Sea wave significant wave height prediction method based on space-time sequence deep learning model

Abstract

The invention provides a sea wave significant wave height prediction method based on a space-time sequence deep learning model, and relates to the technical field of sea wave significant wave height prediction, and the method specifically comprises the steps: obtaining historical wind speed data and historical sea wave significant wave height data of a test area; performing quality screening on the acquired data, and then performing time-space registration to enable historical wind speed data and historical sea wave significant wave height data to be staggered for 1 hour and aligned in a time dimension; generating a space-time sequence pair by using a sliding window algorithm, and constructing a training set; constructing a space-time sequence deep learning model, and inputting the training set into the space-time sequence deep learning model for iterative training to obtain a sea wave significant wave height prediction model based on the space-time sequence deep learning model; and forecasting the wind field according to a numerical value provided externally, and inputting the acquired future wind speed data into the prediction model to obtain a forecasting result. According to the technical scheme, the problem that in the prior art, sea wave forecasting is low in precision and reliability is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of ocean wave significant wave height prediction, and in particular to an ocean wave significant wave height prediction method based on a spatiotemporal sequence deep learning model. Background Art

[0002] Sea waves are fluctuations caused by wind acting on the sea surface, including wind waves and swells. Their formation and development are affected by many factors such as wind direction, wind speed, and topography. The wave height ranges from a few centimeters to more than 20 meters, and can even reach more than 30 meters under rare terrain conditions. The effective wave height of sea waves is an important parameter for measuring ocean conditions, and is crucial for the planning and safety assessment of marine engineering and marine-related activities. Accurate sea wave forecasts are of great significance for responding to climate change, ensuring navigation safety, and coastal protection.

[0003] In the prior art, numerical simulation is the main means of wave forecasting. Although the traditional empirical statistical method has a fast calculation speed, it cannot accurately capture the complex characteristics of waves, especially swells, due to the simplification and assumptions of the model, resulting in large deviations in the forecast results. On the other hand, the wave numerical model based on physical laws can simulate the wave process more accurately, but its calculation amount is huge, the calculation speed is slow, and the phenomenon of error accumulation may occur in long-term forecasts, affecting the accuracy of the forecast.

[0004] To improve the performance of wave simulation and forecasting, researchers are constantly trying to innovate and improve statistical and numerical models. However, current technologies still have significant limitations in quickly and accurately forecasting waves. With the rapid increase of observational and simulation data, we have entered an era of big data. There is an urgent need to adopt new methods to effectively process these data and deeply explore the physical processes and complex relationships behind the waves, so as to improve the accuracy and reliability of wave forecasting.

[0005] Therefore, there is a need for a method for predicting effective wave height based on a spatiotemporal series deep learning model to improve the accuracy and reliability of wave forecasting. Summary of the invention

[0006] The main purpose of the present invention is to provide a method for predicting the effective wave height of waves based on a deep learning model of spatiotemporal sequences, so as to solve the problems of low accuracy and reliability of wave forecasting in the prior art.

[0007] To achieve the above object, the present invention provides a method for predicting the significant wave height of ocean waves based on a deep learning model of spatiotemporal sequences, which specifically comprises the following steps: S1, obtain the historical wind speed data and historical wave significant wave height data of the test area.

[0008] S2, the acquired historical wind speed data and historical wave significant wave height data are quality screened, and then space-time registration is performed to align the historical wind speed data and historical wave significant wave height data with a 1-hour offset in the time dimension.

[0009] S3, use the sliding window algorithm to generate spatiotemporal sequence pairs and construct a training set.

[0010] S4, constructs a spatiotemporal sequence deep learning model, including an encoding layer and a prediction layer. The encoding layer is responsible for encoding and memorizing the historical wave status, and the prediction layer is responsible for forecasting the waves based on the historical status and future wind speed.

[0011] S5, inputting the training set constructed in step S3 into the spatiotemporal sequence deep learning model for iterative training, and obtaining a sea wave effective wave height prediction model based on the spatiotemporal sequence deep learning model.

[0012] S6, according to the numerical forecast wind field provided by the outside, the acquired future wind speed data is input into the trained ocean wave significant wave height prediction model based on the spatiotemporal sequence deep learning model to obtain the forecast result.

[0013] Furthermore, step S2 specifically includes the following steps: S2.1, perform quality screening on historical wind speed data and historical wave significant wave height data, including missing value processing and land marker masking.

[0014] S2.2, perform spatiotemporal registration of historical wind speed data and historical wave significant height data.

[0015] Furthermore, step S2.2 specifically includes the following steps: S2.2.1, perform interpolation processing to make the spatial resolution of historical wind speed data and historical wave significant wave height data consistent.

[0016] S2.2.2, align the historical wind speed data and the historical wave significant wave height data in the time dimension by 1 hour, and integrate the historical wind speed data and the historical wave significant wave height data into a spatiotemporal series data of T×C×H×W, where T, C, H and W represent time, channel, height and width, respectively.

[0017] Furthermore, step S3 is specifically as follows: using a sliding window algorithm, sliding sampling generates training samples, and integrating the training sample data set into N×T×C×H×W, where N, T, C, H and W represent the number of samples, time, channel, height and width respectively.

[0018] Furthermore, the encoding layer in step S4 includes: an encoded space-time sequence of multiple time steps connected in series, and the encoded space-time sequence of each time step includes: a first input layer, a first CBAM layer, and three layers of first ConvLSTM units based on convolutional long short-term memory units connected in series in sequence, and the first input layer inputs the wave effective wave height data SWH and the wind speed data of the next time step.

[0019] Furthermore, the prediction layer in step S4 includes: a prediction spatiotemporal sequence of multiple time steps connected in series, and the prediction spatiotemporal sequence of each time step includes: a second input layer, a second CBAM layer, a second ConvLSTM unit based on convolutional long and short-term memory units connected in series in sequence, and a convolutional layer Conv2D. The hidden layer state of the second ConvLSTM unit based on convolutional long and short-term memory units connected in series in three layers at this time step is formed through the convolutional layer Conv2D to form the effective wave height data SWH of the next time step. The effective wave height data SWH of the wave output by the convolutional layer Conv2D of each time step is used as the input data of the second input layer of the next time step. The second input layer also inputs the wind speed data of the previous time step, and outputs the prediction result after the prediction spatiotemporal sequence connected in series in multiple time steps.

[0020] Furthermore, step S5 is specifically as follows: inputting the training set of historical wind speed data and historical wave data obtained in step S3 into the model, performing iterative training, and obtaining a sea wave effective wave height prediction model based on a spatiotemporal sequence deep learning model.

[0021] The present invention has the following beneficial effects: The present invention combines the advantages of the attention mechanism CBAM, historical wind speed data, historical wave effective wave height data and the coding layer-prediction layer. By introducing future wind speed data, the effective wave height of waves can be predicted quickly and accurately, which plays an important role in ensuring the safety of navigation and coastal activities. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the specific implementation of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the specific implementation or the prior art description. Obviously, the drawings described below are some implementations of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings: Figure 1 A flow chart of a method for predicting effective wave height of ocean waves based on a deep learning model of spatiotemporal sequences according to the present invention is shown.

[0023] Figure 2 The structure diagram of the spatiotemporal sequence deep learning model of the present invention is shown.

[0024] Figure 3 A two-dimensional scatter density plot of the predicted and actual values ​​of the significant wave height for the next 1 hour is shown.

[0025] Figure 4 A two-dimensional scatter density diagram of the predicted and actual values ​​of the significant wave height in the next 3 hours is shown.

[0026] Figure 5 A two-dimensional scatter density diagram of the predicted and actual values ​​of the significant wave height in the next 6 hours is shown.

[0027] Figure 6 A two-dimensional scatter density diagram of the predicted and actual values ​​of the significant wave height in the next 12 hours is shown.

[0028] Figure 7 A two-dimensional scatter density diagram of the predicted and actual values ​​of the significant wave height for the next 24 hours is shown.

[0029] Figure 8 A two-dimensional scatter density diagram of the predicted and actual values ​​of the significant wave height in the next 48 hours is shown. DETAILED DESCRIPTION

[0030] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0031] like Figure 1 A method for predicting the significant wave height of ocean waves based on a deep learning model of spatiotemporal sequences is shown, which specifically includes the following steps: S1, obtain historical wind speed data and historical significant wave height data of the test area. For example, obtain historical sea surface 10m wind speed data and historical significant wave height data of the test area.

[0032] S2, the acquired historical wind speed data and historical wave effective wave height data are quality screened, and then space-time registration is performed to align the historical wind speed data and historical wave effective wave height data with a 1-hour offset in the time dimension, that is, the wind speed data at time T+1 matches the wave data at time T.

[0033] S3, use the sliding window algorithm to generate spatiotemporal sequence pairs and construct a training set.

[0034] S4, build a spatiotemporal sequence deep learning model, including an encoding layer and a prediction layer. The encoding layer is responsible for encoding and memorizing the historical wave status, and the prediction layer is responsible for forecasting the waves based on the historical status and future wind speed;

[0035] S5, inputting the training set constructed in step S3 into the spatiotemporal sequence deep learning model for iterative training, and obtaining a sea wave effective wave height prediction model based on the spatiotemporal sequence deep learning model.

[0036] S6, according to the numerical forecast wind field provided by the outside, the acquired future wind speed data is input into the trained ocean wave significant wave height prediction model based on the spatiotemporal sequence deep learning model to obtain the forecast result.

[0037] Specifically, step S2 includes the following steps: S2.1, perform quality screening on historical wind speed data and historical wave significant wave height data, including missing value processing and land marker masking.

[0038] S2.2, perform spatiotemporal registration of historical wind speed data and historical wave significant height data.

[0039] Specifically, step S2.2 includes the following steps: S2.2.1, perform interpolation processing to make the spatial resolution of historical wind speed data and historical wave significant wave height data consistent.

[0040] S2.2.2, align the historical wind speed data and the historical wave significant wave height data in the time dimension by 1 hour, and integrate the historical wind speed data and the historical wave significant wave height data into a spatiotemporal series data of T×C×H×W, where T, C, H and W represent time, channel, height and width, respectively.

[0041] Specifically, step S3 is as follows: using a sliding window algorithm, sliding sampling generates training samples, and integrating the training sample data set into N×T×C×H×W, where N, T, C, H and W represent the number of samples, time, channel, height and width respectively.

[0042] Specifically, the encoding layer in step S4 includes: a coded space-time sequence of multiple time steps connected in series, and the coded space-time sequence of each time step includes: a first input layer, a first CBAM layer, and three layers of first ConvLSTM units based on convolutional long short-term memory units connected in series in sequence, and the first input layer inputs the wave effective wave height data SWH and the wind speed data of the next time step.

[0043] Specifically, the prediction layer in step S4 includes: a prediction spatiotemporal sequence of multiple time steps connected in series, and the prediction spatiotemporal sequence of each time step includes: a second input layer, a second CBAM layer, a second ConvLSTM unit based on convolutional long and short-term memory units connected in series in sequence, and a convolutional layer Conv2D. The hidden layer state of the second ConvLSTM unit based on convolutional long and short-term memory units connected in series in three layers at this time step is formed through the convolutional layer Conv2D to form the effective wave height data SWH of the next time step. The effective wave height data SWH of the wave output by the convolutional layer Conv2D of each time step is used as the input data of the second input layer of the next time step. The second input layer also inputs the wind speed data of the previous time step, and outputs the prediction result after the prediction spatiotemporal sequence connected in series for multiple time steps.

[0044] Among them, CBAM is used to enhance the characteristics of input data, the convolutional layer Conv2D is responsible for describing the forced motion and spatial correlation of the effective wave height data, and the ConvLSTM unit is responsible for describing the temporal memory of the wave field.

[0045] like Figure 2 As shown, in the encoding layer, at time T-11, the effective wave height data SWH and the wind speed data Wind (T-10) of the next time step are input into the first attention mechanism CBAM layer, and then pass through the first, second and third layers of the first ConvLSTM unit in sequence. The data processing process of other time steps is similar to that of time T-11, and is input into the prediction layer after being encoded by the encoding layer.

[0046] In the prediction layer, at time T, the effective wave height data SWH of the previous time step and the wind speed data Wind(T+1) of the next time step are input into the second attention mechanism CBAM layer, and then pass through the second ConvLSTM unit of the first, second and third layers in sequence. The data processing process of other time steps is similar to that of time T, and the prediction results are output after passing through the prediction layer.

[0047] In the encoding layer stage, the input elements include wind speed data and significant wave height data. In the prediction stage, the update of significant wave height data comes from the previous prediction step, and the update of wind speed data comes from the numerical forecast wind field provided externally. In the prediction layer, the second ConvLSTM unit of the third layer of this time step is formed through the Conv2D convolution layer to form the SWH forecast of the significant wave height data of the next time step and serve as the input of the next prediction.

[0048] Specifically, step S5 is as follows: inputting the training set of historical wind speed data and historical wave data obtained in step S3 into the model, performing iterative training, and obtaining a sea wave effective wave height prediction model based on a spatiotemporal sequence deep learning model.

[0049] The effectiveness of the method provided by the present invention is verified below: To ensure that the training data is representative, the wind and wave element data of a certain area are taken as the training set. The wind and wave element data come from the fifth generation atmospheric reanalysis data set (ERA5) of the European Center for Medium-Range Weather Forecasts. The wind and wave element data are preprocessed, and the data are preprocessed by spatiotemporal interpolation, normalization, land marking, and element vectorization. The element field is interpolated spatiotemporally: the spatial range is 4°N -26°N, 100°E-123°E, the spatial resolution is 0.5°×0.5°, the temporal resolution is 1 hour, and the grid size is 45×45. The maximum and minimum values ​​of each element are counted and the elements are normalized to 0-1. The value of the land area is set to 0. The elements are vectorized into T×C×H×W spatiotemporal series data, where T, C, H, and W represent time, channel, height, and width, respectively. The sliding window algorithm is used to generate training samples by sliding sampling, and the data set is integrated into N×T×C×H×W, where N, T, C, H, and W represent the number of samples, time, channel (number of features), height, and width, respectively.

[0050] The key parameters in the ocean significant wave height prediction model based on the spatiotemporal sequence deep learning model (such as the number of channels in each ConvLSTM unit, the size of the convolution kernel, the hidden layer dimension, etc.) need to be determined through iterative training experience to minimize the prediction error.

[0051] The performance of the ocean significant wave height prediction model based on the spatiotemporal series deep learning model is verified and evaluated by the following formula.

[0052] The first is the root mean square error, which indicates the deviation between the model results and the observed values. The smaller the root mean square error RMSE, the better the model performance. The formula is as follows: ; in, represents the land mask matrix, represents the root mean square error, Represents the amount of data involved in verification, Represents the sample points in the ocean area, To predict the results, For real data.

[0053] The correlation coefficient indicates the correlation between the model results and the observed values. The correlation coefficient CC formula is as follows: ; in, is the average value of the prediction results, is the average value of the real data.

[0054] Mean absolute error It is an objective function that represents the average absolute difference between the predicted value and the true value. It can measure the average error between the predicted value and the true value and has good robustness.

[0055] ;

[0056] The experimental data were processed according to the method of the present invention, and the experimental results were analyzed as follows:

[0057] In the forecast of the next 1 to 48 hours, the root mean square error RMSE, mean absolute error MAE, and mean absolute percentage error MAPE remain at a low level, which are 0.0806m~0.2213m, 0.0469m~0.1515m, and 5.1487%~15.6429% respectively. The correlation coefficient CC also remains between 0.9959 and 0.9728, which shows that the method provided by the present invention can accurately predict the changes in the future wave field.

[0058] Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by technicians in this technical field within the essential scope of the present invention should also fall within the protection scope of the present invention.

Claims

1. A method for predicting significant wave height based on a deep learning model of spatiotemporal sequences, characterized in that: The specific steps include: S1, obtain the historical wind speed data and historical wave significant wave height data of the test area; S2, the quality of the acquired historical wind speed data and historical wave significant wave height data is screened, and then spatiotemporal registration is performed to align the historical wind speed data and historical wave significant wave height data with a 1-hour offset in the time dimension; S3, use the sliding window algorithm to generate spatiotemporal sequence pairs and construct a training set; S4, build a spatiotemporal sequence deep learning model, including an encoding layer and a prediction layer. The encoding layer is responsible for encoding and memorizing the historical wave status, and the prediction layer is responsible for forecasting the waves based on the historical status and future wind speed; S5, inputting the training set constructed in step S3 into the spatiotemporal sequence deep learning model for iterative training, and obtaining a sea wave effective wave height prediction model based on the spatiotemporal sequence deep learning model; S6, according to the numerical forecast wind field provided by the outside, the acquired future wind speed data is input into the trained ocean wave significant wave height prediction model based on the spatiotemporal sequence deep learning model to obtain the forecast result.

2. The method for predicting significant wave height based on a spatiotemporal sequence deep learning model according to claim 1, characterized in that: Step S2 specifically includes the following steps: S2.1, quality screening of historical wind speed data and historical wave significant wave height data, including missing value processing and land marker masking; S2.2, perform spatiotemporal registration of historical wind speed data and historical wave significant height data.

3. The method for predicting significant wave height based on a spatiotemporal sequence deep learning model according to claim 2, characterized in that: Step S2.2 specifically includes the following steps: S2.2.1, perform interpolation processing to make the spatial resolution of historical wind speed data and historical wave significant wave height data consistent; S2.2.2, align the historical wind speed data and the historical wave significant wave height data in the time dimension by 1 hour, and integrate the historical wind speed data and the historical wave significant wave height data into a spatiotemporal series data of T×C×H×W, where T, C, H and W represent time, channel, height and width, respectively.

4. The method for predicting significant wave height based on a spatiotemporal sequence deep learning model according to claim 1, characterized in that: Step S3 is specifically as follows: using the sliding window algorithm, sliding sampling generates training samples, and integrating the training sample data set into N×T×C×H×W, where N, T, C, H and W represent the number of samples, time, channel, height and width respectively.

5. The method for predicting significant wave height based on a spatiotemporal sequence deep learning model according to claim 1, characterized in that: The encoding layer in step S4 includes: an encoded space-time sequence of multiple time steps in series, and the encoded space-time sequence of each time step includes: a first input layer, a first CBAM layer, and three layers of first ConvLSTM units based on convolutional long short-term memory units connected in series in sequence, and the first input layer inputs the wave effective wave height data SWH and the wind speed data of the next time step.

6. The method for predicting significant wave height based on a spatiotemporal sequence deep learning model according to claim 5, characterized in that: The prediction layer in step S4 includes: a prediction spatiotemporal sequence of multiple time steps connected in series, and the prediction spatiotemporal sequence of each time step includes: a second input layer, a second CBAM layer, a second ConvLSTM unit based on convolutional long and short-term memory units connected in series in sequence, and a convolutional layer Conv2D. The hidden layer state of the second ConvLSTM unit based on convolutional long and short-term memory units connected in series in three layers at this time step is formed through the convolutional layer Conv2D to form the effective wave height data SWH of the next time step. The effective wave height data SWH of the wave output by the convolutional layer Conv2D of each time step is used as the input data of the second input layer of the next time step. The second input layer also inputs the wind speed data of the previous time step, and outputs the prediction result after the prediction spatiotemporal sequence connected in series for multiple time steps.

7. The method for predicting significant wave height based on a spatiotemporal sequence deep learning model according to claim 1, characterized in that: Step S5 specifically includes: inputting the training set of historical wind speed data and historical wave data obtained in step S3 into the model, performing iterative training, and obtaining a sea wave effective wave height prediction model based on the spatiotemporal sequence deep learning model.

Images