Method and system for monitoring data of sea wave model based on numerical simulation

Abstract

The application discloses a sea wave mode prediction data monitoring method and system based on numerical simulation and belongs to the technical field of sea wave monitoring. The method comprises the following steps: collecting sea wave mode prediction data, extracting a two-dimensional spectral density function and spectral moments and wind field parameters; calculating local wave steepness and spectral width parameters, inputting a dynamic mapping model to adaptively calculate a dynamic segmentation frequency; dividing a frequency band to separate wind wave and swell wave energy spectrum based on the dynamic segmentation frequency, and smoothing a transition band to suppress energy leakage; calculating effective wave heights respectively and combining overall effective wave heights to perform residual error checking, if the residual error is out of standard, using a golden section search algorithm to iteratively correct the segmentation frequency, and simultaneously combining bilinear interpolation to perform spatial smoothing constraint. The application overcomes the limitation of fixed empirical segmentation frequency, realizes dynamic adjustment of the segmentation frequency with sea conditions, effectively avoids energy leakage caused by hard cutting of the frequency band, and significantly improves the accuracy and monitoring reliability of wind and swell wave separation and identification under complex sea conditions.

Description

Technical Field

[0001] This invention relates to the field of ocean wave monitoring technology, specifically to a method and system for monitoring ocean wave model forecast data based on numerical simulation. Background Technology

[0002] In real marine environments, ocean waves are typically a mixture of wind waves directly driven by local wind fields and swells propagating from distant locations. Because wind waves and swells differ fundamentally in their generation mechanisms, spectral characteristics, and the mechanical mechanisms by which they act on marine engineering structures, accurate separation is a prerequisite for refined ocean wave analysis and disaster early warning.

[0003] Currently, traditional separation methods mainly rely on fixed empirical frequencies as the dividing line between wind waves and swells. However, the wave energy spectrum structure in real ocean environments undergoes continuous and dramatic dynamic evolution with changes in wind field intensity, wind zone length, and wave propagation distance. Static and rigid empirical dividing frequencies are simply unable to adapt to this time-varying characteristic, easily leading to serious misjudgments of wind wave and swell components, resulting in distorted separation results. More significantly, existing technologies generally employ a simple and crude hard truncation method when dividing the wave energy spectrum into frequency bands, completely ignoring the natural gradual change properties of energy within the transition band. This non-physical truncation operation not only easily causes significant energy leakage but also results in a huge residual between the synthesized wave energy after separation and reconstruction and the original total energy, greatly weakening the physical consistency of wave monitoring data and the reliability of engineering applications. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for monitoring wave model forecast data based on numerical simulation, so as to solve the problems mentioned in the background art.

[0005] To address the aforementioned technical problems, this invention provides a method for monitoring wave model forecast data based on numerical simulation, comprising:

[0006] Collect wave model forecast data for the target sea area, analyze the two-dimensional energy spectral density function contained in the wave model forecast data, and extract the spectral moment parameter and wind field parameter from the two-dimensional energy spectral density function.

[0007] Spectral moment parameters include zero-order moment, first-order moment, and spectral peak frequency; wind field parameters include wind speed.

[0008] Ocean wave model forecast data refers to the latitude and longitude gridded forecast data of the target sea area obtained from the output of the ocean wave numerical model.

[0009] Traverse all latitude and longitude grid points in the target sea area. If all elements of the two-dimensional energy spectral density function of a certain grid point are zero or it is marked as a default mask, then the grid point is determined to be an invalid grid point and no action is taken.

[0010] Spectral moment parameters are statistical physical quantities that reflect the energy distribution of ocean waves. Wind field parameters are dynamic parameters that drive the development of ocean waves.

[0011] Acquire collection time Two-dimensional energy spectral density function was extracted from wave model forecast data of the target sea area. ,in For frequency, For wave direction.

[0012] Calculate the two-dimensional energy spectral density function zeroth moment First moment and spectral peak frequency The formula is:

[0013] ;

[0014] ;

[0015] Among them, the spectral peak frequency Two-dimensional energy spectral density function The frequency value corresponding to the maximum value; It is a frequency infinitesimal element; It is a wave-oriented infinitesimal element.

[0016] To reflect the total energy of the wave spectrum, the zero-order moment of the wave spectrum is calculated by performing a double integral on the two-dimensional energy spectral density function over the entire frequency and wave direction range, thereby obtaining the total energy of the waves in the target sea area.

[0017] It is used to reflect the energy-weighted frequency distribution characteristics of the wave spectrum. By multiplying the two-dimensional energy spectral density function by the frequency and then performing a double integral, the first moment of the wave spectrum is calculated, and characteristic parameters such as the average period of the waves are derived, as well as the spectral width parameter is calculated.

[0018] Extracting wave model forecast data Wind speed at a height of meters ,in The preferred value is 10, which means extracting the wind speed at a height of 10 meters above the sea surface. .

[0019] If the forecast data outputs wind speeds not at a height of 10 meters, then the logarithmic law formula for wind shear is used to convert it to a height of 10 meters, and the zero-order moment is calculated. First moment Spectral peak frequency and wind speed The combination forms the current sea state feature vector.

[0020] Based on the spectral moment parameter and wind field parameter, calculate the local wave steepness parameter and spectral width parameter of the current ocean wave.

[0021] A dynamic mapping model is constructed based on local wave steepness parameters and spectral width parameters. These parameters are then input into the dynamic mapping model to calculate the dynamically segmented frequencies. Specifically, this includes:

[0022] Obtain the zeroth moment from the current sea state feature vector First moment and wind speed Substitute the parameters into the preset formula to calculate the local wave steepness parameters of the current ocean waves. Spectral width parameter The formula is:

[0023] ;

[0024] ;

[0025] In the formula, It is the acceleration due to gravity. It is a second-order spectral moment and satisfies: .

[0026] It is used to reflect the steepness of ocean waves relative to wind speed. By using the ratio of total ocean wave energy to driving wind speed, local wave steepness parameters are calculated to quantitatively assess the current growth status and steepness of wind and waves.

[0027] It is used to reflect the degree of dispersion or concentration of wave energy distribution in the frequency domain. Based on the statistical relationship of the zeroth, first, and second moments of the wave spectrum, the spectral width parameter is calculated to measure the width of the wave spectrum, thereby determining the complexity of the wave components.

[0028] A dynamic mapping model based on radial basis functions is constructed to incorporate local wave steepness parameters. Spectral width parameter As an input node, the dynamic segmentation frequency As an output node.

[0029] Obtain historical wave pattern forecast datasets and extract the set of two-dimensional energy spectral density functions that are accurately verified by measured buoys in historical records, along with their corresponding historical dynamic segmentation frequency labels.

[0030] For each record in the historical record set, calculate the corresponding local wave steepness parameter. Spectral width parameter , constitute input sample pairs .

[0031] The K-means clustering algorithm is used to perform cluster analysis on the input sample pairs. During the clustering process, it is necessary to determine the number of hidden layer nodes, i.e., the number of clusters K. Specifically, the elbow rule is used to determine the value of K.

[0032] Calculate the sum of distances from all samples to their respective cluster centers for different K values, and select the K value corresponding to the inflection point where the slope of the curve becomes significantly gentler as the final number of clusters.

[0033] Cluster center points are selected as the center vectors of the dynamic mapping model. The average distance from each sample within a cluster to the cluster center is calculated as the width vector. .

[0034] Construct a system of linear equations, substitute all input sample pairs into the radial basis function to calculate the hidden layer output matrix, and use the least squares method to solve for the weights of the hidden layer nodes. This minimizes the mean square error between the output value of the dynamic mapping model and the historical dynamic segmentation frequency label, thus completing the parameter calibration of the dynamic mapping model.

[0035] The dynamic segmentation frequency is calculated by calling the center vector, width vector, and weight vector of the pre-calibrated dynamic mapping model. :

[0036] ;

[0037] In the formula, For the first The weights of each hidden layer node, As the center vector, It is a width vector. This is the hidden layer node number.

[0038] The critical frequency used to divide the wind wave and swell frequency bands is calculated by inputting the local wave steepness and spectral width features into the network, calculating the distance between the input features and the center of each hidden layer, and then weighting and summing the results after activation by a Gaussian function. This achieves a nonlinear mapping from sea state features to dynamically segmented frequencies, and adaptively calculates the wind and swell segmentation frequency that best matches the current sea state.

[0039] The calculated dynamic segmentation frequency With spectral peak frequency If a comparison is made, Then Forced assignment .

[0040] Based on the laws of physical oceanography, the main energy region of wind and waves is usually concentrated in the spectral peak frequency. The energy of surge waves is concentrated in the vicinity and at high frequencies, while the energy of surge waves is concentrated at low frequencies. (Constant) It is the lower limit of the energy boundary between swell and wind waves, determined based on empirical relationships in classic wind wave spectra such as the JONSWAP spectrum.

[0041] The forced assignment operation is equivalent to applying a physically-based constraint boundary to prevent the dynamic model from failing under extreme sea conditions and to ensure the dynamic segmentation frequency. It is located within a reasonable frequency band.

[0042] If the dynamic mapping model outputs Less than This means that the dividing point is overly biased towards low frequencies, which will incorrectly classify a large amount of high-frequency energy that belongs to wind waves as swells.

[0043] The two-dimensional energy spectral density function is divided into frequency bands based on the dynamic segmentation frequency, and the wind wave energy spectral density function and the surge energy spectral density function are obtained.

[0044] The effective wave heights of the wind wave energy spectral density function and the swell energy spectral density function are calculated separately, and residual verification is performed by combining the overall effective wave height in the wave model forecast data.

[0045] If the residual exceeds a preset threshold, the dynamic segmentation frequency is iteratively corrected, and the final significant wave heights of wind waves and swells are output. Specifically, this includes:

[0046] Obtain the calculated dynamic segmentation frequency For the two-dimensional energy spectral density function Perform frequency band segmentation:

[0047] frequency Less than or equal to the dynamic segmentation frequency Partially divided into surge energy spectral density functions Frequency Greater than the dynamic segmentation frequency Partially divided into wind and wave energy spectral density functions .

[0048] For two-dimensional energy spectral density function At dynamic segmentation frequency The nearby transition frequency band is smoothed to avoid the Gibbs phenomenon caused by frequency band cutoff.

[0049] Set transition bandwidth The value range is preferably 0.03Hz to 0.08Hz, more preferably 0.05Hz. This width can typically cover 3-5 frequency discrete points to balance smoothness and separation accuracy.

[0050] In the interval Internally, a smoothing weight coefficient is introduced. :

[0051] ;

[0052] Surge energy spectral density function Energy spectral density function of wind and waves The values ​​within the transition frequency band are replaced with smoothed calculation results. The formula is:

[0053] ;

[0054] ;

[0055] Within the transition frequency band near the dynamic segmentation frequency, a weighting coefficient for smooth transition is generated, so that the energy spectrum of wind waves and swells transitions smoothly at the segmentation frequency, avoiding energy leakage and Gibbs oscillation caused by hard band cutoff.

[0056] Outside the transition frequency band, the values ​​of the surge energy spectral density function and the wind wave energy spectral density function remain unchanged according to the original frequency band division rules.

[0057] The energy contained in the surge is calculated based on the separated surge energy spectral density function, and the effective wave height of the surge is calculated using the conversion relationship between effective wave height and spectral energy.

[0058] The energy contained in the wind and wave is calculated based on the separated wind and wave energy spectral density function, and the effective wave height is calculated using the conversion relationship between effective wave height and spectral energy.

[0059] Based on the surge energy spectral density function Energy spectral density function of wind and waves Calculate the significant wave height of the swell separately. Effective wave height The formula is:

[0060] ;

[0061] ;

[0062] Extracting the total significant wave height from wave model forecast data Calculate the significant wave height of wind waves And swell effective wave height The synthetic wave height of The formula is: .

[0063] Based on the physical principle that wind wave and swell energy are independent and superimposed with the sum of squares, the separated wind wave significant wave height and swell significant wave height are combined into a theoretical total significant wave height, and the consistency is verified with the total significant wave height of the original forecast data.

[0064] Calculate the total significant wave height with synthetic wave height The residual index between The formula is:

[0065] ;

[0066] The relative residual between the theoretically synthesized wave height and the total effective wave height of the original model is calculated as a quantitative indicator to assess the accuracy of the current dynamic frequency segmentation.

[0067] If the residual index If the residual value is less than or equal to the preset residual threshold, then output the current effective wave height. And swell effective wave height As the final monitoring result.

[0068] If this value exceeds the preset threshold, it indicates that the segmentation frequency is unreasonable, thus triggering an iterative correction mechanism to adjust the dynamic segmentation frequency. Perform iterative corrections:

[0069] An optimization function is constructed with the dynamic segmentation frequency as the independent variable and the goal of minimizing the squared residual. This provides a mathematical basis for finding the optimal segmentation frequency and ensures that the iterative process converges to the frequency point with the minimum separation residual.

[0070] Establish based on residual index Minimize the objective function And set the energy non-negativity constraint conditions for the wind wave energy spectral density function and the swell energy spectral density function:

[0071] ;

[0072] Using the golden section search algorithm in the interval Internal search objective function The minimum point; the physical basis for setting this interval is:

[0073] Based on the physical characteristics of the wave spectrum, the energy boundary frequency between wind waves and swells is usually located at the peak frequency of the spectrum. The region from near the low-frequency side to slightly above the spectral peak frequency.

[0074] The lower limit is set to Consistent with the aforementioned physical constraints, the upper limit is set to This is because if the segmentation frequency exceeds 1.5 times the peak frequency of the spectrum, most of the wind and wave energy will be missed in the calculation.

[0075] Therefore, the golden section method is used for one-dimensional optimization within the physically feasible finite interval. Each iteration can reduce the search interval by a factor of 0.618, which can quickly approach the global optimal solution with minimal computational cost.

[0076] Set a preset search accuracy threshold and iteratively calculate the interval reduction rate.

[0077] When the search interval length is less than the search precision threshold, the midpoint of the output interval is used as the corrected dynamic segmentation frequency. .

[0078] The corrected dynamic segmentation frequency Recalculate the corrected effective wave height using steps S301 to S302. And swell effective wave height This will be used as the final monitoring result and output.

[0079] The significant wave height of wind waves, significant wave height of swell waves, and the final dynamic segmentation frequency in the monitoring results are spatially matched with the latitude and longitude grid points in the wave model forecast data.

[0080] A three-dimensional monitoring matrix is ​​constructed, in which latitude and longitude grid points are two-dimensional planar coordinates, and wind wave significant wave height, swell significant wave height and dynamic segmentation frequency are respectively used as attribute values ​​of the third dimension.

[0081] Since wave model forecast data is spatially continuous, the wave state of adjacent grid points is usually gradual. If the dynamic segmentation frequency of each grid point is calculated in isolation, the segmentation frequency of adjacent grid points may experience unreasonable and drastic jumps when there is local numerical noise in the model data.

[0082] Therefore, a bilinear interpolation algorithm is used to perform spatial smoothing on the dynamic segmentation frequency in the three-dimensional monitoring matrix. The weighted average of the dynamic segmentation frequency of each grid point and its four adjacent grid points is calculated, and the weights are determined by the inverse ratio of the distance between grid points.

[0083] Meanwhile, in order to preserve the true spatial variation characteristics of sea state, the preset spatial threshold is preferably 0.02Hz.

[0084] If the difference between the dynamic segmentation frequency of a grid point and the weighted average value after spatial smoothing exceeds a preset spatial threshold, it is determined to be an abnormal jump, and the dynamic segmentation frequency of that grid point is replaced with the weighted average value. Steps S301 to S302 are then re-executed to calculate the significant wave height of wind waves and significant wave height of swells corresponding to that grid point.

[0085] The present invention also provides a wave model forecast data monitoring system based on numerical simulation, including a data acquisition module, a parameter estimation module and a separation and verification module.

[0086] The data acquisition module is used to collect wave model forecast data for the target sea area, analyze the two-dimensional energy spectral density function contained in the wave model forecast data, and extract the spectral moment parameter and wind field parameter from the two-dimensional energy spectral density function.

[0087] Spectral moment parameters include zero-order moment, first-order moment, and spectral peak frequency; wind field parameters include wind speed.

[0088] The parameter estimation module is used to calculate the local wave steepness and spectral width parameters of the current ocean waves based on the spectral moment parameters and wind field parameters.

[0089] A dynamic mapping model is constructed based on the local wave steepness parameter and the spectral width parameter. The dynamic segmentation frequency is calculated by inputting the local wave steepness parameter and the spectral width parameter into the dynamic mapping model.

[0090] The separation and verification module is used to divide the two-dimensional energy spectral density function into frequency bands according to the dynamic segmentation frequency, and separate the wind wave energy spectral density function and the surge energy spectral density function.

[0091] The effective wave heights of the wind wave energy spectral density function and the swell energy spectral density function are calculated separately, and residual verification is performed by combining the overall effective wave height in the wave model forecast data.

[0092] If the residual exceeds the preset threshold, the dynamic segmentation frequency is iteratively corrected, and the final effective wave height of wind waves and effective wave height of swell waves are output.

[0093] Compared with the prior art, the beneficial effects achieved by the present invention are:

[0094] Overcoming the limitations of existing technologies that rely on fixed empirical segmentation frequencies, this method calculates local wave steepness and spectral width characteristics by extracting spectral moment parameters of the energy spectrum and wind field parameters, and constructs a dynamic mapping model to achieve adaptive dynamic calculation of the segmentation frequency as the sea state evolves, ensuring accurate matching of the segmentation threshold under complex and variable wave spectrum structures.

[0095] To address the issues of energy leakage and Gibbs phenomenon that can easily occur with hard band truncation in existing technologies, a transition band is set near the dynamically segmented frequency, and a smoothing weighting coefficient based on a window function is introduced to perform a gradual transition in the band junction area, replacing the original hard truncation method. This effectively ensures a smooth transition of wind and wave energy and can suppress energy leakage caused by band truncation.

[0096] A residual verification mechanism based on the comparison between synthetic wave height and total effective wave height is introduced. When the residual index of the separation result exceeds the preset threshold, an objective function is established and a one-dimensional search algorithm is used to iteratively optimize and correct the segmentation frequency, forming a closed-loop feedback optimization path, which significantly improves the accuracy of wind wave and swell separation and identification under complex sea conditions.

[0097] To address the local computational noise in the numerical model output data, a spatial constraint mechanism based on interpolation algorithms is introduced. By calculating the weighted average of the segmentation frequencies of adjacent grid points, abnormal frequency jumps are identified and corrected, thereby enhancing the spatial coherence of the overall regional monitoring results and improving the reliability of forecast data monitoring. Attached Figure Description

[0098] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0099] Figure 1 This is a flowchart illustrating the wave model forecast data monitoring method based on numerical simulation of the present invention.

[0100] Figure 2 This is a schematic diagram comparing the separation residual index under different sea conditions according to the present invention;

[0101] Figure 3 This is a schematic diagram comparing the effective wave height separation accuracy under different sea conditions according to the present invention;

[0102] Figure 4 This is a schematic diagram comparing the wave energy spectrum separation effect of the present invention. Detailed Implementation

[0103] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0104] Please see Figure 1 This invention provides a method for monitoring wave model forecast data based on numerical simulation, comprising:

[0105] Collect wave model forecast data for the target sea area, analyze the two-dimensional energy spectral density function contained in the wave model forecast data, and extract the spectral moment parameter and wind field parameter from the two-dimensional energy spectral density function.

[0106] Spectral moment parameters include zero-order moment, first-order moment, and spectral peak frequency; wind field parameters include wind speed.

[0107] Ocean wave model forecast data refers to the latitude and longitude gridded forecast data of the target sea area obtained from the output of the ocean wave numerical model (NetCDF or GRIB2 format data generated by the WW3 or SWAN ocean wave numerical model).

[0108] Traverse all latitude and longitude grid points in the target sea area. If all elements of the two-dimensional energy spectral density function of a certain grid point are zero or it is marked as a default mask, then the grid point is determined to be an invalid grid point (such as land or a completely shaded area) and no further processing is performed.

[0109] Spectral moment parameters are statistical physical quantities that reflect the energy distribution of ocean waves. Wind field parameters are dynamic parameters that drive the development of ocean waves.

[0110] Acquire collection time Two-dimensional energy spectral density function was extracted from wave model forecast data of the target sea area. ,in For frequency, For wave direction.

[0111] Calculate the two-dimensional energy spectral density function zeroth moment First moment and spectral peak frequency The formula is:

[0112] ;

[0113] ;

[0114] Among them, the spectral peak frequency Two-dimensional energy spectral density function The frequency value corresponding to the maximum value; It is a frequency infinitesimal element; It is a wave-oriented infinitesimal element.

[0115] To reflect the total energy of the wave spectrum, the zero-order moment of the wave spectrum is calculated by performing a double integral on the two-dimensional energy spectral density function over the entire frequency and wave direction range, thereby obtaining the total energy of the waves in the target sea area.

[0116] It is used to reflect the energy-weighted frequency distribution characteristics of the wave spectrum. By multiplying the two-dimensional energy spectral density function by the frequency and then performing a double integral, the first moment of the wave spectrum is calculated, and characteristic parameters such as the average period of the waves are derived, as well as the spectral width parameter is calculated.

[0117] Extracting wave model forecast data Wind speed at a height of meters ,in The preferred value is 10, which means extracting the wind speed at a height of 10 meters above the sea surface. .

[0118] If the forecast data outputs wind speeds not at a height of 10 meters, then the logarithmic law formula for wind shear is used to convert it to a height of 10 meters, and the zero-order moment is calculated. First moment Spectral peak frequency and wind speed The combination forms the current sea state feature vector.

[0119] Based on the spectral moment parameter and wind field parameter, calculate the local wave steepness parameter and spectral width parameter of the current ocean wave.

[0120] A dynamic mapping model is constructed based on local wave steepness parameters and spectral width parameters. These parameters are then input into the dynamic mapping model to calculate the dynamically segmented frequencies. Specifically, this includes:

[0121] Obtain the zeroth moment from the current sea state feature vector First moment and wind speed Substitute the parameters into the preset formula to calculate the local wave steepness parameters of the current ocean waves. Spectral width parameter The formula is:

[0122] ;

[0123] ;

[0124] In the formula, It is the acceleration due to gravity. It is a second-order spectral moment and satisfies: .

[0125] It is used to reflect the steepness of ocean waves relative to wind speed. By using the ratio of total ocean wave energy to driving wind speed, local wave steepness parameters are calculated to quantitatively assess the current growth status and steepness of wind and waves.

[0126] It is used to reflect the degree of dispersion or concentration of wave energy distribution in the frequency domain. Based on the statistical relationship of the zeroth, first, and second moments of the wave spectrum, the spectral width parameter is calculated to measure the width of the wave spectrum, thereby determining the complexity of the wave components.

[0127] A dynamic mapping model based on radial basis functions is constructed to incorporate local wave steepness parameters. Spectral width parameter As an input node, the dynamic segmentation frequency As an output node.

[0128] Obtain historical wave pattern forecast datasets and extract the set of two-dimensional energy spectral density functions that are accurately verified by measured buoys in historical records, along with their corresponding historical dynamic segmentation frequency labels.

[0129] For each record in the historical record set, calculate the corresponding local wave steepness parameter. Spectral width parameter , constitute input sample pairs .

[0130] The K-means clustering algorithm is used to perform cluster analysis on the input sample pairs. During the clustering process, it is necessary to determine the number of hidden layer nodes, i.e., the number of clusters K. Specifically, the elbow rule is used to determine the value of K.

[0131] Calculate the sum of distances from all samples to their respective cluster centers for different K values, and select the K value corresponding to the inflection point where the slope of the curve becomes significantly gentler as the final number of clusters (usually, the K value is preferably an integer between 5 and 15).

[0132] Cluster center points are selected as the center vectors of the dynamic mapping model. The average distance from each sample within a cluster to the cluster center is calculated as the width vector. .

[0133] Construct a system of linear equations, substitute all input sample pairs into the radial basis function to calculate the hidden layer output matrix, and use the least squares method to solve for the weights of the hidden layer nodes. This minimizes the mean square error between the output value of the dynamic mapping model and the historical dynamic segmentation frequency label, thus completing the parameter calibration of the dynamic mapping model.

[0134] The dynamic segmentation frequency is calculated by calling the center vector, width vector, and weight vector of the pre-calibrated dynamic mapping model. :

[0135] ;

[0136] In the formula, For the first The weights of each hidden layer node, As the center vector, It is a width vector. This is the hidden layer node number.

[0137] The critical frequency used to divide the wind wave and swell frequency bands is calculated by inputting the local wave steepness and spectral width features into the network, calculating the distance between the input features and the center of each hidden layer, and then weighting and summing the results after activation by a Gaussian function. This achieves a nonlinear mapping from sea state features to dynamically segmented frequencies, and adaptively calculates the wind and swell segmentation frequency that best matches the current sea state.

[0138] The calculated dynamic segmentation frequency With spectral peak frequency If a comparison is made, Then Forced assignment .

[0139] Based on the laws of physical oceanography, the main energy region of wind and waves is usually concentrated in the spectral peak frequency. The energy of surge waves is concentrated in the vicinity and at high frequencies, while the energy of surge waves is concentrated at low frequencies. (Constant) It is the lower limit of the energy boundary between swell and wind waves, determined based on empirical relationships in classic wind wave spectra such as the JONSWAP spectrum.

[0140] The forced assignment operation is equivalent to applying a physically-based constraint boundary to prevent the dynamic model from failing under extreme sea conditions and to ensure the dynamic segmentation frequency. It is located within a reasonable frequency band.

[0141] If the dynamic mapping model outputs Less than This means that the dividing point is overly biased towards low frequencies, which will incorrectly classify a large amount of high-frequency energy that belongs to wind waves as swells.

[0142] The two-dimensional energy spectral density function is divided into frequency bands based on the dynamic segmentation frequency, and the wind wave energy spectral density function and the surge energy spectral density function are obtained.

[0143] The effective wave heights of the wind wave energy spectral density function and the swell energy spectral density function are calculated separately, and residual verification is performed by combining the overall effective wave height in the wave model forecast data.

[0144] If the residual exceeds a preset threshold, the dynamic segmentation frequency is iteratively corrected, and the final significant wave heights of wind waves and swells are output. Specifically, this includes:

[0145] Obtain the calculated dynamic segmentation frequency For the two-dimensional energy spectral density function Perform frequency band segmentation:

[0146] frequency Less than or equal to the dynamic segmentation frequency Partially divided into surge energy spectral density functions Frequency Greater than the dynamic segmentation frequency Partially divided into wind and wave energy spectral density functions .

[0147] For two-dimensional energy spectral density function At dynamic segmentation frequency The nearby transition frequency band is smoothed to avoid the Gibbs phenomenon caused by frequency band cutoff.

[0148] Set transition bandwidth The value range is preferably 0.03Hz to 0.08Hz, more preferably 0.05Hz. This width can typically cover 3-5 frequency discrete points to balance smoothness and separation accuracy.

[0149] In the interval Internally, a smoothing weight coefficient is introduced. :

[0150] ;

[0151] Surge energy spectral density function Energy spectral density function of wind and waves The values ​​within the transition frequency band are replaced with smoothed calculation results. The formula is:

[0152] ;

[0153] ;

[0154] Within the transition frequency band near the dynamic segmentation frequency, a weighting coefficient for smooth transition (smoothly transitioning from 0 to 1) is generated, so that the energy spectrum of wind waves and swells transitions smoothly at the segmentation frequency, avoiding energy leakage and Gibbs oscillation caused by hard band cutoff.

[0155] Outside the transition frequency band, the values ​​of the surge energy spectral density function and the wind wave energy spectral density function remain unchanged according to the original frequency band division rules.

[0156] The energy (zero moment) contained in the surge is calculated based on the separated surge energy spectral density function, and the effective wave height of the surge is calculated using the conversion relationship between effective wave height and spectral energy (4 times the square root of the zero moment).

[0157] The energy (zero-order moment) contained in the wind and wave is calculated based on the separated wind and wave energy spectral density function, and the effective wave height is calculated using the conversion relationship between effective wave height and spectral energy.

[0158] Based on the surge energy spectral density function Energy spectral density function of wind and waves Calculate the significant wave height of the swell separately. Effective wave height The formula is:

[0159] ;

[0160] ;

[0161] Extracting the total significant wave height from wave model forecast data Calculate the significant wave height of wind waves And swell effective wave height The synthetic wave height of The formula is: .

[0162] Based on the physical principle that wind wave and swell energy are independent and superimposed with the sum of squares, the separated wind wave significant wave height and swell significant wave height are combined into a theoretical total significant wave height, and the consistency is verified with the total significant wave height of the original forecast data.

[0163] Calculate the total significant wave height with synthetic wave height The residual index between The formula is:

[0164] ;

[0165] The relative residual between the theoretically synthesized wave height and the total effective wave height of the original model is calculated as a quantitative indicator to assess the accuracy of the current dynamic frequency segmentation.

[0166] If the residual index If the residual value is less than or equal to the preset residual threshold, then output the current effective wave height. And swell effective wave height As the final monitoring result.

[0167] If this value exceeds the preset threshold, it indicates that the segmentation frequency is unreasonable, thus triggering an iterative correction mechanism to adjust the dynamic segmentation frequency. Perform iterative corrections:

[0168] An optimization function is constructed with the dynamic segmentation frequency as the independent variable and the goal of minimizing the squared residual. This provides a mathematical basis for finding the optimal segmentation frequency and ensures that the iterative process converges to the frequency point with the minimum separation residual.

[0169] Establish based on residual index Minimize the objective function And set the energy non-negativity constraint conditions for the wind wave energy spectral density function and the swell energy spectral density function:

[0170] ;

[0171] Using the golden section search algorithm in the interval Internal search objective function The minimum point; the physical basis for setting this interval is:

[0172] Based on the physical characteristics of the wave spectrum, the energy boundary frequency between wind waves and swells is usually located at the peak frequency of the spectrum. The region from near the low-frequency side to slightly above the spectral peak frequency.

[0173] The lower limit is set to Consistent with the aforementioned physical constraints, the upper limit is set to This is because if the segmentation frequency exceeds 1.5 times the peak frequency of the spectrum, most of the wind and wave energy will be missed in the calculation.

[0174] Therefore, the golden section method is used for one-dimensional optimization within the physically feasible finite interval. Each iteration can reduce the search interval by a factor of 0.618, which can quickly approach the global optimal solution with minimal computational cost.

[0175] Set a preset search accuracy threshold and iteratively calculate the interval reduction rate.

[0176] When the search interval length is less than the search precision threshold, the midpoint of the output interval is used as the corrected dynamic segmentation frequency. .

[0177] The corrected dynamic segmentation frequency Recalculate the corrected effective wave height using steps S301 to S302. And swell effective wave height This will be used as the final monitoring result and output.

[0178] The significant wave height of wind waves, significant wave height of swell waves, and the final dynamic segmentation frequency in the monitoring results are spatially matched with the latitude and longitude grid points in the wave model forecast data.

[0179] A three-dimensional monitoring matrix is ​​constructed, in which latitude and longitude grid points are two-dimensional planar coordinates, and wind wave significant wave height, swell significant wave height and dynamic segmentation frequency are respectively used as attribute values ​​of the third dimension.

[0180] Since wave model forecast data is spatially continuous, the wave state of adjacent grid points is usually gradual. If the dynamic segmentation frequency of each grid point is calculated in isolation, the segmentation frequency of adjacent grid points may experience unreasonable and drastic jumps when there is local numerical noise in the model data.

[0181] Therefore, a bilinear interpolation algorithm is used to perform spatial smoothing on the dynamic segmentation frequency in the three-dimensional monitoring matrix. The weighted average of the dynamic segmentation frequency of each grid point and its four adjacent grid points is calculated, and the weights are determined by the inverse ratio of the distance between grid points.

[0182] Meanwhile, in order to preserve the true characteristics of spatial abrupt changes in sea conditions (such as the difference in wind and waves on both sides of a front), the preset spatial threshold is preferably 0.02Hz.

[0183] If the difference between the dynamic segmentation frequency of a grid point and the weighted average value after spatial smoothing exceeds a preset spatial threshold, it is determined to be an abnormal jump, and the dynamic segmentation frequency of that grid point is replaced with the weighted average value. Steps S301 to S302 are then re-executed to calculate the significant wave height of wind waves and significant wave height of swells corresponding to that grid point.

[0184] The present invention also provides a wave model forecast data monitoring system based on numerical simulation, including a data acquisition module, a parameter estimation module and a separation and verification module.

[0185] The data acquisition module is used to collect wave model forecast data for the target sea area, analyze the two-dimensional energy spectral density function contained in the wave model forecast data, and extract the spectral moment parameter and wind field parameter from the two-dimensional energy spectral density function.

[0186] Spectral moment parameters include zero-order moment, first-order moment, and spectral peak frequency; wind field parameters include wind speed.

[0187] The parameter estimation module is used to calculate the local wave steepness and spectral width parameters of the current ocean waves based on the spectral moment parameters and wind field parameters.

[0188] A dynamic mapping model is constructed based on the local wave steepness parameter and the spectral width parameter. The dynamic segmentation frequency is calculated by inputting the local wave steepness parameter and the spectral width parameter into the dynamic mapping model.

[0189] The separation and verification module is used to divide the two-dimensional energy spectral density function into frequency bands according to the dynamic segmentation frequency, and separate the wind wave energy spectral density function and the surge energy spectral density function.

[0190] The effective wave heights of the wind wave energy spectral density function and the swell energy spectral density function are calculated separately, and residual verification is performed by combining the overall effective wave height in the wave model forecast data.

[0191] If the residual exceeds the preset threshold, the dynamic segmentation frequency is iteratively corrected, and the final effective wave height of wind waves and effective wave height of swell waves are output.

[0192] Example 1: This example uses WAVEWATCH III wave model forecast data from January to December 2024 for a certain sea area, as well as synchronous measured buoy observation data, as the experimental dataset.

[0193] A total of 1,200 valid samples were obtained, covering four typical sea conditions: pure wind and waves, wind and waves as the dominant force, swells as the dominant force, and pure swells.

[0194] The sample included wind speeds ranging from 2 m / s to 18 m / s and significant wave heights ranging from 0.5 m to 8.2 m, covering common sea state ranges for routine monitoring and disaster early warning. The following control and experimental groups were established:

[0195] Control group: The fixed empirical segmentation frequency + hard truncation separation method (traditional PM separation method) widely used in existing technologies is adopted. This method determines the segmentation frequency with a fixed empirical proportional coefficient and divides the frequency band by hard truncation.

[0196] Experimental group: The wave model forecast data monitoring method based on numerical simulation proposed in this invention is adopted, which includes all technical steps such as dynamic segmentation frequency calculation, smooth transition frequency band processing, residual iterative correction and spatial smoothing constraints.

[0197] The evaluation metrics used in the experiment included residual index, root mean square error of effective wave height, and energy leakage rate.

[0198] The residual index, as the composite value of the effective wave heights of wind waves and swells after separation, and the relative error between the total effective wave height of the original model, is used to measure the energy conservation of the separation results.

[0199] The root mean square error of significant wave height (RMSE) is the root mean square error between the significant wave heights of the separated wind waves and swells and the measured values ​​of the buoy. It is used to measure the accuracy of the separation results.

[0200] Energy leakage rate, as the proportion of wave energy lost during the separation process to the original total wave energy, is used to measure the impact of band cutoff operations on energy conservation.

[0201] The experimental results comparing the two methods under different sea conditions are as follows:

[0202] Sea conditions (wind, waves, and sea state):

[0203] Existing technical methods: residual exponent is 2.8%, root mean square error of effective wave height is 0.16m, and energy leakage rate is 2.1%;

[0204] The method of this invention has the following characteristics: residual exponent of 0.5%, root mean square error of effective wave height of 0.04m, and energy leakage rate of 0.1%.

[0205] Mixed sea conditions dominated by wind and waves:

[0206] Existing technical methods: residual index is 6.5%, root mean square error of effective wave height is 0.25m, and energy leakage rate is 4.3%;

[0207] The method of this invention has the following characteristics: residual exponent is 0.7%, root mean square error of effective wave height is 0.05m, and energy leakage rate is 0.2%.

[0208] Swell-dominated mixed sea conditions:

[0209] Existing technical methods: residual index is 11.2%, root mean square error of effective wave height is 0.38m, and energy leakage rate is 7.8%;

[0210] The method of this invention has the following characteristics: residual index of 0.9%, root mean square error of effective wave height of 0.06m, and energy leakage rate of 0.2%.

[0211] Pure swell sea conditions:

[0212] Existing technical methods: residual index is 8.3%, root mean square error of effective wave height is 0.29m, and energy leakage rate is 5.6%;

[0213] The method of this invention has the following characteristics: residual exponent of 0.6%, root mean square error of effective wave height of 0.04m, and energy leakage rate of 0.1%.

[0214] Please see Figure 2 This paper presents a comparison of the separation residual indices of the two methods under different sea conditions.

[0215] As can be seen from the figure, the residual index of the existing technical method fluctuates greatly under different sea conditions, reaching a maximum of 11.2% under mixed sea conditions dominated by swells.

[0216] This is because the fixed empirical frequency division used in existing technologies cannot adapt to the dynamic changes in the spectral structure under swell-dominated sea conditions, leading to serious misjudgments in frequency band division and consequently causing large energy residuals.

[0217] The method of this invention constructs a dynamic mapping model by using local wave steepness and spectral width parameters, which can adaptively adjust the segmentation frequency. At the same time, combined with the residual iterative correction mechanism, the residual index is controlled within 1% under all sea states, which significantly improves the energy conservation of the separation results and ensures the physical consistency of the monitoring data.

[0218] Please see Figure 3 This paper presents a comparison of the effective wave height separation accuracy of the two methods under different sea conditions.

[0219] Existing technologies and methods exhibit significantly increased separation errors under complex mixed sea conditions, with the root mean square error reaching 0.38m under swell-dominated sea conditions, failing to meet the requirements for high-precision wave monitoring and marine engineering design.

[0220] The method of this invention suppresses energy leakage due to frequency band cutoff through smooth transition processing, and corrects abnormal jumps caused by numerical noise through spatial constraints, thereby controlling the separation error under all sea states to within 0.06m.

[0221] Compared with existing technologies, the separation accuracy of this invention is improved by 6-8 times, and it can accurately extract the effective wave height parameters of wind waves and swells, providing reliable data support for refined analysis of ocean waves.

[0222] Please see Figure 4 The study demonstrates a comparison of wave energy spectrum separation effects under a typical bimodal mixed sea state, in which the original wave energy spectrum simultaneously contains low-frequency peaks of swells and high-frequency peaks of wind waves, representing a typical complex scenario for separating wind waves and swells.

[0223] Existing technologies employ a hard-cut frequency band division method, which directly cuts off the energy spectrum at the segmentation frequency. This causes a sharp abrupt change in the surge energy spectrum and the wind wave energy spectrum at the segmentation point. This non-physical truncation operation not only destroys the natural gradual change characteristics of the energy spectrum, but also causes significant energy leakage and Gibbs oscillations, resulting in distorted separation results.

[0224] The method of this invention sets a smooth transition frequency band of 0.05Hz near the dynamic segmentation frequency, and achieves a smooth transition between the surge energy spectrum and the wind wave energy spectrum through a gradual weighting coefficient. This ensures the accuracy of the frequency band division and avoids the energy loss caused by hard truncation, making the separated energy spectrum more consistent with the real physical laws and effectively solving the energy leakage problem of the prior art.

[0225] To address the issue of local numerical noise in wave pattern data, this invention further verifies the effectiveness of spatial smoothing constraints.

[0226] Experimental results show that in noisy grid areas, the maximum jump in dynamic segmentation frequency between adjacent grid points using existing techniques reaches 0.08 Hz, resulting in unreasonable spatial abrupt changes that affect the reliability of regional monitoring results.

[0227] After spatial smoothing processing, the maximum jump in the segmentation frequency of adjacent grid points is reduced to less than 0.01Hz. While preserving the spatial abrupt change characteristics of the real sea state, it corrects the abnormal jump caused by numerical noise and enhances the spatial coherence of the monitoring results.

[0228] In summary, the monitoring method proposed in this invention, compared with existing technologies, can adaptively adapt to the spectral structure changes of different sea states, and solves problems such as misjudgment of fixed segmentation frequencies, energy leakage of hard cutoff, and spatial jumps in numerical noise. It achieves high-precision separation of wind, waves, and swells under all typical sea states, and can meet the application needs of refined monitoring of ocean waves and disaster early warning.

[0229] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0230] 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 foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for monitoring wave model forecast data based on numerical simulation, characterized in that: The method includes: S100. Collect wave model forecast data of the target sea area, analyze the two-dimensional energy spectral density function contained in the wave model forecast data, and extract the spectral moment parameter and wind field parameter from the two-dimensional energy spectral density function. The spectral moment parameter includes the zero-order moment, the first-order moment and the spectral peak frequency, and the wind field parameter includes the wind speed. S200. Based on the spectral moment parameter and wind field parameter, calculate the local wave steepness parameter and spectral width parameter of the current sea wave; construct a dynamic mapping model based on the local wave steepness parameter and spectral width parameter, input the local wave steepness parameter and spectral width parameter into the dynamic mapping model, and calculate the dynamic segmentation frequency; S300. Divide the two-dimensional energy spectral density function into frequency bands according to the dynamic segmentation frequency to separate the wind wave energy spectral density function and the swell energy spectral density function; calculate the effective wave height of the wind wave energy spectral density function and the swell energy spectral density function respectively, and perform residual verification by combining the overall effective wave height in the wave model forecast data. If the residual exceeds the preset threshold, iteratively correct the dynamic segmentation frequency and output the final effective wave height of the wind wave and the effective wave height of the swell.

2. The method for monitoring wave model forecast data based on numerical simulation according to claim 1, characterized in that: In S100, wave model forecast data refers to the latitude and longitude gridded forecast data of the target sea area obtained from the output of the wave numerical model. Spectral moment parameters are statistical physical quantities that reflect the energy distribution of ocean waves; wind field parameters are dynamic parameters that drive the development of ocean waves. Acquire collection time Two-dimensional energy spectral density function was extracted from wave model forecast data of the target sea area. ,in For frequency, For wave direction; Calculate the two-dimensional energy spectral density function zeroth moment First moment and spectral peak frequency The formula is: ； ； Among them, the spectral peak frequency Two-dimensional energy spectral density function The frequency value corresponding to the maximum value; It is a frequency infinitesimal element; For wave-to-micro element; Extracting wave model forecast data Wind speed at a height of meters , the zeroth moment First moment Spectral peak frequency and wind speed The combination forms the current sea state feature vector.

3. The method for monitoring wave model forecast data based on numerical simulation according to claim 2, characterized in that: S200 includes: S201. Obtain the zeroth moment from the current sea state feature vector. First moment and wind speed Substitute the parameters into the preset formula to calculate the local wave steepness parameters of the current ocean waves. Spectral width parameter The formula is: ； ； In the formula, It is the acceleration due to gravity. It is a second-order spectral moment and satisfies: ; S202. Construct a dynamic mapping model based on radial basis functions, and incorporate the local wave steepness parameter. Spectral width parameter As an input node, the dynamic segmentation frequency As an output node; S203. Calculate the dynamic segmentation frequency by calling the pre-calibrated center vector, width vector, and weight vector of the dynamic mapping model. : ； In the formula, For the first The weights of each hidden layer node, As the center vector, It is a width vector. The hidden layer node number; S204. Calculate the dynamic segmentation frequency. With spectral peak frequency If a comparison is made, Then Forced assignment .

4. The method for monitoring wave model forecast data based on numerical simulation according to claim 3, characterized in that: The pre-calibration process of the dynamic mapping model includes: Acquire historical wave pattern forecast datasets and extract the set of two-dimensional energy spectral density functions that are accurately verified by measured buoys in historical records, along with their corresponding historical dynamic segmentation frequency labels. For each record in the historical record set, calculate the corresponding local wave steepness parameter. Spectral width parameter , constitute input sample pairs ; The K-means clustering algorithm is used to perform cluster analysis on the input sample pairs, and the cluster centroids are selected as the center vectors of the dynamic mapping model. The average distance from each sample within a cluster to the cluster center is calculated as the width vector. ; Construct a system of linear equations, substitute all input sample pairs into the radial basis function to calculate the hidden layer output matrix, and use the least squares method to solve for the weights of the hidden layer nodes. This minimizes the mean square error between the output value of the dynamic mapping model and the historical dynamic segmentation frequency label, thus completing the parameter calibration of the dynamic mapping model.

5. The method for monitoring wave model forecast data based on numerical simulation according to claim 3, characterized in that: The S300 includes: S301. Obtain the calculated dynamic segmentation frequency. For the two-dimensional energy spectral density function Perform frequency band segmentation: frequency Less than or equal to the dynamic segmentation frequency Partially divided into surge energy spectral density functions ; frequency Greater than the dynamic segmentation frequency Partially divided into wind and wave energy spectral density functions ; S302, Based on the surge energy spectral density function Energy spectral density function of wind and waves Calculate the significant wave height of the swell separately. Effective wave height The formula is: ； ； S303. Extract the total significant wave height from wave model forecast data. Calculate the significant wave height of wind waves And swell effective wave height The synthetic wave height of The formula is: ; S304. Calculate the total significant wave height. with synthetic wave height The residual index between The formula is: ； If the residual index If the residual value is less than or equal to the preset residual threshold, then output the current effective wave height. And swell effective wave height As the final monitoring result.

6. The method for monitoring wave model forecast data based on numerical simulation according to claim 5, characterized in that: In S301, for the two-dimensional energy spectral density function At dynamic segmentation frequency The nearby transition frequency band is smoothed: Set transition bandwidth In the interval Internally, a smoothing weight coefficient is introduced. : ； Surge energy spectral density function Energy spectral density function of wind and waves The values ​​within the transition frequency band are replaced with smoothed calculation results; the formula is: ； ； Outside the transition frequency band, the values ​​of the surge energy spectral density function and the wind wave energy spectral density function remain unchanged according to the original frequency band division rules.

7. The method for monitoring wave model forecast data based on numerical simulation according to claim 5, characterized in that: In S304, if the residual exponent If the residual value exceeds a preset residual threshold, the segmentation frequency correction mechanism is triggered to adjust the dynamic segmentation frequency. Perform iterative corrections; Establish based on residual index Minimize the objective function And set the energy non-negativity constraint conditions for the wind wave energy spectral density function and the swell energy spectral density function: ； Using the golden section search algorithm in the interval Internal search objective function The minimum point; Set a preset search accuracy threshold and iteratively calculate the interval reduction rate; When the search interval length is less than the search precision threshold, the midpoint of the output interval is used as the corrected dynamic segmentation frequency. ; The corrected dynamic segmentation frequency Recalculate the corrected effective wave height using steps S301 to S302. And swell effective wave height This will be used as the final monitoring result and output.

8. The method for monitoring wave model forecast data based on numerical simulation according to claim 1, characterized in that: Following the S300 are: The significant wave height of wind waves, significant wave height of swell waves, and the final dynamic segmentation frequency in the monitoring results are spatially matched with the latitude and longitude grid points in the wave model forecast data. A three-dimensional monitoring matrix is ​​constructed, in which latitude and longitude grid points are two-dimensional planar coordinates, and wind wave significant wave height, swell significant wave height and dynamic segmentation frequency are respectively used as attribute values ​​of the third dimension; A bilinear interpolation algorithm is used to perform spatial smoothing on the dynamic segmentation frequency in the three-dimensional monitoring matrix. The weighted average of the dynamic segmentation frequency of each grid point and its four neighboring grid points is calculated, and the weights are determined by the inverse ratio of the distance between grid points. If the difference between the dynamic segmentation frequency of a certain grid point and the weighted average value after spatial smoothing exceeds the preset spatial threshold, the dynamic segmentation frequency of the grid point is replaced with the weighted average value, and steps S301 to S302 are re-executed to calculate the wind wave significant wave height and swell significant wave height corresponding to the grid point.

9. A wave model forecast data monitoring system based on numerical simulation, applied to the wave model forecast data monitoring method based on numerical simulation as described in claim 1, characterized in that: The system includes a data acquisition module, a parameter estimation module, and a separation and verification module; The data acquisition module is used to collect wave model forecast data of the target sea area, analyze the two-dimensional energy spectral density function contained in the wave model forecast data, and extract the spectral moment parameters and wind field parameters from the two-dimensional energy spectral density function. The spectral moment parameters include the zero-order moment, the first-order moment and the spectral peak frequency, and the wind field parameters include the wind speed. The parameter estimation module is used to calculate the local wave steepness and spectral width parameters of the current ocean waves based on the spectral moment parameters and wind field parameters. A dynamic mapping model is constructed based on the local wave steepness parameter and the spectral width parameter. The dynamic segmentation frequency is calculated by inputting the local wave steepness parameter and the spectral width parameter into the dynamic mapping model. The separation and verification module is used to divide the two-dimensional energy spectral density function into frequency bands according to the dynamic segmentation frequency, and separate the wind wave energy spectral density function and the swell energy spectral density function. The effective wave heights of the wind wave energy spectral density function and the swell energy spectral density function are calculated respectively. The residuals are verified by combining the overall effective wave height in the wave model forecast data. If the residuals exceed the preset threshold, the dynamic segmentation frequency is iteratively corrected, and the final effective wave heights of the wind wave and the swell are output.

Images