A typhoon path similarity analysis method based on path and environment field fusion

Abstract

The application discloses a typhoon path similarity analysis method based on path and environmental field fusion, and relates to the technical field of meteorological information, and comprises the following steps: S1, obtaining real-time path data and future forecast path data of a current typhoon, life history path data of a plurality of historical typhoons, environmental field forecast data corresponding to each moment of the current typhoon, and environmental field reanalysis data corresponding to each moment of the plurality of historical typhoons; the application fuses path and environmental field information, constructs a current typhoon environmental space-time graph and a historical typhoon environmental space-time graph, effectively alleviates the problem that a traditional method only depends on path geometry and ignores the influence of an environmental field, leading to inaccurate similarity judgment, and significantly improves the accuracy and physical consistency of typhoon path similarity analysis.

Description

Technical Field

[0001] This invention relates to the field of meteorological information technology, and in particular to a method for typhoon path similarity analysis based on the fusion of path and environmental field. Background Technology

[0002] In typhoon track forecasting and risk assessment, accurately measuring the similarity between two typhoon tracks is crucial. Existing typhoon track similarity analysis methods primarily rely on the geometric shape of the tracks, measuring their spatial proximity by calculating indicators such as Euclidean distance, Hausdorff distance, and dynamic time warping between track points. However, these methods, which measure spatial proximity by calculating Euclidean distance, Hausdorff distance, and dynamic time warping, depend solely on the typhoon center's position sequence, neglecting key environmental factors influencing typhoon movement and evolution, such as wind field, geopotential height field, temperature field, humidity field, vorticity field, and divergence field.

[0003] Currently, methods for measuring the spatial proximity of paths by calculating metrics such as Euclidean distance, Hausdorff distance, and dynamic time warping between path points struggle to effectively integrate environmental field information into path similarity metrics. This is primarily because environmental field data is distributed across latitude and longitude grids, which are coupled with the location of the typhoon center. Directly comparing grid data from different geographical locations fails to reflect the local distribution characteristics of the environmental field relative to the typhoon center. Furthermore, the direction and speed of a typhoon's movement are closely related to its large-scale environmental field; two geometrically similar paths may correspond to drastically different environmental field configurations, leading to significant differences in their actual trajectories and resulting in inaccurate similarity assessments. Summary of the Invention

[0004] To address the technical problems existing in the background art, this invention proposes a typhoon path similarity analysis method based on the fusion of path and environmental field.

[0005] This invention proposes a typhoon path similarity analysis method based on path and environmental field fusion, comprising the following steps: S1. Obtain the current typhoon's real-time path data and future forecast path data, the life history path data of multiple historical typhoons, the environmental field forecast data corresponding to each moment of the current typhoon, and the environmental field reanalysis data corresponding to each moment of multiple historical typhoons. S2. The real-time path data, future forecast path data and life history path data of multiple historical typhoons are processed by isochronization with a fixed time step to obtain the current typhoon isochronous path sequence and multiple historical typhoon isochronous path sequences. S3. For each moment in the current typhoon isochronous path sequence, a grid generation mechanism is used to generate the corresponding first local environmental field grid; for each moment in multiple historical typhoon isochronous path sequences, a grid generation mechanism is used to generate the corresponding second local environmental field grid. S4. Generate a spatiotemporal map of the current typhoon environment based on the first local environmental field grid corresponding to the current typhoon isochronous path sequence; generate a spatiotemporal map of the historical typhoon environment based on the second local environmental field grid corresponding to multiple historical typhoon isochronous path sequences. S5. Based on the first local environmental field grid and the second local environmental field grid, generate multiple feature vectors to form a feature vector set; S6. Generate the current sequence embedding representation based on the feature vector set and the current typhoon environment spatiotemporal map; S7. Generate a set of historical sequence embedding representations based on the feature vector set and multiple historical typhoon environmental spatiotemporal maps; S8. Obtain the similarity between the current sequence embedding representation and each historical sequence embedding representation in the set of historical sequence embedding representations, and use it as the path similarity between the current typhoon and the historical typhoon corresponding to that historical sequence embedding representation.

[0006] Preferably, in S1, the environmental field forecast data and the environmental field reanalysis data both include meteorological variable data at multiple different altitude levels and latitude and longitude coordinate data corresponding to the meteorological variable data at multiple different altitude levels; Associate meteorological variable data at multiple different altitude levels with their corresponding latitude and longitude coordinates; Meteorological variable data include wind field data, geopotential height field data, temperature field data, humidity field data, eddy current field data, and divergence field data; As an explanation, meteorological variable data from multiple different altitude levels need to be standardized before use, or the meteorological variable data from multiple different altitude levels are themselves standardized data.

[0007] Preferably, in S2, the actual path data, the future forecast path data, and the life history path data of multiple historical typhoons are processed by isochronization with a fixed time step to obtain the current typhoon isochronous path sequence and multiple historical typhoon isochronous path sequences, as follows: Based on a fixed time step, a linear interpolation method is used on the real-time path data to obtain the interpolated real-time path sequence. Based on a fixed time step, a linear interpolation method is used on the future forecast track data to obtain the interpolated future forecast track sequence; As an explanation, the purpose is to use linear interpolation with a fixed time step as the target to convert the original non-isochronous real track data and future forecast track data into real track sequences and future forecast track sequences with equal time intervals. The actual track sequence and the future forecast track sequence are spliced ​​together in chronological order to form the current typhoon isochronous track sequence; Based on a fixed time step, linear interpolation is used on the life history path data of each historical typhoon to obtain the interpolated isochronous path sequence of historical typhoons. Obtain all historical typhoon isochronous path sequences and form multiple historical typhoon isochronous path sequences that correspond one-to-one with the life history path data of multiple historical typhoons. As an explanation, the fixed time step can be determined based on the original data time interval of the current typhoon's real-time path data. For example, the commonly selected original data time interval is 6 hours. In order to ensure that the data is aligned on a unified time grid, this embodiment can set the fixed time step to 6 hours.

[0008] Preferably, in S3, for each moment in the current typhoon isochronous path sequence, a grid generation mechanism is used to generate the corresponding first local environmental field grid, as follows: For each moment in the current typhoon isochronous path sequence, establish a local tangential plane coordinate system with the latitude and longitude coordinates of the typhoon center corresponding to that moment in the current typhoon isochronous path sequence as the origin. The local tangential plane coordinate system is a station-centered coordinate system with the latitude and longitude coordinates of the typhoon center as the origin. The latitude and longitude coordinates of the typhoon center are converted into plane rectangular coordinates under the station-centered coordinate system. Thus, in the local tangential plane coordinate system, the spatial description of the typhoon environmental field is decoupled from the geographical location of the typhoon center, and the spatial perturbation invariance of the environmental field representation is achieved. Based on the latitude and longitude coordinates in the environmental field forecast data corresponding to that moment, the latitude and longitude coordinates are mapped to the local tangential plane coordinate system corresponding to that moment through map projection transformation, resulting in a set of plane coordinates corresponding to meteorological variable data at multiple different altitude levels. The set of plane coordinates includes multiple plane coordinates. In the local tangential plane coordinate system corresponding to that moment, a circular region is drawn with the origin as the center and a fixed physical distance R as the radius. All plane coordinates within this circular region are extracted as the selected plane coordinates. Obtain meteorological variable data for each altitude layer corresponding to all selected plane coordinates, and use this data as the selected meteorological variable data. A grid generation technique is used to construct an N-row × N-column regular grid covering the circular area; the selected meteorological variable data is interpolated to the N-row × N-column regular grid to obtain the first local environmental field grid corresponding to that moment. N is a positive integer greater than 1; The fixed physical distance R is greater than or equal to 1 km; As an explanation, in typhoon research and operational forecasting, the radius of 500 km to 1500 km is generally considered the key area affecting typhoon movement and changes. Therefore, the fixed physical distance R can be set within this range of 500 km to 1500 km, such as 500 km, 800 km, or 1000 km, or the fixed physical distance R can be used as a configurable parameter.

[0009] Preferably, in S3, for each moment in the sequence of multiple historical typhoon isochronous paths, a grid generation mechanism is used to generate a corresponding second local environmental field grid, as follows: For each moment in a historical typhoon isochronous path sequence, a local tangential plane coordinate system is established with the latitude and longitude coordinates of the typhoon center corresponding to that moment in the historical typhoon isochronous path sequence as the origin. The local tangential plane coordinate system is a station-centered coordinate system with the latitude and longitude coordinates of the typhoon center as the origin. The latitude and longitude coordinates of the typhoon center are converted into plane rectangular coordinates under the station-centered coordinate system. Thus, in the local tangential plane coordinate system, the spatial description of the typhoon environmental field is decoupled from the geographical location of the typhoon center, and the spatial perturbation invariance of the environmental field representation is achieved. Based on the latitude and longitude coordinates in the environmental field reanalysis data corresponding to that moment, the latitude and longitude coordinates are mapped to the local tangential plane coordinate system corresponding to that moment through map projection transformation, resulting in a set of plane coordinates corresponding to meteorological variable data at multiple different altitude levels. The set of plane coordinates includes multiple plane coordinates. In the local tangential plane coordinate system corresponding to that moment, a circular region is drawn with the origin as the center and a fixed physical distance R as the radius. All plane coordinates within this circular region are extracted as the selected plane coordinates. Obtain meteorological variable data for each altitude layer corresponding to all selected plane coordinates, and use this data as the selected meteorological variable data. A grid generation technique is used to construct an N-row × N-column regular grid covering the circular area; the selected meteorological variable data is interpolated into the N-row × N-column regular grid using an interpolation method to obtain the second local environmental field grid corresponding to that moment.

[0010] Preferably, in S4, a spatiotemporal map of the current typhoon environment is generated based on the first local environmental field grid corresponding to the current typhoon isochronous path sequence, as follows: For a given moment in the current typhoon isochronous path sequence, obtain the first local environmental field grid corresponding to that moment; For the first local environmental field grid, each grid point in the first local environmental field grid is taken as a grid node, and each grid node carries the meteorological variable data of that grid node, forming a set of grid nodes; For the local tangential plane coordinate system corresponding to the first local environmental field grid, extract the origin in this local tangential plane coordinate system, and add this origin as the typhoon center node to the grid node set to obtain the graph node set; Obtain the row indices and column indices of all grid nodes in the graph node set in the local environmental field grid; for any two grid nodes in the graph node set, if the row indices of the two grid nodes are the same and the column indices differ by 1, or the column indices are the same and the row indices differ by 1, or both the row indices and column indices differ by 1, then determine that the two grid nodes are adjacent, and establish a local adjacent edge between the two grid nodes; obtain all local adjacent edges to form a local adjacent edge set; Preset two radius thresholds as r1 and r2, and 0 < r1 < r2 < fixed physical distance R; For each grid node in the graph node set, obtain the plane coordinates of this grid node in the corresponding local tangential plane coordinate system as the node coordinates; Obtain the distance value from the node coordinates to the typhoon center node; As an illustration, the distance value from the node coordinates to the typhoon center node can be obtained through the Euclidean distance formula; According to the relationship between the distance value and r1, r2, divide all grid nodes in the graph node set into multiple annuli: According to the relationship between the distance value and r1, r2, divide all grid nodes in the graph node set into multiple annuli as follows: If the distance value ≤ r1, then this grid node belongs to the first annulus; If r1 < distance value < r2, then this grid node belongs to the second annulus; If the distance value ≥ r2, then this grid node belongs to the third annulus; For any two grid nodes in the graph node set, if the two grid nodes belong to the same annulus or adjacent annuli, then establish a circumferential multi-scale adjacent edge between the two grid nodes; As an illustration, adjacent annuli refer to two adjacent annuli, for example, the first annulus and the second annulus are adjacent annuli, and the second annulus and the third annulus are adjacent annuli.

[0011] Obtain all circumferential multi-scale adjacent edges to form a circumferential multi-scale adjacent edge set; In the graph node set, establish an interaction edge between the typhoon center node and each grid node; for each grid node in the graph node set, according to the distance value from the node coordinates to the typhoon center node, use the Gaussian radial basis function to obtain the interaction weight between this grid node and the typhoon center node as the edge weight of the interaction edge; Obtain all interaction edges to form an interaction edge set; Take the union of the local adjacent edge set, the circumferential multi-scale adjacent edge set, and the interaction edge set to obtain the spatial edge set at this moment; Obtain the typhoon center nodes at adjacent moments in the current typhoon isochronous path sequence at this moment; Establish a time edge between the typhoon center node at this moment and the typhoon center nodes at adjacent moments; Obtain all time edges to form a time edge set; Obtain all moments corresponding to the current typhoon isochronous path sequence; Regard all grid nodes and typhoon center nodes in the graph node set at each moment as spatio-temporal nodes; Regard the local adjacent edges, circumferential multi-scale adjacent edges, and interaction edges in the spatial edge set at each moment as the first spatio-temporal edges between spatio-temporal nodes; Regard the time edges in the time edge set as the second spatio-temporal edges between two typhoon center nodes at adjacent moments; Form the current typhoon environmental spatio-temporal graph through spatio-temporal nodes, the first spatio-temporal edges, and the second spatio-temporal edges.

[0012] Preferably, in S4, generate a historical typhoon environmental spatio-temporal graph according to the second local environmental field grids corresponding to multiple historical typhoon isochronous path sequences, as follows: For a certain moment in a certain historical typhoon isochronous path sequence, obtain the second local environmental field grid corresponding to this moment; For this second local environmental field grid, take each grid point in this second local environmental field grid as a grid node, and each grid node carries meteorological variable data of this grid node to form a grid node set; For the local tangential plane coordinate system corresponding to this second local environmental field grid, extract the origin in this local tangential plane coordinate system, and add this origin as a typhoon center node to the grid node set to obtain a graph node set; Obtain the row index and column index of all grid nodes in the graph node set in the local environmental field grid; for any two grid nodes in the graph node set, if the row indices of the two grid nodes are the same and the column indices differ by 1, or the column indices are the same and the row indices differ by 1, or both the row index and the column index differ by 1, then determine that the two grid nodes are adjacent, and establish a local adjacent edge between the two grid nodes; obtain all local adjacent edges to form a local adjacent edge set; Preset two radius thresholds as r1 and r2, and 0 < r1 < r2 < fixed physical distance R; For each grid node in the graph node set, obtain the plane coordinate of this grid node in the corresponding local tangential plane coordinate system as the node coordinate; Obtain the distance value from the node coordinate to the typhoon center node; As an illustration, the distance value from the node coordinates to the typhoon center node can be obtained through the Euclidean distance formula; According to the relationship between the distance value and r1, r2, all grid nodes in the graph node set are divided into multiple annular zones: According to the relationship between the distance value and r1, r2, all grid nodes in the graph node set are divided into multiple annular zones as follows: If the distance value ≤ r1, then this grid node belongs to the first annular zone; If r1 < distance value < r2, then this grid node belongs to the second annular zone; If the distance value ≥ r2, then this grid node belongs to the third annular zone; For any two grid nodes in the graph node set, if these two grid nodes belong to the same annular zone or adjacent annular zones, then a circumferential multi-scale adjacent edge is established between these two grid nodes; As an illustration, adjacent annular zones refer to two adjacent annular zones. For example, the first annular zone and the second annular zone are adjacent annular zones, and the second annular zone and the third annular zone are adjacent annular zones.

[0013] Obtain all circumferential multi-scale adjacent edges to form a circumferential multi-scale adjacent edge set; In the graph node set, an interaction edge is established between the typhoon center node and each grid node; for each grid node in the graph node set, according to the distance value from the node coordinates to the typhoon center node, the Gaussian radial basis function is used to obtain the interaction weight between this grid node and the typhoon center node, which is used as the edge weight of the interaction edge; Obtain all interaction edges to form an interaction edge set; Take the union of the local adjacent edge set, the circumferential multi-scale adjacent edge set and the interaction edge set to obtain the spatial edge set at this moment; Obtain the typhoon center node at the adjacent moment in the historical typhoon isochronous path sequence at this moment; Establish a time edge between the typhoon center node at this moment and the typhoon center node at the adjacent moment; Obtain all time edges to form a time edge set; Obtain all moments corresponding to this historical typhoon isochronous path sequence; Take all grid nodes and the typhoon center node in the graph node set at each moment as spatio-temporal nodes; Take the local adjacent edges, circumferential multi-scale adjacent edges, and interaction edges in the spatial edge set at each moment as the first spatio-temporal edges between spatio-temporal nodes; Take the time edges in the time edge set as the second spatio-temporal edges between two typhoon center nodes at adjacent moments; Form a historical typhoon environment spatio-temporal graph through spatio-temporal nodes, first spatio-temporal edges and second spatio-temporal edges.

[0014] Preferably, in S5, multiple feature vectors are generated based on the first local environmental field grid and the second local environmental field grid to form a feature vector set, as follows: Obtain all grid nodes corresponding to the first local environmental field grid and the second local environmental field grid; Extract wind field data from meteorological variable data of a specific grid node; The geometric projection method is used to convert wind field data into radial and tangential components relative to the origin in the local coordinate system. The radial and tangential components are concatenated with geopotential height, temperature, humidity, vorticity, and divergence data from meteorological variable data to form the feature vector of the grid node. Obtain the feature vectors of all grid nodes to form a feature vector set.

[0015] Preferably, in S6, based on the feature vector set and the current typhoon environment spatiotemporal map, the current sequence embedding representation is generated as follows: Input the current typhoon environment spatiotemporal map into the graph neural network, and aggregate the feature vectors of all grid nodes connected to the typhoon center node through interactive edges at each moment to obtain the environmental field features at each moment. For each moment in the current typhoon isochronous path sequence, cubic spline interpolation is used to fit the current typhoon isochronous path sequence to obtain a continuous path function; based on the path function, the position, tangential direction, and curvature of the typhoon center node at each moment in the current typhoon isochronous path sequence are obtained as the path features at each moment. By concatenating the environmental field features and path features at each moment, the embedding vector of the typhoon center node at each moment is obtained. The embedding vectors of the typhoon center nodes at each time point are arranged in temporal order and then input into a long short-term memory network to obtain the current sequence embedding representation for similarity retrieval.

[0016] Preferably, in step S7, a set of historical sequence embedding representations is generated based on the feature vector set and multiple historical typhoon environmental spatiotemporal maps, as follows: Based on multiple historical typhoon environmental spatiotemporal maps, a certain historical typhoon environmental spatiotemporal map is input into a graph neural network. The graph neural network aggregates the feature vectors of all grid nodes connected to the typhoon center node through interactive edges at each moment to obtain the environmental field features at each moment. For each moment in the historical typhoon isochronous path sequence corresponding to the historical typhoon environmental spatiotemporal map, cubic spline interpolation is used to fit the historical typhoon isochronous path sequence to obtain a continuous path function; based on the path function, the position, tangential direction, and curvature of the typhoon center node at each moment in the historical typhoon isochronous path sequence are obtained as the path features at each moment. By concatenating the environmental field features and path features at each moment, the embedding vector of the typhoon center node at each moment is obtained. The embedding vectors of the typhoon center nodes at each time point are arranged in temporal order and then input into a long short-term memory network to obtain the historical sequence embedding representation for similarity retrieval. Obtain all historical sequence embeddings to form a set of historical sequence embeddings.

[0017] The typhoon path similarity analysis method based on path and environmental field fusion proposed in this invention has the following beneficial technical effects: This application obtains current and historical typhoon path data and corresponding environmental field reanalysis data, performs isochronization processing at a fixed time step to obtain the current typhoon isochronous path sequence and multiple historical typhoon isochronous path sequences, and establishes a local tangential plane coordinate system with the typhoon center as the origin at each moment. A local environmental field grid is generated by interpolating a circular region at a fixed physical distance, achieving spatial perturbation invariance in the environmental field representation. Furthermore, based on the local environmental field grid, it constructs a current typhoon environmental spatiotemporal map and historical typhoon environmental spatiotemporal maps containing grid nodes and typhoon center nodes, through local adjacency edges, circumferential multi-scale adjacency edges, and interactive... This paper describes the correlation between edges and time edges, and constructs a feature vector by splicing the radial and tangential components of the wind field and other meteorological variable data. An environmental field feature is obtained by aggregating the feature vectors of all grid nodes connected to the typhoon center node through interactive edges using a graph neural network. Simultaneously, cubic spline interpolation is used to extract position, tangential direction, and curvature as path features from isochronous path sequences or historical typhoon isochronous path sequences. These two features are then spliced ​​to obtain the embedding vector of the typhoon center node at each time point. This vector is then processed by a long short-term memory network to obtain a sequence embedding representation. Finally, the similarity between the current sequence embedding representation and the historical sequence embedding representation is used as the path similarity. This application, by fusing path and environmental field information, constructs current and historical typhoon environmental spatiotemporal maps, effectively alleviating the problem of inaccurate similarity judgments caused by traditional methods that rely solely on path geometry while ignoring the influence of the environmental field. This significantly improves the accuracy and physical consistency of typhoon path similarity analysis. Attached Figure Description

[0018] Figure 1 This is a flowchart of a typhoon path similarity analysis method based on path and environmental field fusion according to the present invention. Detailed Implementation

[0019] Embodiments of the present invention are described in detail below. Examples of these embodiments are illustrated in the accompanying drawings, wherein the same or similar symbols denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0020] like Figure 1 The method for typhoon path similarity analysis based on path and environmental field fusion, as shown, includes the following steps: S1. Obtain the current typhoon's real-time path data and future forecast path data, the life history path data of multiple historical typhoons, the environmental field forecast data corresponding to each moment of the current typhoon, and the environmental field reanalysis data corresponding to each moment of multiple historical typhoons. In S1, the actual track data, the future forecast track data, and the life history track data of multiple historical typhoons all include the latitude and longitude coordinates of the typhoon center. In an optional embodiment, in S1, the environmental field forecast data and the environmental field reanalysis data both include meteorological variable data at multiple different altitude levels and latitude and longitude coordinate data corresponding to the meteorological variable data at multiple different altitude levels. Associate meteorological variable data at multiple different altitude levels with their corresponding latitude and longitude coordinates; Meteorological variable data include wind field data, geopotential height field data, temperature field data, humidity field data, eddy current field data, and divergence field data; As an explanation, meteorological variable data from multiple different altitude levels need to be standardized when used, or the meteorological variable data from multiple different altitude levels are already standardized data. S2. The real-time path data, future forecast path data and life history path data of multiple historical typhoons are processed by isochronization with a fixed time step to obtain the current typhoon isochronous path sequence and multiple historical typhoon isochronous path sequences. In an optional embodiment, in S2, the actual path data, the future forecast path data, and the life history path data of multiple historical typhoons are processed by isochronization at a fixed time step to obtain the current typhoon isochronous path sequence and multiple historical typhoon isochronous path sequences, as follows: Based on a fixed time step, a linear interpolation method is used on the real-time path data to obtain the interpolated real-time path sequence. Based on a fixed time step, a linear interpolation method is used on the future forecast track data to obtain the interpolated future forecast track sequence; As an explanation, the purpose is to use linear interpolation with a fixed time step as the target to convert the original non-isochronous real track data and future forecast track data into real track sequences and future forecast track sequences with equal time intervals. The actual track sequence and the future forecast track sequence are spliced ​​together in chronological order to form the current typhoon isochronous track sequence; Based on a fixed time step, linear interpolation is used on the life history path data of each historical typhoon to obtain the interpolated isochronous path sequence of historical typhoons. Obtain all historical typhoon isochronous path sequences and form multiple historical typhoon isochronous path sequences that correspond one-to-one with the life history path data of multiple historical typhoons. As an explanation, the fixed time step can be determined based on the original data time interval of the current typhoon's real-time path data. For example, the commonly selected original data time interval is 6 hours. In order to ensure that the data is aligned on a unified time grid, this embodiment can set the fixed time step to 6 hours.

[0021] S3. For each moment in the current typhoon isochronous path sequence, a grid generation mechanism is used to generate the corresponding first local environmental field grid; For each moment in multiple historical typhoon isochronous path sequences, a grid generation mechanism is used to generate the corresponding second local environmental field grid; In an optional embodiment, in S3, for each moment in the current typhoon isochronous path sequence, a grid generation mechanism is used to generate the corresponding first local environmental field grid, as follows: For each moment in the current typhoon isochronous path sequence, establish a local tangential plane coordinate system with the latitude and longitude coordinates of the typhoon center corresponding to that moment in the current typhoon isochronous path sequence as the origin. The local tangential plane coordinate system is a station-centered coordinate system with the latitude and longitude coordinates of the typhoon center as the origin. The latitude and longitude coordinates of the typhoon center are converted into plane rectangular coordinates under the station-centered coordinate system. Thus, in the local tangential plane coordinate system, the spatial description of the typhoon environmental field is decoupled from the geographical location of the typhoon center, and the spatial perturbation invariance of the environmental field representation is achieved. Based on the latitude and longitude coordinates in the environmental field forecast data corresponding to that moment, the latitude and longitude coordinates are mapped to the local tangential plane coordinate system corresponding to that moment through map projection transformation, resulting in a set of plane coordinates corresponding to meteorological variable data at multiple different altitude levels. The set of plane coordinates includes multiple plane coordinates. In the local tangential plane coordinate system corresponding to that moment, a circular region is drawn with the origin as the center and a fixed physical distance R as the radius. All plane coordinates within this circular region are extracted as the selected plane coordinates. Obtain meteorological variable data for each altitude layer corresponding to all selected plane coordinates, and use this data as the selected meteorological variable data. A grid generation technique is used to construct an N-row × N-column regular grid covering the circular area; the selected meteorological variable data is interpolated to the N-row × N-column regular grid to obtain the first local environmental field grid corresponding to that moment. In an optional embodiment, the selected meteorological variable data is interpolated to an N-row × N-column regular grid using an interpolation method, which may be bilinear interpolation, inverse distance weighted interpolation, or radial basis function interpolation. N is a positive integer greater than 1; The fixed physical distance R is greater than or equal to 1 km; As an explanation, in typhoon research and operational forecasting, the radius of 500 km to 1500 km is generally considered the key area affecting typhoon movement and changes. Therefore, the fixed physical distance R can be set within this range of 500 km to 1500 km, such as 500 km, 800 km, or 1000 km, or the fixed physical distance R can be used as a configurable parameter.

[0022] In an optional embodiment, in S3, for each moment in the multiple historical typhoon isochronous path sequences, a grid generation mechanism is used to generate a corresponding second local environmental field grid, as follows: For each moment in a historical typhoon isochronous path sequence, a local tangential plane coordinate system is established with the latitude and longitude coordinates of the typhoon center corresponding to that moment in the historical typhoon isochronous path sequence as the origin. The local tangential plane coordinate system is a station-centered coordinate system with the latitude and longitude coordinates of the typhoon center as the origin. The latitude and longitude coordinates of the typhoon center are converted into plane rectangular coordinates under the station-centered coordinate system. Thus, in the local tangential plane coordinate system, the spatial description of the typhoon environmental field is decoupled from the geographical location of the typhoon center, and the spatial perturbation invariance of the environmental field representation is achieved. Based on the latitude and longitude coordinates in the environmental field reanalysis data corresponding to that moment, the latitude and longitude coordinates are mapped to the local tangential plane coordinate system corresponding to that moment through map projection transformation, resulting in a set of plane coordinates corresponding to meteorological variable data at multiple different altitude levels. The set of plane coordinates includes multiple plane coordinates. In the local tangential plane coordinate system corresponding to that moment, a circular region is drawn with the origin as the center and a fixed physical distance R as the radius. All plane coordinates within this circular region are extracted as the selected plane coordinates. Obtain meteorological variable data for each altitude layer corresponding to all selected plane coordinates, and use this data as the selected meteorological variable data. A grid generation technique is used to construct an N-row × N-column regular grid covering the circular area; the selected meteorological variable data is interpolated to the N-row × N-column regular grid to obtain the second local environmental field grid corresponding to that moment. As an explanation, this means that each moment in the current typhoon isochronous path sequence and multiple historical typhoon isochronous path sequences has its own corresponding local tangential plane coordinate system and local environmental field grid.

[0023] S4. Generate the current typhoon environment spatiotemporal map based on the first local environmental field grid corresponding to the current typhoon isochronous path sequence; Based on the second local environmental field grid corresponding to multiple historical typhoon isochronous path sequences, a spatiotemporal map of historical typhoon environment is generated. In an optional embodiment, in S4, according to the first local environmental field grid corresponding to the current typhoon isochronous path sequence, a current typhoon environmental spatio-temporal map is generated as follows: For a certain moment in the current typhoon isochronous path sequence, obtain the first local environmental field grid corresponding to this moment; For this first local environmental field grid, take each grid point in this first local environmental field grid as a grid node, and each grid node carries meteorological variable data of this grid node, forming a grid node set; For the local tangential plane coordinate system corresponding to this first local environmental field grid, extract the origin in this local tangential plane coordinate system, and add this origin as the typhoon center node to the grid node set to obtain a graph node set; Obtain the row index and column index of all grid nodes in the graph node set in the local environmental field grid; for any two grid nodes in the graph node set, if the row indices of these two grid nodes are the same and the column indices differ by 1, or the column indices are the same and the row indices differ by 1, or both the row index and the column index differ by 1, then determine that these two grid nodes are adjacent, and establish a local adjacent edge between these two grid nodes; obtain all local adjacent edges to form a local adjacent edge set; Preset two radius thresholds as r1 and r2, and 0 < r1 < r2 < fixed physical distance R; For each grid node in the graph node set, obtain the plane coordinate of this grid node in the corresponding local tangential plane coordinate system as the node coordinate; Obtain the distance value from the node coordinate to the typhoon center node; As an illustration, the distance value from the node coordinate to the typhoon center node can be obtained through the Euclidean distance formula; According to the relationship between the distance value and r1, r2, divide all grid nodes in the graph node set into multiple annuli: According to the relationship between the distance value and r1, r2, divide all grid nodes in the graph node set into multiple annuli as follows: If the distance value ≤ r1, then this grid node belongs to the first annulus; If r1 < distance value < r2, then this grid node belongs to the second annulus; If the distance value ≥ r2, then this grid node belongs to the third annulus; For any two grid nodes in the graph node set, if these two grid nodes belong to the same annulus or belong to adjacent annuli, then establish a circumferential multi-scale adjacent edge between these two grid nodes; As an illustration, adjacent annuli refer to two adjacent annuli, for example, the first annulus and the second annulus are adjacent annuli, and the second annulus and the third annulus are adjacent annuli.

[0024] Obtain all circumferential multi-scale adjacent edges to form a circumferential multi-scale adjacent edge set; In the graph node set, an interaction edge is established between the typhoon center node and each grid node; for each grid node in the graph node set, the interaction weight between the grid node and the typhoon center node is obtained by using the Gaussian radial basis function based on the distance value from the node coordinates to the typhoon center node, and is used as the edge weight of the interaction edge. Obtain all interaction edges to form an interaction edge set; The spatial edge set at that moment is obtained by taking the union of the local adjacent edge set, the circumferential multi-scale adjacent edge set, and the interactive edge set; Obtain the typhoon center node at the current moment in the typhoon isochronous path sequence; Establish a time edge between the typhoon center node at this moment and the typhoon center node at the adjacent moment; Obtain all time edges to form a time edge set; Obtain all times corresponding to the current typhoon isochronous path sequence; All grid nodes and the typhoon center node in the graph node set at each moment are taken as spatiotemporal nodes; Local adjacent edges, circumferential multi-scale adjacent edges, and interactive edges in the spatial edge set at each time point are taken as the first spatiotemporal edges between spatiotemporal nodes; The time edges in the set of time edges are used as the second spatiotemporal edges between two typhoon center nodes at adjacent times. A spatiotemporal map of the current typhoon environment is formed by spatiotemporal nodes, the first spatiotemporal edge, and the second spatiotemporal edge.

[0025] In an optional embodiment, in S4, a spatiotemporal map of the historical typhoon environment is generated based on the second local environmental field grid corresponding to multiple historical typhoon isochronous path sequences, as follows: For a specific moment in a historical typhoon isochronous path sequence, obtain the second local environmental field grid corresponding to that moment; For the second local environmental field grid, each grid point in the second local environmental field grid is taken as a grid node, and each grid node carries the meteorological variable data of that grid node, forming a set of grid nodes; For the local tangential plane coordinate system corresponding to the second local environmental field grid, the origin of the local tangential plane coordinate system is extracted, and the origin is added to the grid node set as the typhoon center node to obtain the graph node set; Obtain the row indices and column indices of all grid nodes in the local environmental field grid for the set of graph nodes; for any two grid nodes in the set of graph nodes, if the row indices of the two grid nodes are the same and the column indices differ by 1, or the column indices are the same and the row indices differ by 1, or both the row indices and column indices differ by 1, then determine that the two grid nodes are adjacent, and establish a local adjacent edge between the two grid nodes; obtain all local adjacent edges to form a set of local adjacent edges; Preset two radius thresholds as r1 and r2, and 0 < r1 < r2 < the fixed physical distance R; For each grid node in the set of graph nodes, obtain the plane coordinates of the grid node in the corresponding local tangential plane coordinate system as the node coordinates; Obtain the distance value from the node coordinates to the typhoon center node; As an illustration, the distance value from the node coordinates to the typhoon center node can be obtained through the Euclidean distance formula; According to the relationship between the distance value and r1, r2, divide all grid nodes in the set of graph nodes into multiple annuli: According to the relationship between the distance value and r1, r2, divide all grid nodes in the set of graph nodes into multiple annuli as follows: If the distance value ≤ r1, then the grid node belongs to the first annulus; If r1 < distance value < r2, then the grid node belongs to the second annulus; If the distance value ≥ r2, then the grid node belongs to the third annulus; For any two grid nodes in the set of graph nodes, if the two grid nodes belong to the same annulus or adjacent annuli, then establish a circumferential multi-scale adjacent edge between the two grid nodes; As an illustration, adjacent annuli refer to two adjacent annuli. For example, the first annulus and the second annulus are adjacent annuli, and the second annulus and the third annulus are adjacent annuli.

[0026] Obtain all circumferential multi-scale adjacent edges to form a set of circumferential multi-scale adjacent edges; In the set of graph nodes, establish an interaction edge between the typhoon center node and each grid node; for each grid node in the set of graph nodes, according to the distance value from the node coordinates to the typhoon center node, use the Gaussian radial basis function to obtain the interaction weight between the grid node and the typhoon center node as the edge weight of the interaction edge; Obtain all interaction edges to form a set of interaction edges; Take the union of the set of local adjacent edges, the set of circumferential multi-scale adjacent edges, and the set of interaction edges to obtain the spatial edge set at this moment; Obtain the typhoon center node at the adjacent moment in the historical typhoon isochrone path sequence at this moment; Establish a time edge between the typhoon center node at this moment and the typhoon center node at the adjacent moment; Obtain all time edges to form a time edge set; Obtain all times corresponding to the isochronous path sequence of this historical typhoon; All grid nodes and the typhoon center node in the graph node set at each moment are taken as spatiotemporal nodes; Local adjacent edges, circumferential multi-scale adjacent edges, and interactive edges in the spatial edge set at each time point are taken as the first spatiotemporal edges between spatiotemporal nodes; The time edges in the set of time edges are used as the second spatiotemporal edges between two typhoon center nodes at adjacent times. A historical typhoon environment spatiotemporal map is formed by spatiotemporal nodes, the first spatiotemporal edge, and the second spatiotemporal edge.

[0027] S5. Based on the first local environmental field grid and the second local environmental field grid, generate multiple feature vectors to form a feature vector set; In an optional embodiment, in S5, multiple feature vectors are generated based on the first local environmental field grid and the second local environmental field grid to form a feature vector set, as follows: Obtain all grid nodes corresponding to the first local environmental field grid and the second local environmental field grid; Extract wind field data from meteorological variable data of a specific grid node; The geometric projection method is used to convert wind field data into radial and tangential components relative to the origin in the local coordinate system. The radial and tangential components are concatenated with geopotential height, temperature, humidity, vorticity, and divergence data from meteorological variable data to form the feature vector of the grid node. Obtain the feature vectors of all grid nodes to form a feature vector set; S6. Generate the current sequence embedding representation based on the feature vector set and the current typhoon environment spatiotemporal map; In an optional embodiment, in S6, the current sequence embedding representation is generated based on the feature vector set and the current typhoon environment spatiotemporal map, as follows: Input the current typhoon environment spatiotemporal map into the graph neural network, and aggregate the feature vectors of all grid nodes connected to the typhoon center node through interactive edges at each moment to obtain the environmental field features at each moment. For each moment in the current typhoon isochronous path sequence, cubic spline interpolation is used to fit the current typhoon isochronous path sequence to obtain a continuous path function; based on the path function, the position, tangential direction, and curvature of the typhoon center node at each moment in the current typhoon isochronous path sequence are obtained as the path features at each moment. By concatenating the environmental field features and path features at each moment, the embedding vector of the typhoon center node at each moment is obtained. The embedding vectors of the typhoon center nodes at each time point are arranged in temporal order and then input into a long short-term memory network to obtain the current sequence embedding representation for similarity retrieval. S7. Generate a set of historical sequence embedding representations based on the feature vector set and multiple historical typhoon environmental spatiotemporal maps; In an optional embodiment, in S7, a set of historical sequence embedding representations is generated based on the feature vector set and multiple historical typhoon environmental spatiotemporal maps, as follows: Based on multiple historical typhoon environmental spatiotemporal maps, a certain historical typhoon environmental spatiotemporal map is input into a graph neural network. The graph neural network aggregates the feature vectors of all grid nodes connected to the typhoon center node through interactive edges at each moment to obtain the environmental field features at each moment. For each moment in the historical typhoon isochronous path sequence corresponding to the historical typhoon environmental spatiotemporal map, cubic spline interpolation is used to fit the historical typhoon isochronous path sequence to obtain a continuous path function; based on the path function, the position, tangential direction, and curvature of the typhoon center node at each moment in the historical typhoon isochronous path sequence are obtained as the path features at each moment. By concatenating the environmental field features and path features at each moment, the embedding vector of the typhoon center node at each moment is obtained. The embedding vectors of the typhoon center nodes at each time point are arranged in temporal order and then input into a long short-term memory network to obtain the historical sequence embedding representation for similarity retrieval. Obtain all historical sequence embeddings to form a set of historical sequence embeddings; S8. Obtain the similarity between the current sequence embedding representation and each historical sequence embedding representation in the set of historical sequence embedding representations, and use it as the path similarity between the current typhoon and the historical typhoon corresponding to that historical sequence embedding representation.

[0028] As an explanation, cosine similarity or Euclidean distance can be used to obtain the similarity between the current sequence embedding representation and each historical sequence embedding representation in the set of historical sequence embedding representations.

[0029] This application obtains current and historical typhoon path data and corresponding environmental field reanalysis data, performs isochronization processing at a fixed time step to obtain the current typhoon isochronous path sequence and multiple historical typhoon isochronous path sequences, and establishes a local tangential plane coordinate system with the typhoon center as the origin at each moment. A local environmental field grid is generated by interpolating a circular region at a fixed physical distance, achieving spatial perturbation invariance in the environmental field representation. Furthermore, based on the local environmental field grid, it constructs a current typhoon environmental spatiotemporal map and historical typhoon environmental spatiotemporal maps containing grid nodes and typhoon center nodes, through local adjacency edges, circumferential multi-scale adjacency edges, and interactive... This paper describes the correlation between edges and time edges, and constructs a feature vector by splicing the radial and tangential components of the wind field and other meteorological variable data. An environmental field feature is obtained by aggregating the feature vectors of all grid nodes connected to the typhoon center node through interactive edges using a graph neural network. Simultaneously, cubic spline interpolation is used to extract position, tangential direction, and curvature as path features from isochronous path sequences or historical typhoon isochronous path sequences. These two features are then spliced ​​to obtain the embedding vector of the typhoon center node at each time point. This vector is then processed by a long short-term memory network to obtain a sequence embedding representation. Finally, the similarity between the current sequence embedding representation and the historical sequence embedding representation is used as the path similarity. This application, by fusing path and environmental field information, constructs current and historical typhoon environmental spatiotemporal maps, effectively alleviating the problem of inaccurate similarity judgments caused by traditional methods that rely solely on path geometry while ignoring the influence of the environmental field. This significantly improves the accuracy and physical consistency of typhoon path similarity analysis.

[0030] For clarification, "acquisition" in this application refers to obtaining the required content or data using existing technical means.

[0031] Furthermore, any content not described in detail in this specification is existing technology known to those skilled in the art.

[0032] In the embodiments provided by this invention, it should be understood that the disclosed system or method can be implemented in other ways. For example, the embodiments of the invention described above are merely illustrative; for instance, the division of modules is only a logical functional division, and there may be other division methods in actual implementation.

[0033] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0034] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated module can be implemented in hardware or in the form of hardware plus software functional modules.

[0035] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the basic characteristics of the present invention.

[0036] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A typhoon path similarity analysis method based on path and environmental field fusion, characterized in that, Includes the following steps: S1. Obtain the current typhoon's real-time path data and future forecast path data, the life history path data of multiple historical typhoons, the environmental field forecast data corresponding to each moment of the current typhoon, and the environmental field reanalysis data corresponding to each moment of multiple historical typhoons. S2. The real-time path data, future forecast path data and life history path data of multiple historical typhoons are processed by isochronization with a fixed time step to obtain the current typhoon isochronous path sequence and multiple historical typhoon isochronous path sequences. S3. For each moment in the current typhoon isochronous path sequence, a grid generation mechanism is used to generate the corresponding first local environmental field grid; for each moment in multiple historical typhoon isochronous path sequences, a grid generation mechanism is used to generate the corresponding second local environmental field grid. S4. Generate the current typhoon environment spatiotemporal map based on the first local environmental field grid; generate the historical typhoon environment spatiotemporal map based on the second local environmental field grid. S5. Based on the first local environmental field grid and the second local environmental field grid, generate multiple feature vectors to form a feature vector set; S6. Generate the current sequence embedding representation based on the feature vector set and the current typhoon environment spatiotemporal map; S7. Generate a set of historical sequence embedding representations based on the feature vector set and multiple historical typhoon environmental spatiotemporal maps; S8. Obtain the similarity between the current sequence embedding representation and each historical sequence embedding representation in the set of historical sequence embedding representations, and use it as the path similarity between the current typhoon and the historical typhoon corresponding to that historical sequence embedding representation; In S6, based on the feature vector set and the current typhoon environment spatiotemporal map, the current sequence embedding representation is generated as follows: Input the current typhoon environment spatiotemporal map into the graph neural network, and aggregate the feature vectors of all grid nodes connected to the typhoon center node through interactive edges at each moment to obtain the environmental field features at each moment. For each moment in the current typhoon isochronous path sequence, cubic spline interpolation is used to fit the current typhoon isochronous path sequence to obtain a continuous path function; based on the path function, the position, tangential direction, and curvature of the typhoon center node at each moment in the current typhoon isochronous path sequence are obtained as the path features at each moment. By concatenating the environmental field features and path features at each moment, the embedding vector of the typhoon center node at each moment is obtained. The embedding vectors of the typhoon center nodes at each time point are arranged in temporal order and then input into a long short-term memory network to obtain the current sequence embedding representation for similarity retrieval. In S7, based on the feature vector set and multiple historical typhoon environmental spatiotemporal maps, a set of historical sequence embedding representations is generated, as follows: Based on multiple historical typhoon environmental spatiotemporal maps, a certain historical typhoon environmental spatiotemporal map is input into a graph neural network. The graph neural network aggregates the feature vectors of all grid nodes connected to the typhoon center node through interactive edges at each moment to obtain the environmental field features at each moment. For each moment in the historical typhoon isochronous path sequence corresponding to the historical typhoon environmental spatiotemporal map, cubic spline interpolation is used to fit the historical typhoon isochronous path sequence to obtain a continuous path function; based on the path function, the position, tangential direction, and curvature of the typhoon center node at each moment in the historical typhoon isochronous path sequence are obtained as the path features at each moment. By concatenating the environmental field features and path features at each moment, the embedding vector of the typhoon center node at each moment is obtained. The embedding vectors of the typhoon center nodes at each time point are arranged in temporal order and then input into a long short-term memory network to obtain the historical sequence embedding representation for similarity retrieval. Obtain all historical sequence embeddings to form a set of historical sequence embeddings.

2. The typhoon path similarity analysis method based on path and environmental field fusion according to claim 1, characterized in that, In S1, the environmental field forecast data and the environmental field reanalysis data both include meteorological variable data at multiple different altitude levels and latitude and longitude coordinate data corresponding to the meteorological variable data at multiple different altitude levels; Associate meteorological variable data at multiple different altitude levels with their corresponding latitude and longitude coordinates; Meteorological variable data include wind field data, geopotential height field data, temperature field data, humidity field data, eddy field data, and divergence field data.

3. The typhoon path similarity analysis method based on path and environmental field fusion according to claim 1, characterized in that, In S2, the real-time path data, future forecast path data, and life history path data of multiple historical typhoons are processed at a fixed time step to obtain the current typhoon isochronous path sequence and multiple historical typhoon isochronous path sequences, as follows: Based on a fixed time step, a linear interpolation method is used on the real-time path data to obtain the interpolated real-time path sequence. Based on a fixed time step, a linear interpolation method is used on the future forecast track data to obtain the interpolated future forecast track sequence; The actual track sequence and the future forecast track sequence are spliced ​​together in chronological order to form the current typhoon isochronous track sequence; Based on a fixed time step, linear interpolation is used on the life history path data of each historical typhoon to obtain the interpolated isochronous path sequence of historical typhoons. Obtain all historical typhoon isochronous path sequences to form multiple historical typhoon isochronous path sequences that correspond one-to-one with the life history path data of multiple historical typhoons.

4. The typhoon path similarity analysis method based on path and environmental field fusion according to claim 2, characterized in that, In S3, for each moment in the current typhoon isochronous path sequence, a grid generation mechanism is used to generate the corresponding first local environmental field grid, as follows: For each moment in the current typhoon isochronous path sequence, establish a local tangential plane coordinate system with the latitude and longitude coordinates of the typhoon center corresponding to that moment in the current typhoon isochronous path sequence as the origin. The local tangential plane coordinate system is a station-centered coordinate system with the latitude and longitude coordinates of the typhoon center as the origin, and the latitude and longitude coordinates of the typhoon center are converted into plane rectangular coordinates under the station-centered coordinate system; Based on the latitude and longitude coordinates in the environmental field forecast data corresponding to that moment, the latitude and longitude coordinates are mapped to the local tangential plane coordinate system corresponding to that moment through map projection transformation, resulting in a set of plane coordinates corresponding to meteorological variable data at multiple different altitude levels. The set of plane coordinates includes multiple plane coordinates. In the local tangential plane coordinate system corresponding to that moment, a circular region is drawn with the origin as the center and a fixed physical distance R as the radius. All plane coordinates within this circular region are extracted as the selected plane coordinates. Obtain meteorological variable data for each altitude layer corresponding to all selected plane coordinates, and use this data as the selected meteorological variable data. A regular grid of N rows × N columns covering the circular region is constructed using grid generation technology; The selected meteorological variable data are interpolated to an N-row × N-column regular grid to obtain the first local environmental field grid corresponding to that moment. N is a positive integer greater than 1; The fixed physical distance R is greater than or equal to 1 km.

5. The typhoon path similarity analysis method based on path and environmental field fusion according to claim 4, characterized in that, In S3, for each moment in multiple historical typhoon isochronous path sequences, a grid generation mechanism is used to generate the corresponding second local environmental field grid, as follows: For each moment in a historical typhoon isochronous path sequence, a local tangential plane coordinate system is established with the latitude and longitude coordinates of the typhoon center corresponding to that moment in the historical typhoon isochronous path sequence as the origin. The local tangential plane coordinate system is a station-centered coordinate system with the latitude and longitude coordinates of the typhoon center as the origin, and the latitude and longitude coordinates of the typhoon center are converted into plane rectangular coordinates under the station-centered coordinate system; Based on the latitude and longitude coordinates in the environmental field reanalysis data corresponding to that moment, the latitude and longitude coordinates are mapped to the local tangential plane coordinate system corresponding to that moment through map projection transformation, resulting in a set of plane coordinates corresponding to meteorological variable data at multiple different altitude levels. The set of plane coordinates includes multiple plane coordinates. In the local tangential plane coordinate system corresponding to that moment, a circular region is drawn with the origin as the center and a fixed physical distance R as the radius. All plane coordinates within this circular region are extracted as the selected plane coordinates. Obtain meteorological variable data for each altitude layer corresponding to all selected plane coordinates, and use this data as the selected meteorological variable data. A regular grid of N rows × N columns covering the circular region is constructed using grid generation technology; The selected meteorological variable data are interpolated to an N-row × N-column regular grid to obtain the second local environmental field grid corresponding to that moment.

6. The typhoon path similarity analysis method based on path and environmental field fusion according to claim 5, characterized in that, In S4, based on the first local environmental field grid corresponding to the current typhoon isochronous path sequence, a spatiotemporal map of the current typhoon environment is generated, as follows: For a given moment in the current typhoon isochronous path sequence, obtain the first local environmental field grid corresponding to that moment; For the first local environmental field grid, each grid point in the first local environmental field grid is taken as a grid node, and each grid node carries the meteorological variable data of that grid node, forming a set of grid nodes; For the local tangential plane coordinate system corresponding to the first local environmental field grid, the origin of the local tangential plane coordinate system is extracted, and the origin is added to the grid node set as the typhoon center node to obtain the graph node set; Obtain the row indices and column indices of all grid nodes in the local environmental field grid for the set of graph nodes; for any two grid nodes in the set of graph nodes, if the row indices of the two grid nodes are the same and the column indices differ by 1, or the column indices are the same and the row indices differ by 1, or both the row indices and column indices differ by 1, then determine that the two grid nodes are adjacent, and establish a local adjacent edge between the two grid nodes; obtain all local adjacent edges to form a set of local adjacent edges; Preset two radius thresholds as r1 and r2, and 0 < r1 < r2 < the fixed physical distance R; For each grid node in the set of graph nodes, obtain the plane coordinates of the grid node in the corresponding local tangential plane coordinate system as the node coordinates; Obtain the distance value from the node coordinates to the typhoon center node; According to the relationship between the distance value and r1, r2, divide all grid nodes in the set of graph nodes into multiple annuli: According to the relationship between the distance value and r1, r2, divide all grid nodes in the set of graph nodes into multiple annuli as follows: If the distance value ≤ r1, then the grid node belongs to the first annulus; If r1 < distance value < r2, then the grid node belongs to the second annulus; If the distance value ≥ r2, then the grid node belongs to the third annulus; For any two grid nodes in the set of graph nodes, if the two grid nodes belong to the same annulus or adjacent annuli, then establish a circumferential multi-scale adjacent edge between the two grid nodes; Obtain all circumferential multi-scale adjacent edges to form a set of circumferential multi-scale adjacent edges; In the set of graph nodes, establish an interaction edge between the typhoon center node and each grid node; for each grid node in the set of graph nodes, according to the distance value from the node coordinates to the typhoon center node, use the Gaussian radial basis function to obtain the interaction weight between the grid node and the typhoon center node as the edge weight of the interaction edge; Obtain all interaction edges to form a set of interaction edges; Take the union of the set of local adjacent edges, the set of circumferential multi-scale adjacent edges, and the set of interaction edges to obtain the spatial edge set at this moment; Obtain the typhoon center nodes at adjacent moments in the current typhoon isochronous path sequence at this moment; Establish a time edge between the typhoon center node at this moment and the typhoon center node at the adjacent moment; Obtain all time edges to form a set of time edges; Obtain all moments corresponding to the current typhoon isochronous path sequence; Take all grid nodes and typhoon center nodes in the set of graph nodes at each moment as spatio-temporal nodes; Take the local adjacent edges, circumferential multi-scale adjacent edges, and interaction edges in the spatial edge set at each moment as the first spatio-temporal edges between spatio-temporal nodes; Take the time edges in the set of time edges as the second spatio-temporal edges between two typhoon center nodes at adjacent moments; Form the current typhoon environmental spatio-temporal graph through spatio-temporal nodes, the first spatio-temporal edges, and the second spatio-temporal edges.

7. The typhoon path similarity analysis method based on path and environmental field fusion according to claim 6, characterized in that, In S4, generate a historical typhoon environmental spatio-temporal graph according to the second local environmental field grid corresponding to multiple historical typhoon isochronous path sequences, as follows: For a certain moment in a certain historical typhoon isochronous path sequence, obtain the second local environmental field grid corresponding to this moment; For the second local environmental field grid, each grid point in the second local environmental field grid is taken as a grid node, and each grid node carries meteorological variable data of the grid node, forming a grid node set; For the local tangential plane coordinate system corresponding to the second local environmental field grid, the origin in the local tangential plane coordinate system is extracted, and the origin is added as the typhoon center node to the grid node set to obtain a graph node set; Obtain the row index and column index of all grid nodes in the graph node set in the local environmental field grid; for any two grid nodes in the graph node set, if the row indices of the two grid nodes are the same and the column indices differ by 1, or the column indices are the same and the row indices differ by 1, or both the row index and the column index differ by 1, it is determined that the two grid nodes are adjacent, and a local adjacent edge is established between the two grid nodes; obtain all local adjacent edges to form a local adjacent edge set; Preset two radius thresholds as r1 and r2, and 0 < r1 < r2 < fixed physical distance R; For each grid node in the graph node set, obtain the plane coordinate of the grid node in the corresponding local tangential plane coordinate system as the node coordinate; Obtain the distance value from the node coordinate to the typhoon center node; According to the relationship between the distance value and r1, r2, divide all grid nodes in the graph node set into multiple annuli: According to the relationship between the distance value and r1, r2, divide all grid nodes in the graph node set into multiple annuli as follows: If the distance value ≤ r1, then the grid node belongs to the first annulus; If r1 < distance value < r2, then the grid node belongs to the second annulus; If the distance value ≥ r2, then the grid node belongs to the third annulus; For any two grid nodes in the graph node set, if the two grid nodes belong to the same annulus or belong to adjacent annuli, a circumferential multi-scale adjacent edge is established between the two grid nodes; Obtain all circumferential multi-scale adjacent edges to form a circumferential multi-scale adjacent edge set; In the graph node set, an interaction edge is established between the typhoon center node and each grid node; for each grid node in the graph node set, according to the distance value from the node coordinate to the typhoon center node, the Gaussian radial basis function is used to obtain the interaction weight between the grid node and the typhoon center node as the edge weight of the interaction edge; Obtain all interaction edges to form an interaction edge set; Take the union of the local adjacent edge set, the circumferential multi-scale adjacent edge set and the interaction edge set to obtain the spatial edge set at this moment; Obtain the typhoon center nodes at adjacent moments in the historical typhoon isochronous path sequence at this moment; Establish a time edge between the typhoon center node at this moment and the typhoon center node at the adjacent moment; Obtain all time edges to form a time edge set; Obtain all moments corresponding to the historical typhoon isochronous path sequence; Take all grid nodes and the typhoon center node in the graph node set at each moment as spatio-temporal nodes; Take the local adjacent edges, the circumferential multi-scale adjacent edges and the interaction edges in the spatial edge set at each moment as the first spatio-temporal edges between spatio-temporal nodes; Take the time edges in the time edge set as the second spatio-temporal edges between two typhoon center nodes at adjacent moments; A historical typhoon environment spatiotemporal map is formed by spatiotemporal nodes, the first spatiotemporal edge, and the second spatiotemporal edge.

8. The typhoon path similarity analysis method based on path and environmental field fusion according to claim 7, characterized in that, In S5, multiple feature vectors are generated based on the first and second local environmental field grids, forming a feature vector set as follows: Obtain all grid nodes corresponding to the first local environmental field grid and the second local environmental field grid; Extract wind field data from meteorological variable data of a specific grid node; The geometric projection method is used to convert wind field data into radial and tangential components relative to the origin in the local coordinate system. The radial and tangential components are concatenated with geopotential height, temperature, humidity, vorticity, and divergence data from meteorological variable data to form the feature vector of the grid node. Obtain the feature vectors of all grid nodes to form a feature vector set.

Images