Wave prediction method considering evolution characteristics of marine environment

Abstract

The present application belongs to the technical field of sea wave height prediction, and discloses a wave height prediction method considering the evolution characteristics of marine environment. The method comprises the following steps: (1) collecting the historical data of effective wave height, average period and wind speed of a target sea area, and processing the data; (2) performing seasonal trend decomposition on the effective wave height sequence to obtain a trend item, a seasonal item and a residual item; (3) calculating a wind wave response index according to the wind speed variation, the wave average period variation and the effective wave height variation; (4) generating a wind wave response state weight, and adjusting the action intensity of different historical time steps, different wave height components and meteorological marine elements in the prediction process by using the wind wave response state weight; (5) performing weight correction on the prediction contribution of the trend item, the seasonal item and the residual item, and reconstructing to obtain a future effective wave height prediction value. The present application can improve the stability of sea wave effective wave height prediction in the stage of obvious wind speed change or sea state conversion.

Description

Technical Field

[0001] This invention relates to the field of ocean wave height prediction technology, and in particular to a wave prediction method that takes into account the evolution characteristics of the marine environment. Background Technology

[0002] Ocean waves are a crucial marine environmental factor affecting ship navigation safety, the stress state of marine engineering structures, and the operational reliability of nearshore facilities. Significant wave height, as a key parameter characterizing wave strength, directly impacts the safety and economy of various marine activities, including route planning, platform operations, structural design, and disaster early warning. Therefore, accurate prediction of wave changes under complex sea conditions has significant engineering application value.

[0003] In actual sea conditions, wave evolution is simultaneously influenced by multiple factors, including local wind fields, air-sea interactions, and historical wave states. Due to the superposition of wind waves and swells, energy transfer across frequency bands, and continuous changes in environmental conditions, wave effective height sequences typically exhibit strong non-stationarity, randomness, and nonlinearity. Meanwhile, meteorological and oceanographic elements such as wave mean period and wind speed are closely related to wave development; changes in meteorological conditions at different times can significantly impact wave states. Therefore, relying solely on a single wave height sequence is often insufficient to fully reflect the inherent laws governing wave formation and evolution.

[0004] Existing deep learning-based wave prediction methods typically use historical significant wave height as the primary input. Some methods further incorporate meteorological and oceanographic elements such as wind speed and mean period to enhance the model's ability to perceive changes in the external environment. However, in actual sea conditions, wind speed changes are not always fully reflected in significant wave height changes at the same time, and mean period changes are often related to wave energy structure and sea state persistence. If wind speed and mean period are simply used as ordinary input variables and directly concatenated with the wave height sequence, the model struggles to distinguish the differences in the roles of various features during stable sea states, periods of increased wind and waves, and periods of decreased waves. Furthermore, wave sequences typically contain long-term variations, periodic fluctuations, and short-term disturbances simultaneously. While existing methods can model the wave height sequence holistically, they do not adequately consider the differences in the contributions of different variation components under different sea states. For example, when wind speed increases rapidly, the residual and periodic terms typically have a more significant impact on short-term wave height changes; while when wind speed is relatively stable, the trend term has a more significant constraining effect on the prediction results.

[0005] Therefore, it is necessary to propose a wave prediction method that can combine the relationship between wind speed, mean period and significant wave height variation, and correct the contribution of different wave height components according to sea state changes during the prediction process, thereby improving prediction accuracy and engineering applicability. Summary of the Invention

[0006] This invention addresses the problems of existing wave prediction methods, such as the relatively simple utilization of marine environmental factors, the difficulty in distinguishing the contribution of wave height variation components, and the insufficient prediction stability during sea state transition stages. It proposes a wave prediction method that considers the evolution characteristics of the marine environment to improve the accuracy and stability of effective wave height prediction.

[0007] This invention provides a wave prediction method that considers the evolutionary characteristics of the marine environment, comprising the following steps:

[0008] (1) Data acquisition and preprocessing: Collect the ocean wave significant wave height sequence, average period sequence and wind speed sequence of the target sea area, perform time alignment, missing value imputation and outlier correction on the data, and divide the data into training set, validation set and test set according to time order;

[0009] (2) Effective wave height decomposition: Seasonal trend decomposition is performed on the effective wave height sequence to obtain trend term D1, seasonal term D2 and residual term D3. The trend term D1 is used to represent the long-term change of effective wave height, the seasonal term D2 is used to represent the periodic fluctuation of effective wave height, and the residual term D3 is used to represent the short-term disturbance of effective wave height.

[0010] (3) Calculation of wind and wave response index: The wind and wave response index R is calculated based on the changes in the effective wave height, average period, and wind speed within a preset time interval. t The wind and wave response index R t Used to indicate the degree of influence of current changes in meteorological and oceanographic conditions on changes in significant wave height;

[0011] (4) Feature input construction: The trend term D1, seasonal term D2, residual term D3, average period of ocean waves, wind speed, and wind and wave response index R are constructed. t These are used together as model inputs, and input features with temporal sequence information are obtained through linear mapping and positional encoding;

[0012] (5) Wave response feature extraction: The input features obtained in step (4) are fed into the wave response attention module. The wave height change features between different historical time steps are extracted through the attention mechanism, and the wave response index R is used as the basis for the extraction. t Generate wind and wave response state weights G t Using the wind and wave response state weight G t Adjust the intensity of the effects of different historical time steps, different wave height components, average period, and wind speed in the forecasting process;

[0013] (6) Component contribution correction: Based on the wind and wave response index R obtained in step (3) t Determine the contribution weights of the trend term D1, seasonal term D2, and residual term D3 in the prediction of future effective wave height; when the wind and wave response index R tWhen the value is large, increase the contribution weights of the seasonal term D2 and the residual term D3; when the wind and wave response index R... t When the value is small, increase the contribution weight of trend term D1;

[0014] (7) Wave height prediction and reconstruction: The trend term D1, seasonal term D2 and residual term D3 with contribution weights after step (6) are input into the component prediction module; the component prediction module includes a trend component prediction branch, a seasonal component prediction branch and a residual component prediction branch, which output the trend prediction component, seasonal prediction component and residual prediction component respectively; the three prediction components are weighted and reconstructed according to the contribution weights to obtain the future effective wave height prediction value;

[0015] (8) Model training and prediction: Update the model parameters based on the error between the predicted future effective wave height and the actual effective wave height, determine the model parameters using the validation set, and input the test set into the trained model to obtain the final wave height prediction result.

[0016] The above steps are further defined as follows.

[0017] In step (1), the training set, validation set, and test set are divided in chronological order, with a ratio of 6:2:2.

[0018] In step (2), the cycle length of the seasonal trend decomposition is determined according to the data sampling interval; when the data is hourly data, the cycle length is 24 or 168; when the data is daily data, the cycle length is 7 or 30.

[0019] The wind and wave response index R in step (3) t Calculate as follows:

[0020] ;

[0021] Wherein, ΔW t =W t -W t-τW ΔT represents the change in wind speed. t =T t -T t-τt ΔH represents the average periodic change. t =H t -H t-τh W represents the change in significant wave height. t T represents the wind speed at time t. t H represents the average wave period at time t. t τ represents the significant wave height at time t. W τ t τ hβ1, β2, β3 and β4 represent the lag step size corresponding to wind speed, mean period and significant wave height respectively, ε is a constant to prevent the denominator from being zero, and β1, β2, β3 and β4 are weighting coefficients;

[0022] When the data is hourly, the wind speed lag step τ W Take 3, average period lag step τ t Take 6, effective wave height hysteresis step τ h Take 3; where wind speed lag step is used to reflect the impact of local wind speed changes on short-term wind and wave development, and mean period lag step is used to reflect the impact of wave period changes on sea state persistence.

[0023] The wind and wave response attention module in step (5) includes basic temporal attention calculation and wind and wave response weight calculation; wherein, the basic temporal attention calculation is used to extract the temporal variation characteristics between historical significant wave height components, and the wind and wave response weight calculation is used to calculate the wind and wave response index R. t Generate wind and wave response state weights G t The influence of the average period change, wind speed change, and historical wave height component on the prediction of future effective wave height is adjusted by using the wind and wave response state weights.

[0024] The calculation method for the i-th attention head in the wind and wave response attention module is as follows:

[0025] ;

[0026] in:

[0027] ;

[0028] Among them, Q i K i V i Let d represent the query vector, key vector, and value vector of the i-th attention head, respectively. k G represents the dimension of the key vector. t Indicated by the wind and wave response index R t The generated wind and wave response state weights, where ⊙ represents element-wise multiplication and f(·) represents a nonlinear mapping function;

[0029] The wind and wave response state weight G t R-wave response index t It changes with the change of R; t When R is large, the weights of the wind and wave response state increase the influence of the seasonal term, residual term, wind speed variation, and mean period variation on the prediction process; when R... t When the wind and wave response state weight G is relatively small, t Increase the influence of trend terms and stable historical wave height features in the prediction process.

[0030] In step (6), the contribution weights of the trend term D1, seasonal term D2, and residual term D3 are determined as follows: when the wind and wave response index R t When the value is greater than the first threshold r1, increase the contribution weights of the seasonal term D2 and the residual term D3; when the wind and wave response index R t When the value is less than the second threshold r2, increase the contribution weight of the trend term D1; when the wind and wave response index R t When the value is between the second threshold r2 and the first threshold r1, the balanced contributions of the trend term D1, the seasonal term D2, and the residual term D3 are maintained. The first threshold r1 is set to 0.65, and the second threshold r2 is set to 0.35; when R... t When R is greater than 0.65, the current sea state is determined to be in a stage of increasing wind and waves or sea state transition; when R... t When the value is less than 0.35, the current sea state is considered to be in a relatively stable phase.

[0031] The predicted future effective wave height in step (7) is reconstructed as follows:

[0032] ;

[0033] Among them, H t+1:t+L Let represent the effective wave height prediction value for the next L time steps; D1, D2, and D3 represent the trend prediction component, seasonal prediction component, and residual prediction component, respectively; α1, α2, and α3 represent the contribution weights of the three prediction components, and satisfy α1+α2+α3=1.

[0034] The above method is based on historical data of ocean wave significant wave height, mean period, and wind speed. It further calculates the wind and wave response index based on the significant wave height decomposition to describe the response relationship between wind speed changes, mean period changes, and significant wave height changes. Based on this response index, the contribution ratios of the trend term, seasonal term, and residual term in future wave height prediction are adjusted, enabling the model to adopt different component combinations under different conditions such as relatively stable conditions, enhanced wind and waves, and sea state changes, thereby improving the accuracy and stability of wave significant wave height prediction.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] First, this invention not only introduces wind speed and average period of ocean waves as external meteorological and oceanographic information, but also uses wind speed change, average period change and significant wave height change to construct wind and wave response indicators, enabling the model to identify the degree of influence of meteorological conditions on wave height change under different sea states.

[0037] Second, this invention decomposes the effective wave height sequence into a trend term D1, a seasonal term D2, and a residual term D3, and corrects the contribution of the three components in the prediction reconstruction based on the wind and wave response index. Compared with methods that directly predict the original wave height sequence as a whole, this invention can better distinguish the impact of long-term changes, periodic fluctuations, and short-term disturbances on future wave heights.

[0038] Third, this invention introduces wind and wave response state weights in the attention calculation process, and adjusts the influence of historical wave height components, average period and wind speed in the prediction process according to the wind and wave response index, so that the prediction model can simultaneously utilize the historical wave height variation pattern and meteorological and oceanographic element change information, thereby improving the prediction stability of sea waves in the stages of wind speed enhancement, sea state transformation and relatively stable phases. Attached Figure Description

[0039] Figure 1 This is a system schematic diagram of the wave prediction method that takes into account the evolution characteristics of the marine environment, as described in this invention.

[0040] Figure 2 This is a schematic diagram illustrating the calculation process of the wind and wave response index in this invention.

[0041] Figure 3 This is a schematic diagram of the wave height reconstruction process based on component contribution correction in this invention. Detailed Implementation

[0042] like Figure 1 As shown, the wave prediction method proposed in this invention, which considers the evolution characteristics of the marine environment, mainly includes data acquisition and preprocessing, significant wave height decomposition, wind and wave response index calculation, input feature construction, wind and wave response feature extraction, component contribution correction, wave height prediction and reconstruction, and model training and prediction steps, which are described in detail below.

[0043] (1) Data acquisition and preprocessing;

[0044] Historical data on significant wave height, average period, and wind speed of ocean waves within a continuous time range were collected for the target sea area. Since ocean environmental data from different sources may have inconsistent time records, missing data, or anomalous spikes, this embodiment first performs time alignment on the above data to ensure that significant wave height, average period, and wind speed correspond at the same time step; then, missing values ​​are filled in, and outliers that significantly deviate from the variation patterns of adjacent time points are corrected. After preprocessing, the data is divided into training, validation, and test sets according to time order, with a ratio of 6:2:2.

[0045] (2) Effective wave height decomposition;

[0046] Seasonal trend decomposition is required to obtain a trend term D1, a seasonal term D2, and a residual term D3. Trend term D1 represents the long-term background characteristics of significant wave height changes over time, seasonal term D2 represents the fluctuation characteristics formed by periodic sea state changes, and residual term D3 represents short-term disturbances and fluctuation characteristics. In this embodiment, the period length of the seasonal trend decomposition is determined based on the data sampling interval. When the input data is hourly data, the period length can be 24 or 168, where 24 is used to characterize intraday variations and 168 is used to characterize weekly scale variations; when the input data is daily data, the period length can be 7 or 30 to characterize weekly and monthly scale fluctuations, respectively. This processing can separate the variation components at different time scales in the significant wave height sequence, avoiding the problem of mixed trend changes, periodic fluctuations, and short-term disturbances when directly modeling the original wave height sequence as a whole.

[0047] (3) Calculation of wind and wave response indicators;

[0048] like Figure 2 As shown, this embodiment calculates the wind and wave response index R based on the changes in wind speed, the changes in average period, and the changes in significant wave height. t This indicator does not directly use the raw values ​​of wind speed and mean period, but rather calculates using changes within adjacent moments or a preset lag time window. This is because the absolute values ​​of wind speed or mean period only reflect environmental conditions at a specific moment, while changes better reflect whether the sea state is in a development, transition, or attenuation phase. Wind and Wave Response Index R t The calculation method is as follows:

[0049] ;

[0050] Wherein, ΔW t =W t -W t-τW ΔT represents the change in wind speed. t =T t -T t-τt ΔH represents the average periodic change. t =H t -H t-τh W represents the change in significant wave height. t T represents the wind speed at time t. t H represents the average wave period at time t. t τ represents the significant wave height at time t. W τ t τ h Let represent the lag step size corresponding to wind speed, mean period, and significant wave height, respectively. ε is a constant to prevent the denominator from being zero, and β1, β2, β3, and β4 are weighting coefficients. Under hourly data conditions, the wind speed lag step size τ in this embodiment is... WTake 3, average period lag step τ t Take 6, effective wave height hysteresis step τ h We choose 3. The wind speed uses a shorter lag step because local wind speed changes are usually reflected in short-term wind and wave disturbances first; the mean period uses a longer lag step because the mean period changes better reflect the changes in wave energy structure and sea state persistence.

[0051] (4) Input feature construction;

[0052] The trend term D1, seasonal term D2, and residual term D3 obtained in step (2) are compared with the average period of ocean waves, wind speed, and the wind and wave response index R obtained in step (3). t These variables are collectively used as input variables. After linear mapping, they enter a unified feature space and are encoded with positional information to preserve the chronological relationships between time steps. The resulting input features contain both decomposed information on historical significant wave heights and information on the impact of marine environmental changes on wave height development.

[0053] (5) Extraction of wind and wave response features:

[0054] The input features are fed into the wave response attention module. This module includes basic temporal attention calculation and wave response weight calculation. Basic temporal attention calculation is used to extract the wave height variation relationship between different historical time steps; wave response weight calculation is used to calculate the wave response weight based on the wave response index R. t Generate wind and wave response state weights G t The weights are then used to adjust the influence of different historical time steps, different wave height components, average period, and wind speed in the prediction process. In this embodiment, the i-th attention head in the wind and wave response attention module is calculated as follows:

[0055] ;

[0056] in:

[0057] ;

[0058] Among them, Q i K i V i Let d represent the query vector, key vector, and value vector of the i-th attention head, respectively. k G represents the dimension of the key vector. t Indicated by the wind and wave response index R t The generated wind and wave response state weights, where ⊙ represents element-wise multiplication, and f(·) represents a nonlinear mapping function. When R t When R is large, the weights of the wind and wave response state increase the influence of the seasonal term, residual term, wind speed variation, and mean period variation on the prediction process; when R... tWhen the wave response state weights are relatively small, the influence of the trend term and stable historical wave height features on the prediction process is increased.

[0059] (6) Component contribution correction;

[0060] like Figure 3 As shown, the wind and wave response index R calculated according to step (3) is... t The contributions of the trend term D1, seasonal term D2, and residual term D3 to the forecast reconstruction are adjusted. When the wind and wave response index R... t When the sea state exceeds the first threshold r1, it is considered that the current sea state is significantly affected by wind speed changes or sea state transitions. In this case, the contribution weights of the seasonal term D2 and the residual term D3 are increased to enhance the model's response to periodic changes and short-term disturbances. When the wind and wave response index R... t When the sea state is less than the second threshold r2, it is considered relatively stable. In this case, the contribution weight of the trend term D1 is increased to maintain the smoothness and long-term consistency of the prediction results. When the wind and wave response index R... t When the value is between the second threshold r2 and the first threshold r1, a balanced weighted method using the trend term D1, seasonal term D2, and residual term D3 is used for prediction. In this embodiment, the first threshold r1 is set to 0.65, and the second threshold r2 is set to 0.35. When R... t When R is greater than 0.65, the current sea state is determined to be in a stage of increasing wind and waves or sea state transition; when R... t When R is less than 0.35, the current sea state is considered to be in a relatively stable phase; when R... t When the value is between 0.35 and 0.65, the current sea state is considered to be in a general change phase.

[0061] (7) Wave height prediction and reconstruction;

[0062] The trend term D1, seasonal term D2, and residual term D3, after component contribution correction and with contribution weights, are input into the component prediction module. The component prediction module includes a trend component prediction branch, a seasonal component prediction branch, and a residual component prediction branch, which output trend prediction component D1, seasonal prediction component D2, and residual prediction component D3, respectively. Then, a weighted reconstruction is performed according to the following formula to obtain the effective wave height prediction values ​​for the next L time steps:

[0063] ;

[0064] Among them, H t+1:t+L Let α1, α2, and α3 represent the effective wave height prediction values ​​for the next L time steps; α1, α2, and α3 represent the contribution weights of the three prediction components, and satisfy α1 + α2 + α3 = 1.

[0065] (8) Model training and prediction;

[0066] The training set is input into the model for training, and the model parameters are updated based on the error between the predicted and actual significant wave heights. During training, the model parameters are adjusted using the validation set, and the model's convergence is assessed. After training is complete, the test set is input into the trained model to obtain the predicted future significant wave heights.

[0067] In this embodiment, the training objective of the model is to reduce the error between the predicted significant wave height and the actual significant wave height. Since the input features simultaneously include significant wave height decomposition components, wind speed, wave mean period, and wind and wave response indices, the model can comprehensively utilize historical wave height variation patterns and marine environmental change information under different sea states to predict the significant wave height.

Claims

1. A wave prediction method considering the evolutionary characteristics of the marine environment, characterized in that, Includes the following steps: (1) Data acquisition and preprocessing: Collect the ocean wave significant wave height sequence, average period sequence and wind speed sequence of the target sea area, perform time alignment, missing value imputation and outlier correction on the data, and divide the data into training set, validation set and test set according to time order; (2) Effective wave height decomposition: Seasonal trend decomposition is performed on the effective wave height sequence to obtain trend term D1, seasonal term D2 and residual term D3. The trend term D1 is used to represent the long-term change of effective wave height, the seasonal term D2 is used to represent the periodic fluctuation of effective wave height, and the residual term D3 is used to represent the short-term disturbance of effective wave height. (3) Calculation of wind and wave response index: The wind and wave response index R is calculated based on the changes in the effective wave height, average period, and wind speed within a preset time interval. t The wind and wave response index R t Used to indicate the degree of influence of current changes in meteorological and oceanographic conditions on changes in significant wave height; (4) Feature input construction: The trend term D1, seasonal term D2, residual term D3, average period of ocean waves, wind speed, and wind and wave response index R are constructed. t These are used together as model inputs, and input features with temporal sequence information are obtained through linear mapping and positional encoding; (5) Wave response feature extraction: The input features obtained in step (4) are fed into the wave response attention module. The wave height change features between different historical time steps are extracted through the attention mechanism, and the wave response index R is used as the basis for the extraction. t Generate wind and wave response state weights G t Using the wind and wave response state weight G t Adjust the intensity of the effects of different historical time steps, different wave height components, average period, and wind speed in the forecasting process; (6) Component contribution correction: Based on the wind and wave response index R obtained in step (3) t Determine the contribution weights of trend term D1, seasonal term D2, and residual term D3 in the prediction of future effective wave height; (7) Wave height prediction and reconstruction: The trend term D1, seasonal term D2 and residual term D3 with contribution weights after step (6) are input into the component prediction module; the component prediction module includes a trend component prediction branch, a seasonal component prediction branch and a residual component prediction branch, which output the trend prediction component, seasonal prediction component and residual prediction component respectively; the three prediction components are weighted and reconstructed according to the contribution weights to obtain the future effective wave height prediction value; (8) Model training and prediction: Update the model parameters based on the error between the predicted future effective wave height and the actual effective wave height, determine the model parameters using the validation set, and input the test set into the trained model to obtain the final wave height prediction result.

2. The wave prediction method considering the evolution characteristics of the marine environment according to claim 1, characterized in that, In step (1), the training set, validation set, and test set are divided in chronological order, with a ratio of 6:2:

2.

3. The wave prediction method considering the evolution characteristics of the marine environment according to claim 1, characterized in that, In step (2), the cycle length of the seasonal trend decomposition is determined according to the data sampling interval; when the data is hourly data, the cycle length is 24 or 168; when the data is daily data, the cycle length is 7 or 30.

4. The wave prediction method considering the evolution characteristics of the marine environment according to claim 1, characterized in that, The wind and wave response index R in step (3) t Calculate as follows: ； Wherein, ΔW t =W t -W t-τW ΔT represents the change in wind speed. t =T t -T t-τt ΔH represents the average periodic change. t =H t -H t-τh W represents the change in significant wave height. t T represents the wind speed at time t. t H represents the average wave period at time t. t τ represents the significant wave height at time t. W τ t τ h β1, β2, β3 and β4 represent the lag step lengths corresponding to wind speed, mean period and significant wave height, respectively. ε is a constant to prevent the denominator from being zero, and β1, β2, β3 and β4 are weighting coefficients.

5. The wave prediction method considering the evolution characteristics of the marine environment according to claim 4, characterized in that, When the data is hourly, the wind speed lag step τ W Take 3, average period lag step τ t Take 6, effective wave height hysteresis step τ h Take 3; where wind speed lag step is used to reflect the impact of local wind speed changes on short-term wind and wave development, and mean period lag step is used to reflect the impact of wave period changes on sea state persistence.

6. The wave prediction method considering the evolution characteristics of the marine environment according to claim 1, characterized in that, The wind and wave response attention module in step (5) includes basic temporal attention calculation and wind and wave response weight calculation; wherein, the basic temporal attention calculation is used to extract the temporal variation characteristics between historical significant wave height components, and the wind and wave response weight calculation is used to calculate the wind and wave response index R. t Generate wind and wave response state weights G t The influence of the average period change, wind speed change, and historical wave height component on the prediction of future effective wave height is adjusted by using the wind and wave response state weights.

7. The wave prediction method considering the evolution characteristics of the marine environment according to claim 6, characterized in that, The calculation method for the i-th attention head in the wind and wave response attention module is as follows: ； in: Among them, Q i K i V i Let d represent the query vector, key vector, and value vector of the i-th attention head, respectively. k G represents the dimension of the key vector. t Indicated by the wind and wave response index R t The generated wind and wave response state weights, where ⊙ represents element-wise multiplication and f(·) represents a nonlinear mapping function.

8. The wave prediction method considering the evolution characteristics of the marine environment according to claim 1, characterized in that, In step (6), the contribution weights of the trend term D1, seasonal term D2, and residual term D3 are determined as follows: when the wind and wave response index R t When the value is greater than the first threshold r1, increase the contribution weights of the seasonal term D2 and the residual term D3; when the wind and wave response index R t When the value is less than the second threshold r2, increase the contribution weight of the trend term D1; when the wind and wave response index R t When the value is between the second threshold r2 and the first threshold r1, the balanced contribution of the trend term D1, the seasonal term D2, and the residual term D3 is maintained.

9. The wave prediction method considering the evolution characteristics of the marine environment according to claim 8, characterized in that, The first threshold r1 is 0.65, and the second threshold r2 is 0.35; when R t When R is greater than 0.65, the current sea state is determined to be in a stage of increasing wind and waves or sea state transition; when R... t When the value is less than 0.35, the current sea state is considered to be in a relatively stable phase.

10. The wave prediction method considering the evolution characteristics of the marine environment according to claim 1, characterized in that, The predicted future effective wave height in step (7) is reconstructed as follows: ； Among them, H t+1:t+L Let represent the effective wave height prediction value for the next L time steps; D1, D2, and D3 represent the trend prediction component, seasonal prediction component, and residual prediction component, respectively; α1, α2, and α3 represent the contribution weights of the three prediction components, and satisfy α1+α2+α3=1.

Images