Method for seismic data denoising, control device and storage medium

Abstract

This invention provides a method, control device, and storage medium for seismic data denoising, belonging to the field of oil and gas exploration technology. The method includes: acquiring actual seismic data of a preset area; and performing noise suppression processing on the actual seismic data using a dual-channel five-dimensional denoising model to obtain a denoised seismic data volume. The dual-channel five-dimensional denoising model is obtained by training a dual-channel five-dimensional denoising neural network, which includes a Fourier transform layer, a five-dimensional denoising neural network layer, and an inverse Fourier transform layer. The output channels of the Fourier transform layer are dual-channel, and the input channels of the inverse Fourier transform layer are dual-channel. By fully utilizing the high-dimensional information of the "height, width, and height" data, and based on five-dimensional space (three spatial dimensions, offset, and azimuth) and the real and imaginary parts of the frequency, the dual-channel five-dimensional denoising neural network is trained to learn the characteristics of noise samples, thereby improving the noise recognition accuracy of the dual-channel five-dimensional denoising model.

Description

Technical Field

[0001] This invention relates to the field of oil and gas exploration technology, and more specifically to a method, control device, and storage medium for denoising seismic data. Background Technology

[0002] In recent years, oil and gas exploration has increasingly focused on complex geological structures, stratigraphic lithology, and carbonate rocks. The application of "broad azimuth, broad frequency band, and high density" exploration technologies has grown rapidly. However, this has led to a situation where the volume of seismic data has increased exponentially, while data quality (e.g., signal-to-noise ratio) has stagnated. While massive amounts of seismic data contain richer subsurface information, they also present unprecedented challenges to processing technologies. Influenced by acquisition methods and complex subsurface structures, seismic noise exhibits diverse and complex characteristics. This noise severely affects the imaging quality of seismic data, thereby misleading reservoir prediction and well location deployment.

[0003] Based on noise morphology, seismic noise can be broadly classified into random noise and coherent noise. For random noise, existing noise suppression algorithms include fx-domain predictive filtering, wavelet transform, and curvelet transform; for coherent noise such as surface waves and linear interference, existing algorithms include FK filtering, FKK domain filtering, anomalous amplitude attenuation, KL transform, and tilt filtering. Existing noise suppression algorithms abstract information such as the propagation law, physical characteristics, and morphological features of noise into mathematical language, and complete the separation of signal and noise in the data space (e.g., frequency domain, curvelet domain, etc.). Although this type of noise suppression method based on mathematical models has a certain noise suppression effect, it has drawbacks for the widely used "two-wide-one-high" data, including: 1. In the process of mathematical abstraction, theoretical assumptions are usually required; although the conditions of theoretical assumptions can characterize certain features of seismic data to a certain extent, they cannot be accurately met in actual data processing, especially for low signal-to-noise ratio data, thus causing limitations in denoising algorithms. 2. Conventional denoising algorithms are usually designed for a specific type of noise. In actual production, multiple denoising algorithms are needed to complete the task. Each denoising algorithm requires parameter tuning. The noise suppression process is highly subjective. When processing massive amounts of data, such as seismic data with "two widths and one height", the efficiency is relatively low. Summary of the Invention

[0004] The purpose of this invention is to provide a method for denoising seismic data, which addresses the problems of low noise identification accuracy, poor adaptability, and inability to fully utilize "wide and high" seismic data information in existing noise suppression technologies.

[0005] To achieve the above objectives, embodiments of the present invention provide a method for denoising seismic data. The method includes: acquiring actual seismic data of a preset area; and performing noise suppression processing on the actual seismic data using a dual-channel five-dimensional denoising model to obtain a denoised seismic data volume. The dual-channel five-dimensional denoising model is obtained by training a dual-channel five-dimensional denoising neural network. The dual-channel five-dimensional denoising neural network includes a Fourier transform layer, a five-dimensional denoising neural network layer, and an inverse Fourier transform layer. The output channel of the Fourier transform layer is dual-channel, and the input channel of the inverse Fourier transform layer is dual-channel.

[0006] Optionally, training the dual-channel five-dimensional denoising neural network includes: constructing a five-dimensional dataset based on the spatial three-dimensional, offset, and azimuth five-dimensional attributes of seismic data using sample seismic data and / or simulated seismic data, and dividing the five-dimensional dataset into a five-dimensional training set and a five-dimensional validation set; and constructing the dual-channel five-dimensional denoising neural network, training the dual-channel five-dimensional denoising neural network using the five-dimensional training set, and validating the dual-channel five-dimensional denoising neural network using the five-dimensional validation set, to obtain a trained dual-channel five-dimensional denoising model.

[0007] Optionally, the data used to construct the five-dimensional dataset from the sample seismic data includes: denoising the sample seismic data using a preset seismic noise suppression strategy, using the denoised sample seismic data as label signals, using the sample seismic data as input data, and constructing the five-dimensional dataset using the input data and the label data.

[0008] Optionally, the data used to construct the five-dimensional dataset from the simulated seismic data includes: calculating the arrival time of the seismic signal from the shot point to the receiver point and the P-wave reflection coefficient based on the simulated seismic data to generate a simulated seismic signal; generating simulated noise data by simulating random noise, using the simulated seismic signal as a tag signal, and using the simulated seismic signal and the simulated random noise as input data, and constructing the five-dimensional dataset using the input data and the tag data.

[0009] Optionally, the preset seismic noise suppression strategy includes one or a combination of abnormal amplitude suppression, FKK domain filtering, and tilt angle filtering.

[0010] Optionally, the calculation of the arrival time of the seismic signal from the shot point to the receiver point includes: acquiring the coordinates of the shot point, the coordinates of the receiver point, the P-wave velocity, and the azimuth angle in the simulated seismic data; and calculating the arrival time of the seismic signal based on a given time using the acquired coordinates of the shot point, the coordinates of the receiver point, the P-wave velocity, and the azimuth angle.

[0011] Optionally, generating a simulated seismic signal based on the arrival time of the seismic signal and the P-wave reflection coefficient includes: calculating the reflection coefficient r(t) of the stratum at the corresponding time point of the seismic signal arrival time t using the following formula:

[0012]

[0013] Where a(t) is the amplitude fitting vector; e(t) is the dynamically adjusted parameter, R PP θ is the P-wave reflection coefficient; θ is the incident angle of the seismic P-wave. This refers to the azimuth of the seismic P-wave.

[0014] The simulated seismic signal is calculated by the convolution of the reflection coefficient r(t) and the seismic wavelet.

[0015] Optionally, generating simulated noise data based on simulated random noise includes: performing low-pass filtering on Gaussian noise data generated from simulated random noise using a Ricker wavelet with a preset main frequency to generate the simulated noise data.

[0016] Optionally, the five-dimensional denoising neural network layer includes multiple five-dimensional convolutional layers, and the construction of each five-dimensional convolutional layer includes: defining the five-dimensional convolutional layer based on a given five-dimensional array f and g:

[0017]

[0018] Where j1, j2, j3, j4, j5 represent the elements of array f, and i1-j1, i2-j2, i3-j3, i4-j4, i5-j5 represent the elements of array g;

[0019] The defined five-dimensional convolutional layer is deformed as follows:

[0020]

[0021] Optionally, the five-dimensional denoising neural network layer includes seven five-dimensional convolutional layers. The first layer of the five-dimensional convolutional layer includes a set of five-dimensional convolutions with 2 input channels, 64 output channels, and a filter size of 3. The second to sixth layers each include a set of five-dimensional convolutions with 64 input channels, 64 output channels, and a filter size of 3, and a ReLU activation function. The seventh layer includes a set of five-dimensional convolutional layers with 64 input channels, 2 output channels, and a filter size of 3.

[0022] Optionally, the loss function of the dual-channel five-dimensional denoising neural network can be represented by the following formula:

[0023]

[0024] Where K is the number of samples in the five-dimensional training set, {x k ,y k} represents a set of training sample data, x k To simulate noisy data, y k For label data, Θ represents network parameters; ||·|| F It is the Frobenius norm.

[0025] This invention also provides a control device for seismic data denoising. The control device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the computer program to implement the above-described method for seismic data denoising.

[0026] This invention also provides a machine-readable storage medium storing instructions that cause a machine to perform the above-described method for denoising seismic data.

[0027] Through the above technical solution, this embodiment of the invention uses a dual-channel five-dimensional denoising model to suppress noise in actual seismic data, fully utilizing the high-dimensional information of the seismic data and effectively improving the noise recognition accuracy. By constructing a dual-channel five-dimensional denoising neural network, it fully utilizes the high-dimensional information of the "height, width, and height" data. Based on five-dimensional space (three spatial dimensions, offset, and azimuth) and the real and imaginary frequency channels, it trains and learns the features of noise samples in the five-dimensional training set to improve the noise recognition accuracy of the dual-channel five-dimensional denoising model. This provides high-quality basic data for subsequent seismic imaging and reservoir prediction, contributing to improved efficiency and effectiveness in oil and gas exploration and processing.

[0028] Other features and advantages of the embodiments of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0029] The accompanying drawings are provided to further illustrate embodiments of the present invention and form part of the specification. They are used together with the following detailed description to explain the embodiments of the present invention, but do not constitute a limitation thereof. In the drawings:

[0030] Figure 1 This is a schematic flowchart of a method for denoising seismic data provided in an embodiment of the present invention;

[0031] Figure 2 This is a schematic diagram of an example dual-channel five-dimensional denoising neural network architecture;

[0032] Figure 3 This is a schematic diagram illustrating the composition of an example five-dimensional dataset;

[0033] Figure 4This is a schematic diagram of the noisy data from a single shot in Example 1;

[0034] Figure 5 This is a schematic diagram of an effective seismic signal in Example 1;

[0035] Figure 6 This is a schematic diagram of the denoising result based on a two-dimensional convolutional neural network in Example 1;

[0036] Figure 7 This is a schematic diagram of the denoising result based on an embodiment of the present invention;

[0037] Figure 8 This is a schematic diagram of Example 1, which uses a two-dimensional convolutional neural network to identify noise.

[0038] Figure 9 This is a schematic diagram illustrating noise identification based on an embodiment of the present invention;

[0039] Figure 10 Example 2: Schematic diagram of the noisy portion of the data for a single shot;

[0040] Figure 11 Example 2: A schematic diagram of some valid seismic signals;

[0041] Figure 12 This is a partial schematic diagram of the denoising result based on a two-dimensional convolutional neural network in Example 2;

[0042] Figure 13 This is a partial schematic diagram of the denoising result based on an embodiment of the present invention in Example 2;

[0043] Figure 14 This is a partial schematic diagram of Example 2, which uses a two-dimensional convolutional neural network to identify noise; and

[0044] Figure 15 This is a partial schematic diagram of noise identification based on an embodiment of the present invention, as shown in Example 2. Detailed Implementation

[0045] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the scope of the present invention.

[0046] Figure 1 This is a schematic flowchart of a method for denoising seismic data provided in an embodiment of the present invention. Please refer to it. Figure 1 The method for denoising seismic data may include the following steps:

[0047] Step S110: Obtain actual seismic data for the preset area.

[0048] Step S120: The actual seismic data is subjected to noise suppression processing using a dual-channel five-dimensional denoising model to obtain the denoised seismic data volume.

[0049] The dual-channel five-dimensional denoising model is obtained by training a dual-channel five-dimensional denoising neural network, which includes a Fourier transform layer, a five-dimensional denoising neural network layer, and an inverse Fourier transform layer. The output channel of the Fourier transform layer is dual-channel, and the input channel of the inverse Fourier transform layer is dual-channel.

[0050] For seismic data with "two widths and one height" (i.e., "wide azimuth, wide bandwidth, and high density") characteristics, a five-dimensional data volume can be formed according to spatial three dimensions, offset, and azimuth to extract rich subsurface information. However, conventional seismic data denoising algorithms are usually developed from traditional two-dimensional or three-dimensional data denoising processing, which makes it difficult to fully utilize the information in the five dimensions of the data, resulting in a waste of seismic data information and a reduction in the accuracy of seismic denoising.

[0051] Therefore, this embodiment of the invention constructs a five-dimensional dataset and trains a dual-channel five-dimensional denoising neural network to obtain a dual-channel five-dimensional denoising model, which is used to suppress noise in actual seismic data, making full use of the high-dimensional information of seismic data and effectively improving the accuracy of noise recognition.

[0052] Preferably, training the dual-channel five-dimensional denoising neural network may include steps S121-S122.

[0053] Step S121: Based on the spatial three-dimensional, offset, and azimuth five-dimensional attributes of seismic data, construct a five-dimensional dataset using sample seismic data and / or simulated seismic data, and divide the five-dimensional dataset into a five-dimensional training set and a five-dimensional validation set.

[0054] Please refer to Figure 3 A five-dimensional dataset (e.g., denoted as N1×N2×N3×N4×N5) can be constructed from sample seismic data and / or simulated seismic data. The data (or elements) of the five-dimensional dataset may include label data and input data. Embodiments of this invention provide three methods for obtaining five-dimensional dataset data.

[0055] I. The data used to construct the five-dimensional dataset is composed of the sample seismic data.

[0056] Preferably, the data used to construct the five-dimensional dataset from the sample seismic data may include: denoising the sample seismic data using a preset seismic noise suppression strategy, using the denoised sample seismic data as label signals, using the sample seismic data as input data, and constructing the five-dimensional dataset using the input data and the label data.

[0057] The preferred earthquake noise suppression strategy in the embodiments of the present invention includes one or a combination of abnormal amplitude suppression, FKK domain filtering, and tilt angle filtering.

[0058] For example, spatial three-dimensional, offset, and azimuth information is obtained from sample seismic data. After denoising the sample seismic data using the aforementioned noise suppression techniques, the sample seismic data is arranged in the order of "shot line-shot point-receiver line-receiver point" to form a five-dimensional dataset. The denoised five-dimensional dataset can be used as (pseudo)label data, with the original sample seismic data as input data. The five-dimensional dataset is constructed using the input data and the label data.

[0059] II. The data used to construct the five-dimensional dataset is derived from the simulated earthquake data.

[0060] Preferably, the data used to construct the five-dimensional dataset from the simulated seismic data includes: calculating the arrival time of the seismic signal from the shot point to the receiver point and the P-wave reflection coefficient based on the simulated seismic data to generate a simulated seismic signal; and generating simulated noise data by simulating random noise. The simulated seismic signal is used as a tag signal, and the simulated seismic signal and the simulated random noise are used as input data to construct the five-dimensional dataset using the input data and the tag data.

[0061] Preferably, the calculation of the arrival time of the seismic signal from the shot point to the receiver point includes: acquiring the coordinates of the shot point, the coordinates of the receiver point, the P-wave velocity, and the azimuth angle in the simulated seismic data; and calculating the arrival time of the seismic signal based on a given time using the acquired coordinates of the shot point, the coordinates of the receiver point, the P-wave velocity, and the azimuth angle.

[0062] As an example, an orthogonal observation system is designed according to the "shot line-shot point-receiver line-receiver point" method to obtain the coordinates of all shot points and receiver points in the simulated seismic data. Given a time t0, for example, t0 = 0, 1, 2, ... (ms), the arrival time of the simulated seismic signal (i.e., the seismic signal arrival time) is calculated according to the propagation law of seismic signals in anisotropic media, as shown in the following formula:

[0063]

[0064] Where α is the P-wave velocity, with a default value of 2300, and x is the offset distance (or shot-receiver distance). It is the azimuth angle.

[0065] The offset x can be calculated based on the coordinates of the shot point and receiver point in the seismic trace, as shown in the following formula:

[0066]

[0067] Among them, (x src y src ) and (x rec y rec The coordinates of the shot point and receiver point are respectively obtained through an orthogonal observation system.

[0068] Continuing with the above example, for instance, the (longitudinal wave) reflection coefficient R can be calculated using the formula for the longitudinal wave reflection coefficient in anisotropic media derived by Ruger (1998). PP The formula is as follows:

[0069]

[0070] Among them, R PP Here, θ is the longitudinal wave reflection coefficient, and θ is the incident angle. Let α be the azimuth angle, β be the P-wave velocity, β be the S-wave velocity, Z (Z = ρα) be the wave impedance, ρ be the formation density, and G (G = ρβ) be the wave density. 2 ) represents the transverse wave tangential modulus, γ, δ, and ε are anisotropy parameters, and Δ[·] represents the difference of a certain parameter between the upper and lower interfaces. This represents the average value of a certain parameter between the upper and lower interfaces.

[0071] The incident angle θ can be calculated using the following formula:

[0072]

[0073] Here, t0 can be regarded as the arrival time of the seismic signal at offset x = 0.

[0074] The azimuth angle φ can be calculated using the following formula:

[0075]

[0076] Parameters such as α, β, ρ, γ, δ, and ε can be set based on geological information. Default values ​​for each parameter at the interface are as follows: α (km / s) is 2.300 on the interface and 2.580 below the interface; β (km / s) is 1.060 on the interface and 1.256 below the interface; ρ (g / cm³)... 3 For example, γ is 2.20 on the interface and 2.30 on the bottom; δ is 0 on the interface and 0 on the bottom; ε is 0.05 on the interface and 0.15 on the bottom.

[0077] Preferably, generating a simulated seismic signal based on the arrival time of the seismic signal and the P-wave reflection coefficient includes: calculating the arrival time t of the seismic signal using the following formula, and placing the reflection coefficient r(t) of the strata at the corresponding time point:

[0078]

[0079] Where a(t) is the amplitude fitting vector; e(t) is the dynamic adjustment parameter; RPP is the P-wave reflection coefficient; and θ is the incident angle of the seismic P-wave. This refers to the azimuth of the seismic P-wave.

[0080] The simulated seismic signal is calculated by the convolution of the reflection coefficient r(t) and the seismic wavelet.

[0081] Following the example above, based on equation (6), the reflection coefficients of the strata are placed at the corresponding time points according to the arrival time of the seismic signal. Here, a(t) decreases with time t, with a default value of, for example, 1; e(t) has a default value of, for example, a random number between (-0.5, 0.5). The generated seismic signal s is obtained using the following formula:

[0082] s=w(t)*r(t) (7)

[0083] Where w(t) is the seismic wavelet, and the default value is preferably the Ricker wavelet (as shown in the following formula); * is the convolution operator.

[0084]

[0085] Where f0 is the dominant frequency of the wavelet.

[0086] Preferably, step S1214 may include: using a Ricker wavelet with a preset frequency to perform low-pass filtering on Gaussian noise data generated by simulated random noise, thereby generating the simulated noise data.

[0087] As an example, we simulate low-frequency random noise n and generate simulated noise data d. To simulate realistic seismic random noise, a Ricker wavelet w with a dominant frequency of 60Hz is preferred. f0 =60 performs a low-pass filter on the Gaussian noise to generate low-frequency random noise n, as shown in the following formula:

[0088] n = randn() * w f0＝60 (9)

[0089] Here, randn() is the Gaussian noise generation function.

[0090] The data used to construct the five-dimensional dataset is obtained through numerical simulation. Specifically, the simulated seismic signal is used as the label signal, and the simulated seismic signal and the simulated random noise are used as the input data. The data used to construct the five-dimensional dataset is composed of the input data and the label data. This can enrich the features of the training data and improve the denoising ability of the network training.

[0091] Third, the data used to construct the five-dimensional dataset is composed of the sample earthquake data and the simulated earthquake data.

[0092] Through the above steps one and two, we can obtain denoised sample earthquake data, simulated earthquake signals, and simulated random noise. Using the denoised sample earthquake data and simulated earthquake signals as label signals, and the undenoised sample earthquake data, simulated earthquake signals, and simulated random noise as input data, we can construct the five-dimensional dataset by combining the input data and the label data, which can further enrich the features of the training data.

[0093] Step S122: Construct the dual-channel five-dimensional denoising neural network, train the dual-channel five-dimensional denoising neural network using the five-dimensional training set, and validate the dual-channel five-dimensional denoising neural network using the five-dimensional validation set to obtain the trained dual-channel five-dimensional denoising model.

[0094] In the field of seismic exploration, deep learning technology has been rapidly applied in areas such as seismic fault identification, first arrival picking, and attribute recognition. In addition, the application of deep learning in seismic denoising is also gradually expanding. However, current deep learning-based denoising techniques typically rely on two-dimensional or three-dimensional convolutions, and mainstream deep learning frameworks do not support high-dimensional (four-dimensional or higher) convolutions. Therefore, this invention provides a dual-channel five-dimensional denoising neural network to suppress noise in seismic data. Please refer to... Figure 2 The preferred dual-channel five-dimensional denoising neural network of this invention includes a Fourier transform layer, a five-dimensional denoising neural network layer, and an inverse Fourier transform layer. The output channel of the Fourier transform layer is dual-channel, and the input channel of the inverse Fourier transform layer is dual-channel.

[0095] The preferred embodiment of the present invention describes a five-dimensional denoising neural network layer comprising multiple five-dimensional convolutional layers. The construction of each five-dimensional convolutional layer includes: defining the five-dimensional convolutional layer based on given five-dimensional arrays f and g.

[0096]

[0097] Where j1, j2, j3, j4, j5 represent the elements of array f, and i1-j1, i2-j2, i3-j3, i4-j4, i5-j5 represent the elements of array g;

[0098] To ensure compatibility with common deep learning frameworks and enable GPU acceleration, the defined five-dimensional convolutional layer is modified as follows:

[0099]

[0100] The pseudocode explanation for formula (11) is as follows:

[0101] Initialization: Input a five-dimensional array f (of size N1×N2×N3×N4×N5) and g, sum = 0

[0102] Loop i4 executes from 1 to N4:

[0103] Loop i5 executes from 1 to N5:

[0104] sum ← sum + Conv3d(i1, i2, i3)

[0105] Output sum.

[0106] The preferred embodiment of the present invention provides a five-dimensional denoising neural network layer comprising seven five-dimensional convolutional layers. The first layer of the five-dimensional convolutional layer comprises a set of five-dimensional convolutions with 2 input channels, 64 output channels, and a filter size of 3. The second to sixth layers each comprise a set of five-dimensional convolutions with 64 input channels, 64 output channels, and a filter size of 3, and a ReLU activation function. The seventh layer comprises a set of five-dimensional convolutional layers with 64 input channels, 2 output channels, and a filter size of 3.

[0107] Please refer to Figure 2 The dual-channel five-dimensional denoising neural network includes a Fourier transform layer (FFT), an inverse Fourier transform layer (iFFT), and seven five-dimensional convolutional layers. The Fourier Transform (FFT) layer has 1 input channel and 2 output channels (i.e., two channels: Real and Imaginary), with no training parameters. It performs a Fourier transform on the input data of the five-dimensional training set along the time direction to obtain the seismic data in the frequency domain. The first layer of the five-dimensional convolutional layer includes a set of five-dimensional convolutions with 2 input channels, 64 output channels, and a filter size of 3. Layers 2-6 each include a set of five-dimensional convolutions with 64 input channels, 64 output channels, and a filter size of 3, along with a ReLU activation function. The seventh layer includes a set of five-dimensional convolutions with 64 input channels, 2 output channels, and a filter size of 3. The Inverse Fourier Transform (iFFT) layer has 2 input channels (i.e., two channels: Real and Imaginary), 1 input channel, and no training parameters. It performs an Inverse Fourier Transform on the two-channel data output by the network along the time direction to recover the seismic data in the time domain.

[0108] Preferably, the loss function of the dual-channel five-dimensional denoising neural network is represented by the following formula:

[0109]

[0110] Where K is the number of samples in the five-dimensional training set, {x k ,y k} represents a set of training sample data, x k To simulate noisy data, y k For label data, Θ represents network parameters; ||·|| F It is the Frobenius norm.

[0111] Preferably, step S122 may further include: randomly initializing the parameters of the dual-channel five-dimensional denoising neural network to obtain an initial model of the dual-channel five-dimensional denoising neural network; training the model parameters of the dual-channel five-dimensional denoising neural network using the training set; and validating the dual-channel five-dimensional denoising neural network using the five-dimensional validation set.

[0112] The seismic data from the five-dimensional training set constructed using the simulated noise data and the labeled data is input into the initial model of the dual-channel five-dimensional denoising neural network. The output layer data is obtained, and the loss function shown in Equation (12) is calculated. It is then determined whether the current iteration number meets the maximum training number (the default value is, for example, 50). If not, the network parameters of the current network are adjusted using the backpropagation algorithm until the maximum training number is reached. The network model that has reached the maximum training number is then determined as the dual-channel five-dimensional denoising model. The backpropagation algorithm of the dual-channel five-dimensional denoising neural network is the process of minimizing the loss function (as shown in Equation (12)) to obtain the optimal network parameters. The minimization process of the objective function (as shown in Equation (12)) can be achieved using the Adam optimization algorithm.

[0113] Taking actual earthquake data from a certain region as an example, the actual earthquake data is arranged in the order of "shot line-shot point-receiver line-receiver point" to form a five-dimensional data volume, which is then input into a dual-channel five-dimensional denoising model to obtain the data after noise suppression. Figure 4 The data shows the noisy data of a single shot in this area (all receiver lines of the shot point are selected), which contains a lot of random noise and is difficult to denoise. Figure 5 This shows an effective seismic signal without added noise; Figure 6 The denoising results using a two-dimensional convolutional neural network are shown, with a signal-to-noise ratio of 12.4 dB. Figure 7 The denoising results of the dual-channel five-dimensional denoising model according to the embodiment of the present invention are shown, with a signal-to-noise ratio of 14.1 dB; the embodiment of the present invention can effectively improve the denoising signal-to-noise ratio by 12.1%; Figure 8 The noise identified using a two-dimensional convolutional neural network is shown. Figure 9The noise identified by the dual-channel five-dimensional denoising model according to an embodiment of the present invention is shown. Figure 10 The following is a partial noisy data sample from another single shot (selected from a receiver line at that shot point); Figure 11 The corresponding valid seismic signal without added noise is shown; Figure 12 The denoising results using a two-dimensional convolutional neural network are shown, with a signal-to-noise ratio of 12.3 dB. Figure 13 The denoising result (selecting a certain permutation) of the dual-channel five-dimensional denoising model of the present invention is shown, with a signal-to-noise ratio of 15.6 dB; the present invention can effectively improve the denoising signal-to-noise ratio by 21.2%; Figure 14 The noise identified using a two-dimensional convolutional neural network is shown. Figure 15 The noise identified by the dual-channel five-dimensional denoising model of this invention is shown. This invention can accurately identify noise with no significant residual effective signal. It can efficiently and accurately identify noise while maintaining high amplitude preservation.

[0114] Accordingly, this invention employs a dual-channel five-dimensional denoising model to suppress noise in actual seismic data, fully utilizing the high-dimensional information of the seismic data and effectively improving noise recognition accuracy. By constructing a dual-channel five-dimensional denoising neural network, it fully leverages the high-dimensional information of the "height, width, and height" data. Based on five-dimensional space (three spatial dimensions, offset, and azimuth) and the real and imaginary frequency channels, it trains and learns the features of noise samples in the five-dimensional training set to improve the noise recognition accuracy of the dual-channel five-dimensional denoising model. This provides high-quality basic data for subsequent seismic imaging and reservoir prediction, contributing to improved efficiency and effectiveness in oil and gas exploration and processing.

[0115] This invention also provides a control device for seismic data denoising. The control device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the computer program to implement the above-described method for seismic data denoising.

[0116] This invention also provides a machine-readable storage medium storing instructions that cause a machine to perform the above-described method for denoising seismic data.

[0117] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0118] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0119] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0120] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0121] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0122] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0123] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, 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, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0124] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0125] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for denoising seismic data, characterized in that, The method for denoising seismic data includes: Obtain actual seismic data for the preset area; and The actual seismic data was subjected to noise suppression using a dual-channel five-dimensional denoising model to obtain the denoised seismic data volume. The dual-channel five-dimensional denoising model is obtained by training a dual-channel five-dimensional denoising neural network. This neural network includes a Fourier transform layer, a five-dimensional denoising neural network layer, and an inverse Fourier transform layer. The Fourier transform layer has a dual-channel output, and the inverse Fourier transform layer has a dual-channel input. The five-dimensional denoising neural network layer includes multiple five-dimensional convolutional layers, and the construction of each five-dimensional convolutional layer includes: Based on a given five-dimensional array f and g Define the five-dimensional convolutional layer as follows: ， in, Represents array f elements, Represents array g Element; The defined five-dimensional convolutional layer is deformed as follows: ， The loss function of the dual-channel five-dimensional denoising neural network is represented by the following formula: Where K is the number of samples in the five-dimensional training set. For a set of training sample data, To simulate noise data, For label data, Indicates network parameters; It is the Frobenius norm.

2. The method for denoising seismic data according to claim 1, characterized in that, Training the dual-channel five-dimensional denoising neural network includes: Based on the spatial three-dimensional, offset, and azimuth five-dimensional attributes of seismic data, a five-dimensional dataset is constructed using sample seismic data and / or simulated seismic data, and this dataset is divided into a five-dimensional training set and a five-dimensional validation set; and The dual-channel five-dimensional denoising neural network is constructed, trained using the five-dimensional training set, and validated using the five-dimensional validation set to obtain the trained dual-channel five-dimensional denoising model.

3. The method for denoising seismic data according to claim 2, characterized in that, The data used to construct the five-dimensional dataset from the sample seismic data includes: The sample seismic data is denoised using a preset seismic noise suppression strategy. The sample seismic data after denoising is used as the label data, and the sample seismic data is used as the input data. The input data and the label data are used to construct the five-dimensional dataset.

4. The method for denoising seismic data according to claim 2, characterized in that, The data used to construct the five-dimensional dataset from the simulated earthquake data includes: Based on the simulated seismic data, the arrival time of the seismic signal from the shot point to the receiver point and the P-wave reflection coefficient are calculated to generate the simulated seismic signal. Simulated noise data is generated by simulating random noise. The simulated seismic signal is used as the label data, and the simulated seismic signal and the simulated random noise are used as the input data. The input data and the label data are used to construct the five-dimensional dataset.

5. The method for denoising seismic data according to claim 3, characterized in that, The preset earthquake noise suppression strategy includes one or a combination of abnormal amplitude suppression, FKK domain filtering, and tilt angle filtering.

6. The method for denoising seismic data according to claim 4, characterized in that, The calculation of the arrival time of the seismic signal from the shot point to the receiver point includes: Obtain the coordinates of the shot point, the coordinates of the receiver point, the P-wave velocity, and the azimuth angle from the simulated seismic data; and The arrival time of the seismic signal is calculated based on a given time using the coordinates of the shot point, the coordinates of the receiver point, the P-wave velocity, and the azimuth.

7. The method for denoising seismic data according to claim 6, characterized in that, The step of generating a simulated seismic signal based on the arrival time of the seismic signal and the P-wave reflection coefficient includes: The arrival time of the seismic signal is calculated using the following formula. t Reflectance coefficient of the strata at the corresponding time point : in, This is the amplitude fitting vector; To dynamically adjust parameters, R PP The longitudinal wave reflection coefficient; θ The angle of incidence of the seismic P-wave; This refers to the azimuth of the seismic P-wave. Through the reflection coefficient The simulated seismic signal is obtained by calculating the convolution of the seismic wavelet and the seismic wavelet.

8. The method for denoising seismic data according to claim 4, characterized in that, The process of generating simulated noise data by simulating random noise includes: The simulated noise data is generated by low-pass filtering the Gaussian noise data generated by simulated random noise using a Reichschild wavelet with a preset main frequency.

9. The method for denoising seismic data according to claim 1, characterized in that, The five-dimensional denoising neural network layer includes seven five-dimensional convolutional layers. The first layer of this five-dimensional convolutional layer consists of a set of five-dimensional convolutions with 2 input channels, 64 output channels, and a filter size of 3. Layers 2-6 each consist of a set of five-dimensional convolutions and ReLU activation functions with 64 input channels, 64 output channels, and a filter size of 3. The 7th layer consists of a set of five-dimensional convolutional layers with 64 input channels, 2 output channels, and a filter size of 3.

10. A control device for denoising seismic data, characterized in that, The control device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for denoising seismic data according to any one of claims 1-9.

11. A machine-readable storage medium, characterized in that, The machine-readable storage medium stores instructions that cause the machine to perform the method for denoising seismic data according to any one of claims 1-9.

Images