A method and system for comprehensive evaluation of offshore sea conditions

Abstract

The application discloses a comprehensive evaluation method and system for offshore sea conditions, and relates to the technical field of ocean hydrology and ocean wind energy resource utilization. The method comprises the following steps: acquiring initial data sets of a wind field, a coastline, water depth and a tidal level of a sea area to be researched, and generating a prediction data set through a numerical model of bidirectional coupling of hydrodynamic force and waves; iteratively adjusting model parameters by adopting a layered hyperparameter optimization strategy, and combining measured data to control the relative error within 5%; generating extreme working condition data by fusing a typhoon field and a background wind field, and comprehensively analyzing characteristic parameters of tides, waves, tidal currents and storm surges of prediction data output by the model to generate a sea condition comprehensive evaluation report. The application improves the model parameter calibration efficiency and simulation accuracy, reduces the survey cost of offshore sea condition evaluation, and can be widely applied to site selection and feasibility analysis of offshore engineering such as offshore wind power and port construction.

Description

Technical Field

[0001] This invention belongs to the field of marine hydrology and marine wind energy resource utilization technology, and particularly relates to a comprehensive assessment method and system for nearshore sea conditions. Background Technology

[0002] Coastal areas have high annual total solar radiation, and the sea surface reflection effect can improve the power generation efficiency of photovoltaic panels. Tidal energy technology has large exploitable potential, mainly concentrated in areas with significant tidal ranges. These energy sources have advantages such as zero pollution emissions, renewable resources, and large-scale development, making them an important direction for energy structure transformation. Offshore wind, solar, and tidal energy are abundant, and their pollution-free development and high utilization rates have garnered attention and support from various industries. However, the site selection for offshore wind farms and photovoltaic farms must consider the sea conditions of the area. Failure to effectively assess local sea conditions can lead to increased costs, low efficiency, and in severe cases, structural damage to marine engineering.

[0003] Currently, the collection of nearshore hydrological data mainly relies on two technical approaches: in-situ observation and numerical simulation. In current engineering practice, it is typically necessary to establish local hydrological monitoring stations and conduct continuous observations for decades to obtain marine hydrological data. Acquiring such hydrological data often requires significant investment of human and material resources, resulting in high data costs and long cycles, making it difficult to meet the needs of relevant departments for rapid and efficient assessment of hydrological information. The varying sensitivities of different types of energy development to sea state parameters have not been specifically monitored. For example, photovoltaic farms require attention to water transparency, but existing hydrological stations generally lack relevant sensors, leading to missing input parameters for assessment models. Wind farm turbine location layout relies on accurate wind resource maps; sparse monitoring points cause biases in power curve predictions. Photovoltaic farms suffer from inaccurate calculations of floating body displacement due to ocean currents, requiring additional spacing and reducing utilization. Tidal power plants experience prolonged inefficient turbine rotor operation due to insufficient monitoring of tidal current velocity. Taking offshore wind power as an example, a single monitoring station needs to deploy equipment such as current meters, wave height meters, and temperature, salinity, and depth meters, and overcome technical challenges such as corrosion and ocean current impact. The construction cost is high, and the equipment operation and maintenance requires professional vessels to calibrate it regularly. The maintenance cost of a single deep-sea monitoring station can reach hundreds of thousands of yuan. The contradiction between the economy and timeliness of data acquisition is prominent. In addition, the current hydrological data measurement points are scattered, usually tens of kilometers away from the assessment area, or even further away, which cannot fully reflect the sea conditions of the area to be assessed. The lack of data leads to excessive errors in the nearshore sea condition assessment.

[0004] Numerical simulation technology is an important technical means to solve the problems of high cost and long cycle of in-situ observation. Regarding numerical simulation of nearshore sea conditions, a number of related patent applications have been published in the existing technology.

[0005] For example, patent application CN105824993A uses a hydrodynamic and wave coupled model to numerically simulate nearshore tides, flow fields, and wave processes, and combines measured data to determine model parameters. However, this type of method usually relies on manual experience in parameter adjustment, resulting in low parameter adjustment efficiency and insufficient consistency of results. Furthermore, there is a lack of systematic simulation methods under extreme sea conditions.

[0006] Regarding extreme sea state simulation, patent application CN110287504A simulates based on design waves and historical typhoon processes, but it still has shortcomings in terms of multi-factor coupling and dynamic response. Specifically, it does not adequately consider the nonlinear interactions between tides, currents, and waves, and its dynamic response capability to time-varying typhoon wind fields and boundary conditions is limited, making it difficult to accurately reflect the evolution of extreme sea states. Another patent application, CN108920877A, while introducing typhoon wind field driving, mainly focuses on a single factor and does not adequately consider the coupling effects between tides, currents, and waves. Summary of the Invention

[0007] The purpose of this invention is to provide a comprehensive and efficient method for assessing nearshore sea conditions. In the absence of data or other adverse factors, this method can quickly acquire effective data, including wave height, wave period, wavelength, tide level, and tidal current, and integrate relevant evaluation indicators to rapidly analyze the sea conditions of the area.

[0008] To achieve the above objectives, in a first aspect, the present invention provides a comprehensive assessment method for nearshore sea conditions, based on a numerical model that couples hydrodynamics and waves, comprising the following steps:

[0009] Step 1: Obtain the model grid file and initial dataset of wind field, shoreline, water depth and tide level within the coordinate range of the sea area to be studied, and generate the prediction dataset through numerical model;

[0010] Step 2: Perform error analysis on the predicted data and existing measured data in the prediction dataset, and use a hierarchical hyperparameter optimization strategy to adjust the numerical model parameters and the initial dataset until the relative error between the numerical model prediction data and the existing measured data is less than a set threshold, thus obtaining the adjusted numerical model.

[0011] Step 3: Combine historical typhoon data in the sea area to be studied with the numerical simulation module of coastal and marine hydrodynamics in the numerical model to generate new typhoon data. Input the new typhoon data into the adjusted numerical model to obtain predicted data. Analyze the characteristic parameters of water level, tide, wave, ocean current and storm surge in the predicted data to obtain evaluation results.

[0012] Furthermore, in step 1, obtaining the model mesh file and initial dataset for wind field, shoreline, water depth, and tide level within the coordinate range of the sea area to be studied includes:

[0013] Extract shoreline coordinates and generate .shp format files. Convert the generated .shp files to .xyz format files. Open the generated .xyz format files with the MIKE mesh generate block, remove duplicate points and intersecting lines, and generate the model's mesh file. Extract initial water depth files from public datasets and interpolate the initial water depth data into the model's mesh file. Obtain one year or more of horizontal and vertical average wind speed and sea-level atmospheric pressure data from public datasets. The wind field file of the hydrodynamic module of the numerical model contains wind speed and sea-level atmospheric pressure data, and the wind field file of the wave module contains wind speed data.

[0014] Import global tide level model data, match the time dimension and total simulation duration of the numerical simulation, import it into the model's grid file, and generate tide level boundary data adapted to the sea area under study; input the initial dataset and tide level boundary data into the hydrodynamic and wave bidirectional coupled numerical model, and generate predicted datasets of tide level, current, wave, and water level through numerical model calculations.

[0015] The numerical model calculations for the hydrodynamic module include: calculations of water flow mass and momentum conservation, turbulent closure simulation, wind stress driving calculation, bed resistance dissipation calculation, Coriolis effect calculation, and wave radiation stress coupling calculation; the numerical model calculations for the wave module include: wave action conservation calculation, wind energy input calculation, white cap dissipation calculation, water depth induced breakage control calculation, bottom friction dissipation calculation, and wave diffraction effect simulation.

[0016] The numerical model that couples hydrodynamics and waves is a numerical model with real-time two-way data interaction capability between hydrodynamic and wave processes. Specifically, it adopts the MIKE 21 / 3 Coupled model, Delft3D FM model, or FVCOM hydrodynamic model coupled with the SWAN wave model. The numerical model realizes data interaction between the hydrodynamic module and the wave module through online two-way coupling, and transmits water level, flow velocity, and wave radiation stress parameters in real time.

[0017] Furthermore, step 2 specifically includes the following steps:

[0018] Error analysis is performed on the predicted data and existing measured data in the prediction dataset to calculate the relative error between the two. If the relative error is ≥5%, a hierarchical hyperparameter optimization strategy is adopted to adjust the numerical model parameters and the initial dataset. If the relative error is <5%, the calibrated high-precision numerical model is obtained directly.

[0019] The execution steps of the hierarchical hyperparameter optimization strategy include: generating a parameter importance weight matrix for numerical model parameter correction through dimensionality reduction of data from a neural network autoencoder, and arranging it in descending order of weight; generating a principal component analysis weight matrix for numerical model parameter correction through dimensionality reduction of data from principal component analysis, and arranging it in descending order of weight; merging the descending-ordered principal component analysis weight matrix and the parameter importance weight matrix to obtain a comprehensive weight matrix; using the comprehensive weight matrix data as input and the relative error as output, performing hierarchical selection and computational resource reallocation through a deep learning hierarchical hyperparameter optimization strategy, combined with the parameter importance reflected by the weight matrix; and setting mandatory physical constraints during the parameter optimization process.

[0020] Repeat the above iterative steps until the relative error between the predicted dataset and the measured data is less than 5%, and obtain the calibrated high-precision numerical model.

[0021] Furthermore, the optimization process incorporates mandatory physical constraints on the lower bound of optimization priority and the feasible region of parameters as follows:

[0022] Optimize priority lower bound constraints:

[0023]

[0024] Parametric feasible region physical constraints:

[0025]

[0026] In the formula: It is a comprehensive weight matrix. Indicates the physical parameter index, Ω phy It is a set of key physical parameters; w min It is an optimization priority lower bound. , The first j The lower and upper limits of the feasible domain for each physical parameter.

[0027] Furthermore, in step 3, the numerical model automatically matches the input feature parameter values ​​with the standards in the rule base to determine the corresponding water level, tide, wave, ocean current and storm surge status, level and risk of each feature parameter; combining the physical correlation between feature parameters, and comparing the numerical model prediction data with the measured data, the accuracy of the prediction results is automatically marked.

[0028] Furthermore, in step 3, generating new typhoon data using the coastal and ocean hydrodynamic numerical simulation module in the numerical model includes:

[0029] The typhoon field and background wind field are fused using a dynamic weighting formula. The calculation formula is as follows:

[0030]

[0031]

[0032] in, V C This is the resulting hybrid wind field; V M For typhoon wind field; V Q Background wind field; r 0 represents the radius of maximum wind speed; r It is the distance from the calculation point to the center of the typhoon. e w These are the weighting coefficients.

[0033] Secondly, this invention provides a comprehensive assessment system for nearshore sea conditions, including a data acquisition module, a model calculation module, a parameter optimization and calibration module, and a sea condition assessment output module.

[0034] The output of the data acquisition module is connected to the input of the model calculation module. The output of the model calculation module is connected to the input of the parameter optimization and calibration module and the sea state assessment output module, respectively. The output of the parameter optimization and calibration module is connected to the input of the model calculation module.

[0035] The data acquisition module is used to acquire initial data on wind field, shoreline, water depth, and tide level of the sea area to be studied, complete shoreline data format conversion, deduplication and deintersecting processing to generate model mesh file, interpolate water depth to model mesh file, acquire at least one year of continuous horizontal and vertical average wind speed and sea level atmospheric pressure data to construct initial dataset; at the same time, import global tide level model data to generate tide level boundary data of the sea area to be studied.

[0036] The model calculation module is configured to run a two-way coupled numerical model of hydrodynamics and waves. It is used to receive the initial dataset and tidal boundary data, and generate a prediction dataset through two-way coupled calculations of the hydrodynamic module and the wave module. It is also used to receive the optimized model parameters output by the parameter optimization and calibration module and the typhoon wind field data, and run the model to obtain the full-element prediction data of the typhoon condition. The hydrodynamic module covers the functions of water flow mass and momentum conservation, turbulence closure, wind stress drive, bed resistance dissipation, Coriolis effect, and wave radiation stress coupling calculation. The wave module covers the functions of wave action conservation, wind energy input, white cap dissipation, water depth induced breakage control, bottom friction dissipation, and wave diffraction effect simulation.

[0037] The parameter optimization and calibration module receives the predicted dataset output by the model calculation module, performs error analysis between the predicted dataset and existing measured data of the sea area under study, and calculates the relative error. When the relative error is ≥5%, hierarchical hyperparameter optimization is performed, and a parameter importance weight matrix is ​​generated through a neural network autoencoder. Principal component analysis is used to generate a principal component analysis weight matrix. The combined weight matrix is ​​obtained. Using the comprehensive weight matrix as input and the relative error as output, hierarchical filtering and computational resource reallocation are performed according to the parameter weights. At the same time, mandatory physical constraints of the lower limit of optimization priority and the feasible region of parameters are added during the optimization process. The optimized model parameters are fed back to the model calculation module until the relative error between the predicted dataset and the measured data is <5%, and the model calibration is completed.

[0038] The sea state assessment output module has a built-in sea state assessment rule base. It combines historical typhoon data of the sea area under study with dynamic weighting formulas to fuse typhoon wind fields and background wind fields to generate typhoon operational wind field data, which is then input into the calibrated model calculation module. It receives the typhoon operational condition full-element prediction data output from the model calculation module, performs feature parameter analysis on it across four dimensions: tide, wave, tidal current, and storm surge. The feature parameters are then matched with the sea state assessment rule base to generate and output a comprehensive nearshore sea state assessment that includes tidal attributes, wave element characteristics, tidal current characteristics, and storm surge risk level of the sea area under study. The evaluation report specifically includes: determining the precise conversion relationships between tidal attributes, theoretical minimum tide level, and mean sea level of the sea area under study based on the results of tidal harmonic analysis of the site area; drawing a full-year wave rose diagram based on numerical simulation results, statistically analyzing the interannual and seasonal frequency of different wave classes and directions, and analyzing the influence of wave refraction, diffraction, breaking characteristics, and wave-current interaction; determining the tidal attributes and tidal movement patterns based on the results of tidal harmonic analysis; and outputting the statistical patterns of historical storm surge increases in the site area, the differences in storm surge characteristics of typhoons with different paths, the annual maximum storm surge value, and the seasonal distribution characteristics of storm surge increases.

[0039] Thirdly, the present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the aforementioned comprehensive assessment method for nearshore sea conditions.

[0040] Fourthly, this invention provides a method for site selection of nearshore engineering projects, which uses the output results of the aforementioned comprehensive evaluation method of nearshore sea conditions as the basis for engineering feasibility analysis, including:

[0041] Verification of theoretical depth benchmarks for offshore wind farms or photovoltaic sites;

[0042] Assessment of wave loads and current velocity loads on offshore wind turbine foundations;

[0043] Analysis of the adaptability of tidal range and current velocity in port construction areas.

[0044] Compared with the prior art, the present invention has at least the following beneficial effects:

[0045] This invention employs a hierarchical hyperparameter optimization strategy, a dynamic fusion method for typhoon wind fields, and a multi-dimensional comprehensive sea state evaluation system. The method described in this invention does not depend on a specific numerical model type or solution method; it only requires that the numerical model used has the ability to interact with data between hydrodynamic processes and wave processes to realize the method of this invention.

[0046] This invention addresses the challenges of acquiring fundamental data for hydrodynamic models and the low efficiency of parameter calibration in the fields of marine hydrology and marine wind energy resource utilization. By constructing a weight matrix through principal component analysis and an autoencoder, combined with a deep learning hierarchical hyperparameter optimization strategy, it significantly improves the parameter calibration efficiency and model prediction accuracy of nearshore hydrodynamic models. Furthermore, by utilizing existing software toolkits, it significantly improves the efficiency and accuracy of constructing files related to tidal currents, tidal harmonic analysis and prediction, and storm surges.

[0047] Based on the prediction results file, and taking into account the relationship between the initial dataset, model parameters, and the results file, the method of data dimensionality reduction is used to prioritize the adjustment of parameters with relatively higher weights. Deep learning is then used to accelerate parameter determination and increase the accuracy of the prediction results. Optimization in parameter calibration has substantially improved the accuracy of the acquired wind field, water depth, and tidal current data, providing a solid technical foundation for constructing high-confidence long-term environmental change time series data for the sea area under study.

[0048] Furthermore, the numerical model may be a model with wave-current coupling capability, including but not limited to data interaction models implemented using different grid forms, discretization methods or coupling mechanisms.

[0049] Furthermore, by converting shoreline coordinates to .shp format files and using the DHI.Shape to XYZ tool to generate .xyz format files, compatibility between geographic information data and MIKE modeling software is achieved. Subsequently, duplicate points and intersection lines are removed to eliminate data redundancy and conflicts, generating a high-precision model mesh file without logical errors, ensuring the accuracy of the modeling spatial boundaries, and completing the standardization of shoreline data format and model mesh adaptation. Numerical model interpolation is used to achieve accurate mapping of initial water depth data to the model mesh. At the same time, long-term series of wind speed and sea-level atmospheric pressure data are obtained from public datasets, and multi-source basic data are integrated and model embedded to ensure the integrity and timeliness of input parameters.

[0050] Furthermore, mature MIKE series tools are used to obtain tidal, hydrodynamic, and wave data to ensure the reliability of data sources. Coupled models are used to accurately simulate the interaction between hydrodynamics and waves, thereby improving the accuracy and physical consistency of the predicted data.

[0051] Furthermore, by importing global tidal model data and matching it with the time dimension, a high-precision local tidal boundary adapted to the sea area under study is generated, solving the problems of missing and insufficient accuracy of measured tidal boundary data in local nearshore sea areas, and providing reliable boundary conditions for regional hydrodynamic calculations. The hydrodynamic module fully covers the physical processes of water flow conservation, turbulent closure, wind stress, bed resistance, Coriolis force, and wave radiation stress, and fully restores the core driving and dissipation mechanisms of nearshore water flow, solving the problem of distortion in the simulation of complex nearshore wave-current interactions caused by the simplification of physical processes in traditional models. The wave module fully covers the entire wave evolution calculation process of wave action conservation, wind energy input, white cap dissipation, water depth breaking, bottom friction, and diffraction effects, which helps to accurately reproduce the full life cycle law of wave propagation, deformation, refraction, and breaking in nearshore shallow water areas, and solves the problem of insufficient wave simulation accuracy under complex nearshore topography. Incorporating wave radiation stress into the hydrodynamic module calculation realizes the two-way coupling of hydrodynamics and wave modules, restores the physical processes of nearshore wave-current interaction, and can improve the accuracy of the collaborative simulation of water flow and waves under complex sea conditions.

[0052] Furthermore, based on parameter weighting, hierarchical hyperparameter optimization and computational resource reallocation are achieved. More optimization resources are allocated to high-weight core parameters, while the optimization process is simplified for low-weight parameters. While ensuring the accuracy of parameter optimization, the computational and time costs of parameter optimization are significantly reduced, and the efficiency of model calibration is improved. A data-driven weighting mechanism is introduced for model parameter calibration, which solves the problems of traditional manual parameter tuning relying on engineer experience, strong subjectivity, and poor reproducibility. This helps to achieve automation, standardization, and reproducibility of model parameter calibration.

[0053] Furthermore, by optimizing the lower limit constraint of priority, the minimum optimization priority of key physical parameters of marine hydrodynamics is ensured, avoiding the problem of core physical parameters being marginalized and under-optimized in the pure data-driven optimization process; the upper and lower limits of physical parameters are clarified to solve the overfitting problem that is prone to occur in pure data-driven optimization, where parameters exceed the physical reasonable range and the model loses physical meaning, ensuring that the optimized parameters always conform to the basic physical laws of marine hydrodynamics.

[0054] Furthermore, the built-in standardized rule base enables automatic matching of sea state characteristic parameters with assessment standards, automatically determining sea state status, level, and risk. This solves the problems of traditional sea state assessments relying on manual interpretation, low efficiency, and inconsistent assessment standards. By combining the physical correlation between characteristic parameters for comprehensive judgment, it avoids assessment bias caused by independent interpretation of single indicators, restores the coupling influence mechanism between multiple elements of nearshore sea state, and improves the accuracy of sea state level and risk assessment. Based on the comparison of predicted data and measured data, it automatically labels the accuracy of assessment results, completes the self-verification of the quality of assessment results, and provides a reliable accuracy reference for the engineering application of assessment results.

[0055] Furthermore, a dynamic weighting formula is adopted to achieve smooth fusion of typhoon wind field and background wind field. The weights change dynamically with the distance from the calculation point to the typhoon center, solving the problems of unnatural transition between the near-field and far-field wind fields of the typhoon center and distortion of the far-field wind field simulation in traditional typhoon wind field simulation. The weighting coefficients are dynamically adjusted based on the maximum wind speed radius and the distance from the calculation point to the typhoon center to restore the spatial distribution characteristics of the typhoon wind field. The high-precision fused typhoon wind field data provides a reliable driving boundary for the simulation of extreme sea states such as water level, waves, and storm surge under typhoon conditions, improves the prediction and assessment accuracy of nearshore sea states under typhoon conditions, and enhances the risk assessment capability of this method for typhoon-induced marine disasters.

[0056] The nearshore sea condition comprehensive evaluation report generated based on the method described in this application can be applied to the pre-feasibility study and feasibility study stages of marine and nearshore engineering projects such as ports, offshore wind power, and offshore photovoltaics, providing theoretical basis and data support for project site selection and scheme design. Attached Figure Description

[0057] Figure 1 This is a flowchart of the technology of the present invention.

[0058] Figure 2 This is a flowchart of the automated parameter tuning process for the numerical model of this invention.

[0059] Figure 3 This is a schematic diagram of the deep learning hierarchical hyperparameter optimization strategy in this invention. Detailed Implementation

[0060] 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, not all, of the embodiments of the present invention. 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.

[0061] This invention employs a hierarchical hyperparameter optimization strategy: by constructing a weight matrix through principal component analysis and an autoencoder, and combining it with deep learning to optimize key physical parameters, it significantly improves the model's computational efficiency and accuracy; it uses dynamic weighting to fuse typhoon fields and background wind fields, improving the accuracy of storm surge simulation; and it employs multi-element collaborative analysis: integrating tidal type, tidal strength type, wave level and sea state level, tidal asymmetry, wave propagation type, wave breaking type, tidal current type classification, tidal current movement form, tidal current strength type, and storm surge disaster level analysis, to achieve multi-dimensional evaluation of sea conditions; it is applicable to engineering scenarios such as offshore wind power site selection, port construction, and marine ranching planning, and can significantly reduce on-site survey costs and improve the reliability of marine environmental risk assessment.

[0062] like Figure 1 As shown, a comprehensive and efficient method for assessing nearshore sea conditions includes the following steps:

[0063] Step 1: Provide initial data on wind field, coastline, water depth, and tide level for the sea area to be studied. Complete the format conversion of the coastline data, deduplication and deintersecting the data to generate a model mesh file. Interpolate the water depth into the model mesh file. Obtain at least one year of continuous horizontal and vertical average wind speed and sea-level atmospheric pressure data to construct the initial dataset. At the same time, import the FES2022 global tide level model data to generate the tide level boundary data for the sea area to be studied.

[0064] Step 2: Perform error analysis on the predicted data and existing data in the prediction dataset, and use a hierarchical hyperparameter optimization strategy to adjust the numerical model parameters and the initial dataset until the relative error between the numerical model prediction data and the existing data is less than 5%, thus obtaining the adjusted numerical model; the existing data includes publicly available wind field, wave, water level, water depth, and shoreline data, as well as real-time monitored wind field, wave, water level, water depth, and shoreline data;

[0065] Step 3: Combine historical typhoon data in the sea area to be studied with the numerical simulation module of coastal and marine hydrodynamics in the numerical model to generate new typhoon data. Input the new typhoon data into the adjusted numerical model to obtain predicted data. Analyze the characteristic parameters of water level, tide, wave, ocean current and storm surge in the predicted data to obtain evaluation results.

[0066] refer to Figure 2 In step 1, after determining the latitude and longitude of the sea area to be studied, wind field, shoreline, and water depth data are obtained from the database based on the latitude and longitude matching of the sea area to be studied. The wind field, shoreline, and water depth data are pre-cleaned, missing data is supplemented, and erroneous data is deleted.

[0067] As an example: Using GEODAS-NG (Geophysical Data System-Next Generation) software, initial water depth files are extracted from the publicly available Cmap dataset, and the initial water depth data is interpolated into the model's mesh file as the initial conditions for the model. Generally, the water depth data available from publicly available databases has low accuracy. Obtaining high-precision water depth files from nautical charts can significantly shorten the adjustment time for numerical model parameters in step 2. The shoreline coordinates of the sea area under study are extracted from the GEODAS-NG software and generated as .shp files. These .shp files are then converted to .xyz files. The generated .xyz files can be directly opened using the MIKE mesh generate block. After removing duplicate points and intersecting lines, the model mesh file is generated.

[0068] Obtain horizontal and vertical average wind speed and sea-level atmospheric pressure data for one year or more from NCEP or ECMWF public datasets. Select NetCDF4 data format and use the MIKEIO library in Python to convert the .nc format files to .dfs2 format files that MIKE can directly read. This data can then be used in the MIKE 21 / 3 Coupled model. In the MIKE 21 / 3 Coupled model, the wind field file in the hydrodynamic module includes wind speed and sea-level atmospheric pressure data files, while the wave module includes wind speed files.

[0069] In the numerical model (MIKE 21 / 3 Coupled model), the water level and tidal current conditions in the wave module are provided by the hydrodynamic module, and the wave stress in the hydrodynamic module is provided by the wave module; the wave module and the hydrodynamic module interact in real time and transmit calculation parameters synchronously.

[0070] In step 1, the tidal level and tidal current boundary data required for the model of the sea area under study are generated based on the FES2022 global high-resolution tidal database. The accuracy of its tidal level and tidal current simulation has been fully verified in a large number of previous studies and engineering practices worldwide. Import the FES2022 global tidal level model data, match the time dimension and total simulation duration of the numerical simulation, import the completed model mesh file, and generate boundary tidal level prediction data adapted to the sea area under study.

[0071] The predicted dataset generated using numerical models includes: water surface elevation, still water depth, and total water depth. x Towards flow velocity, y Flow velocity, flow flux, velocity magnitude, flow direction, significant wave height, maximum wave height, peak period, wave period, peak wave direction, average wave direction, standard deviation of wave direction, wave speed, water particle velocity, wave power, wave direction frequency spectrum, wave radiation stress tensor, storm surge increase or decrease, extreme water level during typhoon process, tidal amplitude and lag angle.

[0072] When generating the model's grid file, the model's grid dispersion is determined based on the research area, existing data accuracy, step size, dispersion, and computational time and space complexity, and the optimal parameter combination is selected.

[0073] Tidal data is extracted from publicly available global tidal databases, such as FES2022, and calculated as follows:

[0074] Cosine wave generation formula:

[0075] (1)

[0076] In the formula: x ( t )for tThe forecast tidal value at any given time (tidal level in meters); tidal current velocity in meters per second, providing the final tidal calculation output; This refers to the tidal fraction number, which corresponds to the independent tidal fraction driven by different astronomical tidal forces; N The total number of tidal constituents used in the calculation is dimensionless, and 13 constituents are selected, including Sa, Ssa, O1, K1, P1, Q1, M2, S2, N2, K2, M4, MS4, and M6. for t Time of the first Individual tidal node correction factors; For the first The average amplitude of each tidal constituent (tidal level in m, tidal velocity in m / s). t The target forecast time is in seconds, and the independent time variable is used for tide calculation. t 0 represents the starting reference time (in seconds) for harmonic analysis, and represents the time reference constant for fitting the measured sequence. For the first The angular velocity of each tidal constituent (unit: rad / s) is determined by the astronomical tidal force period corresponding to the tidal constituent. It is a fixed constant and its standard value can be calculated by the Doudson number of the tidal constituent. for t Time of the first The astronomical initial phase angle correction (in rad) for each tidal constituent changes dynamically with the calculation time, and is related to the nodal correction factor. The corresponding calculations are determined by the astronomical parameters of the celestial body's orbit. Reference time t 0 corresponds to the first The astronomical initial phase angle correction for each tidal constituent (unit: rad). For the first The time delay angle for each tidal zone (unit: rad); Reference time t 0 corresponds to the first The nodal correction factors for each tidal fraction are dimensionless and are fixed computational constants.

[0077] The MIKE 21 / 3 Coupled model couples hydrodynamic (HD) and wave interaction (SW). It can accurately simulate the hydrodynamic and wave evolution processes in coastal and estuarine areas. Its core governing equations for the hydrodynamic module are as follows:

[0078] The continuity calculation equation is:

[0079] (2)

[0080] In the formula: hIt is the total water depth from the bottom to the free surface. h=h 0 +n ,in h 0 represents the still water depth in the calculation area. or The water level is relative to the reference surface, i.e., the depth of the water body within the calculation area (unit: m). t It is time (unit: s); S It is the source-sink intensity per unit area (unit: m / s). yes x Average flow velocity per unit water depth (unit: m / s); yes y Average flow velocity per unit water depth (unit: m / s); x It is the direction of horizontal water flow; y It refers to the longitudinal direction of water flow.

[0081] x Equation for calculating directional momentum:

[0082] (3)

[0083] In the formula: or It is the water level relative to the reference surface (unit: m); g It is the acceleration due to gravity (unit: m / s²), taken as 9.81 m / s²; It is the Coriolis force parameter (unit: 1 / s). middle oh The Earth's rotational angular velocity (standard value 7.292 × 10⁻⁶) 5 rad / s f The geographical latitude of the calculation area; r It is the density of water (unit: kg / m³). r 0 is the reference density of water (unit: kg / m³), which is the density of water under standard conditions (usually taken as 1000 kg / m³). p a Atmospheric pressure (unit: Pa); t sx It is the wind stress on the liquid surface. x Component (unit: Pa); t bx It is the frictional stress of the bottom bed surface. x Component (unit: Pa); s xx It is the wave radiation stress tensor x Component (unit: Pa); s xy It is the wave radiation stress tensor xyz Component (unit: Pa); T xxIt is the normal Reynolds stress x Component (unit: Pa); T xy It is the normal Reynolds stress xyz Component (unit: Pa); u s It is Yuanhui x Component input velocity (unit: m / s); yes y direction and x Vertical average value of the product of directional flow velocities (unit: m / s). yes x The average of the squares of the directional flow velocities (unit: m / s).

[0084] y Equation for calculating directional momentum:

[0085] (4)

[0086] In the formula: t sy It is the wind stress on the liquid surface. y Component (unit: Pa), and t sx ( x Corresponding to the directional wind stress component; t by It is the frictional stress of the bottom bed surface. y Component (unit: Pa), and t bx ( x Corresponding to the directional bed surface friction stress component; s yx It is the wave radiation stress tensor x Component (unit: Pa), and s xy (Wave radiation stress tensor) xyz (Components) Corresponding to; s yy It is the wave radiation stress tensor y Component (unit: Pa), and s xx (Wave radiation stress tensor) x (Components) Corresponding to; T xy It is the normal Reynolds stress xyz Quantity (unit: Pa), as mentioned above x In the directional momentum equation T xy The meaning is consistent; T yy It is the normal Reynolds stress y Component (unit: Pa), andT xx (Normal Reynolds stress) x (Components) Corresponding to; v s It is Yuanhui y Component input velocity (unit: m / s), and u s (Yuanhui) x Corresponding to the component input flow rate, yes y Average value of the square of the directional flow velocity (unit: m / s).

[0087] The equation for calculating the turbulent viscosity coefficient is:

[0088] (5)

[0089] In the formula: A It is the turbulent viscosity coefficient (also known as the eddy viscosity coefficient, unit: m² / s); C μ The common value for the empirical constant in the turbulence model is 0.09; k It is turbulent kinetic energy (unit: m² / s²). e It is the turbulent kinetic energy dissipation rate (unit: m² / s³).

[0090] The equation for calculating turbulent kinetic energy transport is:

[0091] (6)

[0092] In the formula: s k It is the Prandtl number of turbulent kinetic energy, and its common value is 1.0; P k It is the turbulent kinetic energy generation term (unit: m² / s³). It is the total turbulent kinetic energy per unit area and per unit density (unit: m³ / s²). 3 / s 2 ); yes x Turbulent kinetic energy advection flux in the direction of flow (unit: m) 4 / s 2 ); yes y Turbulent kinetic energy advection flux in the direction of flow (unit: m) 4 / s 2 ).

[0093] Turbulent kinetic energy dissipation rate e The transport calculation equation is as follows:

[0094] (7)

[0095] In the formula: s ε It is the Prandtl number of turbulent flow, which is the rate of turbulent kinetic energy dissipation. The common value is 1.3. C ε1 、C ε2 These are empirical constants for the turbulence model, and their common values ​​are as follows: C ε1 =1.44、 C ε2 =1.92, It is the total turbulent dissipation rate per unit area and per unit density (unit: m). 3 / s 3 ), yes x Turbulent kinetic energy dissipation rate and advection transport flux in the direction of flow (unit: m³) 4 / s 3 ), yes y Turbulent kinetic energy dissipation rate and advection transport flux in the direction of flow (unit: m³) 4 / s 3 ).

[0096] Turbulent kinetic energy generation term P k The calculation formula is:

[0097] (8)

[0098] The specific meanings and physical significance of each parameter in the formula are consistent with the meanings of the same parameters in the previous formula, and will not be explained again.

[0099] The formula for calculating the Reynolds stress tensor eddy cohesion closure is:

[0100] (9)

[0101] In the formula: It is the Reynolds stress tensor component (unit: Pa). Corresponding to x、y direction; , They are respectively The average velocity of water depth in the direction corresponding to x Direction u , y Direction v ; 、 They are respectively Spatial coordinates of direction, corresponding x , y direction; It is the Kronecker function, which is a dimensionless mathematical parameter. When the value is 1, The value is 0.

[0102] The formula for calculating wind stress vector is:

[0103] (10)

[0104] In the formula: t sf This is the wind stress vector (unit: Pa); r a Air density (unit: kg / m³) 3 Under standard conditions, the value is 1.225 kg / m³. 3 ; U 10 It is the wind speed vector (unit: m / s) at a height of 10m above the sea surface. U 10 | represents the wind speed modulus. U 10,x 、U 10,y for x , y Directional wind speed component; C f It is the wind stress coefficient (dimensionless).

[0105] The formula for calculating the wind stress coefficient in segments is as follows:

[0106] (11)

[0107] In the formula: c a This is the wind stress coefficient for low wind speeds, with a commonly used value of 1.225 × 10⁻⁶. 3 ; c b This is the wind stress coefficient for high wind speeds, with a commonly used value of 2.425 × 10⁻⁶. 3 ; w a This is the low wind speed threshold, with a common value of 7 m / s; w b It is the high wind speed threshold, and the common value is 25m / s.

[0108] The formula for the vector formula of frictional stress on the bed surface is:

[0109] (12)

[0110] In the formula: It is the bed surface friction stress vector (unit: Pa); u It is the average current velocity vector at water depth (unit: m / s). The velocity modulus; c fb It is the coefficient of friction at the bottom. c fb Formula (13), formula (14) or formula (15) can be used for calculation.

[0111] Formula for calculating the bottom friction coefficient:

[0112] (13)

[0113] In the formula: M It is the reciprocal of Manning's coefficient (unit: m) 1 / 3 / s), the general range for offshore engineering is 10~90m. 1 / 3 / s; g This is the acceleration due to gravity, commonly taken as 9.81 m / s². 2 Manning roughness coefficient is commonly used in domestic standards. n It means that the conditions are met. M=1 / n , n The unit is s / m 1 / 3 .

[0114] Calculate the bottom friction coefficient using the Chezy formula:

[0115] (14)

[0116] In the formula: C Chezy coefficient (unit: m) 1 / 2 / s), with a general value range of 20~80m. 1 / 2 / s, which can be converted using Manning's formula. C=h 1 / 6 / n .

[0117] The logarithmic coefficient of friction at the bottom is:

[0118] (15)

[0119] In the formula: k s It is the Kármán constant, with a common value of 0.4; k s It is the equivalent roughness height of the bed surface (unit: m).

[0120] The wave module in the MIKE 21 / 3 Coupled SW (Spectral waves, SW) model performs calculations including wave energy conservation, wave breaking control, wind-driven forces, and wave diffraction.

[0121] Wave energy conservation calculation using the wave action conservation equation:

[0122] (16)

[0123] In the formula: Wave action density (unit: s) m 2 ); t It is time (unit: s); x、y These are horizontal and vertical planar spatial coordinates; s It is the relative angular frequency of the wave (unit: rad / s), which satisfies the dispersion relation. s 2 =gκtanh(κh) ,in k Wave number; i It refers to the direction of wave propagation (unit: rad), indicating the direction of wave crest normal and... x The included angle of the axis; c σ It is the rate of change of relative angular frequency (unit: rad / s) 2 ); c θ It is the rate of change of the wave propagation direction (unit: rad / s). It is the vector form of the transport term in planar space. , c g It is the speed of the wave group. c g,x , c g,y Wave group speed at x、y The directional component (unit: m / s) refers to the propagation speed of wave energy, which determines the transport rate of wave action in planar space. S total It is the sum of wave source and sink terms (unit: m) 2 / s 3 ); D diff Wave diffraction effect source and sink terms (unit: m) 2 / s 2 ).

[0124] Wave source-sink term summation decomposition formula:

[0125] (17)

[0126] In the formula: S in It is a wind energy input item; S ds,w It is a white hat dissipation term; S ds,b It is the bottom friction dissipation term; S ds,bk It is the water depth-induced wave breaking dissipation term;S nl It is a nonlinear wave-wave interaction term; S surf It is a surface flow-induced dissipation term.

[0127] The formula for calculating wind energy input is:

[0128] (18)

[0129] In the formula: β The wind input is an empirical coefficient, set to 0.25; r a Air density (unit: kg / m³) 3 ), take 1.225kg / m 3 ; i w Wind direction (unit: rad); c It is the wave phase velocity (unit: m / s), which satisfies c=s / k ; E Wave energy density (unit: m) 3 / s 3 ).

[0130] The formula for calculating the white hat dissipation term is:

[0131] (19)

[0132] In the formula: C wc This is the white hat dissipation coefficient, taken as 4.5 × 10⁻⁶. -5 ; It is the relative angular frequency (unit: rad / s) corresponding to the peak value of the spectrum. It is the wavenumber corresponding to the peak value of the spectrum (unit: 1 / m). The wave energy density corresponding to the peak value of the spectrum (unit: m). 3 / s 2 ); It is the equilibrium wave energy density corresponding to the peak value of the PM spectrum; m It is a non-linear exponent, and its common value is 2.

[0133] The formula for calculating the bottom friction dissipation term is:

[0134] (20)

[0135] In the formula: C b It is the coefficient of frictional dissipation, with a value ranging from 0.03 to 0.1; k It is a wavenumber that satisfies the dispersion relation. s 2 =gκtanh (κh) .

[0136] Formula for calculating the dissipation term of depth-induced wave breaking:

[0137] (twenty one)

[0138] In the formula: α This is the crushing dissipation efficiency coefficient, taken as 1.0; B B It is a breakage indicator.

[0139] The conversion formula between effective wave height and wave action spectrum is:

[0140] (twenty two)

[0141] The quantitative calculation formula for wave diffraction effect is as follows:

[0142] (twenty three)

[0143] In the formula: c g satisfy c g =0.5 c (1+2 kh / sinh(2 kh )); ⊥ is the transverse gradient operator perpendicular to the main propagation direction of the wave; It is the wave action density.

[0144] The formula for calculating the wave radiation stress tensor is:

[0145] (twenty four)

[0146] All parameters in equation (24) are completely consistent with the meaning of the previous formula. The full spectrum integration is used to obtain all components of the two-dimensional wave radiation stress tensor in the plane, which are then directly substituted into equations (6)-(7) to solve the problem.

[0147] In step 2, error analysis is performed on the predicted data and existing measured data in the prediction dataset to calculate the relative error between the prediction dataset and the measured dataset. If the relative error is less than 5%, the model parameters are not adjusted and typhoon data is introduced to obtain the sea state prediction dataset. If the relative error is greater than 5%, the parameters of the prediction model are adjusted.

[0148] The hierarchical hyperparameter optimization strategy involves adjusting model parameters and the initial dataset, including:

[0149] Principal component analysis (PCA) is used to generate a PCA weight matrix of the influence of model parameters on the results, and the matrix is ​​arranged in descending order of weight. The formula for calculating the PCA weights is as follows:

[0150] PCA (principal components analysis) covariance matrix and eigenvalue decomposition formula:

[0151] (25)

[0152] In the formula: R It is the covariance matrix of the standardized input parameters. R∈R J×J ; X std It is the standardized input parameter sample matrix. X std ∈R N×J ; for X std The transpose of the matrix, It is the first of the covariance matrices. The feature values ​​are arranged in descending order. l 1 ≥ l 2 ≥ ≥λ J ; It is the first The eigenvectors corresponding to each eigenvalue. ∈R J .

[0153] The formulas for the principal component variance contribution rate and the cumulative variance contribution rate are as follows:

[0154] (26)

[0155] Usually taken as: or cum (m)≥0.95, J The total number of input parameters. It is the first The variance contribution rate of each principal component m The final number of principal components selected for PCA dimensionality reduction.

[0156] Principal component analysis weight matrix The extraction formula is:

[0157] (27)

[0158] The formula for weight normalization is:

[0159] (28)

[0160] In the formula: This generates the normalized principal component analysis weight matrix. It is the first The th eigenvector of the th feature vector j p The element represents the element. j p The parameter for the first The contribution of each principal component is used only for the calculation of the second weight matrix.

[0161] An autoencoder generates a parameter importance weight matrix to determine the impact of the initial data on the results, and the parameters are sorted in descending order of weight. The autoencoder calculation formula is as follows:

[0162] The autoencoder calculates the complete loss function:

[0163] (29)

[0164] The calculation formulas for each item are as follows:

[0165] (30)

[0166] (31)

[0167] In the formula: L AE It is a loss function; L rec It is a reconstruction loss function that measures the decoder's ability to reconstruct the original data; L reg It is the Frobenius norm regularization term, which prevents the model from overfitting; L sparse It is a sparsity constraint term. r L This is a sparsity parameter (generally taken as 0.05). It is the first i The sample in the hidden layer is the _th sample. The activation value of each neuron; l 1、 l 2 represents the regularization coefficient and the sparsity coefficient, with common values ​​of 10 and 10 respectively. 4 10 3 ; W 1 、W 2 is the weight matrix of the encoder and decoder. W 1∈R K×J , J The total number of input parameters. K This represents the number of neurons in the hidden layer. b 1 、b2 represents the bias term for the encoder and decoder; subscript F It is the Frobenius norm. For the first i A standardized original input sample, It is the decoder reconstruction of the first i A standardized sample, N 0 represents the total number of samples in the training dataset.

[0168] The activation function formulas for the encoder and decoder are:

[0169] (32)

[0170] in It is the first i The activation values ​​of the hidden layer neurons corresponding to each sample, and the ReLU and Sigmoid activation functions are:

[0171] (33)

[0172] Parameter importance weight matrix (Parameter Importance) Extraction Formula:

[0173] (34)

[0174] The formula for weight normalization is:

[0175] (35)

[0176] In the formula: This generates a normalized parameter importance weight matrix; It is the encoder weight matrix W 1st Line number j z The elements of the column are used only for the calculation of the second parameter weight set.

[0177] The parameter importance weight matrix and the principal component analysis weight matrix, sorted in descending order of weight, are merged to obtain the weight matrix data. Deep learning is then used to adjust the parameter priorities. The formula for synthesizing the weight matrix is ​​as follows:

[0178] (36)

[0179] The combined weight normalization and descending order sorting are as follows:

[0180] (37)

[0181] In the formula: For the first j The initial comprehensive weights of the parameters, It is a comprehensive weight matrix; c It is the fusion weighting coefficient, and the general value is 0.5.

[0182] In the parameter hierarchy settings, the lower limit for the optimization priority of the enforced physical constraints is determined as follows:

[0183] The formula for enforced physical constraints is as follows:

[0184] (1) Optimize the lower bound constraint of priority

[0185] (38)

[0186] (2) Physical constraints of the feasible region of parameters

[0187] (39)

[0188] In the formula: Indicates the index of key physical parameters, Ω phy It is a set of key physical parameters, including bottom roughness coefficient, wind stress drag coefficient, horizontal (vertical) eddy viscosity coefficient, wet and dry dynamic boundary parameters, wave pattern parameters, water depth datum, white cap dissipation coefficient, wave breaking coefficient, wave diffraction coefficient, and boundary reflection coefficient. w min This is the lower limit of optimization priority, generally set to 0.05, to ensure that the overall weight of key physical parameters is not less than 5%; , The first j The lower and upper limits of the feasible domain for each key physical parameter are determined based on industry standards and statistical characteristics of measured data.

[0189] Specifically, when the relative error between the predicted tide level, tidal current, and wave height data and the existing measured data exceeds 5%, a hierarchical hyperparameter optimization step is performed. For the water depth, wind field, and tidal current components, due to the large amount of data, a neural network autoencoder is used for data dimensionality reduction to generate a parameter importance weight matrix for parameter correction, which is then arranged in descending order of weight. For numerical model parameters including bottom roughness, eddy viscosity, wave breaking, diffraction parameters, and white hat parameters, principal component analysis is used for data dimensionality reduction to generate a principal component analysis weight matrix for parameter correction, which is then arranged in descending order of weight.

[0190] The parameter importance weight matrix and the principal component analysis weight matrix are merged to obtain the weight matrix data. This weight matrix data reflects the parameter importance; parameters ranked higher in the weight matrix are considered more important. This weight matrix data is used as the input parameters for a hierarchical hyperparameter optimization strategy, while the corresponding relative error is used as the output parameter. The hierarchical hyperparameter optimization strategy in deep learning is used to adjust parameter priorities, significantly improving the accuracy of parameter tuning by prioritizing the identification of important parameters and reducing training time and parameter tuning step size. The parameter tuning standard is that the relative error between predicted and measured data is less than 5%. Figure 3 The hierarchical hyperparameter optimization strategy uses parameter importance analysis for hierarchical selection and resource reallocation. For example, secondary layer parameters are allocated 5% of computational resources, and the optimization algorithm uses default values ​​or fine-tuning; sensitive layer parameters are allocated 25% of computational resources, and the optimization algorithm is random search; critical layer parameters are allocated 70% of computational resources, and the optimization algorithm uses Bayesian optimization. The computational resources mentioned in this scheme cover four core dimensions: the computing power quota for numerical model iterative simulation, the upper limit of the number of iterations of the optimization algorithm, the allocation of parallel computing cores, and the constraint of parameter search space dimensions. The resource allocation ratio strictly matches the parameter importance output by the comprehensive weight matrix. The parameter hierarchy is determined by sorting based on the normalized comprehensive weight results: core physical parameters with the top 20% cumulative weight are assigned to the critical layer, parameters with 20% to 50% cumulative weight are assigned to the sensitive layer, and low-sensitivity parameters with a cumulative weight below 50% are assigned to the secondary layer. Resource allocation is strongly linked to the contribution of parameters to the relative error of the simulation results, with computational power tilted towards core physical parameters that determine the accuracy of hydrodynamic-wave coupling simulation. The closed-loop error feedback mechanism directs computational resources to key-level parameter adjustments, effectively achieving the optimization goal of dimensionality reduction without compromising accuracy, and providing a new paradigm for ocean model parameter optimization.

[0191] Simultaneously, physical constraints are embedded in the hierarchical hyperparameter optimization strategy. Forced lower limits for the optimization priority of key parameters (such as bottom roughness) are set at the parameter level to prevent them from being overly ignored or simplified. Forced physical constraints include bottom roughness and wind-driven force.

[0192] After the numerical model parameters are determined and the prediction accuracy meets the requirements, historical typhoon data for the sea area under study are acquired. The superposition of the typhoon wind field and the original wind field is referenced from Elsberry's research, with weighting coefficients introduced to significantly improve the accuracy of the composite wind field and ensure the reliability of the typhoon's peripheral wind field and the resolution of the typhoon's central wind field. Typhoon wind field generation references the research of Young or Holland and is embedded in the MIKE toolbox. Based on existing research, weighting coefficients are introduced, calculated using the following formula:

[0193] (40)

[0194] in, , V C For the merged mixed wind field, V M For typhoon wind field; V Q Background wind field; r 0 represents the radius of maximum wind speed; r It is the distance from the calculation point to the center of the typhoon. e w These are the weighting coefficients.

[0195] In step 3, based on historical typhoon data for the sea area under study, typhoon data is generated using the MIKE 21 toolbox and run in the adjusted numerical model to obtain predicted data for the sea area under study. Characteristic parameter analysis is then performed on the tide level, tidal current, wave, and storm surge in the predicted data for the sea area under study output by the numerical model. Specifically, this includes:

[0196] (1) Based on the results of the tidal harmonic analysis of the site area, determine the conversion relationship between the tidal attributes, theoretical lowest tide level, and mean sea level of the sea area under study, including mean high tide level, mean low tide level, tidal range, duration of rising and falling tides, design high water level, design low water level, extreme high water level, extreme low water level, and hourly tidal process lines. (2) Based on the results of numerical simulation, draw the annual wave rose diagram, count the interannual and seasonal occurrence frequencies of different wave classes and directions, analyze the influence of wave refraction, diffraction, breaking characteristics, and wave-current interaction; output the 2-year, 5-year, 25-year, 50-year, and 100-year return periods and design high water level, design low water level, extreme high water level, and extreme low water level, and complete the core parameters of the design waves after the most unfavorable combination, including significant wave height, maximum wave height, spectral peak period, mean period, and design wave direction, and simultaneously supplement the monthly and seasonal design wave elements during the construction period. (3) Based on the results of the tidal harmonic analysis, determine the tidal attributes and tidal movement forms. Output a complete set of tidal characteristic parameters under normal sea conditions, including the tidal current velocity and corresponding direction of the full tidal cycle of spring tide, mid tide, and neap tide, the average tidal current velocity, the vertical velocity profile (surface, middle, and bottom layers), and the duration of the tidal cycle. Output the design current velocity of ocean currents for 25-year, 50-year, and 100-year return periods, and clarify the peak current velocity, corresponding direction, and vertical velocity distribution under extreme conditions driven by typhoons. (4) Output the statistical law of historical storm surge water increase in the site area, the differences in water increase characteristics of typhoons with different paths, the annual maximum water increase value and the seasonal distribution characteristics of water increase; the maximum storm surge water increase value in the return period (50 years, 100 years); the most unfavorable combination extreme value of water increase and astronomical spring tide, and clarify the spatial distribution characteristics of storm surge water increase, the coastal inundation range and water depth under extreme conditions.

[0197] The tidal type determination parameter is the tidal amplitude ratio, which is calculated as follows:

[0198] (41)

[0199] in, , , , Parameters for determining tidal type, corresponding to tidal amplitude. F As shown in Table 1:

[0200] Table 1. Parameters for Determining Tide Type F Corresponding tide type

[0201]

[0202] Tidal strength type identification, multi-year average tidal range ΔH As shown in Table 2.

[0203] Table 2. Multi-year average tidal range ΔH Corresponding tidal strength types

[0204]

[0205] The formula for determining tidal asymmetry is as follows:

[0206] (42)

[0207] In the formula: T f For the duration of high tide, T e The duration of low tide is shown in Table 3.

[0208] Table 3. Discrimination results corresponding to the numerical values ​​of tidal asymmetry

[0209]

[0210] Wave class and sea state classification (corresponding to the statistics of different wave classes in the original text).

[0211] Discrimination indicator: Valid wave height H 13% As shown in Table 4

[0212] Table 4 Significant Wave Height H 13% Corresponding sea state level

[0213]

[0214] Wave propagation type identification:

[0215] Deep or shallow water wave discrimination, discrimination index: relative water depth d / L , d Because of the water depth, L The wavelengths are shown in Table 5:

[0216] Table 5. Wave conditions in deep / shallow water corresponding to relative water depth

[0217]

[0218] Wave breakage type identification:

[0219] Critical breakage conditions ：H 13% / d ≥0.78, the shallow water limiting wave height-to-depth ratio, or H 13% / L ≥1 / 7, critical wave steepness for deep-water breaking. H 13% The effective wave height.

[0220] Iribalon number , The slope of the waterbed. L 0 represents the deep-water wavelength, see Table 6.

[0221] Table 6 Wave breaking types corresponding to different Iribalon numbers

[0222]

[0223] The parameter for classifying current types is the current speed ratio of the sub-currents:

[0224] (43)

[0225] in, , , For the corresponding tidal current velocity, the tidal type determination parameters are... K t As shown in Table 7:

[0226] Table 7 Parameters for Determining Trend Type K Corresponding trend types

[0227]

[0228] Trendy sports forms:

[0229] Current rotation rate k t = W min / W max

[0230] in, W max The maximum velocity along the major axis of the tidal current ellipse. W minThe minimum flow velocity along the short axis is shown in Table 8.

[0231] Table 8. Current Movement Patterns Corresponding to Different Current Rotation Rates

[0232]

[0233] Trend strength type:

[0234] Discrimination indicator: Maximum current velocity during high tides. V max See Table 9 for reference.

[0235] Table 9. Tidal current strength types corresponding to different ranges

[0236]

[0237] Furthermore, numerical prediction results are typically surface velocities, requiring model interpolation to calculate depth velocities. This application employs an exponential velocity distribution:

[0238] (44)

[0239] In the formula: U It is the cross-sectional average velocity; h It is the total water depth from the bottom to the free surface; z It refers to the depth of the water quality point.

[0240] The storm surge disaster level is usually determined by the maximum rise in water level, and the determination criteria are shown in Table 10.

[0241] Table 10 Criteria for Determining Storm Surge Disaster Levels

[0242]

[0243] The MIKE 21 / 3 Coupled FM model's hydrodynamic model automatically matches input feature parameter values ​​with standards in a rule base to determine the state, level, and risk of each feature. It performs a comprehensive analysis by incorporating the physical correlations between feature parameters (e.g., the coupling effect of waves and storm surge during typhoons). Based on a comparison between numerical model predictions and measured data, it automatically labels the accuracy of prediction results (e.g., extreme water levels, maximum wave heights). Using a pre-defined structured report template, it automatically combines and fills in the matched evaluation results, accuracy information, key feature parameter values, and their state descriptions to generate a complete and easy-to-understand regional sea state assessment report.

[0244] This application further conducts sensitivity analysis of model parameters, quantifies the sensitivity of model results to input parameters and boundary conditions, clarifies the influence weight of different parameters on simulation results, and provides a quantitative reference for subsequent parameter adjustment.

[0245] On the other hand, based on the concept of the method described in this invention, a comprehensive assessment system for nearshore sea conditions is also provided, including a data acquisition module, a model calculation module, a parameter optimization and calibration module, and a sea condition assessment output module.

[0246] The output of the data acquisition module is connected to the input of the model calculation module. The output of the model calculation module is connected to the input of the parameter optimization and calibration module and the sea state assessment output module, respectively. The output of the parameter optimization and calibration module is connected to the input of the model calculation module.

[0247] The data acquisition module is used to acquire initial data on wind field, shoreline, water depth, and tide level of the sea area to be studied, complete shoreline data format conversion, deduplication and deintersecting processing to generate model mesh file, interpolate water depth to model mesh file, acquire at least one year of continuous horizontal and vertical average wind speed and sea level atmospheric pressure data to construct initial dataset; at the same time, import global tide level model data to generate tide level boundary data of the sea area to be studied.

[0248] The model calculation module is configured to run a two-way coupled numerical model of hydrodynamics and waves. It is used to receive the initial dataset and tidal boundary data, and generate a prediction dataset through two-way coupled calculations of the hydrodynamic module and the wave module. It is also used to receive the optimized model parameters output by the parameter optimization and calibration module and the typhoon wind field data, and run the model to obtain the full-element prediction data of the typhoon condition. The hydrodynamic module covers the functions of water flow mass and momentum conservation, turbulence closure, wind stress drive, bed resistance dissipation, Coriolis effect, and wave radiation stress coupling calculation. The wave module covers the functions of wave action conservation, wind energy input, white cap dissipation, water depth induced breakage control, bottom friction dissipation, and wave diffraction effect simulation.

[0249] The parameter optimization and calibration module receives the predicted dataset output by the model calculation module, performs error analysis between the predicted dataset and existing measured data of the sea area under study, and calculates the relative error. When the relative error is ≥5%, hierarchical hyperparameter optimization is performed, and a parameter importance weight matrix is ​​generated through a neural network autoencoder. Principal component analysis is used to generate a principal component analysis weight matrix. The combined weight matrix is ​​obtained. Using the comprehensive weight matrix as input and the relative error as output, hierarchical filtering and computational resource reallocation are performed according to the parameter weights. At the same time, mandatory physical constraints of the lower limit of optimization priority and the feasible region of parameters are added during the optimization process. The optimized model parameters are fed back to the model calculation module until the relative error between the predicted dataset and the measured data is <5%, and the model calibration is completed.

[0250] The sea state assessment output module has a built-in sea state assessment rule base. It combines historical typhoon data of the sea area under study with dynamic weighting formulas to fuse typhoon wind fields and background wind fields to generate typhoon operational wind field data, which is then input into the calibrated model calculation module. It receives the typhoon operational condition full-element prediction data output from the model calculation module, performs feature parameter analysis on the typhoon operational condition full-element prediction data across four dimensions: tide, wave, tidal current, and storm surge. It matches the feature parameters with the sea state assessment rule base to generate and output nearshore data including tidal attributes, wave element characteristics, tidal current characteristics, and storm surge risk levels for the sea area under study. The comprehensive sea condition assessment report includes: determining the precise conversion relationships between tidal attributes, theoretical minimum tide level, and mean sea level of the sea area under study based on the results of tidal harmonic analysis of the site area; drawing an annual wave rose diagram based on numerical simulation results, statistically analyzing the interannual and seasonal frequency of different wave classes and directions, and analyzing the influence of wave refraction, diffraction, breaking characteristics, and wave-current interaction; determining the tidal attributes and tidal movement patterns based on the results of tidal harmonic analysis; and outputting the statistical patterns of historical storm surge increases in the site area, the differences in storm surge characteristics of typhoons with different paths, the annual maximum storm surge value, and the seasonal distribution characteristics of storm surge increases.

[0251] A computer-readable storage medium is also provided, storing a computer program that, when executed by a processor, implements the comprehensive assessment method for nearshore sea conditions described in this invention.

[0252] Computer-readable storage media can include computer storage media and communication media. Computer storage media includes volatile and non-volatile, removable and non-removable media implemented using any method or technology for storing information such as computer-readable instructions, data structures, program modules, or other data. Computer-readable storage media can include: read-only memory (ROM), random access memory (RAM), solid-state drives (SSDs), or optical discs, etc. Random access memory can include resistive random access memory (ReRAM) and dynamic random access memory (DRAM).

[0253] This invention can also provide a method for site selection of nearshore engineering projects, using the output results of a comprehensive evaluation method for nearshore sea conditions described in this invention as the basis for engineering feasibility analysis, including: verification of the theoretical depth datum for offshore wind farms or photovoltaic sites; evaluation of wave loads and current velocity loads on offshore wind turbine foundations; and analysis of the adaptability of tidal range and current velocity in port construction areas.

[0254] In summary, this invention provides a comprehensive evaluation method for nearshore sea states based on the MIKE 21 / 3 Coupled FM model, addressing the technical bottlenecks of inefficiency, strong subjectivity, and insufficient accuracy in the parameter setting and selection process of traditional hydrodynamic models in the fields of marine hydrological simulation and marine wind energy resource assessment. This invention proposes an automated parameter setting and selection technical solution, a physical framework for the hydrodynamic model, and its core computational equations. Combined with research by professionals on marine hydrophysical conditions in different sea areas, an automated mechanism for the identification, setting, and selection of model parameters is constructed. Compared to manual setting methods relying on human experience, this improves the efficiency and objective accuracy of parameter selection. In the parameter selection stage, principal component analysis and an autoencoder are introduced to reduce the dimensionality of input parameters and generate a weight matrix. A hierarchical hyperparameter optimization strategy from deep learning is employed to accelerate parameter determination. A comprehensive verification method based on multiple indicators makes the parameter accuracy verification process more scientific, rigorous, and reliable, providing a solid quantitative basis for the effectiveness of model parameters. In the parameter setting and selection process, this invention fully considers multiple key influencing factors such as the characteristics of the simulated target area, the accuracy of the computational grid, boundary conditions, topographic features, and water depth distribution. This multi-factor collaborative optimization mechanism ensures that the selected parameters can more accurately reflect the characteristics of the actual marine environment, thereby significantly improving the overall simulation accuracy and effect of the model.

[0255] The above content is only for illustrating the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made to the technical solution based on the technical concept proposed in this invention shall fall within the scope of protection of the claims of this invention.

Claims

1. A comprehensive assessment method for nearshore sea conditions, characterized in that, The numerical model based on the two-way coupling of hydrodynamics and waves is implemented, including the following steps: Step 1: Obtain the model grid file and initial dataset of wind field, coastline, water depth and tide level within the coordinate range of the sea area to be studied, and generate the prediction dataset through numerical model; Step 2 involves performing error analysis on the predicted data and existing measured data in the prediction dataset. A hierarchical hyperparameter optimization strategy is then used to adjust the numerical model parameters and the initial dataset until the relative error between the numerical model's predicted data and the existing measured data is less than a set threshold, resulting in the adjusted numerical model. Specifically, this includes: Error analysis is performed on the predicted data and existing measured data in the prediction dataset to calculate the relative error between the two. If the relative error is ≥5%, a hierarchical hyperparameter optimization strategy is adopted to adjust the numerical model parameters and the initial dataset. If the relative error is <5%, the calibrated high-precision numerical model is obtained directly. The execution steps of the hierarchical hyperparameter optimization strategy include: generating a parameter importance weight matrix of the initial data's influence on the results through an autoencoder, and arranging them in descending order of weight; generating a principal component analysis weight matrix of the model parameters' influence on the results through principal component analysis, and arranging it in descending order of weight; merging the descending-ordered principal component analysis weight matrix and the parameter importance weight matrix to obtain a comprehensive weight matrix; using the comprehensive weight matrix data as input and the relative error as output, performing hierarchical selection and computational resource reallocation through a deep learning hierarchical hyperparameter optimization strategy, combined with the parameter importance reflected in the weight matrix; and setting mandatory physical constraints during the parameter optimization process. Repeat the above iterative steps until the relative error between the predicted dataset and the measured data is <5%, and obtain the calibrated high-precision numerical model. Step 3: Combine historical typhoon data in the sea area to be studied with the numerical simulation module of coastal and marine hydrodynamics in the numerical model to generate new typhoon data. Input the new typhoon data into the adjusted numerical model to obtain predicted data. Analyze the characteristic parameters of water level, tide, wave, ocean current and storm surge in the predicted data to obtain evaluation results.

2. The comprehensive assessment method for nearshore sea conditions according to claim 1, characterized in that, Step 1 involves obtaining the model mesh file and initial dataset for wind field, shoreline, water depth, and tide level within the coordinate range of the sea area to be studied. Extract shoreline coordinates and generate .shp format files. Convert the generated .shp format files to .xyz format files. Open the generated .xyz format files with the MIKE mesh generate block, remove duplicate points and intersecting lines, and generate the model's mesh file. Extract initial water depth files from public datasets and interpolate the initial water depth data into the model's mesh file. Obtain one year or more of horizontal and vertical average wind speed and sea-level atmospheric pressure data from public datasets. The wind field file of the hydrodynamic module of the numerical model contains wind speed and sea-level atmospheric pressure data, and the wind field file of the wave module contains wind speed data. Import global tide level model data, match the time dimension and total simulation duration of the numerical simulation, import it into the model's grid file, and generate tide level boundary data adapted to the sea area under study; input the initial dataset and tide level boundary data into the hydrodynamic and wave bidirectional coupled numerical model, and generate predicted datasets of tide level, current, wave, and water level through numerical model calculations. The numerical model calculations for the hydrodynamic module include: calculations of water flow mass and momentum conservation, turbulent closure simulation, wind stress driving calculation, bed resistance dissipation calculation, Coriolis effect calculation, and wave radiation stress coupling calculation; the numerical model calculations for the wave module include: wave action conservation calculation, wind energy input calculation, white cap dissipation calculation, water depth induced breakage control calculation, bottom friction dissipation calculation, and wave diffraction effect simulation.

3. The comprehensive assessment method for nearshore sea conditions according to claim 1, characterized in that, The numerical model that couples hydrodynamics and waves is a numerical model with real-time two-way data interaction capability between hydrodynamic and wave processes. Specifically, it adopts the MIKE 21 / 3 Coupled model, Delft3D FM model, or FVCOM hydrodynamic model coupled with the SWAN wave model. The numerical model realizes data interaction between the hydrodynamic module and the wave module through online two-way coupling, and transmits water level, flow velocity, and wave radiation stress parameters in real time.

4. The comprehensive assessment method for nearshore sea conditions according to claim 1, characterized in that, The following mandatory physical constraints, including the lower bound of optimization priority and the feasible region of parameters, are added during the optimization process: Optimize priority lower bound constraints: Parametric feasible region physical constraints: In the formula: It is a comprehensive weight matrix. Indicates the physical parameter index, Ω phy It is a set of key physical parameters; w min It is the lower limit of optimization priority. , The first j The lower and upper limits of the feasible domain for each physical parameter.

5. The comprehensive assessment method for nearshore sea conditions according to claim 1, characterized in that, In step 3, the numerical model automatically matches the input feature parameter values ​​with the standards in the rule base to determine the corresponding water level, tide, wave, ocean current and storm surge status, level and risk of each feature parameter; combining the physical correlation between feature parameters and comparing the numerical model prediction data with the measured data, the accuracy of the prediction results is automatically marked.

6. The comprehensive assessment method for nearshore sea conditions according to claim 1, characterized in that, Step 3 involves generating new typhoon data using the coastal and ocean hydrodynamic numerical simulation module in the numerical model, including: The typhoon field and background wind field are fused using a dynamic weighting formula. The calculation formula is as follows: in, V C This is the resulting hybrid wind field; V M For typhoon wind field; V Q Background wind field; r 0 represents the radius of maximum wind speed; r It is the distance from the calculation point to the center of the typhoon. e w These are the weighting coefficients.

7. A comprehensive assessment system for nearshore sea conditions, characterized in that, It includes a data acquisition module, a model calculation module, a parameter optimization and calibration module, and a sea state assessment output module. The output of the data acquisition module is connected to the input of the model calculation module. The output of the model calculation module is connected to the input of the parameter optimization and calibration module and the sea state assessment output module, respectively. The output of the parameter optimization and calibration module is connected to the input of the model calculation module. The data acquisition module is used to acquire initial data on wind field, shoreline, water depth, and tide level of the sea area to be studied, complete shoreline data format conversion, deduplication and deintersecting processing to generate model mesh file, interpolate water depth to model mesh file, acquire at least one year of continuous horizontal and vertical average wind speed and sea level atmospheric pressure data to construct initial dataset; at the same time, import global tide level model data to generate tide level boundary data of the sea area to be studied. The model calculation module is configured to run a two-way coupled numerical model of hydrodynamics and waves. It is used to receive the initial dataset and tidal boundary data, and generate a prediction dataset through two-way coupled calculations of the hydrodynamic module and the wave module. It is also used to receive the optimized model parameters output by the parameter optimization and calibration module and the typhoon wind field data, and run the model to obtain the full-element prediction data of the typhoon condition. The hydrodynamic module covers the functions of water flow mass and momentum conservation, turbulence closure, wind stress drive, bed resistance dissipation, Coriolis effect, and wave radiation stress coupling calculation. The wave module covers the functions of wave action conservation, wind energy input, white cap dissipation, water depth induced breakage control, bottom friction dissipation, and wave diffraction effect simulation. The parameter optimization and calibration module is used to receive the prediction dataset output by the model calculation module, perform error analysis between the prediction dataset and the existing measured data of the sea area to be studied, and calculate the relative error; when the relative error is ≥5%, hierarchical hyperparameter optimization is performed, and a parameter importance weight matrix is ​​generated through a neural network autoencoder. Principal component analysis generates a principal component analysis weight matrix; the weight matrix is ​​then merged to obtain a comprehensive weight matrix. Using the comprehensive weight matrix as input and the relative error as output, hierarchical filtering and computational resource reallocation are performed according to the parameter weights. At the same time, mandatory physical constraints of the lower limit of optimization priority and the feasible region of parameters are added during the optimization process. The optimized model parameters are fed back to the model calculation module until the relative error between the predicted dataset and the measured data is <5%, thus completing the model calibration. Specifically, it includes: Error analysis is performed on the predicted data and existing measured data in the prediction dataset to calculate the relative error between the two. If the relative error is ≥5%, a hierarchical hyperparameter optimization strategy is adopted to adjust the numerical model parameters and the initial dataset. If the relative error is <5%, the calibrated high-precision numerical model is obtained directly. The execution steps of the hierarchical hyperparameter optimization strategy include: generating a parameter importance weight matrix of the initial data's influence on the results through an autoencoder, and arranging them in descending order of weight; generating a principal component analysis weight matrix of the model parameters' influence on the results through principal component analysis, and arranging it in descending order of weight; merging the descending-ordered principal component analysis weight matrix and the parameter importance weight matrix to obtain a comprehensive weight matrix; using the comprehensive weight matrix data as input and the relative error as output, performing hierarchical selection and computational resource reallocation through a deep learning hierarchical hyperparameter optimization strategy, combined with the parameter importance reflected in the weight matrix; and setting mandatory physical constraints during the parameter optimization process. Repeat the above iterative steps until the relative error between the predicted dataset and the measured data is <5%, and obtain the calibrated high-precision numerical model. The sea state assessment output module has a built-in sea state assessment rule base. It combines historical typhoon data of the sea area under study with dynamic weighting formulas to fuse typhoon wind fields and background wind fields to generate typhoon operational wind field data, which is then input into the calibrated model calculation module. It receives the typhoon operational condition full-element prediction data output from the model calculation module, performs feature parameter analysis on it across four dimensions: tide, wave, tidal current, and storm surge. The feature parameters are then matched with the sea state assessment rule base to generate and output a comprehensive nearshore sea state assessment that includes tidal attributes, wave element characteristics, tidal current characteristics, and storm surge risk level of the sea area under study. The evaluation report specifically includes: determining the precise conversion relationships between tidal attributes, theoretical minimum tide level, and mean sea level of the sea area under study based on the results of tidal harmonic analysis of the site area; drawing a full-year wave rose diagram based on numerical simulation results, statistically analyzing the interannual and seasonal frequency of different wave classes and directions, and analyzing the influence of wave refraction, diffraction, breaking characteristics, and wave-current interaction; determining the tidal attributes and tidal movement patterns based on the results of tidal harmonic analysis; and outputting the statistical patterns of historical storm surge increases in the site area, the differences in storm surge characteristics of typhoons with different paths, the annual maximum storm surge value, and the seasonal distribution characteristics of storm surge increases.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the comprehensive assessment method for nearshore sea conditions as described in any one of claims 1-6.

9. A method for site selection of offshore engineering projects, characterized in that, Using the comprehensive assessment method for nearshore sea conditions as described in any one of claims 1-6 as the basis for engineering feasibility analysis includes: Verification of the theoretical depth benchmark surface for offshore wind farms or photovoltaic sites; Assessment of wave loads and current velocity loads on offshore wind turbine foundations; Analysis of the adaptability of tidal range and current velocity in port construction areas.

Images