A Method and System for Identifying Seismic Data Based on Large-Model Free Gas Structures

Abstract

This invention relates to the field of seismic data detection technology, and more particularly to a method and system for identifying seismic data based on large-scale models of free gas structures. The method includes the following steps: acquiring raw seismic data; performing a short-time Fourier transform on the raw seismic data to obtain time-frequency domain seismic data; performing a Hilbert-Huang transform on the time-frequency domain seismic data to obtain enhanced seismic feature data; extracting features from the enhanced seismic feature data to obtain a multi-scale seismic feature map; performing semantic segmentation on the enhanced seismic feature data to obtain a free gas structure mask; performing uncertainty assessment on the free gas structure mask to obtain a confidence heatmap; and performing probabilistic prediction on the confidence heatmap to obtain a low-confidence region marker map. This invention overcomes the difficulty of accurately extracting effective signals due to the complexity of seismic data in traditional methods.

Description

Technical Field

[0001] This invention relates to the field of seismic data detection technology, and in particular to a method and system for identifying seismic data based on large-scale free gas structures. Background Technology

[0002] In the field of oil and gas exploration, seismic data interpretation is a crucial step in finding oil and gas reservoirs. Free gas structures are one of the important characteristics of oil and gas reservoirs, and their accurate identification is of great significance for the assessment and development of oil and gas resources. However, in the actual process of seismic data processing and interpretation, the identification of free gas structures faces many challenges. Traditional seismic data processing methods mainly rely on human experience, identifying free gas structures by subjectively judging the morphology, amplitude, frequency, and other characteristics of seismic waveforms. This method is not only time-consuming and labor-intensive, but also, due to the complexity and diversity of seismic data, the differences in experience among different interpreters can lead to inconsistencies in interpretation results, thus affecting the accuracy and efficiency of oil and gas exploration.

[0003] In practical seismic data interpretation, seismic data is often interfered with by factors such as noise superposition, making the effective signals in the seismic data blurred and increasing the difficulty of identifying free gas structures. Under such circumstances, traditional seismic data processing methods struggle to accurately extract the effective signals related to free gas structures, thus affecting the accurate identification of free gas structures. Summary of the Invention

[0004] Therefore, it is necessary for the present invention to provide a method and system for identifying free gas structure seismic data based on a large model, in order to solve at least one of the above-mentioned technical problems.

[0005] To achieve the above objectives, a method for identifying free gas tectonic seismic data based on a large model includes the following steps:

[0006] Step S1: Acquire raw seismic data; perform short-time Fourier transform on the raw seismic data to obtain time-frequency domain seismic data; perform Hilbert-Huang transform on the time-frequency domain seismic data to obtain enhanced seismic feature data;

[0007] Step S2: Extract features from the enhanced seismic feature data to obtain a multi-scale seismic feature map; perform semantic segmentation on the enhanced seismic feature data to obtain a free gas structure mask;

[0008] Step S3: Perform uncertainty assessment on the free gas structure mask to obtain a confidence heatmap; perform probability prediction on the confidence heatmap to obtain a low-confidence area marker map; perform data augmentation on the low-confidence area marker map to obtain corrected seismic feature data; input the corrected seismic feature data into a pre-trained visual Transformer large model for feature extraction to obtain an optimized free gas structure mask.

[0009] Step S4: Reconstruct the three-dimensional space based on the optimized free gas structure mask to obtain the initial free gas structure model; perform survey line interpolation on the optimized free gas structure mask to obtain the three-dimensional free gas structure point cloud; perform spatial fitting on the three-dimensional free gas structure point cloud to obtain continuous free gas structure data.

[0010] Step S5: Interactively interpret the continuous free gas structure data to obtain a three-dimensional model of the free gas structure; perform visualization rendering based on the three-dimensional model of the free gas structure to obtain a three-dimensional model of the free gas structure with probability labels.

[0011] Preferably, the present invention also provides a large-model-based free gas structure seismic data identification system for performing the large-model-based free gas structure seismic data identification method described above. The large-model-based free gas structure seismic data identification system includes:

[0012] The data preprocessing module is used to acquire raw seismic data; perform short-time Fourier transform on the raw seismic data to obtain time-frequency domain seismic data; and perform Hilbert-Huang transform on the time-frequency domain seismic data to obtain enhanced seismic feature data.

[0013] The feature extraction module is used to extract features from the enhanced seismic feature data to obtain a multi-scale seismic feature map; and to perform semantic segmentation on the enhanced seismic feature data to obtain a free gas structure mask.

[0014] The optimization iteration module is used to evaluate the uncertainty of the free gas structure mask and obtain a confidence heatmap; perform probability prediction on the confidence heatmap to obtain a low-confidence area marker map; perform data augmentation on the low-confidence area marker map to obtain corrected seismic feature data; input the corrected seismic feature data into a pre-trained visual Transformer large model for feature extraction to obtain the optimized free gas structure mask.

[0015] The 3D modeling module is used to reconstruct the 3D space based on the optimized free gas structure mask to obtain the initial free gas structure model; perform survey line interpolation on the optimized free gas structure mask to obtain the 3D free gas structure point cloud; and perform spatial fitting on the 3D free gas structure point cloud to obtain continuous free gas structure data.

[0016] The visualization module is used to interactively interpret continuous free gas structure data to obtain a three-dimensional model of the free gas structure; based on the three-dimensional model of the free gas structure, a visualization rendering is performed to obtain a three-dimensional model of the free gas structure with probability labels.

[0017] The beneficial effects of this invention are as follows:

[0018] On the one hand, using specific signal processing techniques to preprocess the raw seismic data can significantly improve the quality of the seismic data, making the originally unclear effective signals stand out, laying a solid foundation for subsequent free gas structure identification work, and overcoming the problem that traditional methods are difficult to accurately extract effective signals due to the complexity of seismic data.

[0019] On the other hand, by using advanced pre-trained large models to deeply mine and analyze the features in seismic data, free gas structure masks can be identified in a more efficient and accurate way. Furthermore, by effectively evaluating and correcting the uncertainty of the identification results, the accuracy of free gas structure identification can be greatly improved, overcoming the drawbacks of traditional manual experience judgment, such as large subjective differences and unstable results.

[0020] On the other hand, by performing a series of complex spatial processing and interactive interpretation operations on the optimized data, an intuitive three-dimensional model of free gas structure with probability labels is finally constructed, realizing the visualization and quantitative analysis of free gas structure. Attached Figure Description

[0021] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description taken in conjunction with the accompanying drawings:

[0022] Figure 1 A flowchart illustrating the steps of a large-model-based free gas structure seismic data identification method is shown.

[0023] Figure 2 A detailed flowchart of step S23 of one embodiment is shown.

[0024] Figure 3 A diagram illustrating the dominant frequency distribution characteristics of seismic data in one embodiment is shown. Detailed Implementation

[0025] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0026] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0027] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0028] To achieve the above objectives, please refer to Figures 1 to 3 This invention provides a method for identifying free gas tectonic seismic data based on a large model, comprising the following steps:

[0029] Step S1: Acquire raw seismic data; perform short-time Fourier transform on the raw seismic data to obtain time-frequency domain seismic data; perform Hilbert-Huang transform on the time-frequency domain seismic data to obtain enhanced seismic feature data;

[0030] Step S2: Extract features from the enhanced seismic feature data to obtain a multi-scale seismic feature map; perform semantic segmentation on the enhanced seismic feature data to obtain a free gas structure mask;

[0031] Step S3: Perform uncertainty assessment on the free gas structure mask to obtain a confidence heatmap; perform probability prediction on the confidence heatmap to obtain a low-confidence area marker map; perform data augmentation on the low-confidence area marker map to obtain corrected seismic feature data; input the corrected seismic feature data into a pre-trained visual Transformer large model for feature extraction to obtain an optimized free gas structure mask.

[0032] Step S4: Reconstruct the three-dimensional space based on the optimized free gas structure mask to obtain the initial free gas structure model; perform survey line interpolation on the optimized free gas structure mask to obtain the three-dimensional free gas structure point cloud; perform spatial fitting on the three-dimensional free gas structure point cloud to obtain continuous free gas structure data.

[0033] Step S5: Interactively interpret the continuous free gas structure data to obtain a three-dimensional model of the free gas structure; perform visualization rendering based on the three-dimensional model of the free gas structure to obtain a three-dimensional model of the free gas structure with probability labels.

[0034] Preferably, step S1 includes the following steps:

[0035] Step S11: Acquire raw seismic data using seismic acquisition equipment;

[0036] In this embodiment of the invention, a marine seismic acquisition system is used, equipped with a highly sensitive seismic detector and a precise positioning device. In a specific exploration area, raw seismic data covering the target area is continuously acquired using a seismic acquisition device towed at the stern of a ship, at a set sampling rate (e.g., 250 Hz) and sampling interval (e.g., 4 ms). This data contains information on the propagation of seismic waves in different strata.

[0037] Step S12: Perform wavelet threshold denoising on the original seismic data to obtain denoised seismic data;

[0038] In this embodiment of the invention, a wavelet thresholding denoising method is used for preprocessing. Specifically, the Daubechies wavelet (db4) is selected as the wavelet basis. By performing multi-scale wavelet decomposition on the original seismic data, the signal is decomposed into detail coefficients and approximation coefficients of different frequency components. According to Donoho's general threshold theorem, a suitable threshold is calculated (e.g., based on the standard deviation of the data and the square root of the number of sampling points), and soft thresholding is performed on the detail coefficients to remove noise components. Finally, wavelet reconstruction is used to obtain the denoised seismic data after noise removal.

[0039] Step S13: Divide the denoised seismic data into time windows to obtain segmented seismic data;

[0040] Step S14: Perform short-time Fourier transform on the segmented seismic data to obtain time-frequency domain seismic data;

[0041] In this embodiment of the invention, for the filtered and labeled valid segmented seismic data, the Short Time Fourier Transform (STFT) technique is used to transform it to the time-frequency domain. A Hanning window is selected as the window function, with a window length set to 128 sampling points to accommodate the signal characteristics of the seismic data. By performing a Fourier transform on the seismic data within each time window, the signal is decomposed from the time domain into time-frequency spectrograms of different frequency components.

[0042] Step S15: Perform Hilbert-Huang transform on the time-frequency domain seismic data to obtain instantaneous frequency characteristic data;

[0043] In this embodiment of the invention, after obtaining the time-frequency domain seismic data, the Empirical Mode Decomposition (EMD) component of the Hilbert-Huang Transform (HHT) is further used to process the time-frequency domain seismic data. In practice, the EMD toolbox in MATLAB software is used to decompose the time-frequency domain seismic data point by point. Through multiple iterative filtering, the data is decomposed into multiple IMF components, each representing the characteristics of the seismic signal at different frequencies and time scales. Furthermore, a Hilbert Transform is performed on each IMF component to obtain its instantaneous frequency characteristic data. These instantaneous frequency characteristic data can more accurately describe the propagation characteristics of seismic waves in different strata.

[0044] Step S16: Normalize the instantaneous frequency characteristic data to obtain enhanced seismic characteristic data.

[0045] In this embodiment of the invention, a maximum-minimum normalization method is employed. By calculating the maximum and minimum values ​​in the instantaneous frequency characteristic data, the values ​​of all data points are adjusted to the range of 0 to 1. Specifically, a corresponding normalization script is written in the seismic data processing software to process each instantaneous frequency characteristic data point point by point. After normalization, the resulting enhanced seismic characteristic data is numerically more stable, and the differences between different characteristics are more pronounced.

[0046] Preferably, step S13 includes the following steps:

[0047] Step S130: Identify seismic wave velocity in the denoised seismic data to obtain the seismic wave propagation velocity;

[0048] In this embodiment of the invention, seismic wave velocity identification is performed using the reflection wave time difference method. Specifically, the `semblance` analysis tool in the seismic data processing software SeisSpace is used. A seismic trace with a clear reflection interface is selected as a reference. Two arrival times of reflected waves are chosen along this trace, and the time difference between these two times is calculated. Based on known stratigraphic depth information, the actual distance between the two reflection interfaces is calculated. The average propagation velocity of the seismic wave along this path is calculated using the ratio of distance to time; for example, the calculated result is 2500 m / s.

[0049] Step S131: Calculate the dominant frequency based on the seismic wave propagation velocity to obtain the seismic wave dominant frequency characteristic data; set the time window parameters according to the seismic wave dominant frequency characteristic data to obtain the time window length configuration data;

[0050] In this embodiment of the invention, after determining the propagation speed of the seismic wave to be 2500 m / s, the dominant frequency is further calculated. The FFT (Fast Fourier Transform) function in MATLAB software is used to perform spectral analysis on the denoised seismic data. Specifically, a segment of seismic data with 1024 sampling points is selected, and a Fast Fourier Transform is performed on it to obtain its spectrum. By analyzing the peak positions of the spectrum, the dominant frequency of the seismic data is determined to be 35 Hz. Based on this dominant frequency value, time window parameters are set. To ensure that each time window contains at least one complete dominant frequency cycle, the calculated time window length is 1 / 35 second, or approximately 28.57 milliseconds.

[0051] Step S132: Calculate the overlap rate of the denoised seismic data based on the time window length configuration data to obtain the time window overlap parameters;

[0052] In this embodiment of the invention, after determining the time window length to be 28.57 milliseconds, the overlap rate of the time windows needs to be calculated. A 50% overlap rate is chosen to ensure data continuity and feature integrity. Specifically, the NumPy library in the Python programming language is used to process the seismic data. A simple script is written to calculate the start and end positions of each time window, ensuring a 50% data overlap between adjacent time windows. For example, the first time window starts at 0 milliseconds and ends at 28.57 milliseconds; the second time window starts at 14.285 milliseconds and ends at 42.855 milliseconds. In this way, each time window includes a portion of the data from the previous time window, thus ensuring the continuity of the seismic signal.

[0053] Step S133: Use the time window overlap parameter to perform sliding time window segmentation on the denoised seismic data to obtain the initial segmented seismic data;

[0054] In this embodiment of the invention, after determining the time window length to be 28.57 milliseconds and the overlap rate to be 50%, the `window` command in the SeismicUnix (SU) software is used to perform sliding time window segmentation on the denoised seismic data. Specifically, the entire seismic data sequence is segmented according to the set time window length and overlap rate. For example, for a 10-second segment of seismic data, a 28.57-millisecond time window is extracted starting from 0 milliseconds, and the time window is slid in a step of 14.285 milliseconds to extract the next time window segment until the entire seismic data sequence is divided into multiple time window segments. Through the above steps, initial segmented seismic data is obtained, and each segment contains information about the seismic signal within a specific time interval.

[0055] Step S134: Perform energy equalization processing on the initial segmented seismic data to obtain equalized segmented seismic data;

[0056] In this embodiment of the invention, after obtaining the initial segmented seismic data, the Gain function in the Open dTect software is used to perform energy equalization processing on each segment. Specifically, the Automatic Gain Control (AGC) method is selected. This method can automatically adjust the energy of the seismic signal within each time window, making its energy distribution more uniform. By setting the AGC window length to 28.57 milliseconds (consistent with the time window length), the software automatically calculates the energy of the signal within each time window and adjusts it to a preset average energy level. For example, if the signal energy in a time window is low, AGC will increase its amplitude; conversely, if the signal energy is high, AGC will decrease its amplitude. After energy equalization processing, the obtained balanced segmented seismic data is more energy-stable.

[0057] Step S135: Based on a preset signal-to-noise ratio threshold, perform quality screening on the balanced segmented seismic data to obtain effective segmented seismic data;

[0058] In this embodiment of the invention, a preset signal-to-noise ratio (SNR) threshold of 3 is set, meaning that the signal energy must be at least three times the noise energy to be considered a valid data segment. Specifically, the Signal-to-Noise Ratio (SNR) analysis tool in Pro MAX software is used. By calculating the SNR of each segment of seismic data, segments with an SNR lower than 3 are marked as invalid and discarded. For example, in 100 data segments, 10 segments with an SNR lower than 3 will be excluded. Through these steps, valid segmented seismic data is obtained.

[0059] Step S136: Time-stamp the effective segmented seismic data to obtain segmented seismic data.

[0060] In this embodiment of the invention, after selecting valid segmented seismic data, the TimeMark function in SeisWorks software is used to time-stamp these data. Specifically, a time tag is added to each segment based on its start and end times in the original seismic data. For example, the start time of the first valid segmented seismic data is 0 milliseconds, and the end time is 28.57 milliseconds; the start time of the second valid segmented seismic data is 14.285 milliseconds, and the end time is 42.855 milliseconds, and so on. Through the above steps, each segmented seismic data is assigned specific time information.

[0061] Preferably, step S2 includes the following steps:

[0062] Step S21: Expand the channel dimensions of the enhanced seismic feature data to obtain three-dimensional seismic feature data;

[0063] In this embodiment of the invention, when processing enhanced seismic feature data, the Python programming language and the NumPy library are used to expand the data in terms of channel dimensions. Specifically, two-dimensional seismic feature data, which contains the characteristics of the seismic signal at different times and frequencies, is loaded into memory. NumPy's array manipulation capabilities are then used to expand the two-dimensional data into three-dimensional data. For example, the original two-dimensional data has a shape of (1000, 50), representing 1000 time steps and 50 frequency features. By adding a channel dimension, it is expanded into a three-dimensional array of shape (1, 1000, 50), where the first dimension represents the number of channels.

[0064] Step S22: Use a pre-trained visual Transformer model to extract features from the 3D seismic feature data to obtain a primary seismic feature map;

[0065] In this embodiment of the invention, after obtaining the 3D seismic feature data, a pre-trained Vision Transformer model is used for feature extraction. The PyTorch deep learning framework is selected, and a pre-trained Vision Transformer (ViT) model is loaded. Specifically, the 3D seismic feature data is input into the ViT model. Through its multi-head self-attention mechanism, the ViT model can effectively capture long-range dependencies and local features in the data. The model's output is a higher-level feature representation, namely a primary seismic feature map. For example, after processing the input 3D seismic feature data by the ViT model, the output primary seismic feature map has a shape of (1, 64, 25), representing 64 feature channels and 25 feature points. These primary seismic feature maps contain high-level features of the seismic signal.

[0066] Step S23: Enhance the primary seismic feature map to obtain a multi-scale seismic feature map;

[0067] Step S24: Upsample and reconstruct the multi-scale seismic feature map to obtain a semantic segmentation probability map;

[0068] In this embodiment of the invention, the `torch.nn.functional.interpolate` function in PyTorch is used to upsample and reconstruct multi-scale seismic feature maps. Specifically, bilinear interpolation is selected as the upsampling method. For example, the original multi-scale seismic feature map has a shape of (1, 64, 25), indicating 64 feature channels and 25 feature points. Bilinear interpolation upsamples the feature map to the same resolution as the input data, for example, to (1, 64, 1000). This results in a semantic segmentation probability map that more accurately reflects the free gas structure information in the seismic signal.

[0069] Step S25: Binarize the semantic segmentation probability map based on a preset threshold to obtain a free gas structure mask.

[0070] In this embodiment of the invention, the `cv2.threshold` function from the OpenCV library is used to binarize the semantic segmentation probability map. Specifically, a preset threshold, such as 0.5, is used. Each pixel value in the semantic segmentation probability map is compared to this threshold; pixel values ​​greater than the threshold are set to 1 (representing a free gas structure region), and pixel values ​​less than or equal to the threshold are set to 0 (representing a non-free gas structure region). For example, a pixel in the semantic segmentation probability map has a value of 0.6, which is greater than the threshold of 0.5; therefore, this pixel is marked as 1 in the free gas structure mask. Through the above steps, a clear free gas structure mask is obtained.

[0071] Preferably, step S23 includes the following steps:

[0072] Step S231: Perform pyramid pooling on the primary seismic feature map to obtain multi-scale contextual feature data;

[0073] In this embodiment of the invention, when processing the primary seismic feature map, the `torch.nn.functional.adaptive_avg_pool2d` function in the PyTorch framework is used to implement pyramid pooling. Specifically, four different pooling scales are selected: 1×1, 2×2, 4×4, and 8×8. For example, the shape of the primary seismic feature map is (1, 64, 25, 25), indicating 64 feature channels and 25×25 feature points. By applying these four different scales of adaptive average pooling, feature maps of four different scales are obtained. These feature maps capture contextual information of different ranges, thus obtaining multi-scale contextual feature data.

[0074] Step S232: Calculate feature weights for multi-scale contextual feature data using a spatial attention mechanism to obtain a spatial attention weight map;

[0075] In this embodiment of the invention, after obtaining multi-scale contextual feature data, a spatial attention mechanism is used to calculate feature weights for these feature data. The spatial attention mechanism is implemented using the `torch.nn.Conv2d` and `torch.nn.Sigmoid` functions in the PyTorch framework. Specifically, the multi-scale contextual feature data is passed through a convolutional layer with one output channel to generate a spatial attention map. The values ​​of this spatial attention map are normalized to between 0 and 1 using the Sigmoid activation function to obtain a spatial attention weight map. For example, after convolution and activation function processing, the resulting spatial attention weight map has the shape (1,1,25,25), where the value at each position represents the spatial importance weight at that position.

[0076] Step S233: Recalibrate the primary seismic feature map based on the spatial attention weight map to obtain the spatially enhanced feature map;

[0077] In this embodiment of the invention, the `torch.mul` function in the PyTorch framework is used to recalibrate the primary seismic feature map. Specifically, the primary seismic feature map is multiplied element-wise with the spatial attention weight map. For example, the shape of the primary seismic feature map is (1, 64, 25, 25), and the shape of the spatial attention weight map is (1, 1, 25, 25). Through element-wise multiplication, the resulting spatially enhanced feature map still has the shape (1, 64, 25, 25). This operation reweights the feature points in each feature channel according to their spatial importance, thus obtaining the spatially enhanced feature map.

[0078] Step S234: Perform channel correlation analysis on the spatial augmentation feature map using the channel attention mechanism to obtain the channel attention weight map;

[0079] In this embodiment of the invention, the `torch.nn.AdaptiveAvgPool2d` and `torch.nn.Linear` functions in the PyTorch framework are used to implement the channel attention mechanism. Specifically, a global average pooling operation is performed on the spatial augmented feature map, compressing the feature map of each feature channel into a scalar value to obtain a channel descriptor. These channel descriptors are then processed by a fully connected layer to obtain the channel attention weights. For example, if the shape of the spatial augmented feature map is (1, 64, 25, 25), the channel descriptor obtained after global average pooling will have a shape of (1, 64). After processing by the fully connected layer, the resulting channel attention weights will also have a shape of (1, 64). These channel attention weights represent the importance of each channel.

[0080] Step S235: Perform channel feature enhancement on the spatial enhancement feature map based on the channel attention weight map to obtain a multi-scale seismic feature map.

[0081] In this embodiment of the invention, the `torch.mul` function in the PyTorch framework is used to enhance the channel features of the spatial augmentation feature map. Specifically, each channel of the spatial augmentation feature map is multiplied element-wise with its corresponding channel attention weight. For example, if the shape of the spatial augmentation feature map is (1, 64, 25, 25) and the shape of the channel attention weights is (1, 64), the resulting multi-scale seismic feature map will still have the shape (1, 64, 25, 25) after element-wise multiplication. This operation reweights the features of each channel according to their importance, thus obtaining the multi-scale seismic feature map.

[0082] Preferably, step S3 includes the following steps:

[0083] Step S31: Perform multiple predictions on the free gas structure mask based on the Monte Carlo method to obtain prediction variance data;

[0084] Step S32: Perform Gaussian smoothing on the predicted variance data to obtain a confidence heatmap;

[0085] In this embodiment of the invention, the `cv2.GaussianBlur` function from the OpenCV library is used to perform Gaussian smoothing on the predicted variance data. Specifically, a Gaussian kernel size of (5,5) and a standard deviation of 1 are selected. For example, if the shape of the predicted variance data is (1,1,256,256), after applying Gaussian blurring, smoothed predicted variance data, i.e., a confidence heatmap, is obtained. This heatmap can visually display the confidence levels in different regions.

[0086] Step S33: Perform binary segmentation on the confidence heatmap according to the preset confidence threshold to obtain a low confidence region marking map;

[0087] In this embodiment of the invention, the `cv2.threshold` function from the OpenCV library is used to perform binary segmentation on the confidence heatmap. Specifically, a confidence threshold is preset, for example, 0.3. Each pixel value in the confidence heatmap is compared to this threshold; pixel values ​​greater than the threshold are set to 0 (representing high-confidence regions), and pixel values ​​less than or equal to the threshold are set to 1 (representing low-confidence regions). For example, if the confidence heatmap has a shape of (1,1,256,256), after binary segmentation, a low-confidence region marker map is obtained, which also has a shape of (1,1,256,256). This low-confidence region marker map clearly identifies the areas that require further processing.

[0088] Step S34: Use a pre-defined adversarial generative network to complete the features of the low-confidence region label map to obtain enhanced feature data;

[0089] In this embodiment of the invention, after obtaining the low-confidence region labeling map, a pre-defined Generative Adversarial Network (GAN) is used to complete its features. This process is implemented using the PyTorch framework. Specifically, a pre-trained GAN model is selected, which can generate missing feature data based on the low-confidence region labeling map. For example, if the shape of the low-confidence region labeling map is (1,1,256,256), after inputting it into the GAN model, the model will output enhanced feature data, also with the shape (1,1,256,256). This enhanced feature data fills in the missing information in the low-confidence regions, making the feature map more complete and accurate.

[0090] Step S35: Perform feature fusion between the enhanced feature data and the original seismic feature data to obtain the corrected seismic feature data;

[0091] In this embodiment of the invention, the `torch.cat` function in the PyTorch framework is used to fuse enhanced feature data with original seismic feature data. Specifically, the enhanced feature data and the original seismic feature data are concatenated along the channel dimension. For example, the shape of the enhanced feature data is (1,1,256,256), the shape of the original seismic feature data is (1,64,256,256), and the shape of the corrected seismic feature data obtained after concatenation is (1,65,256,256).

[0092] Step S36: Input the corrected seismic feature data into the pre-trained visual Transformer large model for iterative optimization to obtain the optimized free gas structure mask.

[0093] In this embodiment of the invention, the PyTorch framework is used to input corrected seismic feature data into a pre-trained Vision Transformer large model for iterative optimization. Specifically, a pre-trained Vision Transformer (ViT) model is selected and set to trainable mode. The corrected seismic feature data is used as input, and after processing by the ViT model's multi-head self-attention mechanism and feedforward network, the model outputs an optimized free gas structure mask. For example, if the shape of the corrected seismic feature data is (1, 65, 256, 256), after processing by the ViT model, the resulting optimized free gas structure mask has a shape of (1, 1, 256, 256).

[0094] Preferably, step S31 includes the following steps:

[0095] Step S311: Set the Monte Carlo sampling number parameter to obtain the sampling number configuration data;

[0096] In this embodiment of the invention, the Python programming language and the PyTorch framework are used to set the Monte Carlo sampling count parameter. Specifically, a variable `num_samples` is defined to store the sampling count, for example, set to 100 times.

[0097] Step S312: Perform multiple forward propagation calculations on the free gas structure mask based on the data configured by the number of samplings to obtain multiple sets of predicted probability maps;

[0098] In this embodiment of the invention, after setting the number of Monte Carlo sampling iterations, a pre-trained deep learning model is used to perform multiple forward propagation calculations on the free gas structure mask. A pre-trained convolutional neural network (CNN) model in the PyTorch framework is used, specifically designed to output a probability map of the free gas structure. Specifically, the free gas structure mask is input into the model, and forward propagation is performed according to the previously set number of sampling iterations (100 times). Each forward propagation generates a predicted probability map, all with a shape of (1, 1, 256, 256). After 100 forward propagations, 100 sets of predicted probability maps are collected.

[0099] Step S313: Calculate the pixel-level variance of multiple sets of prediction probability maps to obtain the initial prediction variance data;

[0100] In this embodiment of the invention, after obtaining 100 sets of predicted probability maps, pixel-level variance calculation is needed to obtain initial prediction variance data. This task is accomplished using statistical functions within the PyTorch framework. Specifically, the 100 sets of predicted probability maps are integrated into a four-dimensional tensor with a shape of (100, 1, 256, 256). The variance of each pixel is calculated along the dimension of the number of samples (the first dimension). Through these steps, a variance map with a shape of (1, 1, 256, 256) is obtained, which is the initial prediction variance data. This variance map reflects the changes in each pixel across multiple predictions.

[0101] Step S314: Normalize the initial prediction variance data based on the signal-to-noise ratio of the seismic data to obtain the prediction variance data.

[0102] In this embodiment of the invention, the initial prediction variance data is normalized based on the signal-to-noise ratio (SNR) of the seismic data. Specifically, the SNR of the seismic data is calculated by measuring the ratio of signal power to noise power, resulting in an SNR value of 5. This SNR value is then used to normalize the initial prediction variance data. Specifically, the variance value of each pixel is divided by the maximum value in the variance map, and then multiplied by the SNR value. Through these steps, normalized prediction variance data is obtained, and its value is adjusted to a reasonable range.

[0103] Preferably, step S4 includes the following steps:

[0104] Step S41: Perform a three-dimensional coordinate transformation on the optimized free gas structure mask to obtain the initial free gas structure model;

[0105] In this embodiment of the invention, the `meshgrid` and `surf` functions in MATLAB software are used for three-dimensional coordinate transformation. Specifically, two-dimensional free gas structure mask data (e.g., a 256×256 matrix) is loaded into MATLAB. The `meshgrid` function is used to generate a corresponding three-dimensional coordinate grid, which defines the position of the mask in three-dimensional space. The `surf` function is used to combine these coordinates and the mask data to generate a three-dimensional surface model, i.e., the initial free gas structure model. This model visually displays the morphology of the free gas structure in three-dimensional space.

[0106] Step S42: Perform survey line direction interpolation calculations on the initial free gas structure model to obtain interpolation parameters;

[0107] In this embodiment of the invention, the Python programming language and the `interp2d` function from the SciPy library are used to perform interpolation calculations along the survey line direction. Specifically, the direction of the survey line is determined, for example, along the x-axis. Based on the data points of the initial model, the `interp2d` function is used to perform interpolation calculations along the survey line direction. Interpolation parameters are set, such as the interpolation resolution (e.g., interpolating once every 0.1 units of length). Through the above steps, the interpolation parameters are obtained, which define the distribution of model data points along the survey line direction.

[0108] Step S43: Spatial sampling is performed on the initial free gas structure model according to the interpolation parameters to obtain a three-dimensional free gas structure point cloud;

[0109] In this embodiment of the invention, the Python programming language and the NumPy library are used for spatial sampling. Specifically, data points are uniformly sampled in the three-dimensional space of the model according to interpolation parameters. For example, the sampling interval is set to 0.5 units. Through the above steps, a three-dimensional free gas structure point cloud is obtained, and this point cloud data contains detailed location information of the model in three-dimensional space.

[0110] Step S44: Perform density analysis on the three-dimensional free gas structure point cloud to obtain point cloud distribution characteristic data;

[0111] In this embodiment of the invention, the Python programming language and the gaussian_kde function from the SciPy library are used for density analysis. Specifically, point cloud data is loaded into the program, and the gaussian_kde function is used to estimate the density of the point cloud. A bandwidth parameter for density estimation is set (e.g., bandwidth is 0.2). Through these steps, point cloud distribution characteristic data are obtained, which reflect the density distribution of the point cloud in different regions.

[0112] Step S45: Perform Kriging space interpolation based on point cloud distribution feature data to obtain continuous free gas structure data.

[0113] In this embodiment of the invention, the Python programming language and the PyKrige library are used for kriging spatial interpolation. Specifically, point cloud distribution feature data is loaded into the program, and spatial interpolation is performed using the kriging function in the PyKrige library. A suitable kriging model (e.g., a spherical model) is selected, and the model parameters are set (e.g., range parameter is 10, nugget effect is 0.1). Through these steps, continuous free gas structure data are obtained, which can more accurately reflect the continuous morphology of free gas structures in three-dimensional space.

[0114] Preferably, step S5 includes the following steps:

[0115] Step S51: Extract isosurfaces based on continuous free gas structure data to obtain an initial three-dimensional surface model;

[0116] In this embodiment of the invention, ParaView software is used for isosurface extraction. Specifically, continuous free gas structure data is imported into ParaView. The Contour filter in ParaView is used to extract isosurfaces. The isosurface value is set to 0.5 (representing the probability threshold for free gas structures), and an appropriate number of isosurfaces is selected (e.g., 10 isosurfaces). Through the above steps, an initial three-dimensional surface model is obtained, which visually demonstrates the morphology and distribution of free gas structures in three-dimensional space.

[0117] Step S52: Perform topology correction on the initial 3D surface model to obtain topology correction parameters;

[0118] In this embodiment of the invention, Mesh Lab software is used for topology correction. Specifically, the initial 3D surface model is imported into Mesh Lab. Mesh Lab's Topology Cleaning tool is used to check and correct the model's topology. Topology correction parameters are set, for example, Max Edge Length is set to 0.1 units to ensure that the model's geometric details are preserved. Through the above steps, the topology correction parameters are obtained.

[0119] Step S53: Perform geometric repair on the initial 3D surface model according to the topology correction parameters to obtain an optimized 3D surface model;

[0120] In this embodiment of the invention, Mesh Lab software is used for geometric repair. Specifically, the Remeshing tool in Mesh Lab is used to perform geometric repair on the model based on topology correction parameters. Remeshing parameters are set, for example, the Target Number of Faces is set to 10000. Through the above steps, an optimized 3D surface model is obtained, which is more accurate and stable in both geometry and topology.

[0121] Step S54: Manually correct the optimized 3D surface model to obtain interpretation and correction data;

[0122] In this embodiment of the invention, manual corrections are performed using 3ds Max software. Specifically, the optimized 3D surface model is imported into 3ds Max. The Edit Poly tool in 3ds Max is then used to manually edit and correct the model. Based on the experience and knowledge of geological experts, adjustments are made to certain parts of the model, such as smoothing overly sharp edges and filling in missing parts. Through these steps, interpretable correction data is obtained.

[0123] Step S55: Reconstruct the optimized three-dimensional surface model based on the interpretation and correction data to obtain a three-dimensional model of free gas structure;

[0124] In this embodiment of the invention, Blender software is used to reconstruct the optimized 3D surface model. Specifically, the interpreted correction data is imported into Blender. Using Blender's Mesh tool, the model is reconstructed based on the interpreted correction data. Reconstruction parameters are set, for example, Subdivision Level is set to 3, to ensure the smoothness and detail of the model. Through these steps, a 3D model of the free gas structure is obtained, which is more accurate geometrically and topologically, and can better reflect the actual morphology of the free gas structure.

[0125] Step S56: Assign probability labels to the three-dimensional model of free gas structure to obtain a three-dimensional model of free gas structure with probability labels.

[0126] In this embodiment of the invention, the Python programming language and the NumPy library are used to assign probability labels to a 3D model of free gas structures. Specifically, the data of the 3D model of free gas structures is loaded into Python. Using NumPy array operations, the probability value of each vertex (e.g., a range from 0 to 1) is assigned to the model. A threshold for the probability labels is set, for example, 0.8; regions with probability values ​​higher than 0.8 are marked as high-confidence regions, and regions with probability values ​​lower than 0.8 are marked as low-confidence regions. Through these steps, a 3D model of free gas structures with probability labels is obtained. This model not only displays the morphology of free gas structures but also provides probability information for each region.

[0127] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0128] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for identifying free gas tectonic seismic data based on a large model, characterized in that, Includes the following steps: Step S1: Obtain raw seismic data; Short-time Fourier transform is performed on the original seismic data to obtain time-frequency domain seismic data; Hilbert-Huang transform is performed on the time-frequency domain seismic data to obtain enhanced seismic feature data. Step S2: Extract features from the enhanced seismic feature data to obtain a multi-scale seismic feature map; Semantic segmentation is performed on enhanced seismic feature data to obtain a free gas structure mask; Step S3: Perform uncertainty assessment on the free gas structure mask to obtain a confidence heatmap; perform probabilistic prediction on the confidence heatmap to obtain a low-confidence area marker map; perform data augmentation on the low-confidence area marker map to obtain corrected seismic characteristic data; The corrected seismic feature data is input into a pre-trained visual Transformer large model for feature extraction, resulting in an optimized free gas structure mask. Step S3 includes the following steps: Step S31: Perform multiple predictions on the free gas structure mask based on the Monte Carlo method to obtain prediction variance data; Step S32: Perform Gaussian smoothing on the predicted variance data to obtain a confidence heatmap; Step S33: Perform binary segmentation on the confidence heatmap according to the preset confidence threshold to obtain a low confidence region marking map; Step S34: Use a pre-defined adversarial generative network to complete the features of the low-confidence region label map to obtain enhanced feature data; Step S35: Perform feature fusion between the enhanced feature data and the original seismic feature data to obtain the corrected seismic feature data; Step S36: Input the corrected seismic feature data into the pre-trained visual Transformer large model for iterative optimization to obtain the optimized free gas structure mask; Step S4: Reconstruct the three-dimensional space based on the optimized free gas structure mask to obtain the initial free gas structure model; perform survey line interpolation on the optimized free gas structure mask to obtain the three-dimensional free gas structure point cloud; perform spatial fitting on the three-dimensional free gas structure point cloud to obtain continuous free gas structure data. Step S5: Interactively interpret the continuous free gas structure data to obtain a three-dimensional model of the free gas structure; perform visualization rendering based on the three-dimensional model of the free gas structure to obtain a three-dimensional model of the free gas structure with probability labels.

2. The method for identifying free gas tectonic seismic data based on a large model according to claim 1, characterized in that, Step S1 includes the following steps: Step S11: Acquire raw seismic data using seismic acquisition equipment; Step S12: Perform wavelet threshold denoising on the original seismic data to obtain denoised seismic data; Step S13: Divide the denoised seismic data into time windows to obtain segmented seismic data; Step S14: Perform short-time Fourier transform on the segmented seismic data to obtain time-frequency domain seismic data; Step S15: Perform Hilbert-Huang transform on the time-frequency domain seismic data to obtain instantaneous frequency characteristic data; Step S16: Normalize the instantaneous frequency characteristic data to obtain enhanced seismic characteristic data.

3. The method for identifying free gas tectonic seismic data based on a large model according to claim 2, characterized in that, Step S13 includes the following steps: Step S130: Identify seismic wave velocity in the denoised seismic data to obtain the seismic wave propagation velocity; Step S131: Calculate the dominant frequency based on the seismic wave propagation velocity to obtain the seismic wave dominant frequency characteristic data; set the time window parameters according to the seismic wave dominant frequency characteristic data to obtain the time window length configuration data; Step S132: Calculate the overlap rate of the denoised seismic data based on the time window length configuration data to obtain the time window overlap parameters; Step S133: Use the time window overlap parameter to perform sliding time window segmentation on the denoised seismic data to obtain the initial segmented seismic data; Step S134: Perform energy equalization processing on the initial segmented seismic data to obtain equalized segmented seismic data; Step S135: Based on a preset signal-to-noise ratio threshold, perform quality screening on the balanced segmented seismic data to obtain effective segmented seismic data; Step S136: Time-stamp the effective segmented seismic data to obtain segmented seismic data.

4. The method for identifying free gas tectonic seismic data based on a large model according to claim 1, characterized in that, Step S2 includes the following steps: Step S21: Expand the channel dimensions of the enhanced seismic feature data to obtain three-dimensional seismic feature data; Step S22: Use a pre-trained visual Transformer model to extract features from the 3D seismic feature data to obtain a primary seismic feature map; Step S23: Enhance the primary seismic feature map to obtain a multi-scale seismic feature map; Step S24: Upsample and reconstruct the multi-scale seismic feature map to obtain a semantic segmentation probability map; Step S25: Binarize the semantic segmentation probability map based on a preset threshold to obtain a free gas structure mask.

5. The method for identifying free gas tectonic seismic data based on a large model according to claim 4, characterized in that, Step S23 includes the following steps: Step S231: Perform pyramid pooling on the primary seismic feature map to obtain multi-scale contextual feature data; Step S232: Calculate feature weights for multi-scale contextual feature data using a spatial attention mechanism to obtain a spatial attention weight map; Step S233: Recalibrate the primary seismic feature map based on the spatial attention weight map to obtain the spatially enhanced feature map; Step S234: Perform channel correlation analysis on the spatial augmentation feature map using the channel attention mechanism to obtain the channel attention weight map; Step S235: Perform channel feature enhancement on the spatial enhancement feature map based on the channel attention weight map to obtain a multi-scale seismic feature map.

6. The method for identifying free gas tectonic seismic data based on a large model according to claim 1, characterized in that, Step S31 includes the following steps: Step S311: Set the Monte Carlo sampling number parameter to obtain the sampling number configuration data; Step S312: Perform multiple forward propagation calculations on the free gas structure mask based on the data configured by the number of samplings to obtain multiple sets of predicted probability maps; Step S313: Calculate the pixel-level variance of multiple sets of prediction probability maps to obtain the initial prediction variance data; Step S314: Normalize the initial prediction variance data based on the signal-to-noise ratio of the seismic data to obtain the prediction variance data.

7. The method for identifying free gas tectonic seismic data based on a large model according to claim 1, characterized in that, Step S4 includes the following steps: Step S41: Perform a three-dimensional coordinate transformation on the optimized free gas structure mask to obtain the initial free gas structure model; Step S42: Perform survey line direction interpolation calculations on the initial free gas structure model to obtain interpolation parameters; Step S43: Spatial sampling is performed on the initial free gas structure model according to the interpolation parameters to obtain a three-dimensional free gas structure point cloud; Step S44: Perform density analysis on the three-dimensional free gas structure point cloud to obtain point cloud distribution characteristic data; Step S45: Perform Kriging space interpolation based on point cloud distribution feature data to obtain continuous free gas structure data.

8. The method for identifying free gas tectonic seismic data based on a large model according to claim 1, characterized in that, Step S5 includes the following steps: Step S51: Extract isosurfaces based on continuous free gas structure data to obtain an initial three-dimensional surface model; Step S52: Perform topology correction on the initial 3D surface model to obtain topology correction parameters; Step S53: Perform geometric repair on the initial 3D surface model according to the topology correction parameters to obtain an optimized 3D surface model; Step S54: Manually correct the optimized 3D surface model to obtain interpretation and correction data; Step S55: Reconstruct the optimized three-dimensional surface model based on the interpretation and correction data to obtain a three-dimensional model of free gas structure; Step S56: Assign probability labels to the three-dimensional model of free gas structure to obtain a three-dimensional model of free gas structure with probability labels.

9. A free gas tectonic seismic data identification system based on a large model, characterized in that, For executing the large-model-based free gas structure seismic data identification method as described in claim 1, the large-model-based free gas structure seismic data identification system comprises: The data preprocessing module is used to acquire raw seismic data; perform short-time Fourier transform on the raw seismic data to obtain time-frequency domain seismic data; and perform Hilbert-Huang transform on the time-frequency domain seismic data to obtain enhanced seismic feature data. The feature extraction module is used to extract features from the enhanced seismic feature data to obtain a multi-scale seismic feature map; and to perform semantic segmentation on the enhanced seismic feature data to obtain a free gas structure mask. The optimization iteration module is used to evaluate the uncertainty of the free gas structure mask and obtain a confidence heatmap; perform probability prediction on the confidence heatmap to obtain a low-confidence area marker map; perform data augmentation on the low-confidence area marker map to obtain corrected seismic feature data; input the corrected seismic feature data into a pre-trained visual Transformer large model for feature extraction to obtain the optimized free gas structure mask. The 3D modeling module is used to reconstruct the 3D space based on the optimized free gas structure mask to obtain the initial free gas structure model; perform survey line interpolation on the optimized free gas structure mask to obtain the 3D free gas structure point cloud; and perform spatial fitting on the 3D free gas structure point cloud to obtain continuous free gas structure data. The visualization module is used to interactively interpret continuous free gas structure data to obtain a three-dimensional model of the free gas structure; based on the three-dimensional model of the free gas structure, a visualization rendering is performed to obtain a three-dimensional model of the free gas structure with probability labels.

Images