Floating body drifting trajectory prediction method and system fused with marine environment field

Abstract

The invention discloses a floating body drifting trajectory prediction method and system fused with a marine environment field, and relates to the field of application of artificial intelligence in ocean engineering, and the method comprises the steps: firstly extracting the information of the marine environment field through convolution, then extracting the floating body trajectory information through a time attention mechanism, and finally fusing the two information, therefore, floating body trajectory prediction is effectively carried out. According to the method, the problems of data fusion and utilization can be solved, the marine environment field information is fully utilized, the prediction precision and real-time performance are remarkably improved, and the method has important application value in the fields of ocean engineering and physical oceanography.

Description

Technical Field

[0001] The present invention relates to the field of application of artificial intelligence in marine engineering, and in particular to a method and system for predicting the drift trajectory of a floating body integrating a marine environment field. Background Art

[0002] At present, the data recovery float carried by the underwater platform enters the drifting mode after floating to the sea surface. The drift trajectory prediction of the float refers to predicting the movement path of the float floating on the sea surface, which is affected by wind and ocean currents. The drift trajectory prediction of the float is of great significance in the following aspects. In terms of data collection, the float is used to collect ocean data such as temperature, salinity and ocean currents. Accurate drift trajectory prediction helps to collect data in the right area, thereby obtaining comprehensive ocean condition data. In terms of monitoring and management, the float is often lost or needs to be repositioned. Predicting its drift trajectory can help plan effective search and rescue operations. In terms of navigation safety, understanding the possible drift trajectory of the float helps to avoid obstacles and ensure a safe navigation route. Therefore, predicting the drift trajectory of the float plays a positive role in improving the efficiency and safety of marine operations, as well as environmental protection and resource management.

[0003] The main floating body trajectory prediction methods currently used include methods based on statistical models (such as Bayesian statistical models, ARIMA models, etc.) and methods based on machine learning (such as RNN, LSTM, etc.). Although the existing technical solutions have made certain progress in trajectory prediction in the data recovery floating body drifting mode, these methods perform well when processing a single data source, but when dealing with complex, multi-source heterogeneous data, it is often difficult to effectively integrate information from different data sources. Marine environmental field data contains a variety of environmental variables such as ocean temperature, salinity, and ocean currents, which can reflect the impact of changes in the marine environment on the trajectory of floating bodies. Traditional floating body trajectory prediction methods are difficult to effectively extract feature information from high-dimensional and complex marine environmental field data, and fail to fully consider the impact of marine environmental field information on trajectory changes, which will lead to a large deviation between the predicted value and the actual value of the floating body drifting trajectory. Summary of the invention

[0004] In order to solve the above problems, the purpose of the present invention is to provide a floating body drift trajectory prediction technology that integrates the ocean environment field, aiming to make full use of the multi-dimensional characteristics of the integrated ocean environment field information and historical floating body trajectory information to improve the accuracy of floating body trajectory prediction.

[0005] In order to achieve the above technical objectives, the present application provides a floating body drift trajectory prediction method integrating the ocean environment field, comprising the following steps: The floating body drift trajectory data is collected and preprocessed, and then the features are screened by the random forest method to construct a data set for predicting the floating body drift trajectory. The floating body drift trajectory data consists of time, longitude, latitude, sea temperature, meridional velocity, zonal velocity, combined velocity measured by the floating body, and the variance of sea temperature, meridional current velocity and zonal current velocity. A combined model based on CNN and Informer models is constructed, which is trained through data sets to generate a floating body drift trajectory prediction model to predict the floating body drift trajectory.

[0006] Preferably, in the process of preprocessing the floating body drift trajectory data, the floating body drift trajectory data is cleaned, including: eliminating trajectory data containing missing values; eliminating trajectory data with discontinuous observation time; eliminating trajectory data with observation time less than expected time; eliminating trajectory data that does not match the observation time of the marine environment field.

[0007] Preferably, in the process of constructing the data set, the distance and azimuth of the floating body trajectory are obtained based on the floating body trajectory data at the current moment and the floating body trajectory data at the previous moment, and the distance and azimuth are added to the floating body drifting trajectory data as data features. The floating body drifting trajectory data with distance and azimuth are feature screened through the random forest method to generate the data set.

[0008] Preferably, in the process of performing feature screening by the random forest method, first out-of-bag data is selected, and a first error of the first out-of-bag data is obtained; Randomly add noise interference to the first out-of-bag data, obtain second out-of-bag data, and calculate a second error of the second out-of-bag data; According to the first error and the second error, the importance of all features in the first out-of-bag data is obtained, and features with small importance are deleted, and the process is repeated to select the first out-of-bag data corresponding to the combination with the smallest error to generate a data set.

[0009] Preferably, in the process of constructing a data set, based on the floating body drift trajectory data obtained after feature screening through the random forest method, by acquiring ocean environment field data including meridional current velocity and latitudinal current velocity, and performing time and space matching, a data set for predicting the floating body drift trajectory is obtained.

[0010] Preferably, in the process of time-space matching, the ocean environment field data at the corresponding moment is obtained according to the timestamp of the floating body drift trajectory data to perform time matching, wherein, if there is no corresponding moment data, the data before and after the moment are selected and supplemented by linear interpolation.

[0011] Preferably, in the process of space-time matching, the longitude and latitude positions of the floating body trajectory are spatially matched with the longitude and latitude grids of the ocean environment field data; for each floating body trajectory point, it is indexed to the grid of the ocean environment field according to its longitude and latitude, and extends outward 0.5° in the latitude and longitude directions with this as the center, forming an n*n two-dimensional ocean environment field centered on the trajectory.

[0012] Preferably, in the process of obtaining a data set for predicting the floating body drift trajectory, the floating body drift trajectory data and the ocean environment field data are fused based on the results of time matching and space matching, so that each trajectory point used for prediction will have corresponding environmental field data, including sea surface temperature, salinity, meridional current velocity and latitudinal current velocity, to construct a data set.

[0013] Preferably, in the process of constructing the floating body drift trajectory prediction model, an early stopping mechanism is adopted to prevent overfitting, and the model performance is optimized through cross-validation; the loss function of the model adopts mean square error and is trained through Adam optimizer.

[0014] The present invention discloses a floating body drift trajectory prediction system integrating an ocean environment field, the system is used to implement the above-mentioned floating body drift trajectory prediction method integrating an ocean environment field, the system comprises: A data processing module is used to obtain the collected floating body drift trajectory data, perform preprocessing, and then perform feature screening through a random forest method to construct a data set for predicting the floating body drift trajectory, wherein the floating body drift trajectory data consists of time, longitude, latitude, sea temperature measured by the floating body, meridional velocity, zonal velocity, combined velocity, and variance of sea temperature, meridional current velocity, and zonal current velocity; The trajectory prediction module is used to build a combined model based on CNN and Informer models, train it through the data set, generate a floating body drift trajectory prediction model, and predict the floating body drift trajectory.

[0015] The present invention discloses the following technical effects: The present invention can effectively deal with the problem of fusing multi-source heterogeneous data, and then make full use of the marine environment field information to provide more accurate floating body trajectory prediction results, significantly improve the prediction accuracy and real-time performance, and provide an important reference for research in the fields of marine engineering and physical oceanography. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 is a schematic diagram of a neural network model according to the present invention; Figure 2 It is a schematic diagram of the method flow of the present invention. DETAILED DESCRIPTION

[0018] To make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.

[0019] like Figure 1-2 As shown, the present invention provides a floating body drift trajectory prediction method integrating the ocean environment field, which specifically includes the following contents: First, the information of the ocean environment field is extracted through convolution, and then the floating body trajectory information is extracted through the temporal attention mechanism. Finally, the two information are integrated to effectively predict the floating body trajectory. The specific steps are as follows: Step 1: Preprocessing of floating body trajectory data: Select the floating body drift trajectory data in the South China Sea (10-25°N, 105-125°E), which includes variables such as time, longitude, latitude, sea temperature measured by the floating body, meridional velocity, zonal velocity, combined velocity, and variance of sea temperature, meridional velocity, and zonal velocity. Clean the selected trajectory data, including the following steps: 1. Eliminate trajectory data containing missing values; 2. Eliminate trajectory data with discontinuous observation time; 3. Eliminate trajectory data whose observation time is less than the expected time; 4. Eliminate trajectory data that do not match the observation time of the marine environment field.

[0020] In order to better mine hidden information, a total of 17 path-related features are obtained, including longitude, latitude, meridional speed, zonal speed, combined speed, sea temperature, longitude variance, latitude variance, sea temperature variance in the past 72 hours, and the difference between the longitude, latitude, meridional speed, zonal speed, combined speed, and sea temperature in the past 72 hours and the current moment, including the angle and distance formed. The above features are screened using the random forest method to determine whether all features have an impact on the prediction, and the best feature combination is selected to reduce feature redundancy and improve the generalization ability of the model.

[0021] The calculation steps for calculating distance and angle based on the floating body trajectory data are as follows: 1. Angle radian conversion: , ; in, radLat Indicates the radian value after latitude conversion. lat Represents the latitude value, radLon Indicates the radian value after longitude conversion. lon Indicates the longitude value.

[0022] 2. Calculate the projection latitude: ; Among them, ra represents the radius of the earth in the equatorial direction (equatorial radius), rb represents the radius of the earth in the polar direction (polar radius), and pA and pB represent the calculated projection latitudes.

[0023] 3. Calculate the great circle distance: if : ; otherwise: ; Here, x represents the great circle distance.

[0024] 4. Calculate distance: Intermediate variables: ; Corrected distance: ; Where c1 represents the additive effect based on latitude, c2 represents the differential effect of latitude, and dr represents the calibration distance. Represents the corrected distance parameter.

[0025] 5. Calculate the azimuth (direction): Longitude difference: ; Calculate the direction vector: ; Where dLon represents the longitude difference between two points expressed in radians, yy represents the orthogonal projection component in the direction of the longitude difference, and xx represents the projection component in the latitude direction.

[0026] 6. Calculate the azimuth: ; in, brng Indicates the azimuth.

[0027] Therefore, the present invention can obtain the distance and angle between the current moment and the previous moment of the floating body trajectory data, and add the two features of distance and angle to the trajectory data features. The specific calculation steps of the random forest feature selection method are as follows: 1. Select out-of-bag data and calculate the out-of-bag data error: ; Among them, errorm represents the out-of-bag data error, n represents the number of samples, and i represents the sample index. represents the model's predicted output for sample i, and yi represents the true value of sample i.

[0028] 2. Randomly add noise interference to all sample features of out-of-bag data and calculate the out-of-bag data error again: ; Among them, error represents the out-of-bag data error calculated again, and fi represents the model's predicted output for sample i.

[0029] 3. Calculate the importance of all features I: Among them, I represents the importance of the feature, and j represents the index number of the currently calculated feature.

[0030] 4. Delete the features with less importance in order according to the order of feature importance, and repeat the above three steps to calculate the out-of-bag (OOB) error. Select the combination with the smallest error, which represents the score of model performance. The higher the score, the better the result.

[0031] Among them, R represents the importance of the feature. OOBscore represents the model performance score. represents the predicted value of sample i, and N represents the total number of features.

[0032] The selected features are used as input features of subsequent floating body trajectory feature data in order to reduce model parameters and improve the generalization ability of the model.

[0033] Step 2: Spatiotemporal matching of ocean environment field and trajectory data: Floating body drifting is a nonlinear motion process. Based on a single floating body drifting trajectory information and relying on statistical methods to predict the floating body trajectory, it is impossible to accurately analyze the movement law of the floating body under different background fields. This also leads to the inability to accurately model the floating body trajectory data. If multi-source heterogeneous data such as marine environmental field data and floating body trajectory data are used in combination, it will not only be possible to achieve a deeper level of analysis and capture the nonlinear characteristics of floating body motion, but also effectively improve the prediction accuracy of floating body trajectory. The steps for fusing marine environmental field data and floating body trajectory data are as follows: 1. Marine environment field data preprocessing: The ocean background data with a greater impact on the movement of the floating body drift trajectory are selected, including the meridional and latitudinal current velocities. The ocean reanalysis data (HYCOM) dataset with a temporal and spatial resolution of 3 hours is used, with a spatial resolution of 0.08° and 40 layers of vertical depth. Since the floating body trajectory is highly correlated with the sea surface velocity, the surface data are selected as the marine environmental variables. The trajectory data contains time and space information, which must be matched with the time and space information on the marine environment field.

[0034] (1) Time matching: According to the timestamp of the floating body trajectory data, the ocean environment field data at UTC 00, 06, 12, and 18 are selected. If there is no data at the corresponding time, the data before and after the time are selected and supplemented by linear interpolation.

[0035] In the formula, y1 represents the data at the time point before the interpolation point, x1 represents the time point before the interpolation point, x represents the time point of the interpolation point, y represents the target value of the interpolation point, x2 represents the time point after the interpolation point, and y2 represents the data at the time point after the interpolation point; (2) Spatial matching: The longitude and latitude positions of the floating body trajectory are matched with the longitude and latitude grids of the ocean environment field data. For each floating body trajectory point, according to its longitude and latitude, it is indexed to the grid of the ocean environment field according to the following formula, and extends outward 0.5° in the latitude and longitude directions with this as the center, forming an n×n two-dimensional ocean environment field centered on the trajectory.

[0036] ; Where HYCOM(i,j) is the closest longitude and latitude point, lat(i') represents the latitude value corresponding to index i', lon(j') represents the longitude value corresponding to index j', lattarget represents the search for the closest latitude index in the marine environment data grid, and lontarget represents the search for the closest longitude index in the marine environment data grid.

[0037] 2. Data Fusion: The ocean environment field data is integrated with the floating body trajectory data. Each trajectory point used for forecasting will have corresponding environmental field data (sea surface temperature, salinity, meridional velocity and latitudinal velocity), while ensuring that the extracted environmental field data matches the trajectory in time and space.

[0038] The fused ocean environment field data and floating body trajectory data are used as input features to generate a multi-dimensional input feature vector containing ocean environment variables and trajectory information, providing more comprehensive information for subsequent neural network models: , u represents the meridional velocity in the ocean environment field data, v represents the latitudinal velocity in the ocean environment field data, Salinity represents the sea surface salinity in the ocean environment field data, SST represents the sea surface temperature in the ocean environment field data, and trackfeatures represents the floating body track feature data.

[0039] Step 3: Build a trajectory prediction model based on CNN-Informer Use the fused data prepared in the previous step to create a training data set and build a CNN-Informer floating body drift trajectory prediction model. The specific steps are as follows: 1. Divide the data set and preprocess the data: The data time span is selected from 2010 to 2023. After cleaning the data, mining hidden information, and removing redundant features according to the decision tree algorithm, a sliding window of input-output prediction sequence length is adopted to obtain 98,303 sample data. In order to ensure the balance of samples in each data set and improve the accuracy of model prediction in different time periods, the present invention divides the data into three groups: training set (80%), validation set (10%) and test set (10%).

[0040] The present invention uses fused trajectory information and marine environment field information as input features, and performs preprocessing such as normalization on multi-source data according to the dimensions and units of different data to ensure the consistency between features. First, the trajectory data is standardized to remove the scale differences between different trajectories; secondly, the marine environment field data is normalized to map the data values ​​to the range of [0,1], thereby eliminating the dimensional differences between different marine environmental variables. Through the above preprocessing steps, it is ensured that the scale of the input features is consistent during the model training process, thereby ensuring the stability of the model.

[0041] 2. Build a neural network model: The hybrid model developed in this invention can effectively fuse the two-dimensional ocean environment field information and the one-dimensional vector of the floating body trajectory. The model network structure is as follows: The network is mainly divided into three parts: (1) ocean environment field information extraction; (2) trajectory information extraction; (3) time series prediction. Figure 1 The gray shaded area in the upper left corner represents the CNN model's two-dimensional input layer for the ocean environment. The network's hidden layer alternates between convolutional layers (Conv layers) and maximum pooling layers (Max pooling layers). The global average pooling layer (Averagepooling layer) is used at the end to extract the global features of the ocean environment field. The gray shaded area in the lower right corner is the two-dimensional track data of the trajectory. represents the i-th feature of the trajectory at the j-th time point on the past path, , and Indicates the time and location information of the point. The data is input into the encoder of the Informer model, and a series of operations such as the decoder are performed to generate preliminary data for the prediction of the floating body trajectory, which is then fully connected with the characteristics of the ocean environment field to obtain the prediction result. The present invention uses the ReLU function as the activation function, so that the model can learn features with different numerical ranges.

[0042] Model training and prediction: The present invention adopts an early stopping mechanism to prevent overfitting and optimizes model performance through cross-validation. The loss function of the model adopts mean square error (MSE) and is trained through the Adam optimizer. Ultimately, the model can achieve accurate time series prediction by effectively integrating ocean environment field information and trajectory information.

[0043] Step 4: Model application and integration: The trained and verified neural network integrates the marine environment field information and the floating body drift trajectory information, and can achieve efficient and accurate prediction of the floating body trajectory, and can be deployed to local servers, cloud platforms or edge devices. The deployed model can be integrated into the existing marine decision-making system to achieve efficient transmission and processing of marine environment field data and trajectory data, and the processed data can be input into the model for fusion prediction. Using data visualization tools, the prediction results can be displayed in real time. In addition, the present invention adopts a periodic self-iteration update strategy, which adjusts the model parameters or retrains the model by regularly evaluating the prediction accuracy and computational efficiency to adapt to the changing environment and data characteristics, thereby improving the prediction ability and application scope. Through these steps, the model can be effectively applied to actual marine environmental monitoring and prediction, providing strong support for related research and decision-making.

[0044] The present invention significantly improves the accuracy of floating body trajectory prediction, and realizes accurate prediction of complex nonlinear floating body motion by integrating marine environment field information with historical floating body trajectory data. By using CNN to extract local features and Informer's global modeling capabilities, it is possible to capture the deep relationship between multidimensional data, which makes the CNN-Informer model more robust when dealing with complex marine environments.

[0045] Compared with traditional methods that rely on high-cost physical models and experimental simulations, the present invention uses open-source ocean reanalysis data and efficient deep learning models to significantly reduce the cost of experiments and data acquisition, providing a new and effective method for floating body trajectory prediction. At the same time, the present invention has strong scalability and adaptability, can be integrated with different marine environmental data and trajectory data, and does not require frequent changes for long-term applications, and has a long technical life. The periodic self-iterative update strategy allows the model to automatically optimize according to changes in the marine environment and maintain long-term effectiveness and applicability.

[0046] Based on the above content, it can be seen that the present invention overcomes the following problems: 1. Difficulty in fusing multi-source heterogeneous data: Existing trajectory prediction methods are difficult to effectively fuse data from different sources and types, resulting in inaccurate prediction results. The present invention uses CNN and Informer models to extract the features of floating body trajectory data and marine environment field data respectively, and realizes the effective fusion of multi-source heterogeneous data.

[0047] 2. Insufficient use of marine environmental information: Traditional methods fail to fully consider the impact of the marine environmental field on the trajectory of the floating body, resulting in low prediction accuracy. The present invention improves the accuracy of trajectory prediction by constructing the two-dimensional characteristics of the marine environmental field and making full use of environmental field variables such as ocean temperature, salinity, and ocean currents.

[0048] 3. Insufficient real-time performance and accuracy: The marine environment changes frequently, requiring the trajectory prediction method to have high real-time performance and accuracy. The present invention uses the efficient feature extraction capability of the deep learning method to achieve high-precision real-time prediction of the floating body trajectory.

[0049] It can be seen that the present invention proposes a floating body trajectory prediction technology that integrates multi-source heterogeneous data by combining CNN and Informer models, which can solve the problems of data fusion and utilization, and then make full use of marine environmental field information, and significantly improve the prediction accuracy and real-time performance, which has important application value in the fields of marine engineering and physical oceanography.

[0050] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0051] In the description of the present invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0052] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A floating body drift trajectory prediction method integrating ocean environment field, characterized in that: The following steps are involved: Collect the floating body drift trajectory data, perform feature screening through the random forest method after preprocessing, and construct a data set for predicting the floating body drift trajectory, wherein the floating body drift trajectory data consists of time, longitude, latitude, sea temperature, meridional velocity, zonal velocity, combined velocity and variance of sea temperature, meridional current velocity and zonal current velocity measured by the floating body; A combined model based on CNN and Informer models is constructed, and the floating body drift trajectory prediction model is generated through training with the data set to predict the floating body drift trajectory.

2. The method for predicting floating body drift trajectory integrating ocean environment field according to claim 1 is characterized by: In the process of preprocessing the floating body drift trajectory data, the floating body drift trajectory data is cleaned, including: eliminating trajectory data containing missing values; eliminating trajectory data with discontinuous observation time; eliminating trajectory data with observation time less than expected time; eliminating trajectory data that does not match the observation time of the marine environment field.

3. The method for predicting floating body drift trajectory integrating ocean environment field according to claim 2 is characterized in that: In the process of constructing the data set, the distance and azimuth of the floating body trajectory are obtained based on the floating body trajectory data at the current moment and the floating body trajectory data at the previous moment, and the distance and the azimuth are added to the floating body drifting trajectory data as data features. The floating body drifting trajectory data with the distance and the azimuth are feature screened through the random forest method to generate the data set.

4. The method for predicting floating body drift trajectory integrating ocean environment field according to claim 3 is characterized by: In a process of performing feature screening by using a random forest method, selecting first out-of-bag data, and obtaining a first error of the first out-of-bag data; Randomly adding noise interference to the first out-of-bag data, obtaining second out-of-bag data, and calculating a second error of the second out-of-bag data; According to the first error and the second error, the importance of all features in the first out-of-bag data is obtained, and features with small importance are deleted, and the process is repeated to select the first out-of-bag data corresponding to the combination with the smallest error to generate the data set.

5. The method for predicting floating body drift trajectory integrating ocean environment field according to claim 4 is characterized by: In the process of constructing the dataset, based on the floating body drift trajectory data obtained after feature screening through the random forest method, the ocean environment field data including the meridional current velocity and the latitudinal current velocity are obtained, and after time-space matching, a dataset for predicting the floating body drift trajectory is obtained.

6. The method for predicting floating body drift trajectory integrating ocean environment field according to claim 5 is characterized by: In the process of time-space matching, the ocean environment field data at the corresponding time is obtained according to the timestamp of the floating body drift trajectory data to perform time matching. If there is no corresponding time data, the data before and after the time is selected and supplemented by linear interpolation.

7. The method for predicting floating body drift trajectory integrating ocean environment field according to claim 6 is characterized by: In the process of time-space matching, the longitude and latitude positions of the floating body trajectory are spatially matched with the longitude and latitude grids of the ocean environment field data; for each floating body trajectory point, it is indexed to the grid of the ocean environment field according to its longitude and latitude, and extends outward 0.5° in the latitude and longitude directions with this as the center, forming an n*n two-dimensional ocean environment field centered on the trajectory.

8. The method for predicting floating body drift trajectory integrating ocean environment field according to claim 7 is characterized by: In the process of obtaining a data set for predicting the floating body drift trajectory, the floating body drift trajectory data and the ocean environment field data are fused based on the results of time matching and space matching, so that each trajectory point used for prediction will have corresponding environmental field data, including sea surface temperature, salinity, meridional current velocity and latitudinal current velocity, to construct the data set.

9. The method for predicting floating body drift trajectory integrating ocean environment field according to claim 8, characterized in that: In the process of building the floating body drift trajectory prediction model, the early stopping mechanism is adopted to prevent overfitting, and the model performance is optimized through cross-validation; the loss function of the model adopts mean square error and is trained through Adam optimizer.

10. A floating body drift trajectory prediction system integrating ocean environment field, characterized in that: The system is used to implement a floating body drift trajectory prediction method integrating an ocean environment field as described in any one of claims 1 to 9, and the system comprises: A data processing module is used to obtain the collected floating body drift trajectory data, perform preprocessing, and then perform feature screening through a random forest method to construct a data set for predicting the floating body drift trajectory, wherein the floating body drift trajectory data consists of time, longitude, latitude, sea temperature measured by the floating body, meridional velocity, zonal velocity, combined velocity, and variance of sea temperature, meridional current velocity, and zonal current velocity; The trajectory prediction module is used to construct a combined model based on the CNN and Informer models, train the data set, generate a floating body drift trajectory prediction model, and predict the floating body drift trajectory.

Images