A neural network-based sea wave spectrum intelligent correction method and system

Abstract

The application belongs to the technical field of marine environment prediction, and discloses a sea wave spectrum intelligent correction method and system based on a neural network. The method comprises the following steps: target sea area setting and environment data acquisition; numerical calculation of the sea wave spectrum in the target sea area is performed through a sea wave numerical model; a sea wave spectrum correction model comprising a numerical calculation module, a measured data processing module and an intelligent correction module is constructed through a deep learning method; and the constructed sea wave spectrum correction model is subjected to precision verification and adaptability evaluation. The model solves the error problem caused by the simplification of the physical process in the simulation process of the traditional method, and at the same time, gets rid of the problem that the existing data assimilation method is limited by the observation data. The application can effectively improve the simulation precision of the sea wave model, and has good sea area adaptability.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of marine environment prediction, and particularly relates to a sea wave spectrum intelligent correction method and system based on a neural network. BACKGROUND

[0002] The sea wave spectrum is an important tool for describing the characteristics of sea waves and can reflect the energy distribution of sea waves in different frequency and wave number spaces. At the same time, the sea wave spectrum can reveal the composition of waves, including the generation and development of local wind waves, the propagation of ocean swells, and the interaction between multiple wave systems. In addition, through the information such as the energy peak position and the spectral width in the sea wave spectrum, the typical characteristics of the waves can be further reflected, providing a basis for the prediction of wave height, wave period and the like. Therefore, the accurate description of the sea wave spectrum can provide more accurate results for depicting sea waves, and can further support the research on marine climate change, ship navigation safety and the like.

[0003] The current sea wave data correction method includes traditional data assimilation and machine learning optimization. The traditional data assimilation method corrects the numerical simulation results by introducing the measured sea wave data, thereby reducing the simulation bias and improving the prediction accuracy to a certain extent. Representative methods include variational assimilation, Kalman filtering and optimal interpolation. Among them, the variational assimilation method realizes the optimal estimation of the model state by constructing a cost function and minimizing the difference between the observation and the model; the Kalman filtering method is based on the dynamic update of the state space model and the error covariance, which can gradually improve the reliability of the simulation results in the iterative fusion of prediction and observation; the optimal interpolation method is based on statistical principles, and the observation and model results are linearly combined through weight distribution, so as to realize the consistency correction in space and time. These methods have been widely used in large-scale ocean models and sea wave prediction services, but they have high computational complexity and are dependent on the spatiotemporal coverage of the observation data.

[0004] In recent years, with the development of artificial intelligence, machine learning methods have been gradually introduced into the optimization and correction of marine environmental information to make up for the shortcomings of traditional assimilation methods in computational efficiency and complex nonlinear relationship processing. In sea wave simulation and prediction, machine learning methods usually construct neural networks or other nonlinear regression models to fit and learn the bias between numerical simulation results and buoy observation data. This process can effectively capture the systematic errors of numerical models in specific regions or specific scenarios and make real-time corrections to the results in new predictions. Compared with traditional methods, machine learning methods have strong generalization ability and high computational efficiency, which can improve the simulation and prediction accuracy of wave elements in the sea area under limited observation conditions. Therefore, the correction based on machine learning has gradually become an important supplement to the accuracy improvement of sea wave models.

[0005] Through the above analysis, the problems and defects of the prior art are:

[0006] The traditional data assimilation method has a complex calculation process and relies on marine observation information. When there is no observation data available for assimilation, especially in remote sea areas or extreme weather conditions, the applicability of the traditional assimilation method is challenged. Therefore, there is a problem of difficulty in correcting marine environmental data in the case of lack of observation data.

[0007] The current machine learning-based sea wave data correction method mainly focuses on global wave parameters such as significant wave height and wave period. Such parameters can reflect the overall characteristics of sea waves to some extent, but cannot describe the distribution characteristics of sea wave energy in the frequency and direction space. At present, there is no clear proposal for a systematic correction method for sea wave spectrum. SUMMARY

[0008] In order to overcome the problems in the related art, the present application discloses a neural network-based intelligent sea wave spectrum correction method and system.

[0009] The technical solution is as follows: a neural network-based intelligent sea wave spectrum correction method, comprising the following steps:

[0010] S1, target sea area setting and environmental data acquisition;

[0011] S2, numerical calculation of sea wave spectrum in the target sea area by a sea wave numerical model;

[0012] S3, constructing a sea wave spectrum correction model including a numerical calculation module, a measured processing module and an intelligent correction module by a deep learning method; for the sea wave spectrum calculated by the sea wave numerical model, constructing numerical simulation sea wave spectrum data by the numerical calculation module, constructing buoy observation sea wave spectrum data by the measured processing module, taking the buoy observation sea wave spectrum data as a benchmark, constructing the deviation between the numerical simulation sea wave spectrum data and the buoy observation sea wave spectrum data by the intelligent correction module, and outputting the corrected sea wave spectrum;

[0013] S4, precision verification and adaptability evaluation of the constructed sea wave spectrum correction model.

[0014] In step S1, the target sea area setting and environmental data acquisition includes:

[0015] S101, setting the target sea area according to the requirements;

[0016] S102, acquiring environmental data in the sea area, including two parts of buoy data and open source environmental field data in the target sea area; the buoy data includes buoy position, observation time range, sea wave spectrum data and sea wave height; the open source environmental field data includes wind speed field and water depth information in the target sea area.

[0017] In step S2, the sea wave spectrum in the target sea area is calculated by a sea wave numerical model, including:

[0018] S201, a model calculation domain is selected according to the range of the study sea area, and a model calculation boundary field is set considering the sea wave transmission characteristics to achieve high-precision calculation of the sea wave; in the calculation process, a large calculation domain is set based on the target sea area, and the target sea area is calculated by nesting; the large calculation domain is set as a calculation domain which is 20° larger than the target sea area in the east, west, south and north directions, thereby providing boundary field information for the calculation of the target sea area by the model;

[0019] S202, a model calculation forcing field is generated using the obtained wind speed and other environmental data; the forcing field in the sea area is calculated according to the wind speed information at 10 meters above the sea surface in the east and west and in the south and north directions, the wind field data is interpolated according to the model calculation grid, and the forcing field data unified with the time and spatial dimension steps set in the calculation process is obtained;

[0020] S203, the numerical dispersion method in the calculation process of the sea wave numerical model is set, including the spatial dispersion of frequency and wave direction and the global time step time dispersion; the numerical solution of the sea wave spectrum in the sea area is completed, and the sea wave height at the target position is output.

[0021] In step S3, the numerical calculation module processes the spectrum frequency by using linear interpolation for the sea wave spectrum simulated by the sea wave numerical model, and constructs the numerical simulation sea wave spectrum data consistent with the spectrum frequency of the buoy observation sea wave spectrum;

[0022] The linear interpolation calculation method is as follows:

[0023] ;

[0024] In the formula, is the sea wave spectrum, is the buoy observation frequency, is the sea wave model simulation spectrum frequency, is the sea wave spectrum at the buoy observation frequency, is the sea wave model simulation spectrum at a certain frequency, is the sea wave model simulation spectrum at the adjacent spectrum frequency;

[0025] After frequency unification, the numerical simulation sea wave spectrum data obtained is .

[0026] Further, in view of the problem of missing data and data loss of the buoy data, the time series of the sea wave spectrum is processed with reference to the measured data.

[0027] First, a numerical simulation timeline was created based on the time steps of the wave model numerical calculations. Second, a buoy observation timeline was created in the same format. Due to data gaps caused by weather and equipment limitations, the numerical simulation data was deleted according to the missing time series and the timeline.

[0028] In step S3, the measurement processing module smooths the abnormal fluctuation positions in the original observation data of the buoy-observed wave spectrum using a Gaussian filtering method; and retains the dominant frequency and energy distribution characteristics of the spectrum using a peak recovery method. The specific calculation method is shown below:

[0029] ;

[0030] ;

[0031] ;

[0032] ;

[0033] In the formula, For Gaussian kernel, This is the result after Gaussian filtering. This is the result after peak recovery. For the calculation process of the numerical function, In order to be in At the index position, with The original observation results for the sliding window offset; for The Gaussian kernel calculation results at the location, The peak value of the function. The peak value of the function after Gaussian filtering. For peak index, The peak value in the original signal. The standard deviation of the Gaussian kernel. As the independent variable, Let the radius be the Gaussian kernel. This is the offset of the sliding window. The weighting coefficients for peak recovery. The output signal after Gaussian filtering. For index position, The original signal, This is the final output result after peak recovery;

[0034] Filtered and processed wave spectrum data from actual measurements, resulting in smooth and high-fidelity data. .

[0035] Further, the data missing period caused by extreme weather and equipment failure is removed. The removal is directly deleting the data missing part according to the observation time axis.

[0036] In step S3, the intelligent correction module corrects the numerical simulation sea wave spectrum data constructed by the numerical calculation module as input, the buoy observation sea wave spectrum data constructed by the observation processing module as output; and The data dimensions are both 44*6324; 44 represents the frequency number, and 6324 represents the time number; wherein, the input and output layers both contain 44 neurons, representing the sea wave spectrum results corresponding to 44 frequencies at a time.

[0037] A three-layer fully connected hidden layer is adopted, and the neuron numbers are 200, 200 and 100 respectively; all the hidden layers use ReLU activation function to enhance the nonlinear expression ability; the output layer is linearly activated; in the training process, the input data is first batch-processed, read and sent into the network for forward propagation, and the corrected sea wave spectrum is output; the mean square error MSE is used as the loss function to measure the difference between the output and the observation spectrum; the Adam optimizer is used to adaptively adjust the learning rate in the optimization process, and finally the sea wave spectrum is corrected.

[0038] Further, the loss function is specifically calculated as follows:

[0039] ;

[0040] In the formula, is the loss function, is the sample number, and and are the true value and the model prediction value respectively.

[0041] Another purpose of the present application is to provide a neural network-based intelligent sea wave spectrum correction system, which implements the neural network-based intelligent sea wave spectrum correction method, and the system comprises:

[0042] A data acquisition module is used for target sea area setting and environment data acquisition.

[0043] A sea wave spectrum numerical calculation module is used for numerical calculation of the sea wave spectrum in the target sea area through a sea wave numerical model.

[0044] The sea wave spectrum correction model construction module is configured to construct a sea wave spectrum correction model including a numerical calculation module, a measured data processing module and an intelligent correction module through a deep learning method; for the sea wave spectrum calculated by a sea wave numerical model, the numerical calculation module is used to construct numerical simulation sea wave spectrum data, the measured data processing module is used to construct buoy observed sea wave spectrum data, and the intelligent correction module is used to construct the deviation between the numerical simulation sea wave spectrum data and the buoy observed sea wave spectrum data based on the buoy observed sea wave spectrum data, and output the corrected sea wave spectrum.

[0045] The model verification and evaluation module is configured to verify the accuracy and evaluate the adaptability of the constructed sea wave spectrum correction model.

[0046] In combination with all the above technical solutions, the present application has the following beneficial effects:

[0047] Firstly, the present application proposes a neural network-based intelligent sea wave spectrum correction method and system. The model takes buoy observation data as the benchmark, constructs the deviation between the simulated sea wave spectrum of the sea wave model and the real spectrum, and realizes effective correction of the sea wave spectrum. The model solves the error problem caused by simplification of physical processes in the simulation process of the traditional method. At the same time, it gets rid of the limitation problem of the existing data assimilation method caused by observation data. The present application can effectively improve the simulation accuracy of the sea wave model and has good sea area adaptability. The present application provides a novel and effective method and strategy for numerical calculation of real sea area sea wave spectrum.

[0048] Secondly, the present application can realize optimization of the sea wave in the frequency space, more accurate description of the distribution of sea wave energy, and high-precision description of the sea wave spectrum, which can further support the research on marine climate change, ship navigation safety and other researches, and provide more accurate sea wave data input. The present application can realize effective correction of the numerical simulation sea wave spectrum of the sea wave through a small amount of measured data. The present application solves the error problem caused by simplification of physical processes in the simulation process of the traditional method. At the same time, after the model is constructed, the subsequent correction process of the sea wave spectrum no longer needs to rely on a large amount of real sea observation sea wave spectrum data. Compared with the existing data assimilation method, the present application can get rid of the limitation problem of the observation data. In addition, compared with the single sea wave parameter correction method in the prior art, the present application can realize optimization of the distribution characteristics of the sea wave in the frequency space, and more comprehensively reflect the sea wave characteristics.

[0049] Third, the traditional sea wave model describes the sea wave generation and dissipation process in the form of physical expression in the calculation process, which is simplified compared with the real situation. The present application uses a neural network to correct the numerical simulation results using a small amount of real sea observation data, and solves the difference between the simulated sea wave spectrum and the real sea wave spectrum. At the same time, the existing sea wave data correction method is only for global parameters such as sea wave height and sea wave period, and there is no comprehensive correction of sea wave spectrum. The present application aims to optimize the sea wave spectrum, and uses two dimensions of spectrum shape and spectrum parameter (spectrum distance and spectrum peak); sea wave spectrum, sea wave height, spectrum peak frequency and spectrum peak value are used to comprehensively evaluate the model correction effect, optimize the frequency space result of the sea wave spectrum, and more accurately describe the distribution of sea wave energy. The present application solves the problem of simplification in the calculation of sea wave generation, dissipation and energy transfer in the traditional sea wave numerical simulation method, and realizes effective correction of the numerical simulation results using a small amount of real sea observation data. BRIEF DESCRIPTION OF DRAWINGS

[0050] The accompanying drawings, which are incorporated into and form part of the specification, illustrate embodiments consistent with the present disclosure and, together with the specification, serve to explain the principles of the present disclosure;

[0051] Figure 1 is a neural network-based sea wave spectrum intelligent correction method flowchart provided by the embodiment of the present application;

[0052] Figure 2 is a neural network-based sea wave spectrum intelligent correction method principle diagram provided by the embodiment of the present application;

[0053] Figure 3 is a WW3 sea wave spectrum before and after interpolation in time 1 result comparison result diagram;

[0054] Figure 4 is a WW3 sea wave spectrum before and after interpolation in time 2 result comparison result diagram;

[0055] Figure 5 is a WW3 sea wave spectrum before and after filtering in time 1 result comparison result diagram;

[0056] Figure 6 is a WW3 sea wave spectrum before and after filtering in time 2 result comparison result diagram;

[0057] Figure 7 is a sea wave spectrum comparison result diagram before and after the sea wave spectrum correction model correction in time 1;

[0058] Figure 8 is a sea wave spectrum comparison result diagram before and after the sea wave spectrum correction model correction in time 2;

[0059] Figure 9 is a sea wave spectrum comparison result diagram before and after the sea wave spectrum correction model correction in time 3;

[0060] Figure 10 The sea wave spectrum comparison result chart before and after the sea wave spectrum correction model is corrected at time 4;

[0061] Figure 11 The sea wave spectrum accuracy comparison result chart before and after the sea wave spectrum correction model is corrected;

[0062] Figure 12 The sea wave spectrum correlation result chart before and after the sea wave spectrum correction model is corrected in the sea wave mode numerical simulation;

[0063] Figure 13 The sea wave spectrum correlation comparison result chart before and after the sea wave spectrum correction model is corrected;

[0064] Figure 14 The wave height parameter comparison result chart before and after the model is corrected and the observation value;

[0065] Figure 15 The spectrum peak frequency parameter comparison result chart before and after the model is corrected and the observation value;

[0066] Figure 16 The spectrum peak value parameter comparison result chart before and after the model is corrected and the observation value;

[0067] Figure 17 The sea wave spectrum comparison result chart before and after the model is corrected at the buoy 2;

[0068] Figure 18 The sea wave spectrum comparison result chart before and after the model is corrected at the buoy 3;

[0069] Figure 19 The sea wave spectrum comparison result chart before and after the model is corrected at the buoy 4;

[0070] Figure 20 The sea wave spectrum comparison result chart before and after the model is corrected at the buoy 5;

[0071] Figure 21 The wave height parameter comparison result chart before and after the model is corrected and the observation value at different buoy positions;

[0072] Figure 22 The spectrum peak frequency parameter comparison result chart before and after the model is corrected and the observation value at different buoy positions;

[0073] Figure 23 The spectrum peak value parameter comparison result chart before and after the model is corrected and the observation value at different buoy positions. DETAILED DESCRIPTION

[0074] In order to make the above objectives, characteristics and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application are described in detail below with reference to the drawings. In the following description, a large number of specific details are set forth in order to facilitate a thorough understanding of the present application. However, the present application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without departing from the spirit of the present application, so the present application is not limited to the specific implementations disclosed below.

[0075] The innovation of the present application is that the present application proposes a spectrum deviation correction model based on neural network. The model uses neural network to correct the precision of numerical simulation of sea wave spectrum, which can more accurately describe the distribution of sea wave energy. The model verifies the correction performance of the proposed model more comprehensively through two dimensions of spectral shape and spectral distance, spectral peak, and four quantitative indexes of sea wave spectrum, wave height, spectral peak frequency and spectral peak value. The model solves the error problem caused by simplification of physical process in the simulation process of the traditional method. At the same time, it gets rid of the limitation of observation data in the existing data assimilation method. The present application can effectively improve the simulation precision of the sea wave model, and has good sea area adaptability. The present application provides a novel and effective method and strategy for numerical calculation of real sea area sea wave spectrum.

[0076] Embodiment 1, as shown in Figure 1 The neural network-based intelligent correction method for sea wave spectrum provided by the embodiment of the present application includes:

[0077] S1, target sea area setting and environment data acquisition;

[0078] S2, numerical calculation of sea wave spectrum in the target sea area by a sea wave numerical model;

[0079] S3, constructing a sea wave spectrum correction model including a numerical calculation module, a measured processing module and an intelligent correction module by a deep learning method; for the sea wave spectrum calculated by the sea wave numerical model, constructing numerical simulation sea wave spectrum data by the numerical calculation module, constructing buoy observation sea wave spectrum data by the measured processing module, taking the buoy observation sea wave spectrum data as a benchmark, constructing the deviation between the numerical simulation sea wave spectrum data and the buoy observation sea wave spectrum data by the intelligent correction module, and outputting the corrected sea wave spectrum;

[0080] S4, precision verification and adaptability evaluation of the constructed sea wave spectrum correction model.

[0081] As can be seen from the above embodiment, the traditional data assimilation method has a complex calculation process and depends on marine observation information. The sea wave data correction method based on machine learning currently mainly focuses on global wave parameters and cannot describe the distribution characteristics of sea wave energy in frequency and direction and other spaces.

[0082] The application provides a spectrum deviation correction model based on deep learning. The model can more accurately describe the distribution of sea wave energy by correcting the sea wave spectrum. The model verifies the correction performance of the proposed model from two dimensions of spectrum parameters, including spectrum shape and spectrum distance, spectrum peak, etc. through four quantitative indicators. The model significantly improves the simulation accuracy of the sea wave numerical model, and has good adaptability to sea areas.

[0083] Furthermore, the application provides a neural network-based intelligent sea wave spectrum correction method and system. The model takes buoy observation data as a benchmark to construct the deviation between the sea wave spectrum simulated by the sea wave model and the real spectrum shape, and realizes effective correction of the sea wave spectrum. The method can more accurately describe the distribution of sea wave energy in the frequency space. At the same time, the spectrum shape parameters such as sea wave height, spectrum peak frequency and spectrum peak value can be calculated with high precision. The model can effectively improve the simulation accuracy of the sea wave model, and has good adaptability to sea areas. The application provides a novel and effective method and strategy for fast and accurate calculation of sea wave spectrum.

[0084] In embodiment 2, the application provides a neural network-based intelligent sea wave spectrum correction system, which comprises:

[0085] A data acquisition module is configured to set a target sea area and acquire environmental data.

[0086] A sea wave spectrum numerical calculation module is configured to perform numerical calculation on the sea wave spectrum in the target sea area by using a sea wave numerical model.

[0087] A sea wave spectrum correction model construction module is configured to construct a sea wave spectrum correction model comprising a numerical calculation module, a measured data processing module and an intelligent correction module by using a deep learning method. For the sea wave spectrum calculated by the sea wave numerical model, the numerical calculation module is used to construct numerical simulation sea wave spectrum data, the measured data processing module is used to construct buoy observation sea wave spectrum data, the intelligent correction module is used to construct the deviation between the numerical simulation sea wave spectrum data and the buoy observation sea wave spectrum data based on the buoy observation sea wave spectrum data, and the corrected sea wave spectrum is output.

[0088] A model verification and evaluation module is configured to verify the accuracy and evaluate the adaptability of the constructed sea wave spectrum correction model.

[0089] In embodiment 3, as another embodiment of the application, as shown in Figure 2 The neural network-based intelligent sea wave spectrum correction method provided by the application comprises the following steps:

[0090] S1, setting a target sea area and acquiring environmental data; specifically comprising:

[0091] S101, setting a target sea area according to requirements;

[0092] S102, obtain the environmental data in the sea area, which is specifically divided into two parts of buoy data in the target sea area and open-source environmental field data. The buoy data includes buoy position, observation time range, sea wave spectrum data, and sea wave height. The open-source environmental field data includes wind speed field and water depth information in the target sea area.

[0093] S2, sea wave data calculation based on a sea wave model: the sea wave spectrum in the target sea area is numerically calculated through a sea wave numerical model; specifically including:

[0094] S201, according to the range of the research sea area, a model calculation domain is selected, and the model calculation boundary field is set considering the sea wave transmission characteristics;

[0095] During the model calculation process, the external sea area will have an impact on the sea waves in the target sea area, so a boundary field needs to be set during the calculation to achieve high-precision calculation of the sea waves. During the calculation, the target sea area set in S1 is further set as a large calculation domain for nested calculation of the target sea area. The large calculation domain is set as a calculation domain that is expanded by 20 degrees in the east, west, south, and north directions of the target sea area, thereby providing boundary field information for the model calculation of the target sea area.

[0096] S202, using the obtained wind speed and other environmental data, a model calculation forcing field is generated;

[0097] According to the wind speed information in the east-west and north-south directions at the sea surface 10 meters in the sea area, the forcing field in the sea area is calculated, and the wind field data is interpolated according to the model calculation grid to obtain forcing field data that is unified with the time and spatial dimension steps set in the calculation process.

[0098] S203, the numerical dispersion method in the calculation process of the sea wave numerical model (forcing field, boundary field, numerical dispersion, etc. are all links in the calculation process of the sea wave numerical model) is set, including frequency, wave direction spatial dispersion, and global time step time dispersion. The sea wave spectrum in the sea area is calculated, and the sea wave height at the target position is output.

[0099] S3, construction of a sea wave spectrum correction model: the present application constructs a sea wave spectrum correction model through a deep learning method. For the sea wave spectrum calculated by the sea wave numerical model, the deviation between the observed data and the simulated data is constructed to effectively correct the simulated sea wave spectrum. The sea wave spectrum correction model is divided into three parts of a numerical calculation module, a measured processing module, and an intelligent correction module. Specifically including:

[0100] S301, numerical calculation module: for the sea wave spectrum simulated by the sea wave numerical model in step S2, the spectrum frequency is processed by using linear interpolation to construct the numerical simulation sea wave spectrum data consistent with the spectrum frequency of the observed sea wave spectrum.

[0101] This invention innovatively proposes the following linear interpolation calculation method:

[0102] ;

[0103] In the formula, For the soundtrack of the ocean waves, For buoy observation frequency, and For the simulated spectral frequencies of the wave mode, The wave spectrum at the frequency observed by the buoy. The simulated spectrum of ocean wave patterns at a certain frequency. The simulated spectrum of the wave mode at adjacent spectral frequencies;

[0104] After frequency unification, the numerical simulation wave spectrum data were obtained. .

[0105] The expression for calculating the wave spectrum using a numerical model is:

[0106] ;

[0107] In the formula, For wave action, For wave energy, Relative frequency;

[0108] ;

[0109] In the formula, For wave number, As direction, Wavenumber directional spectrum, For wave action density spectrum, This represents the wave action density spectrum as a function of wave number and direction. The wave number direction spectrum represents the wave number and direction-related information; the wave spectrum is the result of integrating the direction component of the wave direction spectrum. express;

[0110] After frequency unification, the numerical simulation wave spectrum data were obtained. .

[0111] like Figure 3 As shown, the comparison results of the WW3 wave spectrum interpolation before and after at time 1, and as shown in the figure... Figure 4 The results of the wave spectrum interpolation before and after WW3 at time 2 are shown.

[0112] In addition, in view of the problem of missing data and data loss of the buoy data, the time sequence of the sea wave spectrum is processed with respect to the buoy position or any self-selected position, with the measured data as a reference. The specific process is as follows: first, according to the numerical calculation time step of the sea wave mode, the numerical simulation time axis is listed. Second, the buoy observation time axis is listed in the same format. Since the observation data is affected by weather and equipment, etc., there is a certain data loss. Therefore, according to the missing time sequence, the numerical simulation data is deleted according to the time axis.

[0113] S302, the measured processing module: for the buoy observation sea wave spectrum, the abnormal fluctuation position in the original observation data is smoothed by the Gaussian filtering method. And through the peak recovery method, the dominant frequency and energy distribution characteristics of the spectrum shape are reserved, and the loss of key fluctuation information is avoided. The specific calculation method is as follows:

[0114] ;

[0115] ;

[0116] ;

[0117] ;

[0118] In the formula, is the Gaussian kernel, is the result after Gaussian filtering, is the result after peak recovery, is the exponential function calculation process, is the original observation result at the index position, is the sliding window offset, is the Gaussian kernel calculation result at is the function peak, is the function peak result after Gaussian filtering, is the peak index, is the peak in the original signal, is the standard deviation of the Gaussian kernel, is the independent variable, is the Gaussian kernel radius, is the sliding window offset, is the weight coefficient of peak recovery, is the output signal after Gaussian filtering processing, is the index position, is the original signal, is the final output result after peak recovery;

[0119] ​​The buoy observed sea wave spectrum data after filtering and processing is smooth and has high fidelity .

[0120] As shown in Figure 5 , the results of WW3 sea wave spectrum before and after filtering at time 1 are compared, and the results are shown in Figure 6 . At time 2, the results of WW3 sea wave spectrum before and after filtering are compared. In addition, the data missing period caused by extreme weather and equipment failure is removed.

[0121] S303, intelligent correction module: the numerical simulation sea wave spectrum data constructed by the numerical calculation module is input, and the buoy observed sea wave spectrum data constructed by the measured processing module is output. In this model, the and data dimensions are both 44x6324. Specifically, 44 represents the frequency number, and 6324 represents the time number. The input and output layers both contain 44 neurons, representing the sea wave spectrum results corresponding to 44 frequencies at a single time. The model uses 3 layers of fully connected hidden layers, with neuron numbers of 200, 200 and 100 respectively. All hidden layers use ReLU activation function to enhance the nonlinear expression ability. The output layer is linearly activated. During the training process, the model first performs batch processing on the input data, reads and sends it into the network for forward propagation, and outputs the corrected sea wave spectrum . The model uses mean square error (MSE) as the loss function to measure the difference between the model output and the observed spectrum. During the optimization process, the Adam optimizer is used to adaptively adjust the learning rate, improve the training efficiency and convergence speed. Finally, the sea wave spectrum is corrected.

[0122] The specific calculation method of the loss function is as follows:

[0123] ;

[0124] In the formula, is the loss function, is the number of samples, are the true value and model prediction value respectively.

[0125] S4, model accuracy verification includes: using the real observed sea wave spectrum data of the buoy to evaluate the correction effect of the sea wave spectrum correction model. Specifically, it includes two parts of sea wave spectrum accuracy and spectrum parameter accuracy. First, the sea wave spectrum accuracy is verified by mean absolute error MAE, mean relative error MAPE, root mean square error RMSE and determination coefficient r. Then, the wave height , the spectral peak frequency and the spectral peak value of the model corrected sea wave spectrum and the buoy observed sea wave spectrum are calculatedand calculate its error, and multi-index joint verification is performed on the model correction effect. Thus, comprehensive evaluation of the model correction performance is realized.

[0126] Precision verification calculation method:

[0127] ;

[0128] ;

[0129] ;

[0130] ;

[0131] In the formula, represents the sample quantity, and respectively represent the sea wave observation result and the sea wave calculation result.

[0132] The calculation method of the spectral shape parameters , the spectral peak frequency , and the wave height is as follows:

[0133] ;

[0134] ;

[0135] ;

[0136] In the formula, is the sea wave spectral peak value, is the corresponding frequency at the peak value, is the sea wave spectrum, is the frequency.

[0137] Exemplarily, the present application is aimed at training and verifying the model in a certain sea area in the middle of a certain sea, and randomly selecting four time points to show the sea wave spectrum results before and after the sea wave spectrum correction model is corrected. For example, Figure 7 the sea wave spectrum comparison results before and after the sea wave spectrum correction model is corrected at time 1, Figure 8 the sea wave spectrum comparison results before and after the sea wave spectrum correction model is corrected at time 2, Figure 9 the sea wave spectrum comparison results before and after the sea wave spectrum correction model is corrected at time 3, Figure 10 the sea wave spectrum comparison results before and after the sea wave spectrum correction model is corrected at time 4;

[0138] Exemplarily, the specific calculation precision of the method proposed by the present application is as follows: the MAE and RMSE of the sea wave mode to the sea wave spectrum simulation are 0.62 and 1.13 m 2 Hz -1The MAPE was 37.03%. After correction using the wave correction model, MAE, RMSE, and MAPE all decreased significantly compared to the wave model, with MAPE decreasing to 27.55%. Figure 11 The comparison of wave spectrum accuracy before and after wave spectrum correction model correction is shown. The spectral correlation increased from 0.83 to 0.94, as... Figure 12 Correlation results of wave spectrum before and after correction in numerical simulation of wave model Figure 13 The results show the correlation between wave spectrum before and after correction of the wave spectrum correction model of this invention; the results show that the current model effectively improves the simulation accuracy of wave spectrum.

[0139] Furthermore, this invention utilizes the spectral shape parameter wave height Spectral peak frequency and spectral peak The accuracy of both was evaluated. Table 1 shows that the model constructed in this invention accurately reflects wave height. Spectral peak frequency and spectral peak The calculated MAPE values ​​decreased from 16.42%, 16.18%, and 28.18% in WW3 to 11.11%, 9.82%, and 24.12%, respectively. The correlation coefficient r increased from 0.94, 0.645, and 0.88 to 0.97, 0.789, and 0.93, respectively. The correction effect on the wave parameters is shown in Figure 14-. Figure 16 As shown. Figure 14 The wave height parameters before and after model correction are compared with the observed values. Figure 15 The results show a comparison between the spectral peak frequency parameters before and after model correction and the observed values. Figure 16 Comparison of spectral peak parameters with observed values ​​before and after model correction;

[0140] The results show that the model effectively improves the calculation accuracy of spectral parameters and can achieve comprehensive correction of wave information.

[0141] Table 1 Comparison of spectral parameter accuracy before and after model correction

[0142]

[0143] For example, step S4, model adaptability assessment, includes: based on the constructed model, further model adaptability analysis is conducted. Multiple buoys are selected within the target sea area, and the wave spectrum at different buoy locations is corrected. The model correction effect is then verified using measured data. This assesses the model's adaptability. Finally, a wave spectrum correction model with strong adaptability to the sea area is constructed.

[0144] This invention utilizes four other buoys within the sea area to evaluate the model's adaptability, with results shown in Figure 17- Figure 20 And as shown in Table 2. Among them Figure 17Comparison of wave spectra before and after model correction at two locations on the buoy. Figure 18 Comparison of wave spectra before and after model correction at three locations on the buoy. Figure 19 Comparison of wave spectra before and after model correction at four locations on the buoy. Figure 20 Comparison of wave spectra before and after model correction at 5 locations for the buoy;

[0145] Table 2 Wave spectra before and after model correction at different buoy positions ( Precision comparison

[0146]

[0147] After model correction, The MAPE is no higher than 12.93%; The MAPE is not higher than 10.50%; The MAPE is no higher than 24.96%. Specific results are shown in Table 3. The changes in spectral parameters over time are detailed below. Figures 21-23 As shown. Figure 21 The wave height parameters before and after model correction are compared with the observed values ​​at different buoy locations. Figure 22 The results show a comparison between the spectral peak frequency parameters and observed values ​​before and after model correction at different buoy locations. Figure 23 The results show the comparison between the peak spectral parameters and observed values ​​before and after model correction at different buoy locations.

[0148] Table 3 Comparison of wave spectrum parameters accuracy before and after model correction at different buoy locations.

[0149]

[0150] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by those skilled in the art within the scope of the technology disclosed in the present invention and within the spirit and principles of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A neural network-based intelligent sea wave spectrum revision method, characterized in that, The method comprises the following steps: S1, target sea area setting and environment data acquisition; S2, numerical calculation of sea wave spectrum in the target sea area by a sea wave numerical model; S3, constructing a sea wave spectrum correction model comprising a numerical calculation module, a measured processing module and an intelligent correction module by a deep learning method; for the sea wave spectrum calculated by the sea wave numerical model, constructing numerical simulation sea wave spectrum data by the numerical calculation module, constructing buoy observation sea wave spectrum data by the measured processing module, and constructing the deviation between the numerical simulation sea wave spectrum data and the buoy observation sea wave spectrum data by the intelligent correction module based on the buoy observation sea wave spectrum data as a benchmark, and outputting the corrected sea wave spectrum; S4, precision verification and adaptability evaluation of the constructed sea wave spectrum correction model; In step S2, the numerical calculation of the sea wave spectrum in the target sea area by the sea wave numerical model comprises: S201, selecting a model calculation domain according to the range of the research sea area, fully considering the sea wave transmission characteristics to set the model calculation boundary field, so as to realize high-precision calculation of the sea wave; in the calculation process, a large calculation domain is set based on the target sea area for nested calculation; the large calculation domain is set as a calculation domain which is 20° larger than the target sea area in the east, west, south and north directions, thereby providing boundary field information for the calculation of the target sea area by the model; S202, generating a model calculation forcing field by using the acquired environment data; calculating the forcing field in the sea area according to the wind speed information at 10 meters above the sea surface in the east-west and south-north directions, and interpolating the wind field data according to the model calculation grid to obtain forcing field data consistent with the time and spatial dimension step length set in the calculation process; S203, setting the numerical discretization method in the sea wave numerical model calculation process, including frequency, wave direction spatial discretization and global time step time discretization; completing the numerical solution of the sea wave spectrum in the sea area, and outputting the sea wave height at the target position; In step S3, the numerical calculation module processes the spectrum frequency by linear interpolation for the sea wave spectrum simulated by the sea wave numerical model, and constructs the numerical simulation sea wave spectrum data consistent with the spectrum frequency of the buoy observation sea wave spectrum; In step S3, the measured processing module smoothes the abnormal fluctuation position in the original observation data by the Gaussian filtering method for the buoy observation sea wave spectrum, and retains the dominant frequency and energy distribution characteristics of the spectrum shape by the peak value recovery method.

2. The neural network-based sea wave spectrum intelligent correction method according to claim 1, characterized in that, In step S1, the target sea area setting and environment data acquisition comprises: S101, setting the target sea area according to the demand; S102, acquiring the environment data in the sea area, including two parts of buoy data and open source environment field data in the target sea area; the buoy data includes buoy position, observation time range, sea wave spectrum data and sea wave height; the open source environment field data includes wind speed field and water depth information in the target sea area.

3. The neural network-based sea wave spectrum intelligent correction method according to claim 1, characterized in that, In step S3, the linear interpolation calculation method is as follows: ； wherein is the wave spectrum, is the buoy observed frequency, is the wave model simulated spectrum frequency, is the wave spectrum at the buoy observed frequency, is the wave model simulated spectrum at a certain frequency, is the wave model simulated spectrum at an adjacent spectrum frequency; After frequency unification, the numerical simulation sea wave spectrum data is obtained .

4. The neural network-based sea wave spectrum intelligent correction method according to claim 3, characterized in that, For the problem of missing data and data loss in the buoy data, the time series of the sea wave spectrum is processed with reference to the measured data.

5. The neural network-based intelligent sea wave spectrum revision method according to claim 1, characterized in that, In step S3, the specific calculation method is as follows: ； ； ； ； In the formula, For Gaussian kernel, This is the result after Gaussian filtering. This is the result after peak recovery. For the calculation process of the numerical function, In order to be in At the index position, with The original observation results for the sliding window offset; for The Gaussian kernel calculation results at the location, The peak value of the function. The peak value of the function after Gaussian filtering. For peak index, The peak value in the original signal. The standard deviation of the Gaussian kernel. As the independent variable, Let the radius be the Gaussian kernel. This is the offset of the sliding window. The weighting coefficients for peak recovery. The output signal after Gaussian filtering. For index position, The original signal, This is the final output result after peak recovery; Smooth and high-fidelity buoy observed sea spectrum data after filtering processing of actual measurement processing .

6. The neural network-based sea wave spectrum intelligent revision method according to claim 5, characterized in that, The data missing period caused by extreme weather and equipment failure is removed.

7. The neural network-based sea wave spectrum intelligent revision method according to claim 1, characterized in that, In step S3, the intelligent correction module corrects the numerical simulation wave spectrum data constructed by the numerical calculation module For input, the buoy observation wave spectrum data constructed by the measured processing module For output; And The data dimensions are both 44x6324; 44 represents the frequency number, and 6324 represents the time number; wherein, the input and output layers both contain 44 neurons, indicating the wave spectrum results corresponding to 44 frequencies at a single time. The neural network has three fully connected hidden layers with 200, 200 and 100 neurons respectively, and all the hidden layers use ReLU activation function to enhance the non-linear representation ability. The output layer is linearly activated. During the training process, the input data is batched first, and then read and fed into the network for forward propagation at each time step. The corrected sea wave spectrum is output ; The mean square error (MSE) is used as the loss function to measure the difference between the output and the observed spectrum. During the optimization process, the Adam optimizer is used to adaptively adjust the learning rate, and finally the sea wave spectrum is corrected.

8. The neural network-based sea wave spectrum intelligent revision method according to claim 7, characterized in that, The loss function is specifically calculated as follows: ； In the formula, is a loss function, is the number of samples, are the true value and the model prediction value, respectively. 9.A neural network-based intelligent sea wave spectrum revision system, characterized in that, The system implements the intelligent sea wave spectrum correction method based on the neural network according to any one of claims 1-8, and the system comprises: A data acquisition module is configured to set a target sea area and acquire environmental data. A sea wave spectrum numerical calculation module is configured to perform numerical calculation on the sea wave spectrum in the target sea area through a sea wave numerical model. A sea wave spectrum correction model construction module is configured to construct a sea wave spectrum correction model including a numerical calculation module, a measured data processing module and an intelligent correction module through a deep learning method. For the sea wave spectrum calculated by the sea wave numerical model, the numerical calculation module is used to construct the numerical simulation sea wave spectrum data, the measured data processing module is used to construct the buoy observed sea wave spectrum data, and the intelligent correction module is used to construct the deviation between the numerical simulation sea wave spectrum data and the buoy observed sea wave spectrum data, and the corrected sea wave spectrum is output. A model verification and evaluation module is configured to verify the accuracy and evaluate the adaptability of the constructed sea wave spectrum correction model.

Images