Method and device for reconstructing velocity model of underground aquifer gas storage

Abstract

The application relates to the field of geological exploration, in particular to an underground aquifer gas storage speed model reconstruction method and device, electronic equipment and a computer readable storage medium. The method comprises the following steps: obtaining a trained geological exploration model, wherein a loss function of the geological exploration model comprises a directional gradient operator, the directional gradient operator is obtained by weighting a horizontal gradient kernel and a vertical gradient kernel by a horizontal gradient weight and a vertical gradient weight, and the horizontal gradient weight is greater than the vertical gradient weight; obtaining measured seismic source data and measured detection data of a region to be explored; inputting the measured seismic source data and the measured detection data into the geological exploration model to obtain an underground aquifer gas storage speed model output by the geological exploration model. The underground aquifer gas storage speed model reconstruction method and device, electronic equipment and computer readable storage medium provided by the application can improve the boundary definition of the underground aquifer gas storage speed model.

Description

Technical Field

[0001] This application relates to the field of geological exploration, specifically to a method and apparatus for reconstructing a velocity model of an underground aquifer gas reservoir, electronic equipment, and a computer-readable storage medium. Background Technology

[0002] Underground aquifer gas storage facilities, as key infrastructure for natural gas peak shaving and supply guarantee and national energy security, place extremely high demands on the precise characterization of underground space during their construction and operation monitoring. These storage facilities are typically located at deep depths (2000 meters and above), with complex geological structures, often exhibiting anticlines or fault nose structures controlled by faults, and containing a complex distribution of gas, water, and rock phases. In the exploration and dynamic monitoring of gas storage facilities, high-precision velocity model reconstruction is fundamental for identifying the gas-water interface and assessing the scale of the storage, while the identification of micro-faults and fractures is crucial for assessing the trap's sealing performance and preventing gas escape. Seismic exploration, as the primary means of probing deep geological structures, utilizes the dynamics and kinematics of seismic wave propagation in the underground medium to invert stratigraphic velocities and structural morphology, and is currently the core technical approach to solving these problems.

[0003] With the development of deep learning technology, establishing a nonlinear mapping between seismic response and geological models using convolutional neural networks has become a research hotspot. However, when performing detailed inversion on deep aquifer gas reservoirs, "strata artifacts" caused by the anisotropy of geological structures result in significant lateral extension of sedimentary strata, while faults exhibit vertical dislocation. Existing inversion networks often ignore this directional difference, leading to an inability to distinguish between strata and faults when sharpening images. This can easily result in the erroneous sharpening of horizontal strata while enhancing faults, producing a large number of non-physical horizontal stripe artifacts that severely interfere with reservoir interpretation. Summary of the Invention

[0004] In view of this, it is necessary to provide a method and apparatus, electronic equipment and computer-readable storage medium for reconstructing the velocity model of an underground aquifer gas reservoir, so as to achieve the technical effect of improving the boundary clarity in the velocity model of the underground aquifer gas reservoir.

[0005] To address the aforementioned technical problems, firstly, this application provides a method for reconstructing a velocity model of an underground aquifer gas storage reservoir, comprising: Obtain the trained geological exploration model. The loss function of the geological exploration model includes a directional gradient operator. The directional gradient operator is calculated by weighting the horizontal gradient kernel and the vertical gradient kernel with horizontal gradient weights and vertical gradient weights. The horizontal gradient weights are greater than the vertical gradient weights. Obtain measured source data and measured detector data of the area to be explored; The measured seismic source data and the measured detection data are input into the geological exploration model to obtain the underground aquifer gas reservoir velocity model output by the geological exploration model.

[0006] In one possible embodiment, obtaining the trained geological exploration model includes: A basic model is constructed, which includes an encoder and a decoder. The encoder includes multiple attention residual modules. Each attention residual module includes a residual block and an attention module connected in series at the end of the residual block. The attention module includes a channel attention module for suppressing channel noise and a spatial attention module for focusing tomography. Obtain training data; The basic model is trained based on the training data to obtain the trained geological exploration model.

[0007] In one possible embodiment, the channel attention module includes a first global average pooling layer and a first global max pooling layer that are parallel to each other, a multilayer perceptron connected to the first global average pooling layer and the first global max pooling layer, and a first sigmoid activation structure connected to the multilayer perceptron. The spatial attention module includes a second global average pooling layer and a second global max pooling layer that run in parallel, a vector concatenation layer connected to the second global average pooling layer and the second global max pooling layer, a convolutional layer connected to the vector concatenation layer, and a second sigmoid activation structure connected to the convolutional layer.

[0008] In one possible embodiment, the decoder includes a first decoding branch and a second decoding branch, and training the base model based on the training data includes: The first decoding branch is trained based on the output of the channel attention module; The second decoding branch is trained based on the output of the spatial attention module.

[0009] In one possible embodiment, calculating the directional gradient operator includes: Based on formula group ; ; The directional gradient operator is calculated. in, For the directional gradient operator, The horizontal gradient weights, The horizontal gradient kernel, The vertical gradient kernel, The vertical gradient weights are... This is the output of the first decoding branch. for The gradient operator in the horizontal direction, for The gradient operator in the vertical direction, This is a convolution operation.

[0010] In one possible embodiment, the loss function of the geological exploration model further includes a fault loss operator, which is calculated by weighting the fault loss function and the non-fault loss function with fault loss weights and non-fault loss weights, wherein the fault loss weights are greater than the non-fault loss weights.

[0011] In one possible embodiment, calculating the fault loss operator includes: Based on formula The fault loss operator is calculated. in, For the fault loss operator, The fault loss weight is... The non-fault loss weight, This is the output of the second decoding branch. for The normalization result, It is the Sigmoid activation function. Let be the fault loss function. is the non-fault loss function.

[0012] Secondly, this application provides a velocity model reconstruction device for underground aquifer gas storage reservoirs, comprising: The model acquisition module is used to acquire the trained geological exploration model. The loss function of the geological exploration model includes a directional gradient operator. The directional gradient operator is obtained by weighting the horizontal gradient kernel and the vertical gradient kernel with horizontal gradient weights and vertical gradient weights. The horizontal gradient weights are greater than the vertical gradient weights. The data acquisition module is used to acquire measured source data and measured detector data of the area to be explored. A velocity model reconstruction module is used to input the measured seismic source data and the measured detection data into the geological exploration model to obtain the velocity model of the underground aquifer gas reservoir output by the geological exploration model.

[0013] Thirdly, this application also provides an electronic device, including a memory and a processor, wherein, The memory is used to store programs; The processor, coupled to the memory, is used to execute the program stored in the memory to implement the steps in the underground aquifer gas storage velocity model reconstruction method described in any of the above implementations.

[0014] Fourthly, this application also provides a computer-readable storage medium for storing a computer-readable program or instruction, which, when executed by a processor, can implement the steps in the underground aquifer gas storage velocity model reconstruction method described in any of the above implementations.

[0015] The beneficial effects of this application are: Compared with related technologies, the underground aquifer gas reservoir velocity model reconstruction method, apparatus, electronic equipment, and computer-readable storage medium provided in this application involve source excitation in the area to be explored, obtaining measured source data and measured detection data during the source excitation process, and then directly acquiring the trained geological exploration model. The measured source data and measured detection data are then input into the geological exploration model to directly obtain the underground aquifer gas reservoir velocity model output by the geological exploration model, which is highly efficient. Furthermore, since the loss function of the geological exploration model sets horizontal gradient weights and vertical gradient weights to calculate the directional gradient operator by weighting the horizontal gradient kernel and the vertical gradient kernel, and setting the horizontal gradient weight to be greater than the vertical gradient weight, the physical characteristics of the vertical cutting of the strata by the fault can be utilized to force the geological exploration model to prioritize sharpening the vertical boundary, while suppressing the horizontal stratum artifacts caused by the vertical gradient constraint, thereby reducing "stratum artifacts" and improving the boundary clarity in the underground aquifer gas reservoir velocity model. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A schematic flowchart illustrating the velocity model reconstruction method for underground aquifer gas storage provided in this application embodiment; Figure 2 A schematic diagram of the process of training the geological exploration model in the underground aquifer gas storage velocity model reconstruction method provided in the embodiments of this application; Figure 3 A schematic diagram of the structure of the velocity model reconstruction device for underground aquifer gas storage provided in an embodiment of this application; Figure 4 This is a schematic diagram of the structure of an electronic device provided in one embodiment of this application. Detailed Implementation

[0018] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0019] In the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more. "And / or" describes the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone.

[0020] The terms "first," "second," etc., used in the embodiments of this application are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a technical feature defined with "first" or "second" may explicitly or implicitly include at least one of those features.

[0021] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0022] This application provides a method and apparatus for reconstructing a velocity model of an underground aquifer gas storage reservoir, as well as electronic equipment and a computer-readable storage medium, which are described below.

[0023] Please refer to Figure 1 The method for reconstructing the velocity model of an underground aquifer gas storage reservoir provided in this application includes: Step S101: Obtain the trained geological exploration model.

[0024] In this step, the geological exploration model is specifically obtained by training the basic model. Please refer to [the relevant documentation / reference]. Figure 2 Specifically, it includes the following steps: Step S201: Construct the basic model.

[0025] In this embodiment, the basic model specifically includes an input layer, an encoder, and a decoder.

[0026] The input layer receives input data and converts it into a tensor of a set size. In this embodiment, the input layer specifically converts the input data into a tensor of a specified size. The tensor.

[0027] The encoder specifically consists of a preprocessing convolutional layer and multiple attention residual modules. Each attention residual module contains a residual block and an attention module connected in series at the end of the residual block. The specific structure is: preprocessing convolutional layer → attention residual module 1 (residual block 1 → attention module 1) → attention residual module 2 (residual block 2 → attention module 2), ..., where the number of channels in the residual block gradually increases with the data processing flow. In this embodiment, an encoder consisting of one preprocessing convolutional layer and four attention residual modules is used as an example. That is, the encoder specifically includes four residual blocks, and the number of channels in the corresponding residual block gradually increases with the data processing flow, specifically according to... The trend increases layer by layer.

[0028] Furthermore, each residual block contains two convolutional layers, a LeakyReLU activation function, and batch normalization (BatchNorm), and is input into the attention module after the residuals are summed. Specifically, each residual block consists of two consecutive convolutional layers (e.g., ... The main path consists of convolutional layers. The first convolutional layer transforms the input feature map through channels and / or adjusts its spatial dimensions. Then, it passes through the LeakyReLU activation function to impart sparse activation capabilities and is then normalized by batch normalization to introduce nonlinearity and stabilize the feature distribution. The second convolutional layer then performs further feature extraction on the preceding output with the same kernel configuration as the first convolutional layer. Its calculation result is also first given sparse activation capabilities by the LeakyReLU activation function and then normalized by batch normalization. Finally, the final output of the entire double convolutional module is directly added to the original input feature map at the corresponding element positions, and the result of the residual addition is used as the input feature map to the attention module.

[0029] Each attention module specifically includes a channel attention module for suppressing channel noise and a spatial attention module for focusing tomography.

[0030] Specifically, the channel attention module includes a first global average pooling layer and a first global max pooling layer that run in parallel, a multilayer perceptron connected to the first global average pooling layer and the first global max pooling layer, and a first sigmoid activation structure connected to the multilayer perceptron.

[0031] After the residual block inputs the feature map into the attention module, it is first fed into two parallel pooling branches: the first global average pooling layer and the first global max pooling layer. The first global average pooling layer performs global average pooling to capture the overall distribution trend and global semantic information of each channel in the feature map, while the first global max pooling layer performs global max pooling to extract the most significant response intensity and local salient features within each channel. Both layers compress the original two-dimensional spatial information into a one-dimensional feature vector that retains only the channel dimension. Subsequently, these two feature vectors of the same size are fed into a multilayer perceptron consisting of fully connected layers for processing. The multilayer perceptron first maps the number of input channels to a low-dimensional representation with a compression ratio of r=16 to reduce the number of parameters and enhance nonlinear expressive power. Then, it expands the low-dimensional representation back to the original channel dimension, thereby achieving dimensionality reduction and then dimensionality increase transformation of the features. The outputs of the two parallel pooling branches of the first global average pooling layer and the first global max pooling layer are processed by the multilayer perceptron for dimensionality reduction and dimensionality increase, respectively. The multilayer perceptron outputs the vectors corresponding to the two parallel pooling branches. The two vectors are summed element-wise at the corresponding channel positions. The summation result is then normalized to between zero and one by the Sigmoid activation function to generate the final channel attention weight vector. This weight vector is used to recalibrate the response intensity of each channel of the original input feature map in a channel multiplicative manner to obtain the channel response feature map. The specific formula is expressed as: , Where F is the input feature map, AvgPool is global average pooling, MaxPool is global max pooling, MLP is multilayer perceptron, and σ is the sigmoid activation function. For element-wise multiplication, F' is the channel response feature map.

[0032] In this way, the channel attention module can enhance the effective signal channels that are highly relevant to the tomography task, while significantly suppressing channel noise containing background noise or lacking discriminative power, thus achieving the effect of suppressing channel noise.

[0033] The spatial attention module includes a second global average pooling layer and a second global max pooling layer that run in parallel, a vector concatenation layer connected to the second global average pooling layer and the second global max pooling layer, a convolutional layer connected to the vector concatenation layer, and a second sigmoid activation structure connected to the convolutional layer.

[0034] After the channel response feature maps are input into the spatial attention module, they are then fed into the second global average pooling layer and the second global max pooling layer, respectively. These layers perform global average pooling and global max pooling operations along the channel dimension. Specifically, they average the channel values ​​at each spatial location along all channel directions to obtain an overview of the spatial distribution of the overall response intensity. Simultaneously, they extract the maximum value from all channels at each spatial location to retain the most prominent local response information. This results in a spatial description map highlighting stable background features and a spatial description map emphasizing extreme response locations. These two single-channel description maps, with the same spatial size but different semantic emphases, are then concatenated along the channel dimension to form a dual-channel information flow. A spatial feature descriptor (size) is then sent to a larger (e.g., A two-dimensional convolutional kernel (of a certain size) is used for convolution operations. The large kernel provides a wide local receptive field, effectively integrating contextual information within neighboring spatial regions within a single sliding window. This smooths noise interference and enhances the continuity of large-scale structural boundaries. The single-channel feature map output by the convolution is then non-linearly mapped to the zero-to-one range using a sigmoid activation function. This results in a spatial weight mask map perfectly aligned with the original input feature map. Regions with values ​​approaching one correspond to faults (including fault breaks, anticline flanks, etc.), while regions with values ​​approaching zero correspond to non-faulted areas with gentle structures or uniform background lithology (areas other than faults). Applying the spatial weight mask element-wise to the channel response feature map output by the channel attention module yields the spatial response feature map output by the spatial attention module.

[0035] The formula is expressed as: , .in, f 7×7 It is a 7×7 convolution, and F” is the spatial response feature map.

[0036] In this way, the spatial attention module can spontaneously improve the representation accuracy of complex geological geometric shapes and suppress the response redundancy of irrelevant background areas, thereby achieving the effect of focusing on faults in fault identification and structural interpretation tasks.

[0037] In this application, the decoder comprises two branches: a first decoding branch and a second decoding branch. During model training, the first decoding branch is trained based on the output of the channel attention module; the second decoding branch is trained based on the output of the spatial attention module. That is, the first and second decoding branches are trained separately using different data, resulting in different outputs from each branch. Since the first decoding branch is trained based on the output of the channel attention module, its specific output includes a continuous single-channel velocity field. Single-channel velocity field The model represents the velocity field beneath the area to be explored, specifically including the seismic wave propagation velocities at various underground locations. The second decoding branch is trained based on the output of the spatial attention module; therefore, the second decoding branch specifically outputs a single-channel fault probability map. Single-channel fault probability map It represents the probability of fault structures occurring underground, specifically including the probability of faults at various underground locations.

[0038] The first decoding branch specifically includes a 3×3 convolutional layer, a batch normalization layer, a ReLU activation structure with leakage, a 1×1 convolutional layer, and a sigmoid activation structure. The second decoding branch specifically includes a 3×3 convolutional layer, a batch normalization layer, a ReLU activation structure with leakage, and a 1×1 convolutional layer.

[0039] Step S202: Obtain training data.

[0040] In this step, a batch generation algorithm is first written based on the MATLAB platform to construct a two-dimensional P-wave velocity basic model. The model mesh size is set to Grid spacing That is, the total horizontal length of the model is (301-1)×10m=3000m, and the total vertical depth is (201-1)×10m=2000m, which perfectly matches the research scenario of a gas storage facility in a deep aquifer of 2000–3000m. Basic Model The model comprises four basic geological layers: shallow background, tight caprock, aquifer reservoir, and basement. The shallow background consists of low-velocity overlying strata near the surface; the tight caprock is a high-velocity, low-permeability gas-bearing caprock; the aquifer reservoir is a medium-velocity target reservoir; and the basement is the high-velocity bedrock beneath the reservoir. Anticline structures are simulated using Gaussian functions to represent the traps required for natural gas accumulation. Lithological pinch-outs are introduced to simulate the horizontal thickness variations and lateral termination of the reservoir. Furthermore, zero to multiple faults are randomly incorporated into the model, covering various dip angles. normal faults and dip angles The reverse fault was used to simulate reservoir faulting and modification under tension and compression stress. In the area where the reservoir exists at the high point of the anticline structure, the medium velocity was replaced with a low velocity value to simulate the gas cap formed by natural gas accumulation. Taking advantage of the fact that the longitudinal wave velocity of natural gas is significantly lower than that of water and rock, a reservoir velocity anomaly that conforms to reality was constructed.

[0041] A Ricker wavelet with a dominant frequency of 20Hz was used as the seismic source. This wavelet is a zero-phase wavelet commonly used in seismic exploration, and its dominant frequency determines the vertical resolution of the seismic wave. Both the seismic source and detectors were deployed at a depth of 20m near the surface of the model, simulating conventional seismic acquisition methods involving ground excitation and reception. The shot spacing was set to 100m and the trace spacing to 20m, forming a two-dimensional observation system covering the entire model. The recording length was set to 1.8s to ensure complete reception of reflected wave signals within the target depth range. Based on the Devito framework in Python, the finite difference method of the acoustic equation was used to generate seismic records. By solving the two-dimensional acoustic equation, the propagation, reflection, and transmission processes of seismic waves in different velocity media were simulated, obtaining the synthetic seismic shot gather data corresponding to each shot. (including the basic model) Seismic source data and reflection collection holes).

[0042] Furthermore, the generated synthetic seismic shot gather data can also be processed. Time-varying spherical diffusion compensation is performed to correct the energy attenuation of seismic waves as the propagation distance increases, making the energy distribution of reflected waves at different depths more balanced. At the same time, a bilinear interpolation algorithm is used for global resampling to uniformly adjust all data to the same grid size and sampling rate to meet the consistency requirements of the basic model input.

[0043] Step S203: Train the basic model based on the training data to obtain the trained geological exploration model.

[0044] In this step, training the base model based on the training data first requires constructing the loss function based on the base model structure provided in step S201 above.

[0045] In this embodiment, the loss function specifically includes a directional gradient operator, which is calculated by weighting the horizontal gradient weights and vertical gradient weights onto the horizontal gradient kernel and vertical gradient kernel, with the horizontal gradient weights being greater than the vertical gradient weights. The specific formula is as follows: ; ; ; The directional gradient operator is calculated based on the above set of formulas.

[0046] in, For directional gradient operators, For horizontal gradient weights, For horizontal gradient kernel, For vertical gradient kernel, For vertical gradient weights, This is the output of the first decoding branch. for The gradient operator in the horizontal direction, for The gradient operator in the vertical direction, For convolution operations, This refers to the velocity base model in step S202 mentioned above.

[0047] Furthermore, the loss function of the geological exploration model also includes a fault loss operator, which is calculated by weighting the fault loss function and the non-fault loss function with fault loss weights and non-fault loss weights. The fault loss weights are greater than the non-fault loss weights.

[0048] The specific formula is expressed as follows: Based on formula The fault loss operator is calculated. in, For fault loss operator, For fault loss weights, For non-fault loss weights, This is the output of the second decoding branch. for The normalization result, It is the Sigmoid activation function. For fault loss function, This is the non-fault loss function.

[0049] In summary, the loss function formula for the basic model provided in this application embodiment is expressed as follows: .in, For loss function, Based on the loss operator, The weights of the basic loss operator, These are the weights of the directional gradient operator. The weight values ​​for the fault loss operator.

[0050] Step S102: Obtain measured seismic source data and measured detection data of the area to be explored.

[0051] In this step, a blasting test is conducted on the area to be explored. The first predetermined location is used as the seismic source, and a micro-blast is performed at the seismic source. The seismic wave data related to the micro-blast is used as the measured seismic source data. At the same time, the second predetermined location is used as the geophone point, and the seismic waves reflected by the underground structure are detected at the geophone point and used as the measured geophone data.

[0052] Step S103: Input the measured source data and measured detection data into the geological exploration model to obtain the underground aquifer gas reservoir velocity model output by the geological exploration model.

[0053] Compared with related technologies, the underground aquifer gas reservoir velocity model reconstruction method provided in this application involves source excitation in the area to be explored, obtaining measured source data and measured detection data during the source excitation process, and then directly acquiring the trained geological exploration model. The measured source data and measured detection data are then input into the geological exploration model to directly obtain the underground aquifer gas reservoir velocity model output by the geological exploration model, which is highly efficient. Furthermore, since the loss function of the geological exploration model sets horizontal gradient weights and vertical gradient weights to calculate the directional gradient operator by weighting the horizontal gradient kernel and the vertical gradient kernel, and setting the horizontal gradient weight to be greater than the vertical gradient weight, the physical characteristics of the fault longitudinally cutting through the strata, which causes abrupt changes in the horizontal gradient, can be utilized to force the geological exploration model to prioritize sharpening the vertical boundary, while suppressing the horizontal stratum artifacts caused by the vertical gradient constraint, thereby reducing "stratum artifacts" and improving the boundary clarity in the underground aquifer gas reservoir velocity model.

[0054] To better implement the velocity model reconstruction method for underground aquifer gas storage reservoirs in this application embodiment, based on the underground aquifer gas storage reservoir velocity model reconstruction method, correspondingly, as follows: Figure 3 As shown in the embodiment of this application, a velocity model reconstruction device for an underground aquifer gas storage reservoir is also provided. The underground aquifer gas storage reservoir velocity model reconstruction device includes: The model acquisition module 301 is used to acquire the trained geological exploration model. The loss function of the geological exploration model includes a directional gradient operator. The directional gradient operator is calculated by weighting the horizontal gradient weight and the vertical gradient weight on the horizontal gradient kernel and the vertical gradient kernel. The horizontal gradient weight is greater than the vertical gradient weight. Data acquisition module 302 is used to acquire measured source data and measured detection data of the area to be explored; The velocity model reconstruction module 303 is used to input measured source data and measured detection data into the geological exploration model to obtain the velocity model of the underground aquifer gas reservoir output by the geological exploration model. The underground aquifer gas reservoir velocity model reconstruction device provided in the above embodiment can realize the technical solution described in the above underground aquifer gas reservoir velocity model reconstruction method embodiment. The specific implementation principle of each module or unit can be referred to the corresponding content in the above underground aquifer gas reservoir velocity model reconstruction method embodiment, which will not be repeated here.

[0055] Please refer to Figure 4 This application also provides an electronic device 400. The electronic device 400 includes a processor 401, a memory 402, and a display 403. Figure 4 Only some components of the electronic device 400 are shown, but it should be understood that it is not required to implement all the components shown, and more or fewer components may be implemented instead.

[0056] In some embodiments, processor 401 may be a central processing unit (CPU), microprocessor, or other data processing chip, used to run program code stored in memory 402 or process data, such as the underground aquifer gas storage velocity model reconstruction method in this application.

[0057] In some embodiments, processor 401 may be a single server or a group of servers. The server group may be centralized or distributed. In some embodiments, processor 401 may be local or remote. In some embodiments, processor 401 may be implemented on a cloud platform. In one embodiment, the cloud platform may include a private cloud, public cloud, hybrid cloud, community cloud, distributed cloud, intranet, multi-cloud, etc., or any combination thereof.

[0058] In some embodiments, memory 402 may be an internal storage unit of electronic device 400, such as a hard disk or memory of electronic device 400. In other embodiments, memory 402 may also be an external storage device of electronic device 400, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. equipped on electronic device 400.

[0059] Furthermore, the memory 402 may include both internal storage units of the electronic device 400 and external storage devices. The memory 402 is used to store application software and various types of data installed on the electronic device 400.

[0060] In some embodiments, display 403 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. Display 403 is used to display information from electronic device 400 and to display a visual user interface. Components 401-403 of electronic device 400 communicate with each other via a system bus.

[0061] In one embodiment, when the processor 401 executes the underground aquifer gas storage velocity model reconstruction program in the memory 402, the following steps can be implemented: Obtain the trained geological exploration model. The loss function of the geological exploration model includes a directional gradient operator. The directional gradient operator is calculated by weighting the horizontal gradient weight and the vertical gradient weight with respect to the horizontal gradient kernel and the vertical gradient kernel. The horizontal gradient weight is greater than the vertical gradient weight. Obtain measured source data and measured detector data of the area to be explored; By inputting the measured seismic source data and measured detection data into the geological exploration model, the velocity model of the underground aquifer gas reservoir output by the geological exploration model is obtained.

[0062] It should be understood that when the processor 401 executes the underground aquifer gas storage velocity model reconstruction program in the memory 402, in addition to the functions mentioned above, it can also perform other functions, as detailed in the description of the corresponding method embodiments above.

[0063] Furthermore, this application does not specifically limit the type of electronic device 400 mentioned in the embodiments. Electronic device 400 can be a mobile phone, tablet computer, personal digital assistant (PDA), wearable device, laptop computer, or other portable electronic device. Exemplary embodiments of portable electronic devices include, but are not limited to, portable electronic devices running iOS, Android, Microsoft, or other operating systems. The aforementioned portable electronic device can also be other portable electronic devices, such as a laptop computer with a touch-sensitive surface (e.g., a touch panel). It should also be understood that in some other embodiments of this application, electronic device 400 may not be a portable electronic device, but rather a desktop computer with a touch-sensitive surface (e.g., a touch panel).

[0064] Accordingly, this application 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 underground aquifer gas storage velocity model reconstruction method provided in the above-described method embodiments.

[0065] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware (such as a processor, controller, etc.), and the computer program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0066] The above provides a detailed description of the method, apparatus, electronic equipment, and storage medium for reconstructing the velocity model of an underground aquifer gas storage tank provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and its core ideas. At the same time, those skilled in the art will recognize that there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for reconstructing a velocity model of an underground aquifer gas storage reservoir, characterized in that, include: Obtain the trained geological exploration model. The loss function of the geological exploration model includes a directional gradient operator. The directional gradient operator is calculated by weighting the horizontal gradient kernel and the vertical gradient kernel with horizontal gradient weights and vertical gradient weights. The horizontal gradient weights are greater than the vertical gradient weights. Obtain measured source data and measured detector data of the area to be explored; The measured seismic source data and the measured detection data are input into the geological exploration model to obtain the underground aquifer gas reservoir velocity model output by the geological exploration model.

2. The method for reconstructing the velocity model of an underground aquifer gas storage reservoir according to claim 1, characterized in that, The process of obtaining the trained geological exploration model includes: A basic model is constructed, which includes an encoder and a decoder. The encoder includes multiple attention residual modules. Each attention residual module includes a residual block and an attention module connected in series at the end of the residual block. The attention module includes a channel attention module for suppressing channel noise and a spatial attention module for focusing tomography. Obtain training data; The basic model is trained based on the training data to obtain the trained geological exploration model.

3. The method for reconstructing the velocity model of an underground aquifer gas storage reservoir according to claim 2, characterized in that, The channel attention module includes a first global average pooling layer and a first global max pooling layer that are parallel to each other, a multilayer perceptron connected to the first global average pooling layer and the first global max pooling layer, and a first sigmoid activation structure connected to the multilayer perceptron. The spatial attention module includes a second global average pooling layer and a second global max pooling layer that run in parallel, a vector concatenation layer connected to the second global average pooling layer and the second global max pooling layer, a convolutional layer connected to the vector concatenation layer, and a second sigmoid activation structure connected to the convolutional layer.

4. The method for reconstructing the velocity model of an underground aquifer gas storage reservoir according to claim 3, characterized in that, The decoder includes a first decoding branch and a second decoding branch, and training the base model based on the training data includes: The first decoding branch is trained based on the output of the channel attention module; The second decoding branch is trained based on the output of the spatial attention module.

5. The method for reconstructing the velocity model of an underground aquifer gas storage reservoir according to claim 4, characterized in that, The directional gradient operator is calculated as follows: Based on formula group ; ； The directional gradient operator is calculated. in, For the directional gradient operator, The horizontal gradient weights, The horizontal gradient kernel, The vertical gradient kernel, The vertical gradient weights are... This is the output of the first decoding branch. for The gradient operator in the horizontal direction, for The gradient operator in the vertical direction, For convolution operations, This is the velocity base model in the training data.

6. The method for reconstructing the velocity model of an underground aquifer gas storage reservoir according to claim 4, characterized in that, The loss function of the geological exploration model also includes a fault loss operator, which is obtained by weighting the fault loss function and the non-fault loss function with fault loss weight and non-fault loss weight, wherein the fault loss weight is greater than the non-fault loss weight.

7. The method for reconstructing the velocity model of an underground aquifer gas storage reservoir according to claim 6, characterized in that, The calculation of the fault loss operator includes: Based on formula The fault loss operator is calculated. in, For the fault loss operator, The fault loss weight is... The non-fault loss weight, This is the output of the second decoding branch. for The normalization result, It is the Sigmoid activation function. Let be the fault loss function. is the non-fault loss function.

8. A device for reconstructing a velocity model of an underground aquifer gas storage tank, characterized in that, include: The model acquisition module is used to acquire the trained geological exploration model. The loss function of the geological exploration model includes a directional gradient operator. The directional gradient operator is obtained by weighting the horizontal gradient kernel and the vertical gradient kernel with horizontal gradient weights and vertical gradient weights. The horizontal gradient weights are greater than the vertical gradient weights. The data acquisition module is used to acquire measured source data and measured detector data of the area to be explored. A velocity model reconstruction module is used to input the measured seismic source data and the measured detection data into the geological exploration model to obtain the velocity model of the underground aquifer gas reservoir output by the geological exploration model.

9. An electronic device, characterized in that, Including memory and processor, among which, The memory is used to store programs; The processor, coupled to the memory, is used to execute the program stored in the memory to implement the steps in the underground aquifer gas storage velocity model reconstruction method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, Used to store computer-readable programs or instructions, which, when executed by a processor, can implement the steps in the underground aquifer gas storage velocity model reconstruction method according to any one of claims 1 to 7.

Images