A prediction method and system for ocean near-inertial current

Abstract

The present invention discloses a method and system for predicting near-inertial flow in the ocean, the method comprising: obtaining multi-source observation data for data preprocessing to obtain target data; constructing a first wind field predicted by a meteorological model and a second wind field in a tropical cyclone period constructed by fusion; constructing a three-dimensional FVCOM model to generate a flow field simulation result; performing data assimilation processing on the flow field simulation result generated by the three-dimensional FVCOM model and the target data to obtain assimilated ocean current data; extracting near-inertial flow and signal features based on the unassimilated ocean current data and the assimilated ocean current data; extracting near-inertial wind stress; obtaining a pre-trained sequence graph neural network TGCN model, analyzing the correlation between near-inertial wind stress and near-inertial flow, constructing a prediction system for constructing near-inertial flow, and completing the prediction of near-inertial flow in the ocean. The present invention improves prediction efficiency and accuracy, and can be widely used in the field of computer technology.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method and system for predicting ocean near-inertial current. Background Art

[0002] The movement of seawater in the ocean that is close to the local inertial frequency (f0) is usually called near-inertial oscillation. It is widely present in different depths of the global ocean, and the main reason for its generation is the unsteady wind driving force on the sea surface. The passage of tropical cyclones can excite significant near-inertial oscillations. In the process of energy transfer and mixing, near-inertial oscillations play an important role in the vertical and horizontal energy transfer of the ocean, especially their function in mixing heat and nutrients in the surface and deep water bodies. In terms of the impact on the climate system, by affecting the temperature and flow rate of the ocean surface, near-inertial oscillations indirectly affect the energy and material exchange between the ocean and the atmosphere, which has an important impact on the regulation of the global climate system. In terms of improving model predictions, accurately estimating and simulating the energy of near-inertial oscillations can help improve the accuracy of climate models and weather prediction models, especially in predicting storms and other extreme weather events. In environmental monitoring and management, the study of near-inertial oscillations can enhance the monitoring capabilities of changes in the marine ecological environment, especially today when global warming and the impact of human activities are increasing. Near-inertial internal waves have a huge impact on the internal dynamics, heat and ecological environment of the ocean, and play a very important role in the global ocean energy balance. They are also of great significance for understanding changes in the ocean environment, improving tropical cyclone forecasts, and disaster prevention and mitigation. Therefore, the generation, evolution and mechanism of near-inertial oscillations have always been a hot topic in the field of oceanography.

[0003] The current state of research in the field of near-inertial flow in the ocean covers a variety of methods from direct observation to advanced numerical simulation. The following are some of the main methods and their research progress:

[0004] (1) Direct observation method: Direct observation is a direct way to estimate the energy of near-inertial oscillations, including the use of buoys, ARGO buoys, satellite remote sensing technology, and anchoring equipment. These devices can provide data such as water temperature, salinity, and flow velocity, through which the kinetic energy of near-inertial oscillations can be calculated. For example, by analyzing the ocean flow field data recorded by drifting buoys and the wind field data obtained by satellites, the wind-generated near-inertial energy can be directly calculated.

[0005] (2) Spectral analysis method: Spectral analysis is to identify the characteristic frequencies of near-inertial oscillations by analyzing the time series of ocean physical quantities. Fast Fourier transform (FFT) is usually used to obtain the spectrum and identify the energy peak of the near-inertial frequency.

[0006] (3) Numerical simulation method: Numerical simulation is a traditional method that uses computer models to simulate the physical processes of the ocean and estimate the near-inertial oscillation energy. The model is usually based on the basic equations of ocean dynamics, such as the Navier-Stokes equations, and simulates the evolution of the ocean state through time stepping. For example, by simulating the interaction between the wind field and the ocean flow, the stress of the wind on the ocean surface can be estimated, and then the input of the near-inertial energy can be calculated.

[0007] (4) Statistical analysis method: The statistical analysis method estimates the energy of near-inertial oscillations by performing statistical analysis on historical observation data, including regression analysis, variance analysis and other techniques.

[0008] (5) Estimation method considering the impact of tropical cyclones: Tropical cyclones are another important factor affecting the estimation of the near-inertial oscillation energy of the ocean. Studies have shown that tropical cyclones can significantly increase the near-inertial energy of the ocean, especially in low-latitude regions. By incorporating tropical cyclone wind field data into numerical models, the contribution of tropical cyclones to the near-inertial energy can be estimated more accurately.

[0009] Each of these methods has its own specific applicability and limitations, and choosing an appropriate method often requires considering the specific research objectives, available data, and resources.

[0010] The related technology has the following disadvantages:

[0011] (1) Spectral analysis based on observation data: Observation data generally comes from ocean data provided by submerged buoys, buoys and other anchored observation equipment placed in the ocean, and wind field data obtained by satellite and other remote sensing technologies. Bandpass filtering and spectrum analysis methods are used to analyze the spectrum characteristics of ocean observation data, identify the frequency range of near-inertial oscillations, and estimate their intensity accordingly. The spectrum analysis method has a large demand for data samples. If the sampling rate is too low or the data is insufficient, it may not be able to accurately identify near-inertial oscillations. Secondly, the spectrum method cannot show large-scale changes in near-inertial oscillations in space and time.

[0012] (2) Numerical simulation method: This is a method to construct a mathematical model of near-inertial oscillation based on ocean dynamics theory. Its limitations are: the accuracy of this method depends on the accuracy and completeness of the mathematical model, but it is difficult to construct a completely accurate simulation of near-inertial motion. In addition, this method requires a lot of computing resources and requires large-scale or high-resolution simulations, which has a very high computing cost;

[0013] (3) Statistical analysis methods: Although statistical analysis methods can utilize a large amount of historical data and help understand long-term trends and changes, they are limited by data quality and availability;

[0014] Direct observation method: Although the direct observation method can provide direct and accurate ocean physical property data and reduce the impact of model uncertainty, it has limited spatial and temporal coverage and high cost. Summary of the invention

[0015] In view of this, an embodiment of the present invention provides a method and system for predicting ocean near-inertial current to improve the accuracy of prediction.

[0016] An aspect of an embodiment of the present invention provides a method for predicting ocean near-inertial current, comprising:

[0017] Acquire multi-source observation data;

[0018] Performing data preprocessing on the multi-source observation data to obtain target data for a prediction process;

[0019] Construct the first wind field predicted by the meteorological model and the second wind field during the tropical cyclone period by fusion;

[0020] Based on the first wind field and the second wind field as model drivers, a three-dimensional FVCOM model is constructed to generate flow field simulation results;

[0021] Performing data assimilation processing on the flow field simulation result generated by the three-dimensional FVCOM model and the target data to obtain assimilated ocean current data;

[0022] Extracting near-inertial current according to the unassimilated ocean current data and the assimilated ocean current data, and extracting detailed signal features of the near-inertial current;

[0023] According to the detailed signal characteristics, the near-inertial wind stress is extracted;

[0024] A pre-trained sequence graph neural network TGCN model is obtained, the correlation between the near-inertial wind stress and the near-inertial flow is analyzed, a prediction system for the near-inertial flow is constructed, and the prediction of the near-inertial flow of the ocean is completed.

[0025] Optionally, the multi-source observation data includes ocean current data, water depth topography data, wind field data, typhoon path data and coastline data, and the preprocessing of the multi-source observation data to obtain target data for the prediction process includes:

[0026] Using the DBCSCAN clustering algorithm to remove outliers in the multi-source observation data;

[0027] Using the spatiotemporal Kriging interpolation method to fill missing values ​​in the multi-source observation data;

[0028] The inverted flow field is smoothed using the median filter method.

[0029] Optionally, the step of constructing a first wind field predicted by a meteorological model includes:

[0030] The ECMWF global forecast model was used to obtain the predicted wind field;

[0031] Obtain wind farm forecast data from ECMWF’s data service platform;

[0032] By using a data processing tool, wind field data within a research period, research area, and research height range are selected and extracted from the predicted wind field and wind field forecast data to construct the first wind field;

[0033] The second wind field during the tropical cyclone period constructed by fusion includes:

[0034] Get CCMP reanalysis data from NOAA's NCEI Data Center;

[0035] Select and extract reanalysis wind field data for a study period and a study area from the CCMP reanalysis data according to a data processing tool;

[0036] Determine the maximum wind speed and maximum radius of the typhoon based on the acquired typhoon path data, and write the calculation function of the Holland model;

[0037] The CCMP wind field data is used as the background wind field, and the typhoon wind field results calculated by the Holland model are used to correct and superimpose the background wind field to obtain the tropical cyclone period wind field as the second wind field.

[0038] Optionally, constructing a three-dimensional FVCOM model based on the first wind field and the second wind field as model drivers to generate flow field simulation results includes:

[0039] The mesh files were generated by mesh generation software to describe the topography and water structure of the study area using triangular meshes;

[0040] Interpolating the water depth data to the grid points of the grid file by a bilinear interpolation method;

[0041] Configuring three-dimensional FVCOM model parameters, wherein the three-dimensional FVCOM model parameters include seawater density, horizontal mixing coefficient, vertical mixing coefficient, and seawater viscosity coefficient;

[0042] Configuring boundary conditions of the three-dimensional FVCOM model, wherein the boundary conditions include sea surface pressure, wind speed, and land boundary conditions;

[0043] Using the first wind farm and the second wind farm as model drivers, configuring grid files, boundary conditions and driving data files according to model requirements;

[0044] The three-dimensional FVCOM model is started to run, the output and running status of the three-dimensional FVCOM model are monitored, and the flow field simulation results output by the three-dimensional FVCOM model are obtained, wherein the flow field simulation results include ocean current velocity and sea surface wind stress.

[0045] Optionally, performing data assimilation processing on the flow field simulation result generated by the three-dimensional FVCOM model and the target data to obtain assimilated ocean current data includes:

[0046] A set of particles is randomly generated, each particle represents a possible state of the system;

[0047] Use the model output to make state predictions for each particle;

[0048] Perform local perturbations on each prior particle to obtain the proposed particle;

[0049] Calculating proposal weights and normalizing the proposal weights;

[0050] Re-extract particles according to the standardized proposal weights and retain particles with high fit;

[0051] Using the resampled particle collection, an estimate of the state is calculated using a weighted average;

[0052] Quantitatively evaluate the simulation accuracy of the assimilation system.

[0053] Optionally, extracting the near-inertial current according to the unassimilated ocean current data and the assimilated ocean current data, and extracting detailed signal features of the near-inertial current, comprises:

[0054] Define the initial frequency set and telescopic length according to the near-inertial flow frequency range;

[0055] According to the initial frequency set and the telescopic length, a fourth-order Butterworth bandpass filter is configured;

[0056] Performing forward filtering and reverse filtering on the unassimilated ocean current data and the assimilated ocean current data through the fourth-order Butterworth bandpass filter to obtain a two-way filtering result, and then obtaining different sets of near-inertial current extraction results;

[0057] The variational mode decomposition is used to perform mode decomposition and reconstruction on the near-inertial flow extraction results to obtain the near-inertial flow signal.

[0058] Optionally, the method of performing modal decomposition and reconstruction on the near-inertial flow extraction result by using variational modal decomposition to obtain a near-inertial flow signal includes:

[0059] Inputting the near-inertial flow extraction result into the VMD algorithm for decomposition;

[0060] Determine the number of modes to be decomposed, adjust the regularization parameters, and control the smoothness and sparsity between modes during the decomposition process;

[0061] VMD decomposition is performed using an iterative optimization algorithm using the alternating direction multiplier method;

[0062] Set the initial mode function and Lagrange multiplier parameters;

[0063] After the iterative optimization is completed, each mode after decomposition is obtained;

[0064] The energy is calculated for each mode using the square of the modal amplitude as an estimate of the energy of each mode;

[0065] Normalize the energy of all modes;

[0066] The normalized energy is used as a reference for weights, and a weight is assigned to each mode.

[0067] Each modal component is weighted and combined according to its corresponding weight to obtain the final reconstructed signal;

[0068] By plotting the comparison graph between the original signal and the reconstructed signal, the reconstruction error is calculated to evaluate the reconstruction effect.

[0069] Optionally, the obtaining of a pre-trained sequence graph neural network TGCN model, analyzing the correlation between the near-inertial wind stress and the near-inertial flow, constructing a prediction system for the near-inertial flow, and completing the prediction of the near-inertial flow of the ocean includes:

[0070] The extracted near-inertial flow and near-inertial wind stress are converted from the frequency domain to the time domain through inverse Fourier transform;

[0071] Based on tropical cyclone data, the extracted near-inertial flow and near-inertial wind stress data are divided into three parts: before, during, and after the tropical cyclone.

[0072] Obtain a pre-trained TGCN model from the open source library whose model structure and parameters are similar to the application scenario of near-inertial flow prediction;

[0073] According to the selected pre-trained model structure, freeze some layers of the pre-trained model to retain the features and representations learned by the model in the source domain;

[0074] Establishing a diffusion model based on the control equation to simulate the propagation process of near-inertial energy, and adding the diffusion model as a custom layer to the pre-trained model;

[0075] Input existing data to debug the pre-trained model and determine the optimal values ​​of the parameters;

[0076] Add the calculation expression of near inertial kinetic energy into the fully connected layer to obtain the distribution prediction of near inertial kinetic energy;

[0077] Wherein, the calculation expression of the near inertial kinetic energy is:

[0078]

[0079] NIKE represents near inertial kinetic energy; ρ is the density of sea water.

[0080] Another aspect of the embodiment of the present invention further provides a prediction system for ocean near-inertial current, comprising:

[0081] The first module is used to obtain multi-source observation data;

[0082] The second module is used to perform data preprocessing on the multi-source observation data to obtain target data for the prediction process;

[0083] The third module is used to construct the first wind field predicted by the meteorological model and the second wind field during the tropical cyclone period constructed by fusion;

[0084] A fourth module is used to construct a three-dimensional FVCOM model based on the first wind field and the second wind field as model drivers to generate flow field simulation results;

[0085] The fifth module is used to perform data assimilation processing on the flow field simulation result generated by the three-dimensional FVCOM model and the target data to obtain assimilated ocean current data;

[0086] The sixth module is used to extract the near-inertial current according to the unassimilated ocean current data and the assimilated ocean current data, and to extract the detailed signal characteristics of the near-inertial current;

[0087] The seventh module is used to extract the near-inertial wind stress according to the detailed signal characteristics;

[0088] The eighth module is used to obtain a pre-trained sequence graph neural network TGCN model, analyze the correlation between the near-inertial wind stress and the near-inertial flow, construct a prediction system for the near-inertial flow, and complete the prediction of the ocean near-inertial flow.

[0089] Another aspect of an embodiment of the present invention further provides an electronic device, including a processor and a memory;

[0090] The memory is used to store programs;

[0091] The processor executes the program to implement the method described above.

[0092] The embodiment of the present invention also discloses a computer program product or a computer program, which includes a computer instruction stored in a computer-readable storage medium. A processor of a computer device can read the computer instruction from the computer-readable storage medium, and the processor executes the computer instruction, so that the computer device executes the above method.

[0093] The embodiment of the present invention obtains multi-source observation data; performs data preprocessing on the multi-source observation data to obtain target data for the prediction process; constructs a first wind field predicted by a meteorological model and a second wind field during a tropical cyclone period constructed by fusion; uses the first wind field and the second wind field as model drivers to construct a three-dimensional FVCOM model to generate a flow field simulation result; performs data assimilation processing on the flow field simulation result generated by the three-dimensional FVCOM model and the target data to obtain assimilated ocean current data; extracts near-inertial flow based on the unassimilated ocean current data and the assimilated ocean current data, and extracts detailed signal features of the near-inertial flow; extracts near-inertial wind stress based on the detailed signal features; obtains a pre-trained sequence graph neural network TGCN model, analyzes the correlation between the near-inertial wind stress and the near-inertial flow, constructs a prediction system for constructing near-inertial flow, and completes the prediction of ocean near-inertial flow. The present invention improves prediction efficiency and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0094] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0095] Figure 1 An overall step flow chart provided for an embodiment of the present invention;

[0096] Figure 2 A schematic diagram of set partitioning provided by an embodiment of the present invention;

[0097] Figure 3 A schematic diagram of transfer learning provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0098] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is further described in detail below in conjunction with the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the embodiments of the present application. They are only examples of devices and methods consistent with some aspects of the embodiments of the present application as detailed in the attached claims.

[0099] It is understood that the terms "first", "second", etc. used in this application can be used to describe various concepts in this article, but unless otherwise specified, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another concept. For example, without departing from the scope of the embodiment of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the words "if" and "if" as used herein can be interpreted as "at the time of" or "when" or "in response to determination".

[0100] The terms "at least one", "multiple", "each", "any", etc. used in this application, at least one includes one, two or more, multiple includes two or more, each refers to each of the corresponding multiple, and any refers to any one of the multiple.

[0101] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0102] The prediction method of the near-inertial flow of the ocean provided in the embodiment of the present application relates to the field of computers. The prediction method of the near-inertial flow of the ocean provided in the embodiment of the present application can be applied to a terminal, can also be applied to a server, and can also be software running in a terminal or a server. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, and a car terminal, etc., but is not limited to this; the server side can be configured as an independent physical server, or it can be configured as a server cluster or a distributed system composed of multiple physical servers, and can also be configured to provide cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms and other basic cloud computing services. The server can also be a node server in a blockchain network; the software can be an application of the prediction method of the near-inertial flow of the ocean, etc., but is not limited to the above forms.

[0103] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0104] Specifically, the method for predicting ocean near-inertial current provided by the embodiment of the present invention comprises the following steps:

[0105] Acquire multi-source observation data;

[0106] Performing data preprocessing on the multi-source observation data to obtain target data for a prediction process;

[0107] Construct the first wind field predicted by the meteorological model and the second wind field during the tropical cyclone period by fusion;

[0108] Based on the first wind field and the second wind field as model drivers, a three-dimensional FVCOM model is constructed to generate flow field simulation results;

[0109] Performing data assimilation processing on the flow field simulation result generated by the three-dimensional FVCOM model and the target data to obtain assimilated ocean current data;

[0110] Extracting near-inertial current according to the unassimilated ocean current data and the assimilated ocean current data, and extracting detailed signal features of the near-inertial current;

[0111] According to the detailed signal characteristics, the near-inertial wind stress is extracted;

[0112] A pre-trained sequence graph neural network TGCN model is obtained, the correlation between the near-inertial wind stress and the near-inertial flow is analyzed, a prediction system for the near-inertial flow is constructed, and the prediction of the near-inertial flow of the ocean is completed.

[0113] Optionally, the multi-source observation data includes ocean current data, water depth topography data, wind field data, typhoon path data and coastline data, and the preprocessing of the multi-source observation data to obtain target data for the prediction process includes:

[0114] Using the DBCSCAN clustering algorithm to remove outliers in the multi-source observation data;

[0115] Using the spatiotemporal Kriging interpolation method to fill missing values ​​in the multi-source observation data;

[0116] The inverted flow field is smoothed using the median filter method.

[0117] Optionally, the step of constructing a first wind field predicted by a meteorological model includes:

[0118] The ECMWF global forecast model was used to obtain the predicted wind field;

[0119] Obtain wind farm forecast data from ECMWF’s data service platform;

[0120] By using a data processing tool, wind field data within a research period, research area, and research height range are selected and extracted from the predicted wind field and wind field forecast data to construct the first wind field;

[0121] The second wind field during the tropical cyclone period constructed by fusion includes:

[0122] Get CCMP reanalysis data from NOAA's NCEI Data Center;

[0123] Select and extract reanalysis wind field data for a study period and a study area from the CCMP reanalysis data according to a data processing tool;

[0124] Determine the maximum wind speed and maximum radius of the typhoon based on the acquired typhoon path data, and write the calculation function of the Holland model;

[0125] The CCMP wind field data is used as the background wind field, and the typhoon wind field results calculated by the Holland model are used to correct and superimpose the background wind field to obtain the tropical cyclone period wind field as the second wind field.

[0126] Optionally, constructing a three-dimensional FVCOM model based on the first wind field and the second wind field as model drivers to generate flow field simulation results includes:

[0127] The mesh files were generated by mesh generation software to describe the topography and water structure of the study area using triangular meshes;

[0128] Interpolating the water depth data to the grid points of the grid file by a bilinear interpolation method;

[0129] Configuring three-dimensional FVCOM model parameters, wherein the three-dimensional FVCOM model parameters include seawater density, horizontal mixing coefficient, vertical mixing coefficient, and seawater viscosity coefficient;

[0130] Configuring boundary conditions of the three-dimensional FVCOM model, wherein the boundary conditions include sea surface pressure, wind speed, and land boundary conditions;

[0131] Using the first wind farm and the second wind farm as model drivers, configuring grid files, boundary conditions and driving data files according to model requirements;

[0132] The three-dimensional FVCOM model is started to run, the output and running status of the three-dimensional FVCOM model are monitored, and the flow field simulation results output by the three-dimensional FVCOM model are obtained, wherein the flow field simulation results include ocean current velocity and sea surface wind stress.

[0133] Optionally, performing data assimilation processing on the flow field simulation result generated by the three-dimensional FVCOM model and the target data to obtain assimilated ocean current data includes:

[0134] A set of particles is randomly generated, each particle represents a possible state of the system;

[0135] Use the model output to make state predictions for each particle;

[0136] Perform local perturbations on each prior particle to obtain the proposed particle;

[0137] Calculating proposal weights and normalizing the proposal weights;

[0138] Re-extract particles according to the standardized proposal weights and retain particles with high fit;

[0139] Using the resampled particle collection, an estimate of the state is calculated using a weighted average;

[0140] Quantitatively evaluate the simulation accuracy of the assimilation system.

[0141] Optionally, extracting the near-inertial current according to the unassimilated ocean current data and the assimilated ocean current data, and extracting detailed signal features of the near-inertial current, comprises:

[0142] Define the initial frequency set and telescopic length according to the near-inertial flow frequency range;

[0143] According to the initial frequency set and the telescopic length, a fourth-order Butterworth bandpass filter is configured;

[0144] Performing forward filtering and reverse filtering on the unassimilated ocean current data and the assimilated ocean current data through the fourth-order Butterworth bandpass filter to obtain a two-way filtering result, and then obtaining different sets of near-inertial current extraction results;

[0145] The variational mode decomposition is used to perform mode decomposition and reconstruction on the near-inertial flow extraction results to obtain the near-inertial flow signal.

[0146] Optionally, the method of performing modal decomposition and reconstruction on the near-inertial flow extraction result by using variational modal decomposition to obtain a near-inertial flow signal includes:

[0147] Inputting the near-inertial flow extraction result into the VMD algorithm for decomposition;

[0148] Determine the number of modes to be decomposed, adjust the regularization parameters, and control the smoothness and sparsity between modes during the decomposition process;

[0149] VMD decomposition is performed using an iterative optimization algorithm using the alternating direction multiplier method;

[0150] Set the initial mode function and Lagrange multiplier parameters;

[0151] After the iterative optimization is completed, each mode after decomposition is obtained;

[0152] The energy is calculated for each mode using the square of the modal amplitude as an estimate of the energy of each mode;

[0153] Normalize the energy of all modes;

[0154] The normalized energy is used as a reference for weights, and a weight is assigned to each mode.

[0155] Each modal component is weighted and combined according to its corresponding weight to obtain the final reconstructed signal;

[0156] By plotting the comparison graph between the original signal and the reconstructed signal, the reconstruction error is calculated to evaluate the reconstruction effect.

[0157] Optionally, the obtaining of a pre-trained sequence graph neural network TGCN model, analyzing the correlation between the near-inertial wind stress and the near-inertial flow, constructing a prediction system for the near-inertial flow, and completing the prediction of the near-inertial flow of the ocean includes:

[0158] The extracted near-inertial flow and near-inertial wind stress are converted from the frequency domain to the time domain through inverse Fourier transform;

[0159] Based on tropical cyclone data, the extracted near-inertial flow and near-inertial wind stress data are divided into three parts: before, during, and after the tropical cyclone.

[0160] Obtain a pre-trained TGCN model from the open source library whose model structure and parameters are similar to the application scenario of near-inertial flow prediction;

[0161] According to the selected pre-trained model structure, freeze some layers of the pre-trained model to retain the features and representations learned by the model in the source domain;

[0162] Establishing a diffusion model based on the control equation to simulate the propagation process of near-inertial energy, and adding the diffusion model as a custom layer to the pre-trained model;

[0163] Input existing data to debug the pre-trained model and determine the optimal values ​​of the parameters;

[0164] Add the calculation expression of near inertial kinetic energy into the fully connected layer to obtain the distribution prediction of near inertial kinetic energy;

[0165] Wherein, the calculation expression of the near inertial kinetic energy is:

[0166]

[0167] NIKE represents near inertial kinetic energy; ρ is the density of sea water.

[0168] The specific implementation process of the present invention in a specific scenario is described in detail below in conjunction with the accompanying drawings of the specification:

[0169] refer to Figure 1 The method for predicting ocean near-inertial current of the present invention comprises the following steps:

[0170] 1. Acquisition of multi-source observation data

[0171] The data that need to be obtained in this step mainly include ocean current data, water depth and terrain data, wind field data, typhoon path data, and coastline data. Among them, ocean current data need to include satellite observation data, ground wave radar observation data, and buoy observation data, which can be obtained from satellite operators, the National Space Administration, marine meteorological departments, and marine observation stations; water depth and terrain data and coastline data can be obtained from geographic information systems such as GIS; wind field data can be obtained from the European Center for Medium-Range Weather Forecasts (ECMWF) and other places; typhoon path data can be obtained from the Tropical Cyclone Data Center of the China Meteorological Administration.

[0172] 2. Data Preprocessing

[0173] The acquired multi-source data are preprocessed to facilitate the subsequent data assimilation process.

[0174] (1) Outlier processing: Use the DBCSCAN clustering algorithm to remove outliers. DBSCAN is a density-based spatial clustering method for noise applications. It defines a cluster as the largest set of density-connected points and can mark out noise points and outliers in the data set.

[0175] The algorithm steps are as follows:

[0176] ① Set the parameters Epsilon neighborhood radius Eps (hereinafter referred to as r) and the minimum number of adjacent points MinPts (hereinafter referred to as m). The clustering effect is sensitive to parameter settings, and it is necessary to try multiple different parameter settings based on the data set to obtain the best result;

[0177] ② Select a point C1 from the input data set C as the current point and initialize the first cluster label L to 1;

[0178] ③ Find the set of points in the Epsilon neighborhood of the current point. If the number of points in the set is greater than m, mark the current C1 point as a core point belonging to cluster L; if there is no point in the set, mark the current C1 point as an outlier and go to step ⑤; if there is a point in the set, but the number of points is less than m, mark the current C1 point as a noise point and go to step ⑤;

[0179] ④ Go through each neighboring point (i.e., take each neighboring point as the new current point) and repeat step ③ until no new neighboring point that can be marked as belonging to the current cluster L is found;

[0180] ⑤Select the next unlabeled point in data set C as the current point and increase the number of clusters by 1;

[0181] ⑥ Repeat steps ③-⑤ until all points in C are marked;

[0182] ⑦ Eliminate outliers in the data set.

[0183] Note: If the noise point marked first meets the Eps-MinPts condition of the later point, the noise point can be reallocated to the cluster.

[0184] (2) Missing value processing: Check whether there are missing values ​​in the data, determine the location and number of missing values, and use the spatiotemporal Kriging interpolation method to fill in the missing values.

[0185] The specific method is as follows:

[0186] ① Spatial distance calculation: Use Euclidean distance as spatial distance h s Computational methods to analyze the spatial structure of data;

[0187] Euclidean distance:

[0188] ②Time distance calculation: Here, the standardized time distance is used as the time distance h t Calculation method, analyze the time structure of data, need to choose the appropriate basic time unit t0 according to the time span of data:

[0189]

[0190] ③ Based on the results of spatial distance and temporal distance, set the appropriate spatiotemporal variation function to establish the spatiotemporal Kriging interpolation model. After expanding to the spatiotemporal domain, assume that the spatiotemporal regionalized variable Z(s,t)∈S*T satisfies the second-order stationary hypothesis, then:

[0191] Space-time variogram:

[0192] Among them, S represents the spatial domain, T represents the time domain, (s, t) represents the coordinates of the sample point in the spatiotemporal domain; h s and h t They are spatial distance and temporal distance respectively.

[0193] ④Weight coefficient λ i Under the assumption of second-order stationarity, the space-time Kriging equations are used to calculate:

[0194]

[0195] ⑤ Using the established spatiotemporal Kriging interpolation model, the missing values ​​are calculated and filled into the data set according to the spatial and temporal distances of the surrounding known data points and the set semivariogram;

[0196]

[0197] (3) Smoothing: The real flow field is relatively smooth in space, while the acquired flow field data may have many mutation values ​​or abnormal values ​​(such as the sea surface flow field inverted by ground wave radar). In this step, the median filter method is used to smooth the inverted flow field.

[0198] The specific method is as follows:

[0199] ① Use the grayscale algorithm to convert the inverted sea surface flow field into a grayscale image;

[0200] ② Define the filter window size according to the degree of smoothing required. The window size needs to be an odd number. Here, the window size is defined as 3*3;

[0201] ③ Perform filtering on each pixel point, that is, sort the pixel values ​​in the surrounding window and select the median value as the new pixel value of the point;

[0202] ④ Repeat step ③ until all pixels have been filtered.

[0203] 3. Wind farm construction

[0204] Considering the impact of the tropical cyclone formation and disappearance process, two wind fields need to be constructed in this step, namely the wind field predicted by the meteorological model and the tropical cyclone period wind field constructed by fusion.

[0205] The specific method is as follows:

[0206] (1) Wind field prediction by meteorological model:

[0207] ① In this step, the ECMWF Global Forecast Model (GFS) is used to obtain the predicted wind field;

[0208] ② Wind farm forecast data (NetCDF format) can be directly obtained from ECMWF’s data service platform;

[0209] ③ Use data processing tools (Matlab, etc.) to extract wind field data within the research period, research area, and research height range to construct the wind field.

[0210] (2) Integrated tropical cyclone wind field: This step constructs the typhoon wind field based on the Holland typhoon model and reconstructs it on the CCMP wind field to obtain the tropical cyclone wind field of the calculation sea area.

[0211] The specific construction method is as follows:

[0212] ① Obtain CCMP reanalysis data from NOAA's NCEI Data Center in NetCDF format;

[0213] ②Use data processing tools (Matlab, etc.) to select and extract reanalysis wind field data for the study period and study area;

[0214] ③ Determine the maximum wind speed and maximum radius of the typhoon based on the acquired typhoon path data, and write the calculation function of the Holland model based on this;

[0215] ④ The CCMP wind field data is used as the background wind field, and the typhoon wind field results calculated by the Holland model are used to correct and superimpose them to obtain the tropical cyclone wind field.

[0216] 4. Construction of 3D FVCOM model:

[0217] In this step, the FVCOM model is used to simulate the flow field.

[0218] The specific construction method is as follows:

[0219] (1) Prepare the grid: Use triangular grid to describe the topography and water structure of the study area. You can use professional grid generation software such as Triangle to generate grid files. In order to ensure better resolution of the hydrodynamic process near the coastline, it is necessary to set the grid density according to the distance from the coastline and generate more grid nodes near the coastline.

[0220] (2) Import the grid file and interpolate the water depth data to the grid points using the bilinear interpolation method;

[0221] (3) Setting model parameters: seawater density, horizontal mixing coefficient, vertical mixing coefficient, seawater viscosity coefficient, etc.;

[0222] (4) Setting boundary conditions: sea surface pressure, wind speed, and land boundary conditions;

[0223] (5) Set up the grid file, boundary conditions, and drive data file according to the model requirements (use the two wind fields constructed in step 3 as model drivers respectively;

[0224] (6) Start the model operation, monitor the model output and operation status; obtain the ocean current velocity, sea surface wind stress and other data output by the model.

[0225] 5. Data Assimilation

[0226] At present, the commonly used ocean data assimilation methods include optimal interpolation, variational method, ensemble Kalman filter (EnKF), particle filter (PF), etc. Among them, particle filter gets rid of the limitations of linear and Gaussian assumptions, but there may be filtering degradation problems in high-dimensional systems. Therefore, in this step, the particle filter method is selected, and EnKF is used as the proposed density to complete the assimilation of the three-dimensional FVCOM model output constructed in the above step 4 with multi-source observation data.

[0227] The specific method is as follows:

[0228] (1) Randomly generate a set of particles, each of which represents a possible state of the system;

[0229] (2) Use the model output to predict the state of each particle;

[0230] (3) Perform local perturbation EnKF on each prior particle to obtain the proposed particle

[0231] (4) Calculate the proposal weight according to the following formula And use parameter α to adjust;

[0232]

[0233]

[0234] Note: Usually, the value of α is slightly less than 1.

[0235] (5) Yes To standardize;

[0236]

[0237] (6) Re-extract particles according to the calculated weights so that more particles with high fitting degree are retained;

[0238] (7) Using the resampled particle set, calculate the estimated value of the state using weighted average;

[0239] (8) Quantitative evaluation of the simulation accuracy of the assimilation system:

[0240] ① Define the observation set as O, the model output set as M, and the assimilation set as A;

[0241] ② Calculate the difference OM between the observed value and the model output value;

[0242] ③ Calculate the difference OA between the observed value and the assimilated value;

[0243] ④ Calculate the mean and standard deviation of OM and OA respectively and compare them to quantitatively evaluate the simulation accuracy of the assimilation system.

[0244] 6. Near-inertial flow extraction

[0245] For the unassimilated ocean current data and the ocean current data assimilated in the previous step, the bidirectional filtering method of (0.8-1.2)f0 was used respectively, and the fourth-order Butterworth bandpass filter was selected, combined with the ensemble method to extract the near-inertial flow, and the VMD method was used to decompose and reconstruct the extracted near-inertial flow signal to extract more detailed signal features.

[0246] Inertia frequency f0: In the formula, ω≈7.292*10 -5 rad / s, the angular velocity of the Earth's rotation; is the latitude.

[0247] The specific method is as follows:

[0248] (1) Set division: Define the initial set f = [0.8f0, 1.2f0] according to the frequency range of the near-inertial flow, and the telescopic length Δ

[0249] f=0.004f0, then we can get the following set (such as Figure 2 ):

[0250]

[0251] (2) (Take set f as an example, and the same goes for the others. Modify F according to the set range. stop Butterworth bandpass filter design specifications: F stop1 =0.8f0,F stop2 =1.2f0,A pass =3dB,A stop =40dB;

[0252] (3) Perform forward filtering (left to right) and reverse filtering (right to left) on the signal respectively, and obtain the bidirectional filtering result by averaging the two filtering results;

[0253] (4) Repeat step (3) to obtain different sets of near-inertial flow extraction results, and calculate the average value to obtain the near-inertial flow extraction result.

[0254] (5) The variational mode decomposition (VMD) is used to perform modal decomposition and reconstruction on the extracted results to obtain a clear near-inertial flow signal.

[0255] The specific method is as follows:

[0256] ① Input the near-inertial flow signal extracted in (4) into the VMD algorithm for decomposition;

[0257] ② Determine the number of modes K to be decomposed, adjust the regularization parameter, and control the smoothness and sparsity between modes during the decomposition process;

[0258] ③ VMD decomposition is performed using an iterative optimization algorithm using the alternating direction multiplier method;

[0259] ④Set the initial mode function and Lagrange multiplier parameters;

[0260] ⑤After the iterative optimization is completed, the decomposed modes are obtained;

[0261] ⑥ Use the square of the modal amplitude as an estimate of the energy of each mode and calculate the energy for each mode;

[0262] ⑦ Normalize the energy of all modes:

[0263]

[0264] Where k represents the kth mode; K represents the number of decomposed modes; E k represents the energy estimate of the kth mode; E k-normal represents the energy of the kth mode after normalization;

[0265] ⑧Use the normalized energy as a reference for weights. Larger energies correspond to larger weights, and a weight is assigned to each mode.

[0266] ⑨ Perform weighted combination of each modal component according to its corresponding weight to obtain the final reconstructed signal;

[0267] ⑩ By drawing a comparison graph between the original signal and the reconstructed signal, the reconstruction error is calculated to evaluate the reconstruction effect.

[0268] 7. Near-inertial wind stress extraction

[0269] The filtering method is used to extract the wind stress of the near-inertial frequency from the wind stress data output by the model in step 6. Similar to the extraction of the near-inertial flow, the near-inertial frequency wind stress is extracted from the wind stress data using a fourth-order Butterworth bandpass filter.

[0270] Butterworth filter specification design: F stop1 =0.8f0,F stop2 =1.2f0,A pass =3dB,A stop =40dB.

[0271] 8. Construction of near-inertial flow prediction system

[0272] refer to Figure 3 The embodiment of the present invention obtains a corresponding pre-trained sequence graph neural network TGCN model, analyzes the correlation between near-inertial wind stress and near-inertial flow in sections (before, during and after the tropical cyclone passes), and constructs a prediction system for near-inertial flow by combining the transfer learning method and the diffusion model.

[0273] The specific method is as follows:

[0274] (1) The extracted near-inertial flow and near-inertial wind stress are converted from the frequency domain to the time domain through inverse Fourier transform;

[0275] (2) Based on tropical cyclone data, the extracted near-inertial flow and near-inertial wind stress data are divided into three parts: before, during, and after the tropical cyclone passes;

[0276] (3) Obtain pre-trained TGCN models with similar model structures and parameters to the near-inertial flow prediction application scenarios from open source libraries, such as models trained on similar spatiotemporal graph datasets;

[0277] (4) According to the selected pre-trained model structure, some layers of the pre-trained model are frozen, and the features and representations learned by the model in the source domain need to be retained;

[0278] (5) A diffusion model is established based on the following control equation to simulate the propagation process of near-inertial energy and added to the pre-trained model as a custom layer:

[0279]

[0280] Where: u, v are the east-west and north-south velocity components respectively; τ x and τ y are the east-west and north-south wind stress components respectively; H is the depth of the mixed layer; r is the empirical linear attenuation coefficient, which represents the e-fold time scale of velocity attenuation, including the combined effects of horizontal and vertical radiation of energy and local dissipation.

[0281] (6) Input existing data to debug the model and determine the optimal value of parameter r;

[0282] (7) Add a fully connected layer based on the following formula to obtain the distribution prediction of the near inertial kinetic energy NIKE:

[0283]

[0284] Where ρ is the density of seawater, and the reference density is 1024 kg / m 3 .

[0285] In summary, the present invention has the following characteristics in the process of realizing the prediction of ocean near-inertial flow:

[0286] 1. In the acquisition of ocean current data, numerical simulation and multi-source observation (satellite, ground wave radar, buoy) are combined, and data assimilation technology is used for fusion. The influence of the formation and disappearance of tropical cyclones is also taken into account in the process.

[0287] 2. A data assimilation method based on particle filtering is adopted, and EnKF is used as the proposed density to integrate multi-source observation data into the three-dimensional FVCOM model to obtain more accurate ocean current information.

[0288] 3. The CCMP reanalysis wind field data is used as the background wind field, and the typhoon wind field results calculated by the Holland model are used to correct and superimpose them to obtain the fused and constructed tropical cyclone period wind field.

[0289] 4. By driving the FVCOM model using the wind field constructed by the meteorological model and the tropical cyclone wind field constructed by fusion, the ocean current data and wind stress data under the two wind fields are obtained.

[0290] 5. According to the early, middle and late stages of a tropical cyclone, the correlation between near-inertial wind stress and near-inertial flow is analyzed segmentally using a sequence graph neural network, taking into account the impact of the tropical cyclone's formation and disappearance process.

[0291] 6. Combine transfer learning with the diffusion model to predict the distribution of near-inertial kinetic energy and build a prediction system for near-inertial flow.

[0292] 7. After adding the diffusion model as a custom layer to the pre-trained sequence graph neural network, the propagation of near-inertial energy is simulated based on the correlation between near-inertial wind stress and near-inertial flow learned by the neural network, and the distribution of near-inertial energy is output through the fully connected layer.

[0293] 8. Use the pre-trained sequence graph neural network as the source domain, and combine it with transfer learning to transfer the features and representations learned by the neural network on the original dataset to the target domain, and quickly build a near-inertial flow prediction system with good performance.

[0294] 9. A diffusion model is established based on Pollard's mixing layer model to simulate the propagation process of near-inertial energy.

[0295] 10. Combine the DBSCAN clustering algorithm, spatiotemporal Kriging interpolation method and median filtering to complete the preprocessing of multi-source ocean current observation data.

[0296] 11. The missing values ​​are filled by establishing a spatiotemporal Kriging interpolation model, and Euclidean distance and standardized time distance are selected for the calculation of spatial distance and time distance.

[0297] 12. Variational mode decomposition (VMD) is used to perform modal decomposition and reconstruction on the near-inertial flow signal extracted by the fourth-order Butterworth bandpass filter to obtain more accurate near-inertial flow information.

[0298] Compared with the prior art, the present invention has the following advantages:

[0299] The present invention uses data assimilation technology to integrate multi-source ocean current observation data into the three-dimensional FVCOM model. By comparing and adjusting the observation data with the model output, the forecasting ability and accuracy of the model are improved, and the real ocean state can be better reflected. In addition, in the process of model construction, the FVCOM model is driven by the wind field constructed by the meteorological model and the tropical cyclone wind field that integrates the Holland typhoon model and the CCMP reanalysis data, and the correlation between wind stress and near-inertial flow is calculated and analyzed in sections, taking into account the influence of the tropical cyclone generation and disappearance process. In the extraction of near-inertial flow, the ensemble method is used for two-way filtering, while including near-inertial oscillations, it is possible to avoid the influence of other factors such as the full-day tidal current. In addition, the extraction results are modally decomposed and reconstructed in combination with variational mode decomposition VMD to obtain a clearer near-inertial flow signal. By calculating the mean and standard deviation of the difference between the observation set O and the output set M and the assimilation set A, the simulation accuracy of the assimilation model is quantitatively evaluated to enhance the credibility of the assimilation data. Using pre-trained sequence graph neural networks as the source domain for transfer learning avoids the problem of insufficient model training due to limited data and improves prediction efficiency. A diffusion model is established based on Pollard's mixed layer model to simulate the propagation process of near-inertial energy and improve the prediction effect of the model.

[0300] Another aspect of the embodiment of the present invention further provides a prediction system for ocean near-inertial current, comprising:

[0301] The first module is used to obtain multi-source observation data;

[0302] The second module is used to perform data preprocessing on the multi-source observation data to obtain target data for the prediction process;

[0303] The third module is used to construct the first wind field predicted by the meteorological model and the second wind field during the tropical cyclone period constructed by fusion;

[0304] A fourth module is used to construct a three-dimensional FVCOM model based on the first wind field and the second wind field as model drivers to generate flow field simulation results;

[0305] The fifth module is used to perform data assimilation processing on the flow field simulation result generated by the three-dimensional FVCOM model and the target data to obtain assimilated ocean current data;

[0306] The sixth module is used to extract the near-inertial current according to the unassimilated ocean current data and the assimilated ocean current data, and to extract the detailed signal characteristics of the near-inertial current;

[0307] The seventh module is used to extract the near-inertial wind stress according to the detailed signal characteristics;

[0308] The eighth module is used to obtain a pre-trained sequence graph neural network TGCN model, analyze the correlation between the near-inertial wind stress and the near-inertial flow, construct a prediction system for the near-inertial flow, and complete the prediction of the ocean near-inertial flow.

[0309] It can be understood that the contents of the above method embodiments are all applicable to the present system embodiments, the functions specifically implemented by the present system embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0310] The embodiment of the present application also provides an electronic device, the electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the above-mentioned ocean near-inertial current prediction method when executing the computer program. The electronic device can be any intelligent terminal including a tablet computer, a car computer, etc.

[0311] It can be understood that the contents of the above method embodiments are all applicable to the present device embodiments, the functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0312] An embodiment of the present application further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned method for predicting ocean near-inertial current is implemented.

[0313] It can be understood that the contents of the above method embodiments are all applicable to the present storage medium embodiments, the functions specifically implemented by the present storage medium embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0314] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0315] In some selectable embodiments, the function / operation mentioned in the block diagram may not occur in the order mentioned in the operation diagram. For example, depending on the function / operation involved, the two boxes shown in succession can actually be executed substantially simultaneously or the boxes can sometimes be executed in reverse order. In addition, the embodiment presented and described in the flow chart of the present invention is provided by way of example, for the purpose of providing a more comprehensive understanding of technology. The disclosed method is not limited to the operation and logic flow presented herein. Selectable embodiments are expected, wherein the order of various operations is changed and the sub-operation of a part for which is described as a larger operation is performed independently.

[0316] In addition, although the present invention is described in the context of functional modules, it should be understood that, unless otherwise specified, one or more of the functions and / or features described may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in separate physical devices or software modules. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the present invention. More specifically, in view of the properties, functions, and internal relationships of the various functional modules in the device disclosed herein, the actual implementation of the module will be understood within the conventional skills of the engineer. Therefore, those skilled in the art can implement the present invention set forth in the claims without excessive experimentation using ordinary techniques. It is also understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the present invention, which is determined by the full scope of the appended claims and their equivalents.

[0317] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0318] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0319] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0320] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0321] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0322] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.

[0323] The above is a specific description of the preferred implementation of the present invention, but the present invention is not limited to the described embodiments. Those skilled in the art may make various equivalent modifications or substitutions without violating the spirit of the present invention. These equivalent modifications or substitutions are all included in the scope defined by the claims of this application.

Claims

1. A method for predicting near-inertial current in the ocean, characterized in that: include: Acquire multi-source observation data; Performing data preprocessing on the multi-source observation data to obtain target data for a prediction process; Construct the first wind field predicted by the meteorological model and the second wind field during the tropical cyclone period by fusion; Based on the first wind field and the second wind field as model drivers, a three-dimensional FVCOM model is constructed to generate flow field simulation results; Performing data assimilation processing on the flow field simulation result generated by the three-dimensional FVCOM model and the target data to obtain assimilated ocean current data; Extracting near-inertial current according to the unassimilated ocean current data and the assimilated ocean current data, and extracting detailed signal features of the near-inertial current; According to the detailed signal characteristics, the near-inertial wind stress is extracted; Obtaining a pre-trained sequence graph neural network TGCN model, analyzing the correlation between the near-inertial wind stress and the near-inertial flow, constructing a prediction system for the near-inertial flow, and completing the prediction of the near-inertial flow of the ocean; The multi-source observation data includes ocean current data, water depth and topography data, wind field data, typhoon path data and coastline data. The multi-source observation data is preprocessed to obtain target data for the prediction process, including: Using the DBSCAN clustering algorithm to remove outliers in the multi-source observation data; Using the spatiotemporal Kriging interpolation method to fill missing values ​​in the multi-source observation data; The inverted flow field is smoothed using the median filter method; The step of constructing the first wind field predicted by the meteorological model comprises: The ECMWF global forecast model was used to obtain the predicted wind field; Obtain wind farm forecast data from ECMWF’s data service platform; By using a data processing tool, wind field data within a research period, research area, and research height range are selected and extracted from the predicted wind field and wind field forecast data to construct the first wind field; The second wind field during the tropical cyclone period constructed by fusion includes: Get CCMP reanalysis data from NOAA's NCEI Data Center; Select and extract reanalysis wind field data for a study period and a study area from the CCMP reanalysis data according to a data processing tool; Determine the maximum wind speed and maximum radius of the typhoon based on the acquired typhoon path data, and write the calculation function of the Holland model; The CCMP wind field data is used as the background wind field, and the typhoon wind field result calculated by the Holland model is used to correct and superimpose the background wind field to obtain the tropical cyclone period wind field as the second wind field; the three-dimensional FVCOM model is constructed based on the first wind field and the second wind field as the model drive to generate the flow field simulation result, including: The mesh files were generated by mesh generation software to describe the topography and water structure of the study area using triangular meshes; Interpolating the water depth data to the grid points of the grid file by a bilinear interpolation method; Configuring three-dimensional FVCOM model parameters, wherein the three-dimensional FVCOM model parameters include seawater density, horizontal mixing coefficient, vertical mixing coefficient, and seawater viscosity coefficient; Configuring boundary conditions of the three-dimensional FVCOM model, wherein the boundary conditions include sea surface pressure, wind speed, and land boundary conditions; Using the first wind farm and the second wind farm as model drivers, configuring grid files, boundary conditions and driving data files according to model requirements; Starting the operation of the three-dimensional FVCOM model, monitoring the output and operation status of the three-dimensional FVCOM model, and obtaining the flow field simulation results output by the three-dimensional FVCOM model, wherein the flow field simulation results include ocean current velocity and sea surface wind stress; The process of performing data assimilation processing on the flow field simulation result generated by the three-dimensional FVCOM model and the target data to obtain assimilated ocean current data includes: A set of particles is randomly generated, each particle represents a possible state of the system; Use the model output to make state predictions for each particle; Perform local perturbations on each prior particle to obtain the proposed particle; Calculating proposal weights and normalizing the proposal weights; Re-extract particles according to the standardized proposal weights and retain particles with high fit; Using the resampled particle collection, an estimate of the state is calculated using a weighted average; Quantitatively evaluate the simulation accuracy of the assimilation system; The extracting of the near-inertial current according to the unassimilated ocean current data and the assimilated ocean current data, and extracting the detailed signal features of the near-inertial current, includes: Define the initial frequency set and telescopic length according to the near-inertial flow frequency range; According to the initial frequency set and the telescopic length, a fourth-order Butterworth bandpass filter is configured; Performing forward filtering and reverse filtering on the unassimilated ocean current data and the assimilated ocean current data through the fourth-order Butterworth bandpass filter to obtain a two-way filtering result, and then obtaining different sets of near-inertial current extraction results; The variational mode decomposition is used to perform mode decomposition and reconstruction on the near-inertial flow extraction results to obtain the near-inertial flow signal; The method of performing modal decomposition and reconstruction on the near-inertial flow extraction result by using variational modal decomposition to obtain the near-inertial flow signal includes: Inputting the near-inertial flow extraction result into the VMD algorithm for decomposition; Determine the number of modes to be decomposed, adjust the regularization parameters, and control the smoothness and sparsity between modes during the decomposition process; VMD decomposition is performed using an iterative optimization algorithm using the alternating direction multiplier method; Set the initial mode function and Lagrange multiplier parameters; After the iterative optimization is completed, each mode after decomposition is obtained; The energy is calculated for each mode using the square of the modal amplitude as an estimate of the energy of each mode; Normalize the energy of all modes; The normalized energy is used as a reference for weights, and a weight is assigned to each mode. Each modal component is weighted and combined according to its corresponding weight to obtain the final reconstructed signal; By drawing a comparison diagram between the original signal and the reconstructed signal, the reconstruction error is calculated to evaluate the reconstruction effect; the pre-trained sequence graph neural network TGCN model is obtained, the correlation between the near-inertial wind stress and the near-inertial flow is analyzed, and a prediction system for the near-inertial flow is constructed to complete the prediction of the near-inertial flow of the ocean, including: The extracted near-inertial flow and near-inertial wind stress are converted from the frequency domain to the time domain through inverse Fourier transform; Based on tropical cyclone data, the extracted near-inertial flow and near-inertial wind stress data are divided into three parts: before, during, and after the tropical cyclone. Obtain a pre-trained TGCN model from the open source library whose model structure and parameters are similar to the application scenario of near-inertial flow prediction; According to the selected pre-trained model structure, freeze some layers of the pre-trained model to retain the features and representations learned by the model in the source domain; Establishing a diffusion model based on the control equation to simulate the propagation process of near-inertial energy, and adding the diffusion model as a custom layer to the pre-trained model; Input existing data to debug the pre-trained model and determine the optimal values ​​of the parameters; Add the calculation expression of near inertial kinetic energy into the fully connected layer to obtain the distribution prediction of near inertial kinetic energy; Wherein, the calculation expression of the near inertial kinetic energy is: NIKE represents near inertial kinetic energy; ρ is the density of seawater; u represents the east-west velocity component; and v represents the north-south velocity component.

2. A system for implementing the method for predicting ocean near-inertial current according to claim 1, characterized in that: include: The first module is used to obtain multi-source observation data; The second module is used to perform data preprocessing on the multi-source observation data to obtain target data for the prediction process; The third module is used to construct the first wind field predicted by the meteorological model and the second wind field during the tropical cyclone period constructed by fusion; A fourth module is used to construct a three-dimensional FVCOM model based on the first wind field and the second wind field as model drivers to generate flow field simulation results; The fifth module is used to perform data assimilation processing on the flow field simulation result generated by the three-dimensional FVCOM model and the target data to obtain assimilated ocean current data; The sixth module is used to extract the near-inertial current according to the unassimilated ocean current data and the assimilated ocean current data, and to extract the detailed signal characteristics of the near-inertial current; The seventh module is used to extract the near-inertial wind stress according to the detailed signal characteristics; The eighth module is used to obtain a pre-trained sequence graph neural network TGCN model, analyze the correlation between the near-inertial wind stress and the near-inertial flow, construct a prediction system for the near-inertial flow, and complete the prediction of the ocean near-inertial flow.

3. An electronic device, characterized in that: including a processor and a memory; The memory is used to store programs; The processor executes the program to implement the method according to claim 1.

Images