed optical fiber;
[0011] Through time-space segmentation, segment the seismic monitoring data of the distributed optical fiber into multiple data shards of a preset size;
[0012] Input the multiple data shards of the preset size into a pre-established denoising network to obtain multiple denoised data bodies;
[0013] Stitch the multiple denoised data bodies according to the splitting positions during spatio-temporal splitting to obtain the denoised seismic monitoring data.
[0014] In one embodiment, pre-establishing the denoising network includes:
[0015] Obtain noisy distributed optical fiber data;
[0016] Select the optical fiber data with a signal-to-noise ratio exceeding a preset threshold from the noisy distributed optical fiber data as initial data;
[0017] Perform denoising processing on the initial data through wavelet transform to obtain the actual denoised data;
[0018] Perform forward modeling on the microseismic data to obtain forward noise-free data;
[0019] Add noises with various amplitudes to the actual denoised data and the forward noise-free data to obtain sample noise data;
[0020] Record the combination of the actual denoised data and the corresponding sample noise data, and the forward noise-free data and the corresponding sample noise data as a training data set;
[0021] Train the denoising network through the training data set.
[0022] In one embodiment, the denoising network includes: a denoising generator network for denoising and a denoising discriminator network for discriminating the denoising effect. Among them, the denoising generator network includes, connected in sequence: a first convolutional layer, a first batch normalization, a first non-linear activation function, a second convolutional layer, a second batch normalization, a second non-linear activation function, a third convolutional layer, a third batch normalization, a third non-linear activation function, a residual structure composed of multiple residual blocks, a third deconvolutional layer, a second deconvolutional layer, and a first deconvolutional layer. Among them, each residual block includes: two convolutional layers, two batch normalizations, and two activation functions to increase the depth of the neural network;
[0023] The denoising discriminator network includes: a fourth convolutional layer, a fourth batch normalization, a fourth non-linear activation function, a fifth convolutional layer, a fifth batch normalization, a fifth non-linear activation function, a sixth convolutional layer, a sixth batch normalization, a sixth non-linear activation function, a seventh convolutional layer, a seventh batch normalization, and a first linear activation function.
[0024] In one embodiment, during the process of training the denoising network with the training data set, the following loss function loss is adopted All :
[0025] loss All =γ 1 loss G +γ 2 loss D +γ 3 loss F
[0026] where loss G represents the loss function of the generator network, loss D represents the loss function of the discriminator network, loss F represents the loss function for the consistency between the spectrum of the data before denoising and the spectrum of the data after denoising, γ 1 、γ 2 、γ 3 represent weight values.
[0027] In one embodiment, the loss function for spectrum consistency is calculated according to the following formula:
[0028]
[0029] where cov(f(x),Y) represents the covariance between the spectrum prediction value f(x) and the spectrum true value Y, σ f(x) represents the standard deviation of the spectrum prediction value f(x), σ Y represents the standard deviation of the spectrum true value Y.
[0030] In one embodiment, through spatio-temporal segmentation, the seismic monitoring data of the distributed optical fiber is segmented into multiple data shards of a preset size, including:
[0031] Obtain a preset spatial window and a time window, where the spatial window includes 200 sampling points and the time window includes 120 sampling points;
[0032] Pre-segment the seismic monitoring data of the distributed optical fiber through the preset spatial window and time window;
[0033] After pre-segmentation, perform data padding according to the picture size of 256×256 to obtain multiple data slices of a preset size.
[0034] A denoising device for seismic monitoring data includes:
[0035] An acquisition module, configured to acquire the seismic monitoring data of the distributed optical fiber;
[0036] A segmentation module, configured to segment the seismic monitoring data of the distributed optical fiber into multiple data shards of a preset size through spatio-temporal segmentation;
[0037] A denoising module, configured to input the multiple data shards of the preset size into a pre-established denoising network to obtain multiple denoised data bodies;
[0038] A splicing module, configured to splice the multiple denoised data bodies according to the segmentation positions during spatio-temporal segmentation to obtain the denoised seismic monitoring data.
[0039] In one embodiment, the above device further includes: a building module, configured to obtain noisy distributed optical fiber data; select the optical fiber data with a signal-to-noise ratio exceeding a preset threshold from the noisy distributed optical fiber data as initial data; perform denoising processing on the initial data through wavelet transform to obtain actual denoised data; perform forward modeling on the microseismic data to obtain forward noise-free data; add noises of multiple amplitudes to the actual denoised data and the forward noise-free data to obtain sample noise data; combine and record the actual denoised data and the corresponding sample noise data, and the forward noise-free data and the corresponding sample noise data as a training data set; train the denoising network through the training data set.
[0040] The denoising method for seismic monitoring data provided by this application, aiming at the seismic monitoring data of distributed optical fiber, segments the seismic monitoring data of distributed optical fiber into multiple data shards of a preset size through spatio-temporal segmentation; then, inputs the multiple data shards of the preset size into a pre-established denoising network to obtain multiple denoised data bodies; and splices the multiple denoised data bodies according to the segmentation positions during spatio-temporal segmentation to obtain the denoised seismic monitoring data. By the above method, the problem of low denoising efficiency caused by excessive data volume in the existing distributed optical fiber microseismic monitoring data is solved, and the technical effect of effectively improving the denoising efficiency of the distributed optical fiber microseismic monitoring data is achieved. BRIEF DESCRIPTION OF THE DRAWINGS
[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments recorded in the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.
[0042] Figure 1 It is a flowchart of a method of an embodiment of the denoising method for seismic monitoring data provided by this application;
[0043] Figure 2It is a flowchart of a method for establishing a model and denoising seismic data provided by this application;
[0044] Figure 3 It is a schematic diagram of a loss function provided by this application;
[0045] Figure 4 It is a block diagram of the hardware structure of an electronic device for a method of denoising seismic monitoring data provided by this application;
[0046] Figure 5 It is a schematic diagram of the module structure of an embodiment of a device for denoising seismic monitoring data provided by this application. Detailed implementation manners
[0047] In order to enable those skilled in the art to better understand the technical solutions in this application, the following will clearly and completely describe the technical solutions in the embodiments of this application with reference to the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of this application.
[0048] Figure 1 It is a flowchart of a method of an embodiment of a method for denoising seismic monitoring data provided by this application. Although this application provides method operation steps or device structures as shown in the following embodiments or drawings, based on routine or non-creative labor, more or fewer operation steps or module units may be included in the method or device. In steps or structures where there is no necessary causal relationship logically, the execution order of these steps or the module structure of the device is not limited to the execution order or module structure described in the embodiments of this application and shown in the drawings. When the method or module structure is applied to an actual device or terminal product, it can be executed sequentially or in parallel according to the method or module structure connection shown in the embodiments or drawings (for example, in an environment of parallel processors or multi-threaded processing, or even a distributed processing environment).
[0049] Specifically, as Figure 1 shown, the above method for denoising seismic monitoring data may include the following steps:
[0050] Step 101: Obtain seismic monitoring data of a distributed optical fiber;
[0051] That is, seismic data is collected through distributed optical fibers. The reason for denoising the collected seismic monitoring data of distributed optical fibers is mainly that in microseismic monitoring, the optical cable needs to be laid in the well, and the data quality will be affected by the coupling conditions of the optical cable. If the laying of the optical cable does not meet the specified requirements or there are external interference sources during the collection process, the signal-to-noise ratio in the collected data will be very low, and the collected data needs to be denoised to improve the effects of subsequent processes such as imaging and inversion. For this purpose, the seismic data collected by the distributed optical fiber can be obtained as the seismic data to be denoised for seismic data collection.
[0052] Step 102: Through spatio-temporal segmentation, segment the seismic monitoring data of the distributed optical fiber into multiple data shards of a preset size;
[0053] For example, a preset spatial window and a time window can be obtained. Among them, the spatial window includes 200 sampling points, and the time window includes 120 sampling points; through the preset spatial window and time window, pre-segment the seismic monitoring data of the distributed optical fiber; after pre-segmentation, perform data filling according to the image size of 256×256 to obtain multiple data slices of a preset size.
[0054] However, it should be noted that the number of sampling points in the above-listed spatial window and time window is only an exemplary description. In actual implementation, the spatial window can also choose 100 sampling points, and the time window can also choose 60 sampling points, which can be selected according to the actual data segmentation efficiency requirements.
[0055] Step 103: Input the multiple data shards of a preset size into a pre-established denoising network to obtain multiple denoised data bodies;
[0056] Step 104: According to the segmentation positions during spatio-temporal segmentation, splice the multiple denoised data bodies to obtain the denoised seismic monitoring data.
[0057] That is, the distributed optical fiber monitoring data to be processed is segmented into a size of 256×256 in space and time, input into the trained denoising network to obtain the denoised data body, and spliced into the original size according to the segmentation positions to obtain the denoised distributed optical fiber seismic monitoring data.
[0058] Furthermore, considering that in practical applications, noise-free data is usually very difficult to obtain, especially for distributed optical fiber seismic data, the amount of data with a high signal-to-noise ratio is often small. For this reason, a certain amount of data can be obtained by combining the forward modeling method to improve the generalization ability of the training network. Based on this, the above denoising network can be as follows:
[0059] S1: Obtain noisy distributed optical fiber data;
[0060] S2: Select the optical fiber data with a signal-to-noise ratio exceeding a preset threshold from the noisy distributed optical fiber data as initial data;
[0061] S3: Denoise the initial data through wavelet transform to obtain actual denoised data;
[0062] S4: Forward model the microseismic data to obtain forward noise-free data;
[0063] S5: Add noises with various amplitudes to the actual denoised data and the forward noise-free data to obtain sample noise data;
[0064] S6: Combine and record the actual denoised data and the corresponding sample noise data, and the forward noise-free data and the corresponding sample noise data as a training dataset;
[0065] S7: Train the denoising network through the training dataset.
[0066] That is, the obtained noisy distributed optical fiber data can be selected, and the part of the data with a relatively high signal-to-noise ratio can be selected for wavelet transform denoising to obtain denoised data. Then, different geological models can be designed to forward model the microseismic data to obtain noise-free data, and then noises with different amplitudes can be added to the noise-free data to obtain corresponding noisy data. In this process, different types of noises can also be added to the actual denoised data and the noise-free data obtained by forward modeling to obtain more training datasets.
[0067] The above denoising network may include: a denoising generator network for denoising and a denoising discriminator network for discriminating the denoising effect. Among them, the denoising generator network may include, connected in sequence: a first convolutional layer, a first batch normalization, a first non-linear activation function, a second convolutional layer, a second batch normalization, a second non-linear activation function, a third convolutional layer, a third batch normalization, a third non-linear activation function, a residual structure composed of multiple residual blocks, a third transposed convolutional layer, a second transposed convolutional layer, and a first transposed convolutional layer. Among them, each residual block includes: two convolutional layers, two batch normalizations, and two activation functions to increase the depth of the neural network; the denoising discriminator network may include, connected in sequence: a fourth convolutional layer, a fourth batch normalization, a fourth non-linear activation function, a fifth convolutional layer, a fifth batch normalization, a fifth non-linear activation function, a sixth convolutional layer, a sixth batch normalization, a sixth non-linear activation function, a seventh convolutional layer, a seventh batch normalization, and a first linear activation function.
[0068] The role of the above denoising discriminator network is to discriminate between the denoised data output by the denoising generator network and the data images marked as denoised in the training set. If the network is well-trained, it can effectively denoise. Among them, the input of the denoising discriminator network structure can be a picture with a size of 256×256. In this example, the denoising discriminator network replaces the fully connected layer in the existing fully convolutional neural network with a convolutional layer, and then adds a self-learning discriminant sub-network to distinguish whether the input data is noisy or denoised. Among the four convolutional layers of the discriminator network, the first three convolutional layers are composed of a convolutional layer, a batch normalization, and a non-linear activation function compressed together, while the last convolutional layer uses a linear activation function (Sigmoid). This is mainly to map the discriminant result to a probability score regularized between [0, 1]. The higher the value output by the discriminator network, the closer the input data is to the denoised data, and vice versa, the closer it is to the non-denoised data.
[0069] To improve the denoising accuracy, the adversarial loss function includes: the loss of the generator network G and the loss of the discriminator network D In addition, a loss function for spectral consistency is added F , which acts on the spectra of the data before denoising and the data after denoising. After simultaneously inputting the two into a deep convolutional neural network for feature extraction, the feature loss can be obtained. That is, during the process of training the denoising network through the training data set, the following loss function is used All :
[0070] loss All =γ 1 loss G +γ 2 loss D +γ 3 loss F
[0071] Among them, loss G represents the loss function of the generator network, loss D represents the loss function of the discriminator network, loss F represents the loss function for the spectral consistency of the spectra of the data before denoising and the data after denoising, and γ 1 , γ 2 , γ 3 represent weight values.
[0072] Among them, the loss function for spectral consistency is expressed as:
[0073]
[0074] Among them, cov(f(x), Y) represents the covariance between the spectrum prediction value f(x) and the spectrum true value Y, and σ f(x) represents the standard deviation of the spectrum prediction value f(x), and σ Y represents the standard deviation of the spectrum true value Y.
[0075] The above method will be described below in conjunction with a specific embodiment. However, it should be noted that this specific embodiment is only for better explaining the present application and does not constitute an improper limitation of the present application.
[0076] Considering the existing artificial intelligence denoising method based on microseismic deep learning, it is generally carried out based on a convolutional neural network according to the original spatio-temporal characteristics of the data. However, as the number of layers of the convolutional neural network deepens, there will not only be a problem of excessive training time, but also the microseismic small event part will be lost compared with the original data.
[0077] In this example, a denoising system is provided, which may include: a data acquisition module for acquiring the actual data and simulation data of distributed optical fiber monitoring data and constructing a network training data set; a data training module for training a denoising reinforcement learning network using the established training data set to obtain a trained denoising deep learning model; a denoising module for applying the trained denoising deep learning model to the actual monitoring data to remove noise and obtain denoised data.
[0078] In this example, an intelligent denoising method for distributed optical fiber microseismic monitoring data based on a reinforcement learning network and spectrum consistency is proposed. Through continuous gaming between a generator G (Generator) and a discriminator D (Discriminator), the generator can learn the characteristics of noise data and output a denoising result, thereby enhancing the effective events in the data and retaining the information of small magnitude events. Further, considering that there are obvious spectral distribution differences between the noise in the distributed optical fiber microseismic monitoring data and the microseismic data in terms of spectrum, therefore, in this example, a spectrum consistency loss function is introduced to highlight the spectral differences between the noise and the effective data, thereby further enhancing the denoising effect.
[0079] Specifically, it may be as Figure 2 shown and includes the following steps:
[0080] S1: Establish a training data set:
[0081] Since in practical applications, noise-free data is usually difficult to obtain, especially for distributed optical fiber seismic data, the amount of data with high signal-to-noise ratio is often small. Therefore, a certain amount of data can be obtained by combining forward modeling to improve the generalization ability of the training network.
[0082] Specifically, the obtained noisy distributed optical fiber data can be selected, and part of the data with a relatively high signal-to-noise ratio can be selected for wavelet transform denoising processing to obtain the denoised data. Then, different geological models can be designed to perform forward simulation of microseismic data to obtain noise-free data, and then noises with different amplitudes can be added to the noise-free data to obtain corresponding noisy data. In this process, different types of noises can also be added to the actual denoised data and the noise-free data obtained by forward simulation to obtain more training data sets.
[0083] S2: Establish a denoising generator network:
[0084] The purpose of establishing the denoising generation network is to denoise the noisy distributed optical fiber monitoring data. The input is the noisy data, and the output is the denoised data. Both the input data and the output data of this denoising generation network can be regularized to 256×256. Specifically, the network structure adopted in this example is different from that of a general convolutional neural network. The network structure adopted in this example includes: three convolutional layers, batch normalization, and a non-linear activation function compressed together to form a convolutional module. Then, a residual structure is connected. Each residual block in the residual structure contains two convolutional layers, two batch normalizations, and two activation functions to increase the depth of the neural network. Then, three deconvolutional layers are connected. Each deconvolutional layer corresponds to the convolutional layer at the front of the network, and the size of the image is adjusted from 64×64 to 128×128, and the final image output size is 256×256.
[0085] S3: Establish a denoising discriminator network:
[0086] The role of the denoising discriminator network is to discriminate between the noisy data (the denoised data output by the denoising generator network) and the data images marked as denoised in the training set. If the network is well-trained, it can effectively denoise. Among them, the input of the denoising discriminator network structure is an image with a size of 256×256. The denoising discriminator network replaces the fully connected layer in the existing fully convolutional neural network with a convolutional layer, and then adds a self-learning discriminant sub-network to distinguish whether the input data is noisy or denoised. The discriminant network can include: four convolutional layers. Among them, the first three convolutional layers are composed of a convolutional layer, a batch normalization, and a non-linear activation function compressed together, while the last convolutional layer uses a linear activation function (Sigmoid) to map the discriminant result to a probability score regularized between [0, 1]. The higher the output value, the closer the input data is to the denoised data, and vice versa, the closer it is to the non-denoised data.
[0087] S4: Establish the loss function of the reinforcement learning network:
[0088] In the loss function of the original reinforcement learning network, the adversarial loss function includes the loss of the generator network G and the loss of the discriminator network D in two parts. In this example, as Figure 3 shown, a loss function for spectral consistency is added F , and by applying this loss function to the spectra of the data before denoising and the data after denoising, and then inputting both into a deep convolutional neural network for feature extraction, the feature loss can be obtained. Based on this, the loss function of the network in this example All can be expressed as:
[0089] loss All = γ 1 loss G + γ 2 loss D + γ 3 loss F
[0090] where γ 1 , γ 2 , γ 3 respectively represent the corresponding weights of each loss function and can be set according to the training situation.
[0091] Among them, the loss function for spectral consistency is expressed as:
[0092]
[0093] where cov(f(x),Y) represents the covariance between the spectral prediction value f(x) and the spectral true value Y, σ f(x) represents the standard deviation of the spectral prediction value f(x), and σ Y represents the standard deviation of the spectral true value Y.
[0094] S5: Train the established reinforcement learning network:
[0095] For example, during the training process, a total of 3000 data, including both the actually collected data and the simulated generated data, each data includes: the data before denoising and the corresponding data after denoising. The size of each data body is 1200 channels in space and 2000 time sampling points, and all are spatially and temporally segmented into a size of 256×256. This data before and after denoising is used as the training dataset of the reinforcement learning network. During the training process, the learning rate can be set to 0.00001, the training batch is set to 5, and the total number of iterations is 25k.
[0096] S6: Apply the trained reinforcement learning network model for denoising processing:
[0097] The distributed optical fiber monitoring data to be processed is spatio-temporally segmented into a size of 256×256, and input into the obtained denoising generator network to obtain the denoised data volume. According to the segmented positions, it is stitched into the original size, and the denoised distributed optical fiber seismic monitoring data can be obtained.
[0098] In the above example, aiming at the requirements of particularly large data volume and real-time processing of distributed optical fiber microseismic monitoring data, a deep learning method based on a reinforcement learning network is proposed to achieve fast denoising of distributed optical fiber microseismic monitoring data. Based on the original loss function of the reinforcement learning network, considering the obvious difference in spectrum between noise and effective signals in distributed optical fiber microseismic monitoring data, a spectrum consistency feature extraction loss function is introduced to further improve the network performance. Specifically, aiming at the problems of particularly large data volume, real-time processing required, and low processing efficiency of existing methods for distributed optical fiber microseismic monitoring data, in this example, an artificial intelligence method is used to train a large amount of data through various different network structures, so as to automatically and deeply extract the features contained in the data to effectively represent the effective signals in seismic data. By adding the observed data to the training set and pre-training the model, the requirement of real-time processing can be met.
[0099] The method embodiments provided in the above embodiments of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Taking running on an electronic device as an example, Figure 4 is a hardware structure block diagram of an electronic device for a method of denoising seismic monitoring data provided by the present application. As Figure 4 shown, the electronic device 10 may include one or more (only one is shown in the figure) processors 02 (the processor 02 may include, but is not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 04 for storing data, and a transmission module 06 for communication functions. Those of ordinary skill in the art can understand that Figure 4 the structure shown is only schematic and does not limit the structure of the above electronic device. For example, the electronic device 10 may further include more or fewer components than those Figure 4 shown, or have a different configuration from that Figure 4 shown.
[0100] The memory 04 can be used to store software programs and modules of application software, such as the program instructions / modules corresponding to the denoising method of seismic monitoring data in the embodiments of the present application. The processor 02 executes various functional applications and data processing by running the software programs and modules stored in the memory 04, that is, implements the denoising method of seismic monitoring data of the above application program. The memory 04 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memories, or other non-volatile solid-state memories. In some instances, the memory 04 may further include a memory remotely provided with respect to the processor 02, and these remote memories can be connected to the electronic device 10 through a network. Examples of the above network include but are not limited to the Internet, intranet, local area network, mobile communication network, and combinations thereof.
[0101] The transmission module 06 is used to receive or send data via a network. Specific examples of the above network may include a wireless network provided by a communication provider of the electronic device 10. In one instance, the transmission module 06 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station and thus communicate with the Internet. In one instance, the transmission module 06 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.
[0102] At the software level, the above denoising device for seismic monitoring data can be as Figure 5 shown, including:
[0103] An acquisition module 501, configured to acquire seismic monitoring data of a distributed optical fiber;
[0104] A segmentation module 502, configured to segment the seismic monitoring data of the distributed optical fiber into multiple data shards of a preset size through spatio-temporal segmentation;
[0105] A denoising module 503, configured to input the multiple data shards of the preset size into a pre-established denoising network to obtain multiple denoised data bodies;
[0106] A splicing module 504, configured to splice the multiple denoised data bodies according to the segmentation positions during spatio-temporal segmentation to obtain denoised seismic monitoring data.
[0107] In one embodiment, the denoising device for the above seismic monitoring data may further include: a building module, configured to obtain noisy distributed optical fiber data; select, from the noisy distributed optical fiber data, the optical fiber data with a signal-to-noise ratio exceeding a preset threshold as initial data; perform denoising processing on the initial data through wavelet transform to obtain actual denoised data; perform forward modeling on the microseismic data to obtain forward noise-free data; add noises with various amplitudes to the actual denoised data and the forward noise-free data to obtain sample noise data; combine and record the actual denoised data and the corresponding sample noise data, and the forward noise-free data and the corresponding sample noise data as a training data set; and train the denoising network through the training data set.
[0108] In one embodiment, the above denoising network may include: a denoising generator network for denoising and a denoising discriminator network for discriminating the denoising effect. Among them, the denoising generator network includes, connected in sequence: a first convolutional layer, a first batch normalization, a first non-linear activation function, a second convolutional layer, a second batch normalization, a second non-linear activation function, a third convolutional layer, a third batch normalization, a third non-linear activation function, a residual structure composed of multiple residual blocks, a third transposed convolutional layer, a second transposed convolutional layer, and a first transposed convolutional layer. Each residual block includes: two convolutional layers, two batch normalizations, and two activation functions to increase the depth of the neural network.
[0109] The denoising discriminator network may include, connected in sequence: a fourth convolutional layer, a fourth batch normalization, a fourth non-linear activation function, a fifth convolutional layer, a fifth batch normalization, a fifth non-linear activation function, a sixth convolutional layer, a sixth batch normalization, a sixth non-linear activation function, a seventh convolutional layer, a seventh batch normalization, and a first linear activation function.
[0110] In one embodiment, during the process of training the denoising network through the training data set, the following loss function loss may be adopted All :
[0111] loss All =γ 1 loss G +γ 2 loss D +γ 3 loss F
[0112] Among them, loss G represents the loss function of the generator network, loss D represents the loss function of the discriminator network, loss FA loss function representing the spectral consistency between the data before denoising and the data after denoising, γ 1 、γ 2 、γ 3 represent weight values.
[0113] Among them, the loss function of spectral consistency can be calculated according to the following formula:
[0114]
[0115] Among them, cov(f(x), Y) represents the covariance between the spectral prediction value f(x) and the spectral true value Y, and σ f(x) represents the standard deviation of the spectral prediction value f(x), and σ Y represents the standard deviation of the spectral true value Y.
[0116] In one embodiment, the above-mentioned segmentation module 502 can specifically obtain a preset spatial window and a time window, where the spatial window includes 200 sampling points and the time window includes 120 sampling points; through the preset spatial window and time window, pre-segment the seismic monitoring data of the distributed optical fiber; after pre-segmentation, perform data filling according to the picture size of 256×256 to obtain multiple data slices of preset size.
[0117] The embodiments of the present application also provide a specific implementation manner of an electronic device that can implement all the steps in the above-mentioned seismic monitoring data denoising method. The electronic device specifically includes the following contents: a processor, a memory, a communication interface, and a bus; among them, the processor, the memory, and the communication interface complete mutual communication through the bus; the processor is used to call the computer program in the memory, and when the processor executes the computer program, it implements all the steps in the above-mentioned seismic monitoring data denoising method. For example, when the processor executes the computer program, it implements the following steps:
[0118] Step 1: Obtain the seismic monitoring data of the distributed optical fiber;
[0119] Step 2: Through spatio-temporal segmentation, segment the seismic monitoring data of the distributed optical fiber into multiple data slices of preset size;
[0120] Step 3: Input the multiple data slices of preset size into a pre-established denoising network to obtain multiple denoised data bodies;
[0121] Step 4: Splice the multiple denoised data bodies according to the segmentation positions during spatio-temporal segmentation to obtain the denoised seismic monitoring data.
[0122] An embodiment of the present application further provides a computer-readable storage medium capable of implementing all steps in the above-described seismic monitoring data denoising method. A computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, all steps of the seismic monitoring data denoising method in the above embodiment are implemented. For example, when the processor executes the computer program, the following steps are implemented:
[0123] Step 1: Obtain seismic monitoring data of a distributed optical fiber;
[0124] Step 2: Through spatio-temporal segmentation, segment the seismic monitoring data of the distributed optical fiber into multiple data shards of a preset size;
[0125] Step 3: Input the multiple data shards of the preset size into a pre-established denoising network to obtain multiple denoised data bodies;
[0126] Step 4: According to the segmentation positions during spatio-temporal segmentation, splice the multiple denoised data bodies to obtain denoised seismic monitoring data.
[0127] As can be seen from the above description, in the embodiment of the present application, for the seismic monitoring data of a distributed optical fiber, through spatio-temporal segmentation, the seismic monitoring data of the distributed optical fiber is segmented into multiple data shards of a preset size; then, the multiple data shards of the preset size are input into a pre-established denoising network to obtain multiple denoised data bodies; according to the segmentation positions during spatio-temporal segmentation, the multiple denoised data bodies are spliced to obtain denoised seismic monitoring data. By the above method, the problem of low denoising efficiency caused by excessive data volume in the existing distributed optical fiber microseismic monitoring data is solved, and the technical effect of effectively improving the denoising efficiency of the distributed optical fiber microseismic monitoring data is achieved.
[0128] Each embodiment in this specification is described in a progressive manner. The same or similar parts among the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the hardware + program type embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.
[0129] The above describes specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in a different order than in the embodiments and still achieve the desired result. Additionally, the processes depicted in the figures do not necessarily require the particular order or sequential order shown to achieve the desired result. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0130] Although the present application provides method operation steps as described in the embodiments or flowcharts, more or fewer operation steps may be included based on routine or non-creative labor. The order of steps listed in the embodiments is only one way among the execution orders of numerous steps and does not represent the only execution order. When the actual device or client product is executed, it may be executed in the method order shown in the embodiments or the drawings or in parallel (such as in an environment of parallel processors or multi-threaded processing).
[0131] Although the embodiments of the present specification provide method operation steps as described in the embodiments or flowcharts, more or fewer operation steps may be included based on routine or non-creative means. The order of steps listed in the embodiments is only one way among the execution orders of numerous steps and does not represent the only execution order. When the actual device or terminal product is executed, it may be executed in the method order shown in the embodiments or the drawings or in parallel (such as in an environment of parallel processors or multi-threaded processing, or even in a distributed data processing environment). The term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, product or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, product or device. Without further limitation, there is no exclusion of additional identical or equivalent elements in the process, method, product or device comprising the said elements.
[0132] For convenience of description, the above device is described by dividing it into various modules according to functions. Of course, when implementing the embodiments of the present specification, the functions of each module may be implemented in the same or multiple software and / or hardware, or the modules implementing the same function may be realized by a combination of multiple sub-modules or sub-units, etc. The device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection to each other may be through some interfaces, and the indirect coupling or communication connection of the device or unit may be in electrical, mechanical or other forms.
[0133] Those skilled in the art also know that, in addition to implementing the controller in the form of pure computer-readable program code, the method steps can be logically programmed to enable the controller to be implemented in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, embedded microcontrollers, etc. to achieve the same functions. Therefore, such a controller can be considered a hardware component, and the devices included therein for implementing various functions can also be regarded as the structures within the hardware component. Or even, the devices for implementing various functions can be regarded as either software modules for implementing the method or structures within the hardware component.
[0134] This application is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.
[0135] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that implement the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.
[0136] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.
[0137] In a typical configuration, a computing device includes one or more processors (CPUs), an input / output interface, a network interface, and a memory.
[0138] The memory may include non-permanent memory in the form of computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. The memory is an example of computer-readable media.
[0139] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile discs (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices, or any other non-transitory media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory media such as modulated data signals and carrier waves.
[0140] Those skilled in the art will appreciate that the embodiments of this specification can be provided as a method, system, or computer program product. Accordingly, the embodiments of this specification can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the embodiments of this specification can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.
[0141] Each embodiment in this specification is described in a progressive manner. For the identical or similar parts among the embodiments, reference can be made to each other, and the differences between each embodiment and other embodiments are emphasized. In particular, for system embodiments, since they are basically similar to method embodiments, the description is relatively simple, and reference can be made to the relevant parts of the method embodiments for the relevant content. In the description of this specification, the description with reference to terms such as "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the embodiments of this specification. In this specification, the schematic description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.
[0142] The above is only the embodiment of the embodiments of this specification and is not used to limit the embodiments of this specification. For those skilled in the art, various changes and modifications can be made to the embodiments of this specification. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the embodiments of this specification shall be included within the scope of the claims of the embodiments of this specification.
Claims
1. A method for denoising seismic monitoring data, characterized in that, it includes: Obtain seismic monitoring data of distributed optical fiber; Through spatio-temporal segmentation, segment the seismic monitoring data of the distributed optical fiber into multiple data shards of a preset size; Input the multiple data shards of the preset size into a pre-established denoising network to obtain multiple denoised data bodies; According to the segmentation positions during spatio-temporal segmentation, splice the multiple denoised data bodies to obtain denoised seismic monitoring data.
2. The method according to claim 1, characterized in that, Pre-establishing the denoising network includes: Obtain noisy distributed optical fiber data; From the noisy distributed optical fiber data, select the optical fiber data with a signal-to-noise ratio exceeding a preset threshold as the initial data; Perform denoising processing on the initial data through wavelet transform to obtain actual denoised data; Perform forward modeling on microseismic data to obtain forward noise-free data; Add various amplitudes of noise to the actual denoised data and the forward noise-free data to obtain sample noise data; Combine and record the actual denoised data and the corresponding sample noise data, and the forward noise-free data and the corresponding sample noise data as the training data set; Train the denoising network through the training data set.
3. The method according to claim 2, characterized in that, The denoising network includes: a denoising generator network for denoising and a denoising discriminator network for discriminating the denoising effect. Among them, the denoising generator network includes, connected in sequence: a first convolutional layer, a first batch normalization, a first non-linear activation function, a second convolutional layer, a second batch normalization, a second non-linear activation function, a third convolutional layer, a third batch normalization, a third non-linear activation function, a residual structure composed of multiple residual blocks, a third deconvolutional layer, a second deconvolutional layer, and a first deconvolutional layer. Among them, each residual block includes: two convolutional layers, two batch normalizations, and two activation functions to increase the depth of the neural network; The denoising discriminator network includes, connected in sequence: a fourth convolutional layer, a fourth batch normalization, a fourth non-linear activation function, a fifth convolutional layer, a fifth batch normalization, a fifth non-linear activation function, a sixth convolutional layer, a sixth batch normalization, a sixth non-linear activation function, a seventh convolutional layer, a seventh batch normalization, and a first linear activation function.
4. The method according to claim 2, characterized in that, During the process of training the denoising network using the training data set, the following loss function loss is adopted All : loss All = γ 1 loss G + γ 2 loss D + γ 3 loss F Among them, loss G represents the loss function of the generator network, loss D represents the loss function of the discriminator network, loss F represents the loss function for the spectrum consistency between the data before denoising and the data after denoising, γ 1 γ 2 γ 3 represent weight values.
5. The method according to claim 4, characterized in that, Calculate the loss function of spectral consistency according to the following formula: Among them, cov(f(x), Y) represents the covariance between the spectral prediction value f(x) and the true spectral value Y, and σ f(x) represents the standard deviation of the spectral prediction value f(x), and σ Y represents the standard deviation of the true spectral value Y.
6. The method according to claim 1, characterized in that, Through spatio-temporal segmentation, segment the seismic monitoring data of the distributed optical fiber into multiple data shards of a preset size, including: Obtain a preset spatial window and a time window, where the spatial window includes 200 sampling points and the time window includes 120 sampling points; Perform pre-segmentation on the seismic monitoring data of the distributed optical fiber through the preset spatial window and time window; After pre-segmentation, data filling is performed according to the image size of 256×256 to obtain multiple data slices of a preset size.
7. A denoising device for seismic monitoring data, characterized in that it includes: an acquisition module for acquiring seismic monitoring data of a distributed optical fiber; a segmentation module for segmenting the seismic monitoring data of the distributed optical fiber into multiple data shards of a preset size through spatio-temporal segmentation; a denoising module for inputting the multiple data shards of a preset size into a pre-established denoising network to obtain multiple denoised data bodies; a splicing module for splicing the multiple denoised data bodies according to the segmentation positions during spatio-temporal segmentation to obtain denoised seismic monitoring data.
8. The device according to claim 7, characterized in that it further includes: a building module for acquiring noisy distributed optical fiber data; selecting the optical fiber data with a signal-to-noise ratio exceeding a preset threshold from the noisy distributed optical fiber data as initial data; performing denoising processing on the initial data through wavelet transform to obtain actual denoised data; performing forward modeling on microseismic data to obtain forward noise-free data; adding various amplitudes of noise to the actual denoised data and the forward noise-free data to obtain sample noise data; combining and recording the actual denoised data and the corresponding sample noise data, and the forward noise-free data and the corresponding sample noise data as a training data set; training the denoising network through the training data set.
9. An electronic device, including a processor and a memory for storing processor-executable instructions, characterized in that when the processor executes the instructions, the steps of the method according to any one of claims 1 to 6 are implemented.
10. A computer-readable storage medium, on which a computer program / instructions are stored, characterized in that when the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.