Mesoscale vortex space-time characteristic analysis method and device based on neural network

Abstract

The invention relates to the technical field of ocean information processing, in particular to a mesoscale vortex space-time characteristic analysis method and device based on a neural network, and the method comprises the following steps: obtaining the type, vortex center position, vortex intensity and temperature and sound velocity abnormal characteristics of each layer of mesoscale vortex according to the ocean current three-dimensional speed data of a target region; and the trained neural network predicts the vortex center position, vortex intensity and temperature and sound velocity abnormal characteristics of the mesoscale vortex at the future moment according to the type, vortex center position, vortex intensity and temperature and sound velocity abnormal characteristics of the mesoscale vortex at each layer at the historical moment. The problems that in the prior art, small-scale vortexes or rapidly-changing vortex characteristics are difficult to capture, large-scale data processing requirements are difficult to meet, and robustness is poor are solved.

Description

Technical Field

[0001] This application relates to the technical field of marine information processing, and particularly relates to a method and device for analyzing the spatio-temporal characteristics of mesoscale eddies based on a neural network. Background Art

[0002] Mesoscale eddies are typical manifestations of mesoscale motions in the ocean. Their scales are usually between dozens and hundreds of kilometers, and their life cycles range from several days to several months. They have strong material transport capabilities and energy conversion characteristics, and have important impacts in fields such as the ocean energy closed-loop, ecosystem maintenance, air-sea interaction, and underwater communication and navigation. Therefore, the research on the identification, structure analysis, and spatio-temporal evolution laws of mesoscale eddies has always been a key topic in the fields of physical oceanography and underwater information processing.

[0003] Currently, the analysis methods of mesoscale eddies mainly include image recognition techniques based on remote sensing data and traditional statistical analysis methods based on physical fields. The former relies on remote sensing observations such as satellite altimeter data and sea surface temperature (SST), and uses image processing means such as edge detection, Hough transform, and waveform fitting to identify the vortex boundary. The latter is based on the velocity field or vorticity field, and uses techniques such as empirical orthogonal function decomposition (EOF), spectral analysis, and rotational velocity distribution analysis to construct a vortex identification index system. Although these methods have made certain progress in practical applications, there are still the following technical bottlenecks: The spatial and temporal resolutions of traditional remote sensing data are limited, making it difficult to accurately capture and track mesoscale eddies with small scales or multi-layer structures that change rapidly, difficult to capture small-scale vortices or rapidly changing vortex characteristics, and difficult to handle large-scale data processing requirements. Moreover, it has insufficient adaptability to complex ocean environments and poor applicability and robustness under complex conditions such as multi-vortex interference and strong background currents. Summary of the Invention

[0004] This application provides a method, device / system, equipment, and computer-readable storage medium for analyzing the spatio-temporal characteristics of mesoscale eddies based on a neural network, which can solve the problems in the prior art that it is difficult to capture small-scale vortices or rapidly changing vortex characteristics, difficult to handle large-scale data processing requirements, and poor robustness.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is: On the one hand, the present invention provides a method for analyzing the spatio-temporal characteristics of mesoscale eddies based on a neural network, including the following steps: Obtain the types, vortex center positions, vortex intensities, and temperature and sound speed anomaly characteristics of mesoscale eddies in each layer according to the three-dimensional sea current velocity data of the target area; The trained neural network predicts the vortex center position, vortex intensity, and temperature and sound speed anomaly characteristics of mesoscale vortices at future times based on the types, vortex center positions, vortex intensities, and temperature and sound speed anomaly characteristics of mesoscale vortices in each layer at historical times.

[0006] In some alternative solutions, training the neural network includes the following steps: Based on the types, vortex center positions, vortex intensities, and temperature and sound speed anomaly characteristics of mesoscale vortices in each layer at different times, construct a graph structure, where the graph nodes of the graph structure include the types, vortex center positions, vortex intensities, and temperature and sound speed anomaly characteristics of each mesoscale vortex, and the edges of the graph structure are the adjacency relationships between the graph nodes of each mesoscale vortex; When updating the parameters of the neural network in each iteration step, input the graph nodes and edges of the corresponding graph structure at the previous time to the graph neural network, output the predicted types, vortex center positions, vortex intensities, and temperature and sound speed anomaly characteristics of mesoscale vortices at the next time, and obtain the loss value of this iteration step in combination with the label data. Update the parameters of the graph neural network according to the loss value.

[0007] In some alternative solutions, the loss value is determined according to the loss function where , , , are the intensity regression loss, the structural continuity constraint loss, the temperature regression and sound speed regression loss hyperparameter weights respectively, is the total loss value, is the intensity regression loss regarding the vortex intensity, is the structural continuity constraint loss regarding the vortex center position, is the temperature regression loss regarding the temperature anomaly, is the sound speed regression loss regarding the sound speed anomaly.

[0008] In some alternative solutions, the intensity regression loss regarding the vortex intensity , where is the true value of the vortex intensity, is the predicted value of the vortex intensity, and N is the number of mesoscale vortices; The structural continuity constraint loss regarding the vortex center position , where is the predicted position of the vortex center at the t-th time step, is the predicted position of the vortex center at the (t + 1)-th time step, and T is the total number of time steps in the tracking sequence; The temperature regression loss regarding the temperature anomaly , where is the actually observed temperature anomaly value; is the predicted temperature anomaly value; Sound speed regression loss regarding sound speed anomaly

[0009] wherein, is the actually observed sound speed anomaly; is the predicted sound speed anomaly value.

[0010] In some alternative solutions, obtaining the type, vortex center position, vortex intensity, and temperature and sound speed anomaly characteristics of mesoscale eddies in each layer based on the three-dimensional sea current velocity data of the target area includes: Determining the sea current type and vortex intensity in each layer based on the three-dimensional sea current velocity data of the target area, and screening out the vortex centers of mesoscale eddies based on the velocity field and constraint conditions in the vortex intensity; Equivalent each layer of mesoscale eddies to an irregular disk, and constructing a three-dimensional vortex structure based on the depth; Identifying the evolution trend prediction of mesoscale eddies based on the relative displacement, maximum tangential velocity, and circulation area of the three-dimensional vortex structure at consecutive moments; Based on the flow velocity and temperature data on the day when a certain moment is located, subtracting the flow velocity and temperature data on the same day in the adjacent set number of years, and extracting the temperature and sound speed anomaly characteristics of mesoscale eddies.

[0011] In some alternative solutions, determining the sea current type and vortex intensity in each layer based on the three-dimensional sea current velocity data of the target area includes: Marking the positions where the east-west direction velocity component v and the north-south direction velocity component u in the sea current velocity data of each layer change from positive to negative or from negative to positive as the mesoscale eddy boundary; Determining the type of mesoscale eddy according to whether the velocity vector inside the mesoscale eddy boundary is clockwise or counterclockwise; Determining the mesoscale eddy velocity field according to the east-west direction velocity component v and the north-south direction velocity component u of the data points inside the mesoscale eddy boundary.

[0012] In some alternative solutions, the points that meet the conditions that the direction signs on both sides of the v component along the east-west direction are opposite and the amplitude increases linearly with the distance, the direction signs on both sides of the u component along the north-south direction are opposite and the amplitude increases linearly with the distance, the mesoscale eddy boundary velocity value is the smallest, and the directions of the surrounding adjacent velocity vectors are the same are determined as the vortex centers of mesoscale eddies.

[0013] In some alternative solutions, equivalent each layer of mesoscale eddies to an irregular disk and constructing a three-dimensional vortex structure based on the depth includes: Determining that the adjacent layer disks with the same corresponding vortex type and the central vertical distance less than one-fourth of the vortex radius of the upper layer belong to the same mesoscale eddy; Construct the disks belonging to the same mesoscale eddy into a three-dimensional eddy structure.

[0014] In some alternative solutions, before obtaining the type, eddy center position, eddy intensity, and temperature and sound speed anomaly characteristics of each layer of mesoscale eddies based on the three-dimensional ocean current velocity data of the target area, traverse the three-dimensional ocean current data point by point to remove the points with incomplete or abnormal data.

[0015] In a second aspect, the present invention provides an apparatus for analyzing the spatio-temporal characteristics of mesoscale eddies based on a neural network, including: A historical feature acquisition module, which is used to obtain the type, eddy center position, eddy intensity, and temperature and sound speed anomaly characteristics of each layer of mesoscale eddies based on the three-dimensional ocean current velocity data of the target area; A neural network prediction module, which is used to predict the eddy center position, eddy intensity, and temperature and sound speed anomaly characteristics of the mesoscale eddies at a future moment based on the type, eddy center position, eddy intensity, and temperature and sound speed anomaly characteristics of each layer of mesoscale eddies at a historical moment.

[0016] Compared with the prior art, the advantages of the present invention are as follows: In this solution, based on the three-dimensional ocean current velocity data of the target area, the type, eddy center position, eddy intensity, and temperature and sound speed anomaly characteristics of each layer of mesoscale eddies are obtained, which is equivalent to having performed a round of preprocessing on the three-dimensional ocean current velocity data, can eliminate the influence of invalid or abnormal data, and the data volume will be greatly reduced, which can improve the prediction efficiency. By integrating the mesoscale eddy characteristics driven by ocean current data and the learning and reasoning ability of the neural network, the bottleneck problems existing in the traditional mesoscale eddy identification and analysis methods in terms of accuracy, stability, and dynamic tracking are comprehensively solved. Description of the Drawings

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

[0018] Figure 1 It is a flowchart of the method for analyzing the spatio-temporal characteristics of mesoscale eddies based on a neural network in an embodiment of the present invention; Figure 2 It is a map of the ocean current velocity field in an embodiment of the present invention; Figure 3 It is a three-dimensional structure diagram of mesoscale eddies in an embodiment of the present invention; Figure 4 It is a comparison diagram of sound speed profiles in an embodiment of the present invention; Figure 5 It is a comparison diagram of propagation losses in an embodiment of the present invention; Figure 6 This is a schematic diagram of the hardware structure of the mesoscale eddy spatiotemporal characteristic analysis device involved in the solution of the embodiment of the present application. Specific implementation mode

[0019] In order to enable those skilled in the art to better understand the solution of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0020] As Figure 1 shown, on the one hand, the present invention provides a method for analyzing the spatiotemporal characteristics of mesoscale eddies based on a neural network, including the following steps: S1: According to the three-dimensional velocity data of ocean currents in the target area, obtain the types, vortex center positions, vortex intensities, and temperature and sound speed anomaly characteristics of mesoscale eddies in each layer.

[0021] In this solution, in order to use a neural network to predict the types, vortex center positions, vortex intensities, and temperature and sound speed anomaly characteristics of future mesoscale eddies, eliminate the influence of abnormal ocean current data, and improve the prediction efficiency of the neural network. By obtaining the three-dimensional velocity data of ocean currents at historical moments in the target area, obtain the types, vortex center positions, vortex intensities, and temperature and sound speed anomaly characteristics of mesoscale eddies in each historical moment, and use this as the input of the neural network to predict the types, vortex center positions, vortex intensities, and temperature and sound speed anomaly characteristics of mesoscale eddies at future moments through the neural network. In this way, it is equivalent to having performed a round of preprocessing on the three-dimensional velocity data of ocean currents, which can eliminate the influence of invalid or abnormal data, and the amount of data will be greatly reduced, improving the prediction efficiency.

[0022] Specifically, step S1 includes the following steps: S11: According to the three-dimensional velocity data of ocean currents in the target area, determine the types of ocean currents in each layer and the vortex intensity, and screen out the vortex centers of mesoscale eddies based on the velocity field and constraint conditions in the vortex intensity.

[0023] In this example, the vortex intensity includes the circulation area and the maximum tangential velocity. The circulation area is determined according to the mesoscale eddy boundary; the maximum tangential velocity is determined according to the mesoscale eddy velocity field. Determining the types of ocean currents in each layer and the vortex intensity specifically includes: A: Mark the positions where the east-west direction velocity component v and the north-south direction velocity component u in the ocean current velocity data of each layer change from positive to negative or from negative to positive as the mesoscale eddy boundary.

[0024] In this example, before marking the boundaries of mesoscale vortices, the three-dimensional ocean current data is traversed point by point to eliminate points with incomplete or abnormal data, ensuring the accuracy and robustness of the analysis.

[0025] B: Determine the type of mesoscale vortex according to whether the velocity vector inside the mesoscale vortex boundary is clockwise or counterclockwise.

[0026] Specifically, if the local velocity vector inside the mesoscale vortex boundary shows counterclockwise rotation, it is a cold vortex; if it shows clockwise rotation, it is a warm vortex.

[0027] C: Determine the mesoscale vortex velocity field according to the east-west direction velocity component v and the north-south direction velocity component u of the data points inside the mesoscale vortex boundary.

[0028] As Figure 2 shown, within the identified boundary point range, calculate the velocity field:

[0029] Find the point with the minimum velocity as the candidate vortex center point, and conduct the following four constraint tests on it to confirm the vortex center position: The signs of the v components on both sides along the east-west direction are opposite and the amplitude increases linearly with distance; The signs of the u components on both sides along the north-south direction are opposite and the amplitude increases linearly with distance; There is a global minimum velocity value within the candidate area; The directions of the adjacent velocity vectors around the mesoscale vortex center are the same, falling in the same quadrant or adjacent quadrants.

[0030] In this example, the vortex centers of mesoscale vortices are screened based on the velocity field and constraint conditions in the vortex intensity, including: determining as the vortex centers of mesoscale vortices the points that meet the conditions that the signs of the v components on both sides along the east-west direction are opposite and the amplitude increases linearly with distance, the signs of the u components on both sides along the north-south direction are opposite and the amplitude increases linearly with distance, the velocity value at the mesoscale vortex boundary is the minimum, and the directions of the adjacent velocity vectors around are the same.

[0031] Perform the above calculations of the type, vortex intensity, and vortex center for each layer of three-dimensional ocean current data to obtain the type, vortex intensity, and vortex center of each layer of ocean current.

[0032] As Figure 3 shown, S12: Equivalent each layer of mesoscale vortices to an irregular disk and construct a three-dimensional vortex structure based on the depth.

[0033] In the above steps, the boundaries of each layer of mesoscale vortices have been obtained. According to the boundaries of each layer of mesoscale vortices, each layer of mesoscale vortices can be equivalent to an irregular disk to obtain the two-dimensional structure of each layer of mesoscale vortices. By matching the two-dimensional structures of each layer of mesoscale vortices, a three-dimensional vortex structure can be constructed based on the depth.

[0034] Specifically, step S12 includes: S121: Determine adjacent layers of disks with the same corresponding vortex type and a vertical center distance less than one-fourth of the radius of the vortex in the upper layer as belonging to the same mesoscale vortex.

[0035] Specifically, to determine whether the vortex structures of adjacent layers belong to the same three-dimensional mesoscale vortex, the following conditions need to be met: Type consistency: Both adjacent layers are cold vortices or warm vortices.

[0036] Spatial continuity: The vertical center distance between adjacent disks is less than one-fourth of the radius of the vortex in the upper layer.

[0037] If the above conditions are met, classify the disks of each layer into the same three-dimensional mesoscale vortex structure.

[0038] Based on the vortex type consistency and spatial distance constraints, the three-dimensional structure assembly is completed, realizing the construction of the vortex body from the two-dimensional profile to the three-dimensional structure, breaking through the limitation that the traditional layering method is difficult to characterize the overall structure.

[0039] And a tracking strategy based on the maximum propagation distance and polarity consistency is designed to effectively ensure the continuous expression of vortices in the time series, which is applicable to the multi-vortex interference or complex background flow field environment.

[0040] S122: Construct the disks belonging to the same mesoscale vortex into a three-dimensional vortex structure.

[0041] S13: Based on the relative displacement, maximum tangential velocity, and circulation area of the three-dimensional vortex structures at consecutive moments, identify the evolution trend prediction of the mesoscale vortex.

[0042] At subsequent moments, first determine whether the three-dimensional vortex structures at consecutive moments are the same three-dimensional vortex structure. Specifically, set the maximum possible movement range of the current vortex center, search for the mesoscale vortex center of the same type (polarity) within this range, and find the mesoscale vortex closest to the mesoscale vortex center at the previous moment as the matching object to achieve continuous tracking of the vortex.

[0043] When the three-dimensional vortex structures at consecutive moments are determined to be the same three-dimensional vortex structure, based on the relative displacement, maximum tangential velocity, and circulation area of the three-dimensional vortex structures at consecutive moments, identify the evolution trend prediction of the mesoscale vortex.

[0044] S14: Based on the flow velocity and temperature data on the day when a certain moment is located, subtract the flow velocity and temperature data on the same day in the adjacent set number of years to extract the temperature and sound speed anomaly characteristics of the mesoscale vortex.

[0045] In this example, use the method of subtracting the daily data from the multi-year average (such as five years) of the same day's data to eliminate the influence of seasonal cycles and extract the abnormal velocity field and temperature field of the mesoscale vortex.

[0046] As Figure 4 and Figure 5 shown, construct the temperature field profile and sound speed field profile corresponding to the mesoscale vortex for analyzing the influence of the vortex on the sound propagation environment.

[0047] S2: The trained neural network predicts the vortex center position, vortex intensity, and temperature and sound speed anomaly characteristics of the mesoscale vortex at future moments based on the types, vortex center positions, vortex intensities, and temperature and sound speed anomaly characteristics of the mesoscale vortices at each layer at historical moments.

[0048] In this example, the neural network uses a graph neural network. Specifically, a stacked graph attention network (Graph Attention Network, GAT) is used as the basic inference model. The model structure is as follows: Input layer: Receive the node feature matrix , where N is the number of vortices, the number of mesoscale vortices identified through step S1, and F is the node feature dimension, including node features such as vortex type, vortex center geographical location (longitude, latitude, depth), vortex intensity, temperature anomaly characteristics, and sound speed anomaly characteristics. That is, the N×F matrix is the input node feature, and the node features are obtained through feature extraction of the vortices in the data preprocessing stage of step S1.

[0049] And receive the edge feature matrix , which represents the connection relationship between the node features describing each vortex. The element in the adjacency matrix is used to characterize whether there is an adjacency relationship between node feature i and node feature j, where indicates that there is an edge connection between the two node features. For example, when the vortex intensity is large (>1.2) and the temperature anomaly is high (>1°C), the sound speed anomaly will also increase at the next moment; when the vortex center position shifts northward, it is usually accompanied by a decrease in intensity. indicates that there is no connection between the two nodes.

[0050] If there is a structural connection (such as spatial adjacency or temporal continuity) between node feature i and feature node j, then it is recorded as Aij = 1, otherwise Aij = 0. In the l layer of the graph attention neural network, based on the neighbor set determined by the adjacency matrix A, the neighbor node features are weighted and fused to complete the feature update operation of the target node feature i. Specifically, the multi-head attention mechanism is used to calculate the weighted representation of the relationship between adjacent nodes as:

[0051] Among them, is the attention weight, and is the relationship between node feature i and neighbor node feature j at the lThe attention weight of the layer, indicating the contribution degree of the adjacent node features to the update of node feature i. is a linear transformation matrix used to perform feature mapping on the input node features. is an activation function (such as ELU) used to enhance the expression ability of the model. is the node feature representation of node feature i in the l+1 layer. represents the feature representation of adjacent node feature j in the l layer. is the set of adjacent node features of node feature i, that is, all node features directly connected to node feature i in the graph structure.

[0052] The l-th layer refers to the l-th neural network computing unit in the graph neural network, that is, the l-th propagation layer in the multi-layer graph neural network (GAT).

[0053] The adjacency matrix and the node feature matrix are jointly input into the graph neural network as the basis for the model to learn node feature propagation and feature update. The middle layer can stack 2 - 4 layers of GAT to capture the vortex correlations across time and space.

[0054] Output layer: Output according to the task type: the type, vortex center position, vortex intensity, and temperature and sound speed anomaly features of mesoscale vortices in each layer.

[0055] Before using this graph neural network, it is necessary to train the neural network. The training of the neural network includes the following steps: Based on the type, vortex center position, vortex intensity, and temperature and sound speed anomaly features of mesoscale vortices in each layer at different times, construct a graph structure. Among them, the graph nodes of the graph structure include the type, vortex center position, vortex intensity, and temperature and sound speed anomaly features of each mesoscale vortex, and the edges of the graph structure are the adjacency relationships between the graph nodes of each mesoscale vortex.

[0056] During the processing of the graph neural network, the input graph nodes are the node feature matrix, and the edges of the graph structure are the edge feature matrix.

[0057] In this example, the training samples are obtained from the three-dimensional ocean current data of the set sea area during the set time period. The method of obtaining the type, vortex center position, vortex intensity, and temperature and sound speed anomaly features of mesoscale vortices in each layer at different times is the same as the steps in step S1.

[0058] Connecting the vortex centers that satisfy the spatial distance constraint at adjacent times refers to the vortex center at the current time and the vortex center at the previous time. The distance constraint is to set the maximum possible movement range of the current vortex center, and within this range, search for mesoscale vortex centers of the same type (polarity). The vortex bodies that satisfy the type consistency and vertical continuity constraints in adjacent depth layers at the same time refer to the two-dimensional vortex structures in two adjacent depth layers determined to be the same mesoscale vortex.

[0059] When updating the parameters of the neural network at each iteration step, the graph nodes and edges of the corresponding graph structure at the previous time are input into the graph neural network, and the predicted vortex center position, vortex intensity, and temperature and sound speed anomaly characteristics of the mesoscale vortex at the next time are output. Combining with the label data, the loss value of this iteration step is obtained, and according to the loss value, the parameters of the graph neural network are updated.

[0060] During training, the graph nodes and edges of the corresponding graph structure at the previous time are input into the graph neural network, and the predicted type, vortex center position, vortex intensity, and temperature and sound speed anomaly characteristics of the mesoscale vortex at the next time are output. Since when obtaining the training samples, the type, vortex center position, vortex intensity, and temperature and sound speed anomaly characteristics of the mesoscale vortex corresponding to all acquisition times have been obtained, therefore, the vortex center position, vortex intensity, and temperature and sound speed anomaly characteristics of the actual mesoscale vortex corresponding to the predicted vortex center position, vortex intensity, and temperature and sound speed anomaly characteristics of the mesoscale vortex at the next time by the neural network can be used as label data to obtain the loss value of this iteration step, and according to the loss value, the parameters of the graph neural network are updated.

[0061] In this embodiment, the loss value is determined according to the loss function where , , and are the intensity regression loss, the structural continuity constraint loss, and the temperature and sound speed loss hyperparameter weights respectively, is the total loss value, is the intensity regression loss regarding the vortex intensity, is the structural continuity constraint loss regarding the vortex center position, is the temperature regression loss regarding the temperature anomaly, is the sound speed regression loss regarding the sound speed anomaly.

[0062] The intensity regression loss regarding the vortex intensity ; where is the true value of the vortex intensity, is the predicted value of the vortex intensity, ‖ ‖ represents the norm of the vector. In this example, the vortex intensity specifically takes the maximum tangential velocity.

[0063] The structural continuity constraint loss regarding the vortex center position ; where is the predicted position of the vortex center at the t-th time step, is the predicted position of the vortex center at the (t + 1)-th time step, and T is the total number of time steps in the complete tracking sequence. The purpose is to constrain the smoothness of the trajectory. Instead of fitting a certain annotation, it enables the model to learn to avoid drastic jumps in the prediction results and conform to the physical continuity expectation of ocean dynamics.

[0064] Temperature regression loss regarding temperature anomaly :

[0065] is the actually observed temperature anomaly value; is the temperature anomaly value predicted by the GNN; N is the number of mesoscale vortices.

[0066] Sound speed regression loss regarding sound speed anomaly :

[0067] where: is the actually observed sound speed anomaly; is the predicted value.

[0068] In summary, this solution comprehensively solves the bottleneck problems existing in traditional mesoscale vortex identification and analysis methods in terms of accuracy, stability, three-dimensional modeling, and dynamic tracking by integrating the physical mechanism modeling driven by ocean current data and the learning and reasoning ability of neural networks.

[0069] Model the mesoscale vortex identification results as a dynamic graph structure, combine with the multi-head attention mechanism to achieve intelligent prediction of vortex types, intensities, and evolution trends, and design a composite loss function to balance the classification, regression, and path continuity objectives, significantly improving the automation and adaptability of the system.

[0070] Feature anomaly extraction strategy for removing background field perturbations: By subtracting the multi-year monthly average field strategy, effectively eliminate the seasonal cycle terms in the ocean field data, retain the mesoscale vortex anomaly features, and enhance the physical interpretation ability of environmental factors such as temperature and sound speed.

[0071] End-to-end analysis process integrated design: Construct a complete processing flow from ocean current data preprocessing, vortex identification, three-dimensional modeling, tracking analysis, GNN reasoning to environmental response evaluation, with high modularity and task customization capabilities.

[0072] In the second aspect, a mesoscale vortex spatio-temporal characteristic analysis device based on a neural network includes: a historical feature acquisition module and a neural network prediction module.

[0073] Among them, the historical feature acquisition module is used to obtain the type, vortex center position, vortex intensity, and temperature and sound speed anomaly features of mesoscale vortices in each layer based on the three-dimensional ocean current velocity data of the target area; the neural network prediction module is used to predict the vortex center position, vortex intensity, and temperature and sound speed anomaly features of mesoscale vortices at future times based on the type, vortex center position, vortex intensity, and temperature and sound speed anomaly features of mesoscale vortices in each layer at historical times.

[0074] Among them, the function implementation of each module in the above mesoscale vortex spatio-temporal characteristic analysis device based on a neural network corresponds to each step in the above mesoscale vortex spatio-temporal characteristic analysis method embodiment based on a neural network, and its function and implementation process will not be elaborated here one by one.

[0075] In a third aspect, an embodiment of the present application provides a mesoscale vortex spatio-temporal characteristic analysis device based on a neural network. The mesoscale vortex spatio-temporal characteristic analysis device based on a neural network can be a device with data processing functions such as a personal computer (PC), a laptop computer, a server, etc.

[0076] Refer to Figure 6 , Figure 6 which is a schematic hardware structure diagram of the mesoscale vortex spatio-temporal characteristic analysis device based on a neural network involved in the solution of the embodiment of the present application. In the embodiment of the present application, the mesoscale vortex spatio-temporal characteristic analysis device based on a neural network may include a processor, a memory, a communication interface, and a communication bus.

[0077] Among them, the communication bus can be of any type and is used to interconnect the processor, the memory, and the communication interface.

[0078] The communication interface includes input / output (I / O) interfaces, physical interfaces, and logical interfaces, etc., which are used to implement the interconnection of components inside the mesoscale vortex spatio-temporal characteristic analysis device based on a neural network, and interfaces for implementing the interconnection of the mesoscale vortex spatio-temporal characteristic analysis device based on a neural network with other devices (such as other computing devices or user devices). The physical interface can be an Ethernet interface, an optical fiber interface, an ATM interface, etc.; the user device can be a display screen (Display), a keyboard (Keyboard), etc.

[0079] The memory can be various types of storage media, such as random access memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), flash memory, optical memory, hard disk, programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), etc.

[0080] The processor can be a general-purpose processor, which can call the program for analyzing the spatio-temporal characteristics of mesoscale vortices based on neural networks stored in the memory and execute the method for analyzing the spatio-temporal characteristics of mesoscale vortices based on neural networks provided in the embodiments of the present application. For example, the general-purpose processor can be a central processing unit (CPU). Among them, the method executed when the program for analyzing the spatio-temporal characteristics of mesoscale vortices based on neural networks is called can refer to the various embodiments of the method for analyzing the spatio-temporal characteristics of mesoscale vortices based on neural networks in the present application, which will not be elaborated here.

[0081] Those skilled in the art can understand that Figure 6 the hardware structure shown in does not constitute a limitation to the present application, and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0082] In a fourth aspect, the embodiments of the present application further provide a computer-readable storage medium.

[0083] The computer-readable storage medium of the present application stores a program for analyzing the spatio-temporal characteristics of mesoscale vortices based on neural networks. When the program for analyzing the spatio-temporal characteristics of mesoscale vortices based on neural networks is executed by a processor, the steps of the method for analyzing the spatio-temporal characteristics of mesoscale vortices based on neural networks as described above are implemented.

[0084] Among them, the method implemented when the program for analyzing the spatio-temporal characteristics of mesoscale vortices based on neural networks is executed can refer to the various embodiments of the method for analyzing the spatio-temporal characteristics of mesoscale vortices based on neural networks in the present application, which will not be elaborated here.

[0085] It should be noted that the serial numbers of the above embodiments of the present application are only for description and do not represent the advantages or disadvantages of the embodiments.

[0086] In the description of the specification, claims and the above-mentioned drawings of this application, the terms "comprising", "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally further includes steps or units not listed, or optionally further includes other steps or units inherent to these processes, methods, products or devices. The descriptions such as "first", "second" and "third" are used to distinguish different objects, etc., and do not represent a sequence, nor do they limit that "first", "second" and "third" are different types.

[0087] In the description of the embodiments of this application, "exemplary", "for example" or "for instance" are used to indicate examples, illustrations or explanations. Any embodiment or design solution described as "exemplary", "for example" or "for instance" in the embodiments of this application should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Rather, the use of words such as "exemplary", "for example" or "for instance" is intended to present relevant concepts in a specific manner.

[0088] In the description of the embodiments of this application, unless otherwise specified, " / " means "or". For example, A / B may mean A or B; "and / or" in the text is merely a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, in the description of the embodiments of this application, "a plurality of" means two or more than two.

[0089] In some processes described in the embodiments of this application, there are multiple operations or steps that appear in a specific order. However, it should be understood that these operations or steps may not be executed in the order in which they appear in the embodiments of this application or may be executed in parallel. The serial numbers of the operations are only used to distinguish different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations or steps may be executed in sequence or in parallel, and these operations or steps may be combined.

[0090] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-described embodiment methods can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium as described above (such as ROM / RAM, magnetic disk, optical disc), and includes several instructions for causing a terminal device to execute the methods described in the various embodiments of this application.

[0091] The above are only the preferred embodiments of the present application, and do not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made by using the content of the specification and drawings of the present application, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present application.

Claims

1. A method for analyzing the spatiotemporal characteristics of mesoscale eddies based on neural networks, characterized in that: The following steps are involved: Based on the three-dimensional velocity data of ocean currents in the target area, the types, vortex center positions, vortex strengths, and temperature and sound velocity anomaly characteristics of each layer of mesoscale eddies are obtained; The trained neural network predicts the vortex center position, vortex intensity, and temperature and sound speed anomaly characteristics of mesoscale vortices at future moments based on the types, vortex center positions, vortex intensity, and temperature and sound speed anomaly characteristics of each layer of mesoscale vortices at historical moments.

2. The method for analyzing the spatiotemporal characteristics of mesoscale eddies based on neural networks according to claim 1, characterized in that: Training a neural network consists of the following steps: Based on the type, vortex center position, vortex intensity, and temperature and sound speed anomaly characteristics of each layer of mesoscale vortex at different times, a graph structure is constructed, wherein the graph nodes of the graph structure include the type, vortex center position, vortex intensity, and temperature and sound speed anomaly characteristics of each mesoscale vortex, and the edges of the graph structure are the adjacency relationships between the graph nodes of each mesoscale vortex; When the parameters of the neural network are iteratively updated in each step, the graph nodes and edges corresponding to the graph structure of the previous moment are input into the graph neural network, and the predicted type, vortex center position, vortex intensity, and temperature and sound speed anomaly characteristics of the mesoscale vortex at the next moment are output. The loss value of the iterative step is obtained in combination with the label data, and the parameters of the graph neural network are updated according to the loss value.

3. The method for analyzing the spatiotemporal characteristics of mesoscale eddies based on neural networks as claimed in claim 2, characterized in that: The loss value is based on the loss function Determine, among which, , , and are the hyperparameter weights of strength regression loss, structural continuity constraint loss, temperature regression and sound speed regression loss, respectively. is the total loss value, is the intensity regression loss with respect to eddy strength, is the structural continuity constraint loss about the vortex center position, is the temperature regression loss for temperature anomalies, About the sound velocity regression loss of sound velocity anomaly.

4. The method for analyzing the spatiotemporal characteristics of mesoscale eddies based on a neural network as claimed in claim 3, characterized in that: Intensity regression loss on eddy strength , in, is the true value of the eddy strength, is the predicted value of eddy strength, N is the number of mesoscale eddies; Structural continuity constraint loss at the vortex center , in, is the predicted position of the vortex center at time step t, is the predicted position of the vortex center at the t+1 time step, and T is the total number of time steps in the tracking sequence; Temperature regression loss with respect to temperature anomalies , in, is the actual observed temperature anomaly value; is the predicted temperature anomaly; About the sound velocity regression loss of sound velocity anomaly in, To actually observe the sound velocity anomaly; To predict the outlier value of sound speed.

5. The method for analyzing the spatiotemporal characteristics of mesoscale eddies based on neural networks according to claim 1, characterized in that: The method of obtaining the type, vortex center position, vortex intensity, and temperature and sound velocity anomaly characteristics of each layer of mesoscale vortexes based on the three-dimensional velocity data of the ocean current in the target area includes: According to the three-dimensional velocity data of the ocean current in the target area, the types of ocean currents and the eddy strength of each layer are determined, and the eddy core of the mesoscale eddy is screened out based on the velocity field and constraint conditions in the eddy strength; The mesoscale vortices in each layer are equivalent to irregular disks, and the three-dimensional vortex structure is constructed based on the depth. Based on the relative displacement, maximum tangential velocity and circulation area of ​​the three-dimensional vortex structure at consecutive moments, the evolution trend of the mesoscale vortex is identified; Based on the velocity and temperature data of a certain day at a certain moment, the velocity and temperature data of the same day in the nearby set year are subtracted to extract the temperature and sound speed anomaly characteristics of the mesoscale vortex.

6. The method for analyzing the spatiotemporal characteristics of mesoscale eddies based on neural networks according to claim 5, characterized in that: Determining the types of currents and eddy strengths of each layer according to the three-dimensional velocity data of the currents in the target area includes: The positions corresponding to when the east-west velocity component v and the north-south velocity component u in each layer of ocean current velocity data change from positive to negative or from negative to positive are marked as mesoscale eddy boundaries; Determine the type of mesoscale vortex according to whether the velocity vector within the mesoscale vortex boundary is clockwise or counterclockwise; The mesoscale eddy velocity field is determined based on the east-west velocity component v and the north-south velocity component u of the data points within the mesoscale eddy boundary.

7. The method for analyzing the spatiotemporal characteristics of mesoscale eddies based on neural networks according to claim 5, characterized in that: The point that meets the conditions that the directions of the v component on both sides in the east-west direction are opposite in sign and the amplitude increases linearly with distance, the directions of the u component on both sides in the north-south direction are opposite in sign and the amplitude increases linearly with distance, the velocity value of the mesoscale vortex boundary is the smallest, and the directions of the surrounding adjacent velocity vectors are consistent is determined as the vortex center of the mesoscale vortex.

8. The method for analyzing the spatiotemporal characteristics of mesoscale eddies based on neural networks according to claim 5, characterized in that: The method of converting each layer of mesoscale vortex into an irregular disk and constructing a three-dimensional vortex structure based on depth includes: Adjacent layers of disks with the same corresponding vortex type and a vertical distance between their centers less than one-fourth of the radius of the previous layer of vortex are judged to belong to the same mesoscale vortex; The disks belonging to the same mesoscale vortex are constructed into a three-dimensional vortex structure.

9. The method for analyzing the spatiotemporal characteristics of mesoscale eddies based on neural networks according to claim 1, characterized in that: Before obtaining the type, eddy center position, eddy intensity, and temperature and sound speed anomaly characteristics of each layer of mesoscale eddies based on the three-dimensional ocean current velocity data in the target area, the three-dimensional ocean current data is traversed point by point to eliminate points with incomplete or abnormal data.

10. A device for analyzing the spatiotemporal characteristics of mesoscale eddies based on neural networks, characterized in that: include: A historical feature acquisition module is used to obtain the type, vortex center position, vortex intensity, and temperature and sound speed anomaly characteristics of each layer of mesoscale vortex based on the three-dimensional velocity data of the ocean current in the target area; The neural network prediction module is used to predict the vortex center position, vortex intensity, and temperature and sound speed anomaly characteristics of the mesoscale vortex at future moments based on the types, vortex center positions, vortex intensity, and temperature and sound speed anomaly characteristics of each layer of mesoscale vortex at historical moments.

Images