Seismic data frequency band expansion method, device, equipment and medium

Abstract

The invention relates to a seismic data frequency band expanding method, device and equipment and a medium, and belongs to the field of seismic exploration data processing.The seismic data frequency band expanding method comprises the steps that a pseudo reflection coefficient sequence is obtained according to an extracted work area logging curve; setting an expanded frequency band according to the frequency band of the seismic data in the work area, generating a Ricker wavelet consistent with the main frequency of the seismic data, constructing a broadband B-spline wavelet based on the expanded frequency band, and respectively performing convolution processing on the broadband B-spline wavelet and the Ricker wavelet with the pseudo reflection coefficient sequence to obtain a narrow frequency band and broadband seismic data; the constructed DAU-Net model is trained based on the narrow-band and wide-band seismic data, the to-be-predicted narrow-band seismic data is predicted based on the completely trained DAU-Net model, the predicted wide-band seismic data is obtained, and the resolution of the seismic data is improved.

Description

Technical Field

[0001] The present invention relates to the field of seismic exploration data processing, and particularly to a method, device, equipment and medium for expanding the frequency band of seismic data. Background Art

[0002] Seismic exploration is one of the most important means in oil and gas exploration and geological research. The frequency band width of seismic data directly affects the resolution and imaging quality of seismic signals. However, limited by factors such as the performance of seismic instruments, formation absorption attenuation and noise interference, the actually collected seismic data usually has problems such as low main frequency and narrow profile frequency band, which will lead to a decrease in the resolution of seismic signals, and further affect the accurate interpretation of geological structures. Expanding the frequency band width of seismic data through frequency band expansion methods can improve the resolution and imaging quality of seismic signals.

[0003] Traditional frequency band expansion methods mainly rely on signal processing-based algorithms, such as time-frequency analysis, deconvolution, wavelet transform and matching filtering, etc.

[0004] However, these methods can expand the frequency band of seismic data to a certain extent, but there are many limitations. Due to the inherent trade-off between time resolution and frequency resolution, the time-frequency analysis method is difficult to accurately capture the rapid changes of complex signals and is extremely sensitive to noise; the deconvolution method depends on the accurate estimation of seismic wavelets, and is prone to result distortion when geological conditions are complex or the model is inaccurate, and usually tends to enhance high-frequency signals while ignoring low-frequency parts; although the wavelet transform method can decompose signals at multiple scales, its effect is significantly affected by the selection of wavelet bases, and its performance degrades when the noise is strong; the matching filtering method requires strong prior knowledge of the target signal, and often introduces pseudo-frequency components due to model errors in actual seismic data processing. These traditional methods are sensitive to noise, lack adaptability to complex geological conditions, and the frequency band expansion range is usually limited to high frequency or low frequency, and it is difficult to take both into account at the same time, so there are obvious deficiencies in high-precision expansion tasks. Summary of the Invention

[0005] In view of this, it is necessary to provide a method, device, equipment and medium for expanding the frequency band of seismic data to solve the technical problem that the traditional frequency band expansion methods have limitations.

[0006] To solve the above problems, in a first aspect, the present invention provides a method for expanding the frequency band of seismic data, including: Obtaining a pseudo-reflection coefficient sequence according to the extracted logging curves of the work area; Expand the frequency band according to the frequency band of the seismic data in the work area, generate a Ricker wavelet consistent with the main frequency of the seismic data, construct a broadband B-spline wavelet based on the expanded frequency band, and convolve the broadband B-spline wavelet and the Ricker wavelet with the pseudo-reflection coefficient sequence respectively to obtain narrow-band and wide-band seismic data; Train the constructed DAU-Net model based on the narrow-band and wide-band seismic data; Predict the narrow-band seismic data to be predicted based on the trained DAU-Net model to obtain predicted wide-band seismic data.

[0007] In a possible implementation manner, the well logging curve in the work area includes well logging data; the obtaining of the pseudo-reflection coefficient sequence according to the extracted well logging curve in the work area includes: Calculate the reflection coefficient sequence corresponding to the well logging data based on the well logging data; Interpolate and resample the reflection coefficient sequence to generate a pseudo-reflection coefficient sequence.

[0008] In a possible implementation manner, the calculation formula of the Ricker wavelet is: , where, is the Ricker wavelet, is the main frequency, is the time.

[0009] In a possible implementation manner, the calculation formula of the broadband B-spline wavelet is: , where, is the broadband B-spline wavelet, is the bandwidth, is the upper limit of the expanded frequency band, is the lower limit of the expanded frequency band, is an integer, is the sinc function.

[0010] In a possible implementation manner, the DAU-Net model includes an encoder, a decoder, a skip connection, and a residual attention module, and the decoder includes a sub-pixel convolutional layer; the training of the constructed DAU-Net model based on the narrow-band and wide-band seismic data includes: Use the narrow-band seismic data as training data and the wide-band seismic data as training labels; Input the training data into the DAU-Net model. Before the encoder downsamples the training data, the feature enhancement process is performed on the features of the training data through the residual attention module. After extracting the features of the training data and generating the encoded feature map through downsampling by the encoder; Transmit the local feature map of the corresponding layer in the encoder after feature enhancement through the residual attention module to the decoder through skip connection; Decode the encoded feature map and the locally feature-enhanced map through the decoder to generate high-resolution broadband seismic data. Among them, the sub-pixel convolutional layer of the decoder is used to convert the low-resolution feature map into a high-resolution feature map; After comparing the high-resolution broadband seismic data with the training labels, adjust the loss function of the DAU-Net model to generate a fully trained DAU-Net model; Input the obtained test set data into the fully trained DAU-Net model to obtain the prediction data of the test set, and evaluate the fully trained DAU-Net model based on the prediction data of the test set.

[0011] In a possible implementation, the residual attention module includes a channel attention module and a spatial attention module combined. The channel attention module dynamically allocates weights for different feature channels through global pooling and fully connected layers, and the spatial attention module optimizes the allocation of spatial weights through maximum convolution.

[0012] In a possible implementation, the calculation formula of the loss function of the DAU-Net model is: , where, is the loss function, is the number of samples, is the true value, is the predicted value, is an integer.

[0013] In a second aspect, the present invention also provides a seismic data bandwidth expansion device, including: A pseudo-reflection coefficient calculation module, configured to obtain a pseudo-reflection coefficient sequence according to the extracted logging curves in the work area; An earthquake data acquisition module, configured to expand a frequency band according to the frequency band setting of seismic data in a work area, generate a Ricker wavelet consistent with the main frequency of the seismic data, construct a broadband B-spline wavelet based on the expanded frequency band, and perform convolution processing on the broadband B-spline wavelet and the Ricker wavelet respectively with the pseudo-reflection coefficient sequence to obtain narrow-band and wide-band seismic data; A training module, configured to train a constructed DAU-Net model based on the narrow-band and wide-band seismic data; A prediction module, configured to predict narrow-band seismic data to be predicted based on a trained DAU-Net model to obtain predicted wide-band seismic data.

[0014] In a third aspect, the present invention further provides an electronic device, including: a processor and a memory; A computer-readable program executable by the processor is stored on the memory; When the processor executes the computer-readable program, the steps in the above-mentioned seismic data frequency band expansion method are implemented.

[0015] In a fourth aspect, the present invention further provides a computer-readable storage medium, where one or more programs are stored on the computer-readable storage medium, and the one or more programs can be executed by one or more processors to implement the steps in the above-mentioned seismic data frequency band expansion method.

[0016] The beneficial effects of the present invention are: training a constructed DAU-Net model based on narrow-band and wide-band seismic data, the DAU-Net model is improved on the basis of the U-Net network, a residual attention module is added on the basis of the U-Net network, combining the pseudo-reflection coefficient generated by well logging data and the broadband B-spline wavelet, fully excavating the complex relationship between narrow-band and wide-band seismic data, realizing high-precision frequency band expansion of seismic data, using the powerful feature extraction ability of deep learning, capable of simultaneously expanding high frequencies and retaining low-frequency information, overcoming the problem that the frequency band expansion range is usually limited to high frequencies or low frequencies, and improving the resolution and interpretation accuracy of seismic data. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those skilled in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0018] Figure 1 It is a flowchart of an embodiment of the seismic data frequency band expansion method provided by the present invention; Figure 2 Schematic diagram of Ricker wavelet and broadband B-spline wavelet in the time domain for the seismic data bandwidth expansion method provided by the present invention; Figure 3 Schematic diagram of the amplitude spectra of the Ricker wavelet and broadband B-spline wavelet for the seismic data bandwidth expansion method provided by the present invention; Figure 4 Schematic diagram of the structure of the DAU-Net model for the seismic data bandwidth expansion method provided by the present invention; Figure 5 Schematic diagram of the original seismic profile input to the DAU-Net model for the seismic data bandwidth expansion method provided by the present invention; Figure 6 Schematic diagram of the seismic profile with improved resolution output by the DAU-Net model for the seismic data bandwidth expansion method provided by the present invention; Figure 7 Schematic diagram of the seismic spectrum comparison for the seismic data bandwidth expansion method provided by the present invention; Figure 8 Schematic diagram of the structure of an embodiment of the seismic data bandwidth expansion device provided by the present invention; Figure 9 Schematic diagram of the structure of an embodiment of the electronic device provided by the present invention. Detailed implementation manners

[0019] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative efforts fall within the protection scope of the present invention.

[0020] The descriptions such as "first" and "second" involved in the embodiments of the present invention are only for descriptive purposes, and cannot be understood as indicating or implying their relative importance or implicitly indicating the quantity of the indicated technical features. Therefore, the technical features defined with "first" and "second" may explicitly or implicitly include at least one of such features.

[0021] Referring to "embodiment" herein means that a specific feature, structure, or characteristic described in connection with the embodiment may be included in at least one embodiment of the present invention. The phrase appears in various places in the specification and does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art will explicitly and implicitly understand that the embodiments described herein may be combined with other embodiments.

[0022] Before presenting the embodiments, the following terms are explained first.

[0023] U-Net network: As a deep learning model specifically designed for image data, its unique encoder-decoder structure combined with the design of skip connections enables it to effectively retain detailed information while capturing spatial information; leveraging its advantages to overcome the limitations of traditional methods, achieving more accurate and stable frequency band extension, improving the resolution and interpretation accuracy of seismic data. Through the deep architecture of the U-Net network, not only can complex seismic data sets be efficiently processed, but also the network parameters can be optimized to adapt to different geological structure characteristics, thereby enhancing the accuracy and reliability of seismic exploration results.

[0024] Reflection coefficient sequence: A key parameter in seismic exploration, used to describe the reflection characteristics of underground media in the depth direction.

[0025] The present invention discloses a method, device, equipment and medium for seismic data frequency band extension, which can be used in a computer. The method, equipment or computer-readable storage medium involved in the present invention can either be integrated with the above-mentioned equipment or be relatively independent.

[0026] A specific embodiment of the present invention discloses a method for seismic data frequency band extension, which can be executed by a computer, specifically by one or more processors of the computer. As Figure 1 shown, the method for seismic data frequency band extension includes: S101. Obtain a pseudo-reflection coefficient sequence according to the extracted logging curves in the work area.

[0027] It should be noted that the pseudo-reflection coefficient sequence, as an important bridge connecting logging data and seismic data, can provide a reliable basis for data extension.

[0028] S102. Set an extended frequency band according to the frequency band of seismic data in the work area, generate a Ricker wavelet consistent with the main frequency of the seismic data, construct a broadband B-spline wavelet based on the extended frequency band, and perform convolution processing on the broadband B-spline wavelet and the Ricker wavelet with the pseudo-reflection coefficient sequence respectively to obtain narrow-band and wide-band seismic data.

[0029] It should be noted that convolution processing is a key technology in seismic data processing for generating synthetic seismic records. Through convolution processing, by utilizing the different characteristics of the broadband B-spline wavelet and the Ricker wavelet, the pseudo-reflection coefficient sequence is filtered and transformed, thereby generating seismic data with different frequency band characteristics; narrow-band seismic data mainly reflects the large-scale structure and low-frequency components of underground media, while wide-band seismic data can reveal more information about the details and high-frequency components of underground media.

[0030] S103. Train the constructed DAU-Net model based on narrow-band and wide-band seismic data; It should be noted that the DAU-Net model is improved on the basis of the U-Net model, mainly introducing a channel attention module and a spatial attention module. The channel attention module dynamically assigns weights to different feature channels through global pooling and fully connected layers, highlighting the feature channels crucial for high-frequency information recovery; the spatial attention module learns the spatial correlation between pixels through maximum convolution operations, optimizes the spatial weight distribution, and enhances the detail recovery ability in sparse regions of seismic data. The attention module and the spatial attention module are combined through a residual structure to form a dual attention residual block (residual attention module), enabling the network to focus on key information simultaneously in the channel and spatial dimensions. In addition, the DAU-Net model removes the batch normalization (BN) layer to reduce computational complexity and adds a sub-pixel convolution layer to improve the detail recovery effect.

[0031] S104. Predict the narrow-band seismic data to be predicted based on the trained DAU-Net model to obtain high-resolution wide-band seismic data.

[0032] In some embodiments of the present invention, in step S101, according to the extracted well logging curves of the work area, a pseudo-reflection coefficient sequence is obtained, and according to the extracted well logging curves of the work area, well logging data is obtained; a reflection coefficient sequence corresponding to the well logging data is calculated based on the well logging data, and the reflection coefficient sequence is a reflection coefficient sequence in the depth domain. The reflection coefficient sequence is interpolated and resampled to generate a pseudo-reflection coefficient sequence, that is, the reflection coefficient sequence is interpolated and resampled by the Piecewise Cubic Hermite Interpolating Polynomial (abbreviated as PCHIP). The Hermite interpolation method is a polynomial interpolation method that requires not only that the function values of the interpolation polynomial at the interpolation nodes are equal to the function values of the function to be interpolated, but also that the first derivative (and even higher-order derivatives) of the interpolation polynomial at these nodes are also equal to the first derivative (or corresponding order derivatives) of the function to be interpolated; this interpolation method can more comprehensively reflect the behavior of the function to be interpolated; through the Piecewise Cubic Hermite interpolation method, a cubic polynomial is obtained on each interval The calculation formula of the cubic polynomial is: , where is the cubic polynomial, , , , are the polynomial coefficients; after interpolation, a large number of pseudo-reflection coefficient sequences of the same length as the original reflection coefficient sequence are generated.

[0033] The real logging data is augmented by the method of interpolation resampling, which effectively alleviates the problem of insufficient labeled data in the training of the deep network and significantly improves the generalization ability of the model.

[0034] In some embodiments of the present invention, in step S102, the extended frequency band is set according to the frequency band of the seismic data in the work area, and a Ricker wavelet consistent with the main frequency of the seismic data is generated. A broadband B-spline wavelet is constructed based on the extended frequency band. According to the actual seismic data frequency band in the work area, a Ricker wavelet with the same main frequency as the seismic data is generated. The calculation formula of the Ricker wavelet is: , where, is the Ricker wavelet, is the main frequency, is the time; The extended frequency band is set according to the actual seismic data in the work area. The extended frequency band is the target frequency band to be extended. According to the actual seismic data in the work area and the extended frequency band, a broadband B-spline wavelet is constructed within the frequency range of the extended frequency band. The calculation formula of the broadband B-spline wavelet is: , where, is the broadband B-spline wavelet, is the bandwidth, is the upper limit of the extended frequency band, is the lower limit of the extended frequency band, is an integer, is the sinc function, and the calculation formula of the sinc function is: , where, is the input value and is the independent variable; For the schematic diagrams of the Ricker wavelet and the broadband B-spline wavelet in the time domain, please refer to Figure 2 , and for the schematic diagrams of the amplitude spectra of the Ricker wavelet and the broadband B-spline wavelet, please refer to Figure 3 , The broadband B-spline wavelet and the Ricker wavelet are respectively convolved with the pseudo-reflection coefficient sequence to obtain narrow-band and wide-band seismic data. That is, the generated Ricker wavelet is convolved with the pseudo-reflection coefficient sequence to obtain the narrow-band synthetic seismic record for training, and the narrow-band synthetic seismic record is the narrow-band seismic data; the broadband B-spline wavelet is convolved with the pseudo-reflection coefficient sequence to obtain the wide-band synthetic seismic record as the training label, and the wide-band synthetic seismic record is the wide-band seismic data. The expression of the synthetic seismic record is: , where, is the pseudo-reflection coefficient sequence in the time domain, is the wavelet. After processing all the reflection coefficient sequences in turn, training data and training labels are obtained.

[0035] The broadband B-spline wavelet has flexible frequency control and good localization characteristics, which can effectively capture the transient characteristics in seismic signals, reduce noise interference, and can simultaneously expand the high and low frequency information of seismic data.

[0036] In some embodiments of the present invention, in step S103, the constructed DAU-Net model is trained based on narrow-band and wide-band seismic data. The DAU-Net model includes an encoder, a decoder, skip connections, and a residual attention module. The residual attention module includes a channel attention module and a spatial attention module. The channel attention module dynamically allocates weights for different feature channels through global pooling and fully connected layers. The spatial attention module optimizes the allocation of spatial weights through max convolution. The decoder includes a sub-pixel convolutional layer. The training process is as follows: using the narrow-band seismic data as training data and the wide-band seismic data as training labels; inputting the training data into the DAU-Net model, and through the encoder, downsampling the training data to extract the features of the training data and generating an encoded feature map. Among them, before the encoder downsamples the training data, the features of the training data are enhanced through the residual attention module; through the skip connections, the corresponding hierarchical local feature maps in the encoder after feature enhancement by the residual attention module are transmitted to the decoder; through the decoder, the encoded feature map and the feature-enhanced local feature map are decoded to generate high-resolution wide-band seismic data. Among them, the sub-pixel convolutional layer of the decoder converts the low-resolution feature map into a high-resolution feature map; after comparing the high-resolution wide-band seismic data with the training labels, the loss function of the DAU-Net model is adjusted to generate a fully trained DAU-Net model. Finally, the obtained test set data is input into the fully trained DAU-Net model to obtain the predicted data of the test set, and the fully trained DAU-Net model is evaluated based on the predicted data of the test set. The test set data is seismic record data that has not participated in the training of the DAU-Net model. The calculation formula of the loss function of the DAU-Net model is: , where, is the loss function, is the number of samples, is the true value, is the predicted value, is an integer.

[0037] For the schematic diagram of the structure of the DAU-Net model, please refer to Figure 4, as Figure 4 shown, the prepared narrow-band seismic data is input into the DAU-Net model as the initial input of the model. The input data enters the encoder part, which consists of multiple convolutional layers and pooling layers. The convolutional layers gradually extract the feature information of the image. As the number of layers increases, the spatial resolution of the feature map gradually decreases, but the number of channels increases to capture higher-level features. Before downsampling at each layer, a residual attention mechanism is introduced. The residual attention module calculates the correlation between feature maps, generates an attention map, and applies it to the feature map to enhance important features and suppress unimportant features. Skip connections directly connect the outputs of the corresponding levels in the encoder to the corresponding levels in the decoder. These connections help retain more detailed information during the decoding process, avoid the vanishing gradient problem, and promote the effective fusion of features. The decoder part corresponds to the encoder and is responsible for gradually restoring the spatial resolution of the image. The decoder gradually restores the spatial details of the feature map through upsampling. In each layer of the decoder, feature fusion is performed by combining the features transmitted by the skip connections. Finally, the prediction result of the wide-band seismic data is output, and the prediction result should be as close as possible to the training label (i.e., the real wide-band seismic data).

[0038] The training process of the DAU-Net model is an iterative optimization process. By extracting features through the encoder, reconstructing images through the decoder, enhancing features through the residual attention mechanism, and retaining details through skip connections, high-precision frequency band expansion of seismic data is achieved.

[0039] The channel attention module of the residual attention module dynamically assigns weights to different feature channels through global pooling and fully connected layers, highlighting the feature channels crucial for high-frequency information recovery; the spatial attention module learns the spatial correlation between pixels through maximum convolution operations, optimizes the spatial weight distribution, and enhances the detail recovery ability in sparse regions of seismic data, enabling the network to focus on key information in both the channel and spatial dimensions. Based on the traditional U-Net, the DAU-Net model enhances the model's learning ability for complex features in seismic data by adding a channel attention module and a spatial attention module before each downsampling step in the encoder part. Using an improved U-Net structure (removing the batch normalization layer), combined with the residual attention module and the sub-pixel convolutional layer, not only improves the resolution of seismic data but also enhances the generalization ability and stability of the model, thus more effectively recovering high-frequency information and suppressing noise, and improving the accuracy and detail performance of seismic data processing.

[0040] In some embodiments of the present invention, in step S104, based on the trained DAU-Net model, the narrow-band seismic data to be predicted is predicted to obtain high-resolution wide-band seismic data. The narrow-band seismic data to be predicted is input into the trained DAU-Net model, and the trained DAU-Net model predicts the narrow-band seismic data to be predicted to obtain high-resolution wide-band seismic data. For the schematic diagram of the original seismic profile input into the DAU-Net model, please refer to Figure 5 , and for the schematic diagram of the seismic profile with improved resolution output by the DAU-Net model, please refer to Figure 6 . By comparing Figure 5 and Figure 6 , it can be seen that after being processed by the DAU-Net model, the resolution of the seismic data is significantly improved; for the schematic diagram of the seismic spectrum comparison, please refer to Figure 7 . As shown in Figure 7 , by comparing the spectrum of the high-resolution seismic data output by the DAU-Net model with the spectrum of the original input seismic data, it can be known that using the DAU-Net model to expand the seismic data frequency band can not only expand the high-frequency data while retaining the low-frequency data, effectively improving the resolution of the seismic data.

[0041] By constructing an improved dual-attention mechanism U-Net network, namely the DAU-Net model, the dual-attention mechanism (Dual Attention Mechanism), that is, the channel attention mechanism and the spatial attention mechanism, combines the pseudo-reflection coefficient sequence generated from well logging data and the broadband B-spline wavelet to achieve high-precision frequency band expansion of seismic data. Utilizing the powerful feature extraction ability of deep learning, it can expand high frequencies and retain low-frequency information simultaneously, significantly improving the resolution and interpretation accuracy of seismic data. Compared with traditional methods, it has higher prediction accuracy and generalization ability, and is suitable for seismic data processing under complex geological conditions, providing an innovative high-resolution data processing solution for the seismic exploration field.

[0042] In summary, the seismic data frequency band expansion method provided by the present invention obtains a pseudo-reflection coefficient sequence according to the extracted well logging curves in the work area; sets an expansion frequency band according to the frequency band of the seismic data in the work area and generates a Ricker wavelet consistent with the main frequency of the seismic data, constructs a broadband B-spline wavelet based on the expansion frequency band, convolves the broadband B-spline wavelet and the Ricker wavelet with the pseudo-reflection coefficient sequence respectively to obtain narrow-band and wide-band seismic data; trains the constructed DAU-Net model based on the narrow-band and wide-band seismic data; predicts the narrow-band seismic data to be predicted based on the trained DAU-Net model to obtain high-resolution wide-band seismic data, improving the resolution of the seismic data.

[0043] To better implement the seismic data bandwidth expansion method in the embodiments of the present invention, correspondingly, based on the seismic data bandwidth expansion method, as Figure 8 shown, the embodiments of the present invention further provide a seismic data bandwidth expansion device. The seismic data bandwidth expansion device 800 includes a pseudo-reflection coefficient calculation module 801, a seismic data acquisition module 802, a training module 803, and a prediction module 804; The pseudo-reflection coefficient calculation module 801 is used to obtain a pseudo-reflection coefficient sequence according to the extracted well logging curves in the work area; The seismic data acquisition module 802 is used to set an expanded bandwidth according to the bandwidth of the seismic data in the work area, generate a Ricker wavelet consistent with the main frequency of the seismic data, construct a broadband B-spline wavelet based on the expanded bandwidth, and perform convolution processing on the broadband B-spline wavelet and the Ricker wavelet with the pseudo-reflection coefficient sequence respectively to obtain narrow-band and wide-band seismic data; The training module 803 is used to train the constructed DAU-Net model based on the narrow-band and wide-band seismic data; The prediction module 804 is used to predict the narrow-band seismic data to be predicted based on the trained DAU-Net model to obtain predicted wide-band seismic data.

[0044] As Figure 9 shown, the present invention also correspondingly provides an electronic device 900. The electronic device 900 can be a computing device such as a mobile terminal, a desktop computer, a notebook, a palm computer, and a server. The electronic device 900 includes a processor 901, a memory 902, and a display 903. Figure 9 Only some components of the electronic device 900 are shown, but it should be understood that it is not required to implement all the shown components, and more or fewer components can be implemented alternatively.

[0045] The memory 902 can be an internal storage unit of the electronic device 900 in some embodiments, such as a hard disk or memory of the electronic device 900. The memory 902 can also be an external storage device of the electronic device 900 in other embodiments, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. equipped on the electronic device 900. Further, the memory 902 can also include both the internal storage unit of the electronic device 900 and external storage devices. The memory 902 is used to store application software installed on the electronic device 900 and various types of data, such as program codes installed on the electronic device 900. The memory 902 can also be used to temporarily store data that has been output or will be output. In one embodiment, a seismic data bandwidth expansion program is stored on the memory 902, and the seismic data bandwidth expansion program can be executed by the processor 901 to implement the seismic data bandwidth expansion method of various embodiments of the present invention.

[0046] The processor 901 can be a central processing unit (CPU), a microprocessor or other data processing chips in some embodiments, and is used to run program codes stored in the memory 902 or process data, such as the seismic data bandwidth expansion method, etc.

[0047] The display 903 can be an LED display, a liquid crystal display, a touch liquid crystal display, and an OLED (Organic Light-Emitting Diode) toucher, etc. in some embodiments. The display 903 is used to display identification information of the seismic data bandwidth expansion program and to display a visual user interface. Components 901 - 903 of the electronic device 900 communicate with each other through a device bus.

[0048] In some embodiments, when the processor 901 executes the seismic data bandwidth expansion program in the memory 902, each step in the seismic data bandwidth expansion method described in the above embodiments is implemented. Since the seismic data bandwidth expansion method has been described in detail above, it will not be elaborated here.

[0049] Furthermore, the embodiments of the present invention do not specifically limit the type of the mentioned electronic device 900. The electronic device 900 may be a mobile phone, a tablet computer, a personal digital assistant (PDA), a wearable device, a laptop, or other portable electronic devices. Exemplary embodiments of the portable electronic device include, but are not limited to, portable electronic devices equipped with IOS, android, microsoft, or other operating devices. The above-mentioned portable electronic device may also be other portable electronic devices, such as a laptop with a touch-sensitive surface (e.g., a touch panel). It should also be understood that in some other embodiments of the present invention, the electronic device 900 may not be a portable electronic device, but a desktop computer with a touch-sensitive surface (e.g., a touch panel).

[0050] Correspondingly, the present invention also provides a computer-readable storage medium for storing computer-readable programs or instructions. When the programs or instructions are executed by a processor, they can implement the steps or functions in the seismic data frequency band expansion method provided by the above-mentioned method embodiments.

[0051] Those skilled in the art can understand that all or part of the processes of implementing the methods in the above embodiments can be completed by instructing relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium. Among them, the computer-readable storage medium is a magnetic disk, an optical disk, a read-only memory, or a random access memory, etc.

[0052] The above is only the preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present invention should be covered by the protection scope of the present invention.

Claims

1. A method for extending the frequency band of seismic data, characterized in that: include: According to the extracted logging curve of the work area, the pseudo reflection coefficient sequence is obtained; An extended frequency band is set according to the frequency band of the seismic data in the work area, and a Ricker wavelet consistent with the main frequency of the seismic data is generated, a broadband B-spline wavelet is constructed based on the extended frequency band, and the broadband B-spline wavelet and the Ricker wavelet are convolved with the pseudo reflection coefficient sequence respectively to obtain narrow-band and broadband seismic data; Training the constructed DAU-Net model based on the narrowband and broadband seismic data; Based on the well-trained DAU-Net model, the narrow-band seismic data to be predicted are predicted to obtain the predicted broadband seismic data.

2. The method for extending the frequency band of seismic data according to claim 1, characterized in that: The work area logging curve includes logging data; the step of obtaining a pseudo reflection coefficient sequence based on the extracted work area logging curve includes: Calculating a reflection coefficient sequence corresponding to the well logging data based on the well logging data; The reflection coefficient sequence is interpolated and resampled to generate a pseudo reflection coefficient sequence.

3. The seismic data frequency band extension method according to claim 1, characterized in that: The calculation formula of the Ricker wavelet is: ， in, is the Ricker wavelet, is the main frequency, For time.

4. The method for extending the frequency band of seismic data according to claim 1, characterized in that: The calculation formula of the broadband B-spline wavelet is: ， in, is a broadband B-spline wavelet, is the bandwidth, To expand the upper limit of the frequency band, To extend the lower limit of the frequency band, is an integer, is the Symgine function.

5. The method for extending the frequency band of seismic data according to claim 1, characterized in that: The DAU-Net model includes an encoder, a decoder, a skip connection, and a residual attention module, and the decoder includes a sub-pixel convolution layer; the DAU-Net model constructed based on the narrow-band and wide-band seismic data is trained, including: Using the narrowband seismic data as training data and the broadband seismic data as training labels; The training data is input into the DAU-Net model, and the training data is downsampled by an encoder, and features of the training data are extracted to generate an encoded feature map, wherein before the encoder downsamples the training data, features of the training data are enhanced by a residual attention module; Transmitting the local feature map of the corresponding level in the encoder after feature enhancement processing by the residual attention module to the decoder through a skip connection; Decoding the encoded feature map and the feature-enhanced local feature map through the decoder to generate high-resolution broadband seismic data, wherein the low-resolution feature map is converted into a high-resolution feature map through a sub-pixel convolution layer of the decoder; After comparing the high-resolution broadband seismic data with the training labels, adjusting the loss function of the DAU-Net model to generate a fully trained DAU-Net model; The acquired test set data is input into the fully trained DAU-Net model to obtain the predicted data of the test set, and the fully trained DAU-Net model is evaluated based on the predicted data of the test set.

6. The method for extending the frequency band of seismic data according to claim 5, characterized in that: The residual attention module includes a channel attention module and a spatial attention module. The channel attention module dynamically allocates weights of different feature channels through global pooling and fully connected layers, and the spatial attention module optimizes the allocation of spatial weights through maximum convolution.

7. The method for extending the frequency band of seismic data according to claim 5, characterized in that: The loss function of the DAU-Net model is calculated as: ， in, is the loss function, is the number of samples, is the true value, is the predicted value, is an integer.

8. A seismic data frequency band expansion device, characterized in that: include: The pseudo reflection coefficient calculation module is used to obtain the pseudo reflection coefficient sequence according to the extracted logging curve of the work area; A seismic data acquisition module is used to set an extended frequency band according to the frequency band of seismic data in the work area, generate a Ricker wavelet consistent with the main frequency of the seismic data, construct a broadband B-spline wavelet based on the extended frequency band, and convolve the broadband B-spline wavelet and the Ricker wavelet with the pseudo reflection coefficient sequence respectively to obtain narrow-band and broadband seismic data; A training module, used for training the constructed DAU-Net model based on the narrowband and broadband seismic data; The prediction module is used to predict the narrow-band seismic data to be predicted based on the well-trained DAU-Net model to obtain the predicted broadband seismic data.

9. An electronic device, characterized in that: including memory and processor; The memory stores a computer-readable program executable by the processor; When the processor executes the computer-readable program, the steps in the seismic data frequency band extension method as described in any one of claims 1-7 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps in the seismic data frequency band extension method as described in any one of claims 1-7.

Images