A Method and System for Marine Geological and Seismic Monitoring Data Analysis Based on Wavelet-PCA

Abstract

This invention relates to the field of seismic data processing and analysis technology, and particularly to a method and system for analyzing marine geological seismic monitoring data based on wavelet-PCA. The invention performs data preprocessing on marine geological seismic monitoring data; performs wavelet decomposition on the preprocessed data; expands the approximation coefficients and detail coefficients obtained from the wavelet decomposition into one-dimensional vectors, and concatenates them to form a feature matrix; performs dimensionality reduction on the feature matrix using principal component analysis; and performs data analysis on the dimensionality-reduced feature data using a graph neural network model. When using the graph neural network model for analysis, it respects the fundamental constraint of spatial distance as a basis for station association, while also introducing the inherent linkage of seismic signals to characterize more essential physical connections, significantly improving the feature expression capability and analysis accuracy of seismic signals, and providing key technical support for disaster early warning, resource exploration, and other applications.

Description

Technical Field

[0001] This invention relates to the field of earthquake data processing and analysis technology, and in particular to a method and system for analyzing marine geological earthquake monitoring data based on wavelet-PCA. Background Technology

[0002] In marine geological and seismic monitoring, data typically exhibits high dimensionality and complexity, making it difficult for traditional analysis methods to effectively extract key information and process it efficiently. Wavelet transform and principal component analysis (PCA) are two commonly used data processing techniques. Wavelet transform can extract multi-scale features of signals, while PCA can reduce data dimensionality while retaining key information. However, using either method alone still has limitations when processing complex data. Therefore, combining wavelet transform and PCA can more effectively process and analyze marine geological and seismic monitoring data. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention provides a method and system for analyzing marine geological and seismic monitoring data based on wavelet-PCA, which solves the problems existing in the prior art.

[0004] This invention provides a method for analyzing marine geological and seismic monitoring data based on wavelet-PCA, the method comprising the following steps:

[0005] S1: Perform data preprocessing operations on marine geological and seismic monitoring data;

[0006] S2: Perform wavelet decomposition on the preprocessed marine geological and seismic monitoring data;

[0007] S3: Expand the approximation coefficients and detail coefficients obtained from wavelet decomposition into one-dimensional vectors and concatenate them to form a feature matrix;

[0008] S4: Principal component analysis is used to reduce the dimensionality of the feature matrix;

[0009] S5: Use a graph neural network model to perform data analysis on the dimensionality-reduced feature data;

[0010] Specifically, S5 consists of: S5.1: Constructing a spatiotemporal feature tensor; S5.2: Calculating the connection weights between monitoring points based on their spatial locations in the seismic data, and constructing an adjacency matrix;

[0011] The calculation process of the connection weight is as follows: calculate the Euclidean spatial distance between the two monitoring points; calculate the Pearson similarity coefficient of the two monitoring points in the wavelet coefficient time series; determine the connection weight of the two monitoring points according to the Euclidean spatial distance and the Pearson similarity coefficient; S5.3: input the spatiotemporal feature tensor and the adjacency matrix into the graph neural network model to realize data analysis.

[0012] Preferably, in step S5.1, the dimensionality-reduced feature data and the corresponding spatial coordinates of the earthquake monitoring points are combined to construct the spatiotemporal feature tensor.

[0013] Preferably, in step S5.2, the coordinates of monitoring point i are (x... i ,y i ,z i The coordinates of monitoring point j are (x) j ,y j ,z j If ), then the Euclidean distance d(i,j) is given by:

[0014] ;

[0015] The formula for calculating the connection weight w(i,j) between two monitoring points is:

[0016] ;

[0017] In the formula, α is a weighting coefficient set by the user, and s(i,j) is the Pearson correlation coefficient of the signals from the two monitoring points.

[0018] Preferably, in step S5.3, the main architecture of the graph neural network model includes: an input layer, a graph convolutional layer, a graph attention layer, a fully connected layer, and an output layer; wherein, the input layer receives the spatiotemporal feature tensor and the adjacency matrix; the graph convolutional layer is used to extract the local topological features of the spatiotemporal feature tensor and the adjacency matrix; the graph attention layer is used to assign weights to each monitoring point; the fully connected layer is used to implement dimension selection; and the output layer is used to output the data analysis results.

[0019] Preferably, the loss function of the graph neural network model is the cross-entropy loss function, the optimizer is the Adam optimizer, and the Dropout regularization mechanism is used.

[0020] Preferably, S2 specifically comprises:

[0021] S2.1: Select the wavelet basis functions for wavelet decomposition of the preprocessed marine geological and seismic monitoring data;

[0022] S2.2: Determine the number of decomposition levels;

[0023] S2.3: Perform multi-scale wavelet decomposition on the preprocessed marine geological and seismic monitoring data.

[0024] Preferably, in step S2.3, the preprocessed marine geological seismic monitoring data is decomposed into sub-bands of different frequencies by applying wavelet transform multiple times. Each decomposition layer divides the signal into low-frequency approximation coefficients and high-frequency detail coefficients.

[0025] The first layer of decomposition is as follows: the preprocessed marine geological and seismic monitoring data is decomposed into low-frequency approximation coefficients A1 and high-frequency detail coefficients D1;

[0026] The second layer of decomposition is as follows: the low-frequency approximation coefficient A1 is further decomposed into low-frequency approximation coefficient A2 and high-frequency detail coefficient D2;

[0027] The third layer of decomposition is as follows: the low-frequency approximation coefficient A2 is further decomposed into low-frequency approximation coefficient A3 and high-frequency detail coefficient D3.

[0028] Preferably, S4 specifically comprises:

[0029] S4.1: Calculate the covariance matrix of the characteristic matrix;

[0030] S4.2: Perform eigenvalue decomposition on the covariance matrix;

[0031] S4.3: Select the number of principal components based on the eigenvalues, and obtain k principal components;

[0032] S4.4: Project the feature matrix onto the k principal components to achieve dimensionality reduction and obtain the dimensionality-reduced feature data.

[0033] Preferably, in step S4.3, the number of principal components k is selected according to the cumulative variance contribution rate.

[0034] According to another aspect of the present invention, a wavelet-PCA-based marine geological and seismic monitoring data analysis system is provided, the system employing the aforementioned wavelet-PCA-based marine geological and seismic monitoring data analysis method, the system comprising:

[0035] The preprocessing module is used to perform data preprocessing operations on marine geological and seismic monitoring data;

[0036] The wavelet decomposition module is used to perform wavelet decomposition on preprocessed marine geological and seismic monitoring data.

[0037] The feature extraction module is used to expand the approximation coefficients and detail coefficients obtained from wavelet decomposition into one-dimensional vectors and then concatenate them to form a feature matrix.

[0038] The dimensionality reduction module is used to reduce the dimensionality of the feature matrix using principal component analysis.

[0039] The data analysis module is used to perform data analysis on the dimensionality-reduced feature data using a graph neural network model.

[0040] The embodiments of the present invention have the following technical effects:

[0041] This invention performs data preprocessing on marine geological and seismic monitoring data; it then performs wavelet decomposition on the preprocessed data; expands the approximation coefficients and detail coefficients obtained from the wavelet decomposition into one-dimensional vectors, and concatenates them to form a feature matrix; it uses principal component analysis to reduce the dimensionality of the feature matrix; and it employs a graph neural network model to analyze the dimensionality-reduced feature data. When using the graph neural network model for analysis, it respects the fundamental constraint of spatial distance as a basis for station association, while also introducing the inherent linkage of seismic signals to characterize more essential physical connections. This significantly improves the feature representation ability and analysis accuracy of seismic signals, providing key technical support for disaster early warning, resource exploration, and other applications. Attached Figure Description

[0042] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0043] Figure 1 This is a flowchart of the wavelet-PCA-based marine geological and seismic monitoring data analysis method provided in this embodiment of the invention;

[0044] Figure 2 This is a flowchart of data analysis of the dimensionality-reduced feature data using a graph neural network model, provided in an embodiment of the present invention. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0046] Example 1, Figure 1 A flowchart of a wavelet-PCA-based method for analyzing marine geological and seismic monitoring data is shown, such as... Figure 1 As shown, the method for analyzing marine geological and seismic monitoring data based on wavelet-PCA includes the following steps:

[0047] S1: Perform data preprocessing operations on marine geological and seismic monitoring data;

[0048] In this step, the preprocessing includes data denoising, data standardization, and data segmentation.

[0049] The data denoising specifically employs a moving average method or a median filtering method. The moving average method specifically involves: selecting an appropriate smoothing window size, calculating the average value of the data points within the smoothing window, replacing the data value at the center point of the smoothing window with the average value, and moving the smoothing window across the entire data sequence to process each data point step by step, thereby achieving data denoising.

[0050] In this step, the smoothing window size is 11 data points.

[0051] The median filtering method specifically involves: selecting an appropriate filter window size, calculating the median of the data points within the filter window, and using the median to replace the data value at the center point of the filter window to achieve data denoising. Median filtering has a good effect on processing impulse noise, effectively removing noise points without affecting the overall trend of the signal.

[0052] In this step, the smoothing window size is 5 data points.

[0053] The data standardization specifically involves: calculating the mean μ and standard deviation σ of the marine geological and seismic monitoring data, and performing standardization processing on each data point x to obtain the standardized data z. The formula for the standardization processing is:

[0054] ;

[0055] Through standardization, the marine geological and seismic monitoring data are made to have a mean of 0 and a standard deviation of 1, thus eliminating dimensional differences between different features.

[0056] The data segmentation specifically involves selecting an appropriate frame length and dividing the long-term series data in the marine geological and seismic monitoring data into multiple data frames according to the frame length. Each data frame can be regarded as an independent sample, which facilitates subsequent wavelet decomposition and PCA processing.

[0057] The length of each frame should be adjusted according to the characteristics of the data and the analysis requirements to ensure that each frame contains enough information for analysis, while avoiding an increase in computational complexity due to too many frames; in this step, the frame length is 1024 data points.

[0058] By employing data preprocessing steps such as denoising, standardization, and segmentation, the quality and consistency of the marine geological and seismic monitoring data can be significantly improved, ensuring the effectiveness and reliability of the data in subsequent wavelet decomposition and PCA processing. These steps lay a solid foundation for subsequent feature extraction and data analysis.

[0059] S2: Perform wavelet decomposition on the preprocessed marine geological and seismic monitoring data;

[0060] The goal of wavelet decomposition is to decompose marine geological and seismic monitoring data into sub-bands of different frequencies, thereby extracting multi-scale features. Through multi-scale wavelet decomposition, the signal can be decomposed into low-frequency approximation coefficients and high-frequency detail coefficients. The low-frequency part represents the overall trend of the signal, while the high-frequency part represents the details and abrupt changes of the signal. Wavelet decomposition is particularly effective in processing non-stationary signals, providing rich time-frequency information.

[0061] Specifically, S2 is as follows:

[0062] S2.1: Select the wavelet basis functions for wavelet decomposition of the preprocessed marine geological and seismic monitoring data;

[0063] The db4 wavelet basis function is suitable for general signals, with good smoothness and tight support. The haar wavelet basis function is suitable for abrupt signals, with a simple structure and fast calculation speed. The sym5 wavelet basis function is suitable for smooth signals, with symmetry and good time-frequency localization characteristics. The Molet wavelet basis function is designed specifically for seismic signals and can effectively extract the features of seismic signals.

[0064] In this step, the Molet wavelet basis function is used as the wavelet basis function for wavelet decomposition of the preprocessed marine geological seismic monitoring data.

[0065] S2.2: Determine the number of decomposition levels;

[0066] The number of decomposition levels determines the number of levels at which the signal is decomposed. Each level divides the signal into low-frequency approximation coefficients and high-frequency detail coefficients. More levels mean a narrower frequency range in the low-frequency component and higher frequency resolution, but also increased computational complexity. Typically, 3-5 levels achieve a good balance between frequency resolution and computational complexity; in this step, 3 levels are used.

[0067] S2.3: Perform multi-scale wavelet decomposition on the preprocessed marine geological and seismic monitoring data;

[0068] By applying wavelet transform multiple times, the preprocessed marine geological and seismic monitoring data is decomposed into sub-bands of different frequencies. Each decomposition layer divides the signal into low-frequency approximation coefficients and high-frequency detail coefficients.

[0069] The first layer of decomposition involves decomposing the preprocessed marine geological and seismic monitoring data into low-frequency approximation coefficients A1 and high-frequency detail coefficients D1.

[0070] The second layer of decomposition is as follows: the low-frequency approximation coefficient A1 is further decomposed into low-frequency approximation coefficient A2 and high-frequency detail coefficient D2;

[0071] The third layer of decomposition is as follows: the low-frequency approximation coefficient A2 is further decomposed into low-frequency approximation coefficient A3 and high-frequency detail coefficient D3.

[0072] Wavelet decomposition, by selecting an appropriate wavelet basis and decomposition level, divides the original data into sub-bands of different frequencies, extracting multi-scale features. Low-frequency approximation coefficients represent the overall trend of the signal, while high-frequency detail coefficients represent the signal's details and abrupt changes. Multi-scale wavelet decomposition provides rich time-frequency information and is particularly suitable for processing non-stationary signals. Wavelet decomposition lays a solid foundation for subsequent feature extraction and data analysis.

[0073] S3: Expand the approximation coefficients and detail coefficients obtained from wavelet decomposition into one-dimensional vectors and concatenate them to form a feature matrix;

[0074] The wavelet coefficients of each layer are expanded into a one-dimensional wavelet coefficient vector from left to right. The one-dimensional wavelet coefficient vectors of all layers are then concatenated by rows or columns to form a feature matrix, which forms the input data suitable for principal component analysis.

[0075] S4: Principal component analysis is used to reduce the dimensionality of the feature matrix;

[0076] Principal Component Analysis (PCA) dimensionality reduction projects high-dimensional feature data into a low-dimensional space through linear transformation, preserving the main information in the data while significantly reducing its dimensionality. PCA calculates the principal components of the data, transforming the original high-dimensional feature matrix into a low-dimensional representation, thus simplifying subsequent data analysis and processing tasks. The dimensionality-reduced data not only reduces computational complexity but also effectively removes noise and redundant information.

[0077] Specifically, S4 is:

[0078] S4.1: Calculate the covariance matrix of the characteristic matrix;

[0079] In this step, the formula for calculating the covariance matrix is:

[0080] ;

[0081] In the formula, Cov is the covariance matrix of the feature matrix, n is the number of features of the feature matrix, and X std Let be the feature matrix.

[0082] S4.2: Perform eigenvalue decomposition on the covariance matrix;

[0083] The decomposition formula is as follows:

[0084] ;

[0085] In the formula, V is the eigenvector matrix, where each column is a principal component direction, and Λ is a diagonal matrix, where the diagonal elements are the eigenvalues ​​λ1, λ2, ..., λ. d , where the eigenvalue λ i This represents the variance contribution of the i-th principal component. The larger the eigenvalue, the more original data information the corresponding principal component contains.

[0086] S4.3: Select the number of principal components based on the eigenvalues, and obtain k principal components;

[0087] The number of principal components k is selected based on the Cumulative Explained Variance (CEV); in this step, the CEV is set to 95%.

[0088] S4.4: Project the feature matrix onto the k principal components to achieve dimensionality reduction and obtain the dimensionality-reduced feature data;

[0089] S5: Use a graph neural network model to perform data analysis on the dimensionality-reduced feature data;

[0090] Seismic signals typically exhibit high-dimensional characteristics, containing a large amount of redundant information and noise. Directly using wavelet coefficients or single PCA dimensionality reduction is insufficient to effectively distinguish between valid signals and noise, resulting in the obscuring of key features. Seismic signals also exhibit significant spatiotemporal distribution characteristics. For example, seismic waves are affected by geological structures during propagation, and signals from different monitoring points exhibit spatial correlation. Existing methods usually only focus on dimensionality reduction of features from a single sample, failing to fully utilize the spatial relationships between monitoring points or the dynamic evolution information of time series, leading to the ineffective capture of key spatiotemporal patterns.

[0091] Based on this, this embodiment uses a graph neural network (GNN) to construct the spatial topological relationship between monitoring points and explicitly model the physical laws of seismic wave propagation.

[0092] Specifically, such as Figure 2 As shown, S5 specifically includes:

[0093] S5.1: Constructing the spatiotemporal feature tensor;

[0094] In this step, the dimensionality-reduced feature data and the corresponding spatial coordinates of earthquake monitoring points are combined to construct a spatiotemporal feature tensor.

[0095] The spatiotemporal feature tensor contains the time series characteristics and spatial location information of the earthquake monitoring data, comprehensively describing the earthquake signal characteristics and geographical location of each monitoring point at different times and scales, providing a high-quality data foundation for subsequent analysis.

[0096] S5.2: Calculate the connection weights between monitoring points based on their spatial locations in the seismic data, and construct an adjacency matrix;

[0097] An adjacency matrix is ​​constructed based on the spatial location of monitoring points, and the connection weights between each monitoring point are defined. The calculation of the connection weights needs to take into account both spatial proximity and seismic wave propagation characteristics, while also ensuring that the entire graph structure is interpretable and practical. In this step, a connection weight calculation method based on the superposition of spatial distance and seismic signal similarity is proposed. This method not only gives higher weights to monitoring points that are geographically close and have similar signals, but also captures anomalous correlations that are geographically distant but have strong signal linkages.

[0098] Specifically, the calculation process for the connection weights is as follows:

[0099] Calculate the Euclidean spatial distance between the two monitoring points;

[0100] Wherein, the coordinates of monitoring point i are (x i ,y i ,z i The coordinates of monitoring point j are (x) j ,y j ,z j If ), then the Euclidean distance d(i,j) is given by:

[0101] ;

[0102] The Euclidean spatial distance reflects the physical distance between two monitoring points in geographic space.

[0103] Calculate the Pearson similarity coefficient between two monitoring points in the wavelet coefficient time series;

[0104] For each monitoring point, after dimensionality reduction, the wavelet coefficients of each layer are selected at a certain time-frequency scale, and the Pearson correlation coefficient is calculated to obtain the similarity s(i,j) at the signal level. The closer the Pearson correlation coefficient is to 1, the more similar the seismic signal fluctuation patterns of the two monitoring points are at that time-frequency scale.

[0105] The connection weight between the two monitoring points is determined based on the Euclidean spatial distance and the Pearson similarity coefficient.

[0106] The formula for calculating the connection weight w(i,j) between two monitoring points is as follows:

[0107] ;

[0108] In the formula, α is a weighting coefficient set by the user to balance the contributions of the two dimensions of spatial proximity and signal similarity. This allows monitoring points that are geographically close and have similar signals to receive higher weights, while also capturing abnormal correlations that are geographically distant but have strong signal linkages.

[0109] S5.3: Input the spatiotemporal feature tensor and the adjacency matrix into the graph neural network model to achieve data analysis;

[0110] Graph Neural Network (GNN) models effectively capture dependencies between nodes by performing message passing and feature aggregation on a graph structure, thereby enhancing the feature representation of nodes. The main architecture of the GNN model includes an input layer, a graph convolutional layer, a graph attention layer, a fully connected layer, and an output layer. The input layer receives the spatiotemporal feature tensor and the adjacency matrix. The graph convolutional layer extracts local topological features from the spatiotemporal feature tensor and the adjacency matrix. The graph attention layer assigns weights to each monitoring point. The fully connected layer performs dimension selection. The output layer outputs the data analysis results.

[0111] The graph neural network model uses cross-entropy loss function, Adam optimizer, and Dropout regularization mechanism as its optimization mechanism.

[0112] Example 2: This invention also provides a wavelet-PCA-based marine geological and seismic monitoring data analysis system. The system employs the wavelet-PCA-based marine geological and seismic monitoring data analysis method of Example 1. The system includes:

[0113] The preprocessing module is used to perform data preprocessing operations on marine geological and seismic monitoring data;

[0114] The wavelet decomposition module is used to perform wavelet decomposition on preprocessed marine geological and seismic monitoring data.

[0115] The feature extraction module is used to expand the approximation coefficients and detail coefficients obtained from wavelet decomposition into one-dimensional vectors and then concatenate them to form a feature matrix.

[0116] The dimensionality reduction module is used to reduce the dimensionality of the feature matrix using principal component analysis.

[0117] The data analysis module is used to perform data analysis on the dimensionality-reduced feature data using a graph neural network model.

[0118] Example 3: The present invention also provides an electronic device, including one or more processors and a memory.

[0119] A processor can be a central processing unit (CPU) or other form of processing unit with data processing and / or instruction execution capabilities, and can control other components in an electronic device to perform desired functions.

[0120] The memory may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and a processor may execute the program instructions to implement the wavelet-PCA-based marine geological seismic monitoring data analysis method described above in any embodiment of this application, and / or other desired functions. Various contents such as initial extrinsic parameters and thresholds may also be stored in the computer-readable storage medium.

[0121] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.

Claims

1. A method for analyzing marine geological and seismic monitoring data based on wavelet-PCA, characterized in that, The method includes the following steps: S1: Perform data preprocessing operations on marine geological and seismic monitoring data; S2: Perform wavelet decomposition on the preprocessed marine geological and seismic monitoring data; S3: Expand the approximation coefficients and detail coefficients obtained from wavelet decomposition into one-dimensional vectors and concatenate them to form a feature matrix; S4: Principal component analysis is used to reduce the dimensionality of the feature matrix; S5: Use a graph neural network model to perform data analysis on the dimensionality-reduced feature data; Specifically, S5 consists of: S5.1: Constructing a spatiotemporal feature tensor; S5.2: Calculating the connection weights between monitoring points based on their spatial locations in the seismic data, and constructing an adjacency matrix; The calculation process of the connection weight is as follows: calculate the Euclidean spatial distance between the two monitoring points; calculate the Pearson similarity coefficient of the two monitoring points in the wavelet coefficient time series; determine the connection weight of the two monitoring points according to the Euclidean spatial distance and the Pearson similarity coefficient; S5.3: input the spatiotemporal feature tensor and the adjacency matrix into the graph neural network model to realize data analysis.

2. The method for analyzing marine geological and seismic monitoring data based on wavelet-PCA according to claim 1, characterized in that, In step S5.1, the dimension-reduced feature data and the spatial coordinates of the corresponding earthquake monitoring points are combined to construct the spatiotemporal feature tensor.

3. The method for analyzing marine geological and seismic monitoring data based on wavelet-PCA according to claim 2, characterized in that, In S5.2, the coordinates of monitoring point i are (x... i ,y i ,z i The coordinates of monitoring point j are (x) j ,y j ,z j If ), then the Euclidean distance d(i,j) is given by: ； The formula for calculating the connection weight w(i,j) between two monitoring points is: ； In the formula, α is a weighting coefficient set by the user, and s(i,j) is the Pearson correlation coefficient of the signals from the two monitoring points.

4. The method for analyzing marine geological and seismic monitoring data based on wavelet-PCA according to claim 2, characterized in that, In S5.3, the main architecture of the graph neural network model includes: an input layer, a graph convolutional layer, a graph attention layer, a fully connected layer, and an output layer; wherein, the input layer receives the spatiotemporal feature tensor and the adjacency matrix; the graph convolutional layer is used to extract the local topological features of the spatiotemporal feature tensor and the adjacency matrix; the graph attention layer is used to assign weights to each monitoring point; the fully connected layer is used to implement dimension selection; and the output layer is used to output the data analysis results.

5. The method for analyzing marine geological and seismic monitoring data based on wavelet-PCA according to claim 4, characterized in that, The graph neural network model uses cross-entropy loss function, Adam optimizer, and Dropout regularization mechanism as its optimization mechanism.

6. The method for analyzing marine geological and seismic monitoring data based on wavelet-PCA according to claim 1, characterized in that, Specifically, S2 is: S2.1: Select the wavelet basis functions for wavelet decomposition of the preprocessed marine geological and seismic monitoring data; S2.2: Determine the number of decomposition levels; S2.3: Perform multi-scale wavelet decomposition on the preprocessed marine geological and seismic monitoring data.

7. The method for analyzing marine geological and seismic monitoring data based on wavelet-PCA according to claim 6, characterized in that, In S2.3, the preprocessed marine geological seismic monitoring data is decomposed into sub-bands of different frequencies by applying wavelet transform multiple times. Each decomposition layer divides the signal into low-frequency approximation coefficients and high-frequency detail coefficients. The first layer of decomposition is as follows: the preprocessed marine geological and seismic monitoring data is decomposed into low-frequency approximation coefficients A1 and high-frequency detail coefficients D1; The second layer of decomposition is as follows: the low-frequency approximation coefficient A1 is further decomposed into low-frequency approximation coefficient A2 and high-frequency detail coefficient D2; The third layer of decomposition is as follows: the low-frequency approximation coefficient A2 is further decomposed into low-frequency approximation coefficient A3 and high-frequency detail coefficient D3.

8. The method for analyzing marine geological and seismic monitoring data based on wavelet-PCA according to claim 1, characterized in that, Specifically, S4 is: S4.1: Calculate the covariance matrix of the characteristic matrix; S4.2: Perform eigenvalue decomposition on the covariance matrix; S4.3: Select the number of principal components based on the eigenvalues, and obtain k principal components; S4.4: Project the feature matrix onto the k principal components to achieve dimensionality reduction and obtain the dimensionality-reduced feature data.

9. The method for analyzing marine geological and seismic monitoring data based on wavelet-PCA according to claim 8, characterized in that, In S4.3, the number of principal components k is selected according to the cumulative variance contribution rate.

10. A marine geological and seismic monitoring data analysis system based on wavelet-PCA, characterized in that, The system employs the wavelet-PCA-based marine geological and seismic monitoring data analysis method as described in any one of claims 1-9, and the system comprises: The preprocessing module is used to perform data preprocessing operations on marine geological and seismic monitoring data; The wavelet decomposition module is used to perform wavelet decomposition on preprocessed marine geological and seismic monitoring data. The feature extraction module is used to expand the approximation coefficients and detail coefficients obtained from wavelet decomposition into one-dimensional vectors and then concatenate them to form a feature matrix. The dimensionality reduction module is used to reduce the dimensionality of the feature matrix using principal component analysis. The data analysis module is used to perform data analysis on the dimensionality-reduced feature data using a graph neural network model.

Images