Cluster buoy sea condition prediction method and system based on digital twinborn technology

Abstract

The invention discloses a cluster buoy sea condition prediction method and system based on a digital twinborn technology. The method comprises the following steps: acquiring sensor space-time sea condition data under different sea conditions in a preset research area; on the basis of the wave measurement principle of an accelerometer, the space-time sea condition data of the sensor is calculated to obtain space-time wave data of each buoy in the buoy cluster; on the basis of a graph embedding-based data fusion method, acquiring a multivariable time sequence vector of each buoy node according to the space-time wave data of each buoy; constructing a spatio-temporal interpolation data sample set according to the multivariable time sequence vector; constructing an inductive directed graph neural network based on Kriging interpolation; training the constructed inductive directed graph neural network based on Kriging interpolation according to the spatio-temporal interpolation data sample set to obtain an IGNNK model used for simulating and generating unsampled buoy sensor nodes; according to the IGNNK model obtained through training, interpolation prediction of the cluster buoy sea condition is achieved, and the problems that cluster buoy measurement data are relatively dispersed and not high in relevance in the actual measurement process at present, and consequently cluster buoy sea condition prediction is insufficient in precision and low in efficiency are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of sea condition prediction for buoy clusters, and particularly to a method and system for predicting sea conditions of cluster buoys based on digital twin technology. Background Technique

[0002] The development of ocean exploitation has been increasingly emphasized by countries around the world. To study the ocean, it is necessary to develop advanced and reliable ocean monitoring technologies to obtain various ocean monitoring elements. Ocean monitoring instruments are the carriers for realizing ocean monitoring technologies and are the infrastructure for constructing a multi-angle, multi-level, continuous, and dynamic ocean stereo monitoring network. Ocean monitoring buoys are an effective means for directly monitoring the ocean environment on-site. Compared with other monitoring systems, they have the advantages of uninterrupted monitoring, multi-function, low cost, strong survivability, etc., and play an irreplaceable role compared with other monitoring means such as ship-based equipment, island stations, and satellite remote sensing. Ocean monitoring buoys can measure, process, store, and communicate data on ocean dynamic parameters and ecological environment parameters for a long time and continuously under harsh environments.

[0003] At present, wave buoys have been widely applied in large-scale operations at sea in recent years and have become one of the most important monitoring devices in China's ocean environment stereo monitoring network. However, in the actual measurement process, the measurement data is relatively scattered and has weak correlation, which leads to insufficient accuracy and low efficiency in predicting the sea conditions of cluster buoys. If a large number of buoys are added, on the one hand, it will bring higher costs, and on the other hand, it will have a certain impact on the ocean environment. Summary of the Invention

[0004] The present invention provides a method and system for predicting sea conditions of cluster buoys based on digital twin technology to overcome the above technical problems.

[0005] To achieve the above object, the technical solution of the present invention is as follows:

[0006] A method for predicting sea conditions of cluster buoys based on digital twin technology includes the following steps:

[0007] S1: Through the buoy sensors in the buoy cluster, collect the sensor spatio-temporal sea condition data under different sea conditions in the preset research area;

[0008] And based on the wave measurement principle of the accelerometer, calculate and obtain the spatio-temporal wave data of each buoy in the buoy cluster from the sensor spatio-temporal sea condition data; and the spatio-temporal wave data at least includes the wave height data, wave period data, and wave direction data of each buoy's location.

[0009] S2: Based on the data fusion method of graph embedding, obtain the multi-variable time series vectors of each buoy node according to the spatio-temporal wave data of each buoy.

[0010] S3: Construct a spatio-temporal interpolation data sample set based on the multi-variable time series vector;

[0011] S4: Construct an inductive directed graph neural network based on Kriging interpolation;

[0012] And train the constructed inductive directed graph neural network based on Kriging interpolation according to the spatio-temporal interpolation data sample set to obtain the IGNNK model for simulating and generating unsampled buoy sensor nodes;

[0013] S5: Implement interpolation prediction of the sea conditions of the cluster buoys according to the trained IGNNK model.

[0014] Furthermore, for the data fusion method based on graph embedding in S2, the method for obtaining the multi-variable time series vector of each buoy node according to the spatio-temporal wave data of each buoy is

[0015] Obtain the time series features of the spatio-temporal wave data;

[0016] Define the buoy as a buoy node, and obtain the buoy cluster graph structure according to the positional relationship of each buoy node in the buoy cluster;

[0017] Convert the spatio-temporal wave data in the buoy cluster graph structure into low-dimensional data;

[0018] And perform embedding fusion of the low-dimensional data and the time series features based on the graph neural network to obtain the multi-variable time series vector.

[0019] Furthermore, the construction method of the spatio-temporal interpolation data sample set in S3 includes:

[0020] S31: Randomly select a group of multi-variable time series vectors as spatio-temporal graph signals;

[0021] And define each buoy sensor in the buoy cluster corresponding to the spatio-temporal graph signal as a buoy node in the fully connected graph, and define the connection relationship between each buoy node as an edge in the fully connected graph to obtain the sea condition spatio-temporal dynamic graph;

[0022] S32: Based on the sea condition spatio-temporal dynamic graph, use the distance between each buoy node as the edge weight;

[0023] S33: And form the adjacency matrix corresponding to the sea condition spatio-temporal dynamic graph according to the edge weight;

[0024] Based on the Kriging interpolation method, define a mask matrix for simulating the interpolation points in the multi-variable time series vector according to the adjacency matrix; the interpolation points in the mask matrix are obtained by randomly deleting the historical spatio-temporal wave data in the multi-variable time series vector and using the corresponding buoy node as the interpolation point for data interpolation;

[0025] S34: Use the spatio-temporal graph signal, adjacency matrix, and mask matrix corresponding to the multi-variable time series vector as the spatio-temporal interpolation data sample set.

[0026] Furthermore, for the inductive directed graph neural network based on Kriging interpolation constructed in S4, its model expression is

[0027]

[0028] In the formula: represents the forward transition matrix and represents the reverse transition matrix and W sample represents the adjacency matrix corresponding to the spatio-temporal dynamic graph of sea conditions; X sample represents the spatio-temporal graph signal; is the learning weight parameter of the l-th layer; sample represents the entire sample domain, and L ∈ sample represents a sample label in the entire sample domain; M sample represents the mask matrix; H l represents the output of the l-th layer of the directed graph neural network; H l-1 represents the output of the (l - 1)-th layer of the directed graph neural network; J represents the error loss function constructed for the observed data to be interpolated and spatial interpolation; T k (x) represents matrix operation and T k (x) = 2xT k-1 (x) - T k-2 (x), T0(x) = I, T1(x) = x; I represents the identity matrix; x represents the input of matrix operation; represents the observed data to be interpolated of the multi-variable time series vector; represents the spatial interpolation of the multi-variable time series vector.

[0029] Furthermore, in S4, train the inductive directed graph neural network based on Kriging interpolation constructed according to the spatio-temporal interpolation data sample set to obtain the IGNNK model for simulating and generating unsampled buoy sensor nodes, specifically including:

[0030] S41: Randomly divide the samples in the spatio-temporal interpolation data sample set into a sample training set and a sample test set;

[0031] S42: Train the inductive directed graph neural network based on Kriging interpolation according to the sample training set to obtain the trained inductive directed graph neural network;

[0032] S43: Based on the constructed error loss function, confirm whether the output of the trained inductive directed graph neural network converges according to the sample test set;

[0033] If it is confirmed that the output of the trained inductive directed graph neural network converges, it is confirmed that the trained inductive directed graph neural network at this time is the IGNNK model obtained for simulating and generating unsampled buoy sensor nodes;

[0034] Otherwise, the weight parameters of the trained inductive directed graph neural network are adaptively adjusted by the backpropagation method, and step S42 is repeated.

[0035] A cluster buoy sea condition prediction system based on digital twin technology includes a cluster buoy system, a link communication system, and a buoy digital twin system;

[0036] The cluster buoy system and the link communication system are connected by wireless communication;

[0037] The link communication system and the buoy digital twin system are connected by wired transmission communication;

[0038] The cluster buoy system includes a power supply unit, a parameter acquisition unit, a parameter processing unit, and a control unit, and the power supply control unit is used to provide power for the parameter acquisition unit, the parameter processing unit, and the control unit;

[0039] The parameter acquisition unit is used to collect and obtain sensor spatio-temporal sea condition data under different sea conditions in a preset research area through the buoy sensors in the current buoy cluster;

[0040] The parameter processing unit is used to calculate the sensor spatio-temporal sea condition data based on the wave measurement principle of the accelerometer to obtain the spatio-temporal wave data of each buoy in the buoy cluster; and the spatio-temporal wave data at least includes the wave height data, wave period data, and wave direction data of each buoy;

[0041] The parameter processing unit is also used to preprocess the spatio-temporal wave data of each buoy to obtain preprocessed data, and fuse the preprocessed data based on the data fusion method of graph embedding to obtain the multi-variable time series vector of each buoy node;

[0042] The control unit is used to control the interaction of data information among the cluster buoy system, the communication link system, and the buoy digital twin system;

[0043] The communication link system includes a cloud database, a wireless communication link, and a wired transmission communication link;

[0044] The cloud database and the wired transmission communication link are connected through a network API;

[0045] The cloud database and the wireless communication link are connected through a serial port;

[0046] The cloud database is used to back up and store the sensor spatio-temporal sea conditions data, the output data of the parameter processing unit, and the control instructions of the control unit transmitted by the cluster buoy system;

[0047] The wireless communication link is used to connect the cluster buoy system and the communication link system;

[0048] And its connection methods include but are not limited to the Beidou system and the GPS system;

[0049] The wired transmission communication link is used to connect the buoy digital twin system and the communication link system;

[0050] And its connection methods include but are not limited to undersea optical cables;

[0051] The buoy digital twin system includes a digital twin construction module, a digital buoy construction module, and a sea condition real-scene simulation model;

[0052] The digital twin construction module is used to construct a spatio-temporal interpolation data sample set according to the multi-variable time series vector, and construct an inductive directed graph neural network based on Kriging interpolation, so as to train the constructed inductive directed graph neural network based on Kriging interpolation according to the spatio-temporal interpolation data sample set, and obtain the IGNNK model used to simulate and generate unsampled buoy sensor nodes;

[0053] The digital buoy construction module is used to obtain the three-dimensional model of the sea condition area where the buoy cluster is located;

[0054] And the 3D modeling software for obtaining the 3D model includes but is not limited to DesignModeler software, UG software, and SOLIDWORKS software;

[0055] The sea condition real-scene simulation model is used to call the IGNNK model obtained by the digital twin construction module to interpolate the spatio-temporal wave data of the sea condition area where the buoy cluster to be interpolated is located, and call the three-dimensional model of the sea condition area where the buoy cluster is located obtained by the digital buoy construction module, so as to obtain the sea condition real-scene simulation model of the sea condition area where the buoy cluster is located after interpolation.

[0056] Beneficial effects: The present invention provides a method and system for predicting the sea conditions of a cluster of buoys based on digital twin technology. By preprocessing the spatio-temporal wave data of each buoy to obtain preprocessed data, and based on a graph embedding-based data fusion method, the preprocessed data is fused to obtain the multivariate time series vectors of each buoy node, so as to construct a spatio-temporal interpolation data sample set, which solves the problem that the measurement data of the buoy cluster is too scattered and has weak correlation during the actual measurement process, and greatly improves the compactness and integrity of the measurement data of the buoy cluster; By constructing an inductive directed graph neural network based on Kriging interpolation and training to obtain the IGNNK model to realize the interpolation prediction of the sea conditions of the cluster of buoys, it is possible to more accurately predict the spatio-temporal wave data in the target area without expanding the buoy cluster; In addition, the present invention can deploy the digital twin model to the cloud server, and efficiently transmit the interaction data between the twin body and the physical entity of the buoy cluster through the wired data interface and the Internet interface, so as to realize the digital twin optimization technology model for predicting the sea conditions of the cluster of buoys, and greatly improve the accuracy and efficiency of predicting the sea conditions of the cluster of buoys. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0058] Figure 1 It is a flowchart of the method for predicting the sea conditions of a cluster of buoys based on digital twin technology of the present invention;

[0059] Figure 2 It is a connection block diagram of the system for predicting the sea conditions of a cluster of buoys based on digital twin technology in this embodiment;

[0060] Figure 3 It is a flowchart of establishing the Kriging interpolation set Y in this embodiment;

[0061] Figure 4 It is a simulation diagram of predicting the sea conditions of a cluster of buoys based on digital twin technology in this embodiment. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0062] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0063] This embodiment provides a method for predicting sea conditions of a cluster of buoys based on digital twin technology, as Figure 1 shown, which includes the following steps:

[0064] S1: Through the buoy sensors in the buoy cluster, collect the sensor spatio-temporal sea condition data under different sea conditions in a preset research area;

[0065] And based on the wave measurement principle of the accelerometer, calculate the sensor spatio-temporal sea condition data to obtain the spatio-temporal wave data of each buoy in the buoy cluster; and the spatio-temporal wave data at least includes the wave height data, wave period data, and wave direction data where each buoy is located.

[0066] S2: Based on the data fusion method of graph embedding, obtain the multivariate time series vectors of each buoy node according to the spatio-temporal wave data of each buoy.

[0067] In a specific embodiment, the data fusion method based on graph embedding is specifically

[0068] Obtain the time series characteristics of the spatio-temporal wave data;

[0069] Define the buoy as a buoy node, and obtain the graph structure of the buoy cluster according to the positional relationship of each buoy node in the buoy cluster;

[0070] Convert the spatio-temporal wave data in the graph structure of the buoy cluster into low-dimensional data;

[0071] And based on a graph neural network, the low-dimensional data and time series features are embedded and fused to obtain a multivariate time series vector; among them, the method of embedding and fusing the low-dimensional data and time series features based on a graph neural network is a well-known existing technology. For example, it includes: graph construction: mapping the low-dimensional data and time series data into a graph structure; for example, each time series variable can be regarded as a node in the graph, and the node features include low-dimensional data and historical time series values, and then the edges between the nodes are defined (such as based on similarity, causality, or domain knowledge); feature embedding: using a graph neural network (such as GCN, GAT, Graph SAGE) to encode the graph structure to obtain node embeddings; using a time series model (such as RNN, LSTM, GRU, Transformer) to encode the time series data to obtain time series embeddings; fusion and output: fusing the graph embeddings and time series embeddings (such as concatenation, weighted summation, attention mechanism, etc.) to generate the final multivariate time series vector, and the specific content will not be elaborated here anymore;

[0072] S3: Construct a spatio-temporal interpolation data sample set according to the multivariate time series vector;

[0073] In a specific embodiment, the construction method of the spatio-temporal interpolation data sample set includes:

[0074] S31: Randomly select a group of multivariate time series vectors as spatio-temporal graph signals;

[0075] And define each buoy sensor in the buoy cluster corresponding to the spatio-temporal graph signal as a buoy node in the fully connected graph, and define the connection relationship between each buoy node as an edge in the fully connected graph to obtain a sea condition spatio-temporal dynamic graph;

[0076] S32: Based on the sea condition spatio-temporal dynamic graph, use the distance between each buoy node as the edge weight;

[0077] S33: And form an adjacency matrix corresponding to the sea condition spatio-temporal dynamic graph according to the edge weight;

[0078] Based on the Kriging interpolation method, define a mask matrix for simulating the points to be interpolated in the multivariate time series vector according to the adjacency matrix; the points to be interpolated in the mask matrix are obtained by randomly deleting the historical spatio-temporal wave data in the multivariate time series vector and using the corresponding buoy node as the interpolation point for data interpolation;

[0079] S34: Use the spatio-temporal graph signal, adjacency matrix, and mask matrix corresponding to the multivariate time series vector as the spatio-temporal interpolation data sample set;

[0080] S4: Construct an inductive directed graph neural network based on Kriging interpolation;

[0081] Specifically, the model expression of the inductive directed graph neural network based on Kriging interpolation is

[0082]

[0083] In the formula: represents the forward transition matrix and represents the reverse transition matrix and W sample represents the adjacency matrix corresponding to the spatio-temporal dynamic graph of sea conditions; X sample represents the spatio-temporal graph signal; is the learning weight parameter of the l-th layer; sample represents the entire sample domain, and L ∈ sample represents a sample label in the entire sample domain; M sample represents the mask matrix; H l represents the output of the l-th directed graph neural network; H l-1 represents the output of the (l - 1)-th directed graph neural network; J represents the error loss function constructed for the interpolated observation data to be interpolated and spatial interpolation; T k (x) represents matrix operation and T k (x) = 2xT k-1 (x) - T k-2 (x), T0(x) = I, T1(x) = x; I represents the identity matrix; x represents the input of matrix operation; represents the interpolated observation data to be interpolated of the multivariate time series vector; represents the spatial interpolation of the multivariate time series vector;

[0084] Train the inductive directed graph neural network based on Kriging interpolation according to the spatio-temporal interpolation data sample set to obtain the IGNNK model for simulating and generating unsampled buoy sensor nodes;

[0085] Specifically, the method for obtaining the IGNNK model for simulating and generating unsampled buoy sensor nodes includes the following steps:

[0086] S41: Randomly divide the samples in the spatio-temporal interpolation data sample set into a sample training set and a sample test set;

[0087] S42: Train the inductive directed graph neural network based on Kriging interpolation according to the sample training set to obtain the trained inductive directed graph neural network;

[0088] S43: Based on the constructed error loss function, confirm whether the output of the trained inductive directed graph neural network converges according to the sample test set;

[0089] If it is confirmed that the output of the trained inductive directed graph neural network converges, it is confirmed that the trained inductive directed graph neural network at this time is the obtained IGNNK model for simulating and generating unsampled buoy sensor nodes;

[0090] Otherwise, the weight parameters of the trained inductive directed graph neural network are adaptively adjusted using the backpropagation method, and step S42 is repeatedly executed;

[0091] S5: Interpolate and predict the sea conditions of the cluster buoys according to the obtained IGNNK model through training;

[0092] Specifically, as shown in Figure 3 the process shown, pre - establish the Kriging interpolation set Y. In the figure, I sample is the sample set, J sample is the sample loss function, M sample is the sample set mask matrix, that is, the set matrix used to determine whether to ignore the sample. In this embodiment, the sample set is created as a graph signal X with dimensions and t and the adjacency matrix W t . At the same time, set the period for Kriging interpolation as [t, t + h) = {t, t + 1,..., t + h - 1}, and given the type of sensor network during [t, t + h]. During this period, there is data from the sensor corresponding to the unsampled nodes, that is, the nodes to be interpolated. The interpolation size is In this embodiment, in order to estimate according to substitute W t and into the trained IGNNK model, and finally obtain indicating that the final Kriging interpolation result is the sea condition prediction result of the cluster buoys based on the digital twin technology;

[0093] A cluster buoy sea condition prediction system based on digital twin technology, as shown in Figure 2 includes a cluster buoy system, a link communication system, and a buoy digital twin system;

[0094] The cluster buoy system and the link communication system are connected by wireless communication;

[0095] The link communication system and the buoy digital twin system are connected by wired transmission communication;

[0096] The cluster buoy system includes a power supply unit, a parameter acquisition unit, a parameter processing unit, and a control unit, and the power control unit is used to supply power to the parameter acquisition unit, the parameter processing unit, and the control unit; among them, in this example, the function of the parameter acquisition unit is to monitor and collect the current environmental state, and the parameter acquisition unit includes but is not limited to sensor devices such as a three-axis accelerometer, a three-axis gyroscope, and a three-axis magnetometer for monitoring the environmental state;

[0097] The parameter acquisition unit is used to collect and obtain the sensor spatio-temporal sea state data under different sea conditions in the preset research area through the buoy sensors in the current buoy cluster;

[0098] The parameter processing unit is used to calculate the sensor spatio-temporal sea state data based on the wave measurement principle of the accelerometer to obtain the spatio-temporal wave data of each buoy in the buoy cluster; and the spatio-temporal wave data at least includes the wave height data, wave period data, and wave direction data of each buoy;

[0099] The parameter processing unit is also used to preprocess the spatio-temporal wave data of each buoy to obtain preprocessed data, and based on the data fusion method of graph embedding, fuse the preprocessed data to obtain the multi-variable time series vectors of each buoy node;

[0100] The control unit is used to control the data information interaction among the cluster buoy system, the communication link system, and the buoy digital twin system;

[0101] The communication link system includes a cloud database, a wireless communication link, and a wired transmission communication link;

[0102] The cloud database is connected to the wired transmission communication link through a network API;

[0103] The cloud database is connected to the wireless communication link through a serial port;

[0104] The cloud database is used to backup and store the sensor spatio-temporal sea state data, the output data of the parameter processing unit, and the control instructions of the control unit transmitted by the cluster buoy system;

[0105] The wireless communication link is used to connect the cluster buoy system and the communication link system;

[0106] And its connection methods include but are not limited to the Beidou system and the GPS system;

[0107] The wired transmission communication link is used to connect the buoy digital twin system and the communication link system;

[0108] And its connection methods include but are not limited to submarine optical cables;

[0109] The buoy digital twin system includes a digital twin construction module, a digital buoy construction module, and a sea condition real-scene simulation model;

[0110] The digital twin construction module is used to construct a spatio-temporal interpolation data sample set according to the multi-variable time series vector, and construct an inductive directed graph neural network based on Kriging interpolation, so as to train the constructed inductive directed graph neural network based on Kriging interpolation according to the spatio-temporal interpolation data sample set, and obtain the IGNNK model for simulating and generating unsampled buoy sensor nodes;

[0111] The digital buoy construction module is used to obtain the three-dimensional model of the sea condition area where the buoy cluster is located;

[0112] And the 3D modeling software for obtaining the 3D model includes but is not limited to DesignModeler software, UG software, and SOLIDWORKS software; the method for obtaining the 3D model through the 3D modeling software is a well-known technical means in the art and will not be elaborated here;

[0113] The sea condition real-scene simulation model is used to call the IGNNK model obtained by the digital twin construction module to interpolate the spatio-temporal wave data of the sea condition area where the buoy cluster to be interpolated is located, as Figure 4 shown, and call the three-dimensional model of the sea condition area where the buoy cluster is located obtained by the digital buoy construction module to obtain the sea condition real-scene simulation model of the sea condition area where the buoy cluster is located after interpolation.

[0114] In this embodiment, the preprocessing data is obtained by preprocessing the spatio-temporal wave data of each buoy, and based on the data fusion method of graph embedding, the preprocessing data is fused to obtain the multi-variable time series vector of each buoy node, so as to construct a spatio-temporal interpolation data sample set, which solves the problem that the measurement data of the buoy cluster is too scattered and has weak correlation in the actual measurement process, and greatly improves the tightness and integrity of the measurement data of the buoy cluster; by constructing an inductive directed graph neural network based on Kriging interpolation and training to obtain the IGNNK model to realize the interpolation prediction of the sea conditions of the cluster buoys, it can more accurately predict the spatio-temporal wave data in the target area without expanding the buoy cluster. In addition, this embodiment can also deploy the digital twin model to the cloud server, and efficiently transmit the interaction data between the twin body and the physical entity of the buoy cluster through the wired data interface and the Internet interface, realizing the digital twin optimization technology model for predicting the sea conditions of the cluster buoys, and greatly improving the accuracy and efficiency of predicting the sea conditions of the cluster buoys.

[0115] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting sea conditions of cluster buoys based on digital twin technology, characterized in that, It includes the following steps: S1: Through the buoy sensors in the buoy cluster, collect the spatio-temporal sea condition data of sensors under different sea conditions in the preset research area; And based on the wave measurement principle of the accelerometer, resolve the spatio-temporal sea condition data of the sensors to obtain the spatio-temporal wave data of each buoy in the buoy cluster; and the spatio-temporal wave data at least includes the wave height data, wave period data, and wave direction data where each buoy is located; S2: Based on the data fusion method of graph embedding, obtain the multi-variable time series vectors of each buoy node according to the spatio-temporal wave data of each buoy; S3: Construct a spatio-temporal interpolation data sample set according to the multi-variable time series vectors; S4: Construct an inductive directed graph neural network based on Kriging interpolation; And train the constructed inductive directed graph neural network based on Kriging interpolation according to the spatio-temporal interpolation data sample set to obtain the IGNNK model for simulating and generating unsampled buoy sensor nodes; S5: Realize the interpolation prediction of the sea conditions of the cluster buoys according to the trained IGNNK model.

2. The method for predicting the sea conditions of a cluster of buoys based on digital twin technology according to claim 1, wherein In the data fusion method of graph embedding described in S2, the method for obtaining the multi-variable time series vectors of each buoy node according to the spatio-temporal wave data of each buoy is Obtain the time series characteristics of the spatio-temporal wave data; Define the buoy as a buoy node, and obtain the graph structure of the buoy cluster according to the positional relationship of each buoy node in the buoy cluster; Convert the spatio-temporal wave data in the graph structure of the buoy cluster into low-dimensional data; And based on the graph neural network, perform embedding fusion on the low-dimensional data and the time series characteristics to obtain the multi-variable time series vectors.

3. The method for predicting sea conditions of a cluster of buoys based on digital twin technology according to claim 2, wherein, The construction method of the spatio-temporal interpolation data sample set described in S3 includes: S31: Randomly select a group of multi-variable time series vectors as spatio-temporal graph signals; And define each buoy sensor in the buoy cluster corresponding to the spatio-temporal graph signal as a buoy node in the fully connected graph, and define the connection relationship between each buoy node as an edge in the fully connected graph to obtain the sea condition spatio-temporal dynamic graph; S32: Based on the sea condition spatio-temporal dynamic graph, use the distance between each buoy node as the edge weight; S33: And form the adjacency matrix corresponding to the sea condition spatio-temporal dynamic graph according to the edge weight; Based on the Kriging interpolation method, define a mask matrix for simulating the interpolation points in the multi-variable time series vectors according to the adjacency matrix; the interpolation points in the mask matrix are obtained by randomly deleting the historical spatio-temporal wave data in the multi-variable time series vectors and using the corresponding buoy node as the interpolation point for data interpolation; S34: Use the spatio-temporal graph signal, adjacency matrix, and mask matrix corresponding to the multi-variable time series vectors as the spatio-temporal interpolation data sample set.

4. The method for predicting sea conditions of a cluster of buoys based on digital twin technology according to claim 3, wherein, For the inductive directed graph neural network based on Kriging interpolation constructed in S4, its model expression is Wherein: represents the forward transition matrix and represents the reverse transition matrix and W sample represents the adjacency matrix corresponding to the spatio-temporal dynamic diagram of sea conditions; X sample represents the spatio-temporal diagram signal; is the learning weight parameter of the l-th layer; sample represents the entire sample domain, and L ∈ sample represents a sample label in the entire sample domain; M sample represents the mask matrix; H l represents the output of the l-th layer of the directed graph neural network; H l-1 represents the output of the (l - 1)-th layer of the directed graph neural network; J represents the error loss function constructed for the interpolated observation data and spatial interpolation; T k (x) represents matrix operation and T k (x) = 2xT k-1 (x)-T k-2 (x), T0(x) = I, T1(x) = x; I represents the identity matrix; x represents the input of matrix operation; represents the interpolated observation data of the multivariate time series vector; represents the spatial interpolation of the multivariate time series vector.

5. The method for predicting sea conditions of a cluster of buoys based on digital twin technology according to claim 4, wherein In S4, training the constructed inductive directed graph neural network based on Kriging interpolation according to the spatio-temporal interpolation data sample set to obtain the IGNNK model for simulating and generating unsampled buoy sensor nodes specifically includes: S41: Randomly divide the samples in the spatio-temporal interpolation data sample set into a sample training set and a sample test set; S42: Train the inductive directed graph neural network based on Kriging interpolation using the sample training set to obtain the trained inductive directed graph neural network; S43: Based on the constructed error loss function, confirm whether the output of the trained inductive directed graph neural network converges according to the sample test set; If it is confirmed that the output of the trained inductive directed graph neural network converges, then confirm that the trained inductive directed graph neural network at this time is the obtained IGNNK model for simulating and generating unsampled buoy sensor nodes; Otherwise, adaptively adjust the weight parameters of the trained inductive directed graph neural network using the backpropagation method, and repeat step S42.

6. A system for a method of predicting sea conditions of a cluster of buoys based on digital twin technology according to any one of claims 1 to 5, characterized in that, It includes a cluster buoy system, a link communication system, and a buoy digital twin system; The cluster buoy system and the link communication system are connected by wireless communication; The link communication system and the buoy digital twin system are connected by wired transmission communication; The cluster buoy system includes a power supply unit, a parameter acquisition unit, a parameter processing unit, and a control unit, and the power control unit is used to supply power to the parameter acquisition unit, the parameter processing unit, and the control unit; The parameter acquisition unit is used to collect and obtain the sensor spatio-temporal sea condition data under different sea conditions in the preset research area through the buoy sensors in the current buoy cluster; The parameter processing unit is used to calculate the sensor spatio-temporal sea condition data based on the wave measurement principle of the accelerometer to obtain the spatio-temporal wave data of each buoy in the buoy cluster; and the spatio-temporal wave data at least includes the wave height data, wave period data, and wave direction data of each buoy; The parameter processing unit is also used to preprocess the spatio-temporal wave data of each buoy to obtain preprocessed data, and fuse the preprocessed data based on the data fusion method of graph embedding to obtain the multi-variable time series vectors of each buoy node; The control unit is used to control the data information interaction among the cluster buoy system, the communication link system, and the buoy digital twin system; The communication link system includes a cloud database, a wireless communication link, and a wired transmission communication link; The cloud database and the wired transmission communication link are connected through a network API; The cloud database and the wireless communication link are connected through a serial port; The cloud database is used to back up and store the sensor spatio-temporal sea condition data transmitted by the cluster buoy system, the output data of the parameter processing unit, and the control instructions of the control unit; The wireless communication link is used to connect the cluster buoy system and the communication link system; And its connection method includes but is not limited to the Beidou system and the GPS system; The wired transmission communication link is used to connect the buoy digital twin system and the communication link system; And its connection method includes but is not limited to submarine optical cables; The buoy digital twin system includes a digital twin construction module, a digital buoy construction module, and a sea condition real-scene simulation model; The digital twin construction module is used to construct a spatio-temporal interpolation data sample set according to the multivariate time series vector, and construct an inductive directed graph neural network based on Kriging interpolation to train the constructed inductive directed graph neural network based on Kriging interpolation according to the spatio-temporal interpolation data sample set, so as to obtain the IGNNK model for simulating and generating unsampled buoy sensor nodes; The digital buoy construction module is used to obtain a three-dimensional model of the sea state area where the buoy cluster is located; And the 3D modeling software for obtaining the 3D model includes but is not limited to DesignModeler software, UG software, and SOLIDWORKS software; The sea state real-scene simulation model is used to call the IGNNK model obtained by the digital twin construction module to interpolate the spatio-temporal wave data of the sea state area where the buoy cluster to be interpolated is located, and call the three-dimensional model of the sea state area where the buoy cluster is located obtained by the digital buoy construction module, so as to obtain the sea state real-scene simulation model of the sea state area where the buoy cluster is located after interpolation.

Images