Wind and wave response modeling and ocean current inversion method based on machine learning

Abstract

A wind and wave response modeling and ocean current inversion method based on machine learning relates to the technical field of artificial intelligence and ocean information services, and comprises the following steps: (1) data acquisition; (2) data processing; (3) constructing a data set; (4) establishing a sea surface flow velocity inversion model by using a particle swarm optimization machine learning algorithm; and (5) obtaining an optimal sea surface flow velocity inversion model through testing and case analysis. Aiming at ocean and Doppler frequency shift characteristics of the sea surface, the invention provides a particle swarm optimization XGBoost model wind wave Doppler contribution prediction method for the first time, and the wind wave Doppler contribution estimation precision and speed can be effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence and marine information service technology, and in particular to a method for wind and wave response modeling and ocean current inversion based on machine learning. Background Art

[0002] Ocean currents are a core element in the ocean dynamic system. By transporting heat and salt on a global scale, they play a decisive role in maintaining the Earth's energy balance and climate stability. Accurate ocean current observations are not only important for a deeper understanding of the coupling process between the ocean and the atmosphere, but can also significantly improve the accuracy of offshore oil spill spread predictions, optimize shipping routes, and provide key support for fishery resource management and maritime search and rescue operations. Ocean current measurements can be obtained through buoys, shore-based or ship-borne radars, but these observation methods have high maintenance costs and limited spatial coverage, making it difficult to achieve continuous monitoring on a global scale. Satellite remote sensing, due to its large-scale and repeated observation capabilities, has become an important means of obtaining global sea surface current field information. Spaceborne Synthetic Aperture Radar (SAR) is currently the only satellite remote sensing method that can directly measure sea surface currents.

[0003] There are generally two methods for inverting ocean currents using spaceborne SAR: Along Track Interferometry (ATI) and Doppler Centroid Anomaly (DCA). The ATI method primarily uses two antennas spaced closely along the satellite's orbit to acquire two SAR images of the same area, and then inverts the sea surface velocity in the radar's direction of view based on the phase difference between the two images. However, this data is difficult to obtain. The DCA method inverts ocean currents by correcting the Doppler contribution of the SAR Doppler shift that is unrelated to the ocean current, thereby obtaining the Doppler shift of the ocean current and ultimately inverting the ocean current. Therefore, accurately removing the Doppler shift not caused by the ocean current is the key to accurately inverting ocean currents using the DCA method.

[0004] The Doppler effect refers to the phenomenon that when there is relative motion between the wave source and the observer, the frequency of the wave received by the observer differs from the frequency actually emitted by the wave source. Spaceborne SAR can transmit electromagnetic waves of a specific frequency to the sea surface and receive its echo signal. By comparing the phase difference between the transmitted and received electromagnetic waves, the Doppler shift caused by the relative motion between the sea surface and the satellite can be calculated. However, this frequency shift does not only include the contribution of the sea surface motion. In the process of satellite observation of the sea surface, the satellite's own motion and the rotation of the earth will also affect the observed Doppler shift. Considering that the Doppler shift of the OCN data product is obtained by processing the SLC data, the Doppler shift it provides Usually includes the following contributions: (01); (02); (03); in, and They are geophysical contribution and non-geophysical contribution, which together constitute the total Doppler shift of SAR observations. ; is the Doppler shift caused by the motion of the sea surface, It is the Doppler shift caused by changes in satellite attitude, orbit and antenna orientation, as well as the relative motion between the satellite and the Earth; Mainly by and Composition, of which is the Doppler shift caused by the sea surface current, is the Doppler shift caused by wind and waves; Depend on 、 、 and Composition, of which It is the Doppler frequency shift caused by the change of satellite orbit altitude and satellite attitude. The Doppler frequency shift is caused by the mis-pointing of the electromagnetic waves emitted by the SAR antenna away from the center of the sea surface element. The Doppler shift is caused by the sector error caused by the movement of the SAR satellite TOPSAR scanning mode antenna. The Doppler shift is caused by other unknown deviations.

[0005] Using SAR Doppler shift to invert ocean currents requires the total Doppler shift observed by SAR to be Correction is performed to remove the Doppler contribution of ocean currents Unrelated Doppler contribution; in non-geophysical terms middle, It has been calculated and stored in the product during SAR satellite observation; It is periodic noise and can be removed by frequency domain filtering; and The land cover data from the same observation track can be used for estimation; It is necessary to establish a wind-wave Doppler contribution estimation model to eliminate it; although the CDOP series models based on BPNN perform well in estimating wind-wave Doppler frequency shift, BPNN relies on gradient descent iterative optimization, has high computational complexity, requires fine-tuning the number of layers, number of neurons and learning rate, is prone to falling into local optimality, and cannot automatically handle missing values; while decision tree-based integrated models, such as XGBoost, support parallel computing, and the training speed is usually 5-10 times faster than BPNN. Its parameters (such as tree depth and learning rate) are more robust, and it has built-in missing value processing and feature binning; therefore, it is necessary to explore the application of invented and efficient machine learning models such as XGBoost in modeling wind-wave Doppler frequency shift contribution. Summary of the Invention

[0006] The present invention provides a method for wind and wave response modeling and ocean current inversion based on machine learning, with the aim of providing a new method that can improve the accuracy and speed of wind and wave Doppler contribution estimation to address the problems existing in the existing technology.

[0007] To achieve the above object, the technical solution of the present invention is: A method for wind and wave response modeling and ocean current inversion based on machine learning includes the following steps: (1) data acquisition; (2) data processing; (3) data set construction; (4) establishing an ocean surface current inversion model using a machine learning algorithm based on particle swarm optimization; and (5) obtaining the optimal ocean surface current inversion model through testing and case analysis.

[0008] Preferably, the step (1) comprises: observing SAR Doppler frequency shift data through the Sentinel-1 satellite; obtaining ocean wave data through WAVERYS; obtaining ocean current data through HYCOM, and obtaining ocean current data through HFRNet.

[0009] Preferably, the step (2) includes: (21) non-geophysical Doppler shift correction; (22) performing time-space matching on the data.

[0010] Preferably, the step (21) includes: satellite attitude Doppler contribution correction; sector error Doppler contribution correction; satellite antenna mispointing Doppler contribution and residual Doppler contribution correction; the step (22) includes: performing spatiotemporal matching of the data corrected in step (21) with wave data, ocean current data and ECMWF wind field data, wherein the wind field data refers to wind direction and wind speed.

[0011] Preferably, in the step (3), the time-space matching data is subjected to coordinate transformation and ocean current Doppler conversion to construct a sea surface multi-parameter data set.

[0012] Preferably, in the step (4), the SAR incident angle, the SAR radar ground-to-ground wind speed and direction, and the ground-to-ground wave speed and direction are used as input parameters to establish a relationship with the wind-wave Doppler shift.

[0013] Preferably, in step (4), the incident angle is selected , wind speed at ground level ,wind direction , orbital velocity of the ground-to-ground wave , wave direction As the model input parameter, the wind-wave induced Doppler shift is expressed as: (1); The ECMWF wind direction is converted to the angle relative to the SAR ground distance direction, which is expressed as: (2); Where, represents the SAR satellite orbit azimuth, Indicates the wind direction relative to true north; after transformation, 0° and 180° represent headwind (i.e., wind blowing toward the radar) and tailwind (i.e., wind blowing away from the radar), and 90° and 270° represent azimuth wind (i.e., wind along the satellite track); the SAR ground range wind speed is expressed as: (3); Where u represents the ECMWF wind speed in the matching dataset; the average wave direction is transformed to the direction relative to the SAR ground off-direction, which is expressed as: (4); Where, Indicates the average wave direction relative to due north; the wave orbital velocity is calculated using WAVERYS significant wave height and average wave period: (5); Where, represents the effective wave height, represents the average angular frequency, represents the average wave period; the wave orbit velocity in the SAR range direction is expressed as: (6); Then, formula (1) is modeled based on two machine learning algorithms: BPNN and XGBoost.

[0014] Preferably, in the step (4), the key parameters in the BPNN and XGBoost models are optimized by using a particle swarm optimization (PSO): for the BPNN model, the optimized parameters include the learning rate, the number of iterations, and the number of batch samples; for the XGBoost model, the optimized parameters include the number of base models, the learning rate, the maximum tree depth, and the minimum leaf weight, and the optimized BPNN model and the optimized XGBoost model are obtained respectively through optimization.

[0015] Preferably, in step (5), the optimized XGBoost model obtained through testing and case analysis is the optimal sea surface current velocity inversion model.

[0016] A device for wind and wave response modeling and ocean current inversion based on machine learning includes a memory and a processor; the memory is used to store a computer program; the processor is used to implement a method for wind and wave response modeling and ocean current inversion based on machine learning when executing the computer program.

[0017] A computer-readable storage medium stores a computer program. When the computer program is executed by a processor, a method for wind and wave response modeling and ocean current inversion based on machine learning is implemented.

[0018] The present invention provides a method for wind and wave response modeling and ocean current inversion based on machine learning, which has the following beneficial effects: Existing BPNN-based CDOP models often require a large amount of data and are time-consuming due to their inherent characteristics. They also have slow convergence and poor ocean current inversion performance, especially in the core areas of ocean currents. This invention uses satellite remote sensing SAR Doppler shift and oceanographic data to construct a new ocean current inversion algorithm based on machine learning. The beneficial effects are as follows: (1) Aiming at the oceanographic and Doppler shift characteristics of the sea surface, this paper proposes for the first time a particle swarm optimized XGBoost model wind-wave Doppler contribution prediction method. The Doppler data of the Sentinel-1 satellite global observation contains multiple Doppler contributions, and accurately estimating the wind-wave Doppler contribution is the key to ocean current inversion. The existing BPNN-based CDOP model has low prediction accuracy and slow convergence speed. How to estimate the wind-wave Doppler contribution more accurately and quickly is a difficult problem in ocean current inversion. This paper proposes a new method that can improve the accuracy and speed of wind-wave Doppler contribution estimation. (2) The new method proposed in this invention will provide new ideas and scientific understanding for SAR ocean current inversion and its assessment of ocean primary productivity. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is the distribution map of Sentinel-1 SAR data in the study area of ​​the algorithm of this invention.

[0020] Figure 2 The changing trend of loss value with the number of iterations during the training process of the two models.

[0021] Figure 3 Middle: (a) Scatter plot of wind wave Doppler contribution estimation results for BPNN optimized by particle swarm optimization; (b) Scatter plot of wind wave Doppler contribution estimation results for XGBoost optimized by particle swarm optimization.

[0022] Figure 4 The scatter plots are compared between the wind wave Doppler shift predicted by BPNN and the SAR observations in different wind speed ranges.

[0023] Figure 5 Scatter plot comparing the wind wave Doppler shift predicted by XGBoost and the SAR observations in different wind speed ranges.

[0024] Figure 6 Middle: (a) is a scatter plot comparing the ocean current velocity inverted by BPNN and the ocean current velocity in HYCOM on a case basis; (b) is a scatter plot comparing the ocean current velocity inverted by XGBoost and the ocean current velocity in a case basis.

[0025] Figure 7 This is the HYCOM ocean current characteristics map.

[0026] Figure 8 This is the BPNN inversion of ocean current characteristics.

[0027] Figure 9 This is the XGBoost inversion ocean current feature map.

[0028] Figure 10 A scatter plot comparing the inverted ocean current velocity and the HF ocean current velocity (BPNN on the left and XGBoost on the right).

[0029] Figure 11 This is the technical roadmap for the method of wind and wave response modeling and ocean current inversion based on machine learning proposed in this invention. DETAILED DESCRIPTION

[0030] The following describes in detail the implementation methods of the present invention in a step-by-step manner. This description is only a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

[0031] In the description of the present invention, it should be noted that the terms "up", "down", "left", "right", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the accompanying drawings. They are only for describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, and a specific orientation structure and operation. Therefore, they cannot be understood as limiting the present invention.

[0032] A method for wind and wave response modeling and ocean current inversion based on machine learning: First, the ocean current is inverted using SAR Doppler shift. The total Doppler shift observed by SAR needs to be Correction is performed to remove the Doppler contribution of ocean currents Unrelated Doppler contribution; in non-geophysical terms middle, It has been calculated and stored in the product during SAR satellite observation; It is periodic noise and can be removed by frequency domain filtering; and Land cover data from the same observation track can be used for estimation; based on land Assuming that the average Doppler shift on is 0 Hz, the non-geophysical term is approximated as: (04); From the total Doppler shift Remove non-geophysical components The steps are as follows: (1) Select an observation point with 100% land coverage. At this time, the total Doppler shift of the SAR observation is ; (2) Using the Doppler shift of the land observation area minus get ; (3) Use frequency domain filtering to remove periodic sector errors (4) The residual Doppler shift in the land area is linearly fitted using the least squares method and modeled as a function of the incident angle; (5) The fitted model is used to calibrate the .

[0033] The Sentinel-1A OCN (Doppler shift, ECMWF wind field), WAVERYS wave data (significant wave height, mean wave direction and mean wave period) and HYCOM current data (surface meridional velocity and surface zonal velocity) are temporally and spatially matched. First, the ECMWF wind field data (wind speed and direction) provided by OCN are spatially interpolated to the latitude and longitude grid corresponding to the Doppler shift data. Then, the OCN Doppler shift and wind field data are temporally and spatially matched with the WAVERYS wave data and HYCOM current data. The specific operation is as follows: using OCN time (UTC) as the reference time, WAVERYS is used to calculate the wind speed and direction of the ECMWF wind field data. The time units of the RYS wave data and the HYCOM current data were unified into OCN time units (hours accumulated from 00:00:00 on January 1, 2000). WAVERYS and HYCOM data within a ±30-minute time window were selected for temporal matching. Bilinear interpolation was used for spatial matching, and all data were interpolated to a Doppler shift grid. In addition, the matching datasets were spatiotemporally matched with the high-frequency radar observation data. Based on HF time and longitude and latitude, a ±30-minute time window and bilinear interpolation were selected. A total of 3404 pairs of samples were successfully matched.

[0034] There is a complex nonlinear coupling relationship between Doppler shift and sea surface wind and wave fields. Among existing artificial intelligence methods, deep learning models represented by Back Propagation Neural Networks (BPNN) and ensemble learning models represented by eXtreme Gradient Boosting (XGBoost) have been widely used due to their excellent performance. This paper selects these two representative intelligent algorithms to establish a relationship model between wind-wave-induced Doppler shift and sea surface wind and wave parameters. By comparing and evaluating their applicability, the paper determines the optimal algorithm solution.

[0035] Select the angle of incidence , wind speed at ground level ,wind direction , orbital velocity of the ground-to-ground wave , wave direction As the model input parameter, the wind-wave induced Doppler shift is expressed as: (1); The ECMWF wind direction is converted to the angle relative to the SAR ground distance direction, which is expressed as: (2); Where, represents the SAR satellite orbit azimuth, Indicates the wind direction relative to true north; after transformation, 0° and 180° represent headwind (i.e., wind blowing toward the radar) and tailwind (i.e., wind blowing away from the radar), and 90° and 270° represent azimuth wind (i.e., wind along the satellite track); the SAR ground range wind speed is expressed as: (3); Where u represents the ECMWF wind speed in the matching dataset; the average wave direction is transformed to the direction relative to the SAR ground off-direction, which is expressed as: (4); Where, Indicates the average wave direction relative to due north; the wave orbital velocity is calculated using WAVERYS significant wave height and average wave period: (5); Where, represents the effective wave height, represents the average angular frequency, represents the average wave period; the wave orbit velocity in the SAR range direction is expressed as: (6); Formula (1) is modeled based on two machine learning algorithms: BPNN and XGBoost.

[0036] (1) BPNN: BPNN adopts the classic four-layer fully connected neural network architecture, which consists of an input layer, two hidden layers and an output layer. The specific configuration of the network topology is as follows: the input layer contains 5 neurons, corresponding to input feature dimensions such as incident angle, wind speed, wind direction, wave speed and wave direction; the first hidden layer has 16 neurons, which are responsible for primary feature extraction and nonlinear transformation; the second hidden layer has 8 neurons for high-level feature abstraction and dimensionality compression; the output layer uses a single neuron to achieve regression prediction; Adaptive Moment Estimation (Adam) is selected as the weight optimizer. Adam combines the advantages of momentum method and RMSprop and can effectively adjust the learning rate; the loss function uses root mean square error (RMSE) to quantify the deviation between the predicted value and the true value; each hidden layer uses rectified linear unit (ReLU) as the activation function. This nonlinear activation mechanism can not only alleviate the gradient vanishing problem, but also enhance the model's ability to represent complex nonlinear relationships; BPNN calculates the predicted output through forward propagation and then iterates based on the error backpropagation algorithm. Its mathematical model is expressed as follows: (7); Where, is the input parameter vector, w is the weight matrix of each layer, and b is the bias vector of each layer; input parameter vector Defined as: (8).

[0037] (2) XGBoost: XGBoost is an ensemble learning algorithm based on the gradient boosting framework. Its core idea is to gradually optimize the model's predictive performance by iteratively adding decision trees. Compared with traditional gradient boosting methods, XGBoost introduces a second-order Taylor expansion and regularization term, which significantly improves the model's accuracy and generalization ability. This second-order approximation not only more accurately approximates various complex loss functions (including custom loss functions), but also provides more reliable curvature information for model optimization. From the perspective of model architecture, XGBoost adopts an additive training strategy, consisting of n base models (decision trees) connected in series, where the optimization process of the tree model in the tth iteration directly depends on the accumulated results of the previous t-1 iterations. Assume that The tree model to be trained in the iteration is , Indicates the training samples, then the model’s The objective function is expressed as: (9); Where, is the sample size, Indicates the The loss function term of the round, is the regularization term of the model, as follows: (10); Among them, γ and λ are custom parameters; the second-order Taylor expansion of equation (9) yields: (11); Among them, g is the first-order derivative and h is the second-order derivative; as follows: (12); (13); Substituting equations (10), (12), and (13) into equation (11) and taking the derivative, we obtain the following solution: (14); (15); in, Represents the value of the loss function. The smaller the value, the better the tree structure. is the weight.

[0038] The performance of BPNN and XGBoost models depends largely on the selection of hyperparameters. Although these models have preset default parameter values ​​that are suitable for most problems, these parameters may not be suitable for all datasets and application scenarios. Therefore, adjusting hyperparameters for specific datasets becomes a key step in optimizing model performance. In this paper, the particle swarm optimization (PSO) algorithm is used to optimize the key parameters in the BPNN and XGBoost models. Specifically, for the BPNN model, the optimized parameters include the learning rate, the number of iterations, and the number of batch samples. For the XGBoost model, the optimized parameters include the number of base models, the learning rate, the maximum tree depth, and the minimum leaf weight.

[0039] First, the dimension of the search space and the speed of the particles are determined according to the characteristics of the parameters to be optimized. The position of each particle in the search space is a multidimensional vector. Each dimension corresponds to a different BPNN or XGBoost parameter, so the initialization range of each dimension is different. Taking XGBoost as an example, The particle in The position vector of the iteration is expressed as: (16); Since all particles move in the same search space, when When , the particle velocity in each dimension is randomly initialized to a value in the range of (0,1); The particle in The velocity vector of the iteration is expressed as: (17); Then assign the position vector to the corresponding parameters of the model, and use the performance on the training set as the fitness. This invention uses the mean square error (RMSE) of the model prediction as the optimization target, so the fitness is set to the mean square error of the model. The particle in The fitness of the iteration is: (18); For a single particle, The historical optimal position in the iteration, that is, the position with the minimum fitness, is expressed as: (19); All particles in front The historical optimal position of the iteration, that is, the global optimal position, is expressed as: (20).

[0040] The present invention uses 80% of the data as a training set to train the model, and the remaining 20% ​​as a test set to evaluate the performance of the model; the BPNN and XGBoost models are optimized on the training set, and the final results are that the learning rate of BPNN is 0.01, the number of batches is 512, and the number of iterations is 100; the number of base models of XGBoost is 152, the learning rate is 0.04, the maximum tree depth is 7, and the minimum leaf weight is 10; Figure 2 The training process of the two models is given. It can be seen that the mean square error of BPNN drops rapidly in the early stage of training, and basically converges at the 80th iteration, with the mean square error converging to around 7.0; the mean square error of XGBoost drops faster in the early stage, and converges to around 4.0 at the 40th iteration. It can be seen that XGBoost has stronger fitting ability than BPNN.

[0041] BPNN and XGBoost models were evaluated on the test set to estimate wind-wave induced Doppler shift. performance; such as Figure 3 As shown in the figure, the prediction results of the BPNN model are relatively scattered, with a root mean square error (RMSE) of approximately 6.941 Hz and a correlation coefficient of approximately 0.903. In comparison, the XGBoost model shows better prediction performance, with its prediction points densely distributed near the 1:1 reference line, the RMSE reduced to 4.043 Hz, and the correlation coefficient increased to 0.971. Compared with the BPNN model, the XGBoost model's mean square error is reduced by 2.898 Hz, showing higher prediction accuracy.

[0042] The prediction performance differences between BPNN and XGBoost models in different wind speed ranges were further analyzed; Figure 4 、 Figure 5 As shown in the figure, the RMSE of the BPNN model is 9.224 Hz in the low wind speed range (0-5 m / s), drops to 6.767 Hz in the medium-low wind speed range (5-10 m / s), and further decreases to 5.848 Hz in the medium-high wind speed range (greater than 10 m / s). In comparison, the XGBoost model shows better prediction stability in all wind speed ranges. In the same wind speed range, its RMSE is 4.835 Hz (0-5 m / s), 3.771 Hz (5-10 m / s), and 3.494 Hz (greater than 10 m / s), respectively.

[0043] In order to further verify the regional effectiveness of the new algorithm proposed in this paper, the flow velocity obtained by the new algorithm based on the Sentinel-1A SAR Doppler shift observation data was compared with the HYCOM flow velocity; Figure 6A scatter plot comparing the ocean current velocities inverted by the BPNN and XGBoost models with the HYCOM ocean current velocities in the validation area is presented. The plot shows that the XGBoost model significantly outperforms the BPNN model. Specifically, the scatter plots of the XGBoost model's inversion results are more closely distributed near the 1:1 baseline, with an RMSE of 0.252 m / s and a MAE of 0.202 m / s, which are 25% and 26% lower than those of the BPNN model (RMSE of 0.336 m / s and MAE of 0.271 m / s), respectively. The mean error (ME) is reduced from 0.127 m / s for the BPNN model to 0.027 m / s for the XGBoost model (an improvement of 79%), and the correlation coefficient R is also improved from 0.71 to 0.787 (an improvement of 11%).

[0044] Figure 7-Figure 9 The spatial distribution of the HYCOM current velocity, the BPNN model inversion velocity, and the XGBoost model inversion velocity in the validation area are shown. Comparative analysis shows that the spatial distribution of the BPNN and XGBoost model inversion velocities is generally consistent with the spatial characteristics of the HYCOM current velocity; however, the BPNN model inversion velocity is significantly overestimated in some areas (see Figure 8 The area marked by the box); In comparison, the spatial characteristics of the velocity inverted by the XGBoost model better overcome the overestimation problem of the BPNN model and show a higher degree of consistency with the HYCOM current in terms of ocean current details.

[0045] Figure 10 The model is further validated using a dataset matching the on-site high-frequency radar in the Gulf Stream region. The comparison results of BPNN and XGBoost on HF data are shown in Figure 2. Figure 10 As shown; the left figure shows the BPNN inversion results, whose root mean square error, average relative error and deviation are 0.252 m / s, 0.182 m / s and -0.147 m / s; the right figure shows the comparison of the XGBoost ocean current inversion results and high-frequency radar data. At this time, the root mean square error, average relative error and deviation are reduced to 0.213 m / s, 0.167 m / s and -0.073 m / s respectively; the results show that the XGBoost ocean current inversion accuracy is better than BPNN, and it can achieve high-precision inversion of ocean currents.

[0046] Those skilled in the art should understand that the embodiments of the present invention may be provided as methods, apparatuses or computer program products. Therefore, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0047] like Figure 11 As shown, the present invention is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products of the embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as the combination of the processes and / or blocks in the flowcharts and / or block diagrams, 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 generate 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 flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A system that specifies the functions of a box or boxes.

[0048] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture including an instruction system that is implemented in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0049] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 The present invention is a block or a plurality of blocks of steps for specifying functions; although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they are aware of the basic creative concepts; therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0050] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art will be able to modify the technical solutions described in the aforementioned embodiments or substitute equivalents for some of the technical features. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A method for wind and wave response modeling and ocean current inversion based on machine learning, characterized by: The following steps are involved: (1) Data acquisition; (2) Data processing; (3) Data set construction; (4) Establishment of a sea surface current velocity inversion model using a machine learning algorithm based on particle swarm optimization; (5) Obtaining the optimal sea surface current velocity inversion model through testing and case analysis.

2. The method of wind and wave response modeling and ocean current inversion based on machine learning according to claim 1, characterized by: The step (1) includes: observing SAR Doppler shift data through the Sentinel-1 satellite; obtaining ocean wave data through WAVERYS; obtaining ocean current data through HYCOM; and obtaining ocean current data through HFRNet. The step (2) includes: (21) non-geophysical Doppler shift correction; and (22) performing spatiotemporal matching on the data.

3. The method of wind and wave response modeling and ocean current inversion based on machine learning according to claim 2, characterized by: The step (21) includes: satellite attitude Doppler contribution correction; sector error Doppler contribution correction; satellite antenna mispointing Doppler contribution and residual Doppler contribution correction; the step (22) includes: performing spatiotemporal matching of the data corrected in step (21) with wave data, ocean current data and ECMWF wind field data, wherein the wind field data refers to wind direction and wind speed.

4. The method of wind and wave response modeling and ocean current inversion based on machine learning according to claim 3, characterized by: In the step (3), the time-space matching data is subjected to coordinate conversion and ocean current Doppler conversion to construct a sea surface multi-parameter data set.

5. The method for wind and wave response modeling and ocean current inversion based on machine learning according to claim 4, characterized in that: In the step (4), the SAR incident angle, the SAR radar ground-to-ground wind speed and direction, and the ground-to-ground wave speed and direction are used as input parameters to establish a relationship with the wind-wave Doppler shift.

6. The method of wind and wave response modeling and ocean current inversion based on machine learning according to claim 5, characterized by: In step (4), the incident angle is selected , wind speed at ground level ,wind direction , orbital velocity of the ground-to-ground wave , wave direction As the model input parameter, the wind-wave induced Doppler shift is expressed as: (1)； The ECMWF wind direction is converted to the angle relative to the SAR ground distance direction, which is expressed as: (2)； Where, represents the SAR satellite orbit azimuth, Indicates the wind direction relative to due north; after transformation, 0° and 180° represent the headwind (i.e., the wind blowing toward the radar) and the tailwind (i.e., the wind blowing away from the radar), and 90° and 270° represent the azimuth wind (i.e., the wind along the satellite track); the SAR ground range wind speed is expressed as: (3)； Where u represents the ECMWF wind speed in the matching dataset; the average wave direction is transformed to the direction relative to the SAR ground off-direction, which is expressed as: (4)； Where, Indicates the average wave direction relative to due north; the wave orbital velocity is calculated using WAVERYS significant wave height and average wave period: (5)； Where, represents the effective wave height, represents the average angular frequency, represents the average wave period; the wave orbit velocity in the SAR range direction is expressed as: (6)； Then, formula (1) is modeled based on two machine learning algorithms: BPNN and XGBoost.

7. The method for wind and wave response modeling and ocean current inversion based on machine learning according to claim 6, characterized by: In the step (4), the key parameters in the BPNN and XGBoost models are optimized by using the particle swarm optimization (PSO): for the BPNN model, the optimized parameters include the learning rate, the number of iterations and the number of batch samples; for the XGBoost model, the optimized parameters include the number of base models, the learning rate, the maximum tree depth and the minimum leaf weight, and the optimized BPNN model and the optimized XGBoost model are obtained through optimization.

8. The method for wind and wave response modeling and ocean current inversion based on machine learning according to claim 7, characterized by: In step (5), the optimized XGBoost model obtained through testing and case analysis is the optimal sea surface current velocity inversion model.

9. A device for wind and wave response modeling and ocean current inversion based on machine learning, characterized by: It comprises a memory and a processor; the memory is used to store a computer program; the processor is used to implement a method for wind and wave response modeling and ocean current inversion based on machine learning as described in any one of claims 1 to 8 when executing the computer program.

10. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by the processor, the method of wind and wave response modeling and ocean current inversion based on machine learning as described in any one of claims 1 to 8 is implemented.

Images