Wave short-term prediction method and system based on support vector machine

Abstract

The application discloses a wave short-term prediction method and system based on a support vector machine, and belongs to the technical field of wave prediction. The method comprises the following steps: collecting a wave surface rising time sequence, and preprocessing to obtain a cleaned wave rising time sequence signal sequence; performing frequency spectrum conversion processing, screening a main frequency band, designing a band-pass filter suitable for the main frequency band, dividing the main frequency band into a plurality of sub-frequency bands, performing inverse transformation reconstruction, and obtaining a plurality of sub-frequency band wave rising time sequences; respectively constructing independent least squares support vector regression models and completing training, independently carrying out time sequence prediction, and obtaining prediction time sequences corresponding to the sub-frequency bands; performing linear superposition processing, simultaneously smoothing edges of the sub-frequency bands by using a preset window function, completing time sequence reconstruction, and obtaining a final prediction wave rising sequence, and completing prediction. The application can combine wave frequency domain information, improve the strong generalization ability of the model, effectively improve the precision and effective length of short-time prediction, and be relatively suitable for dynamic complex sea conditions.

Description

Technical Field

[0001] This invention belongs to the field of wave prediction technology, specifically relating to a short-term wave prediction method and system based on support vector machines. Background Technology

[0002] With the rapid development of industries such as marine engineering, offshore wind power, and marine ranching, accurate short-term wave forecasting is playing an increasingly important role in ensuring safe equipment operation, optimizing energy harvesting, and providing disaster early warning. Traditional wave forecasting mainly relies on numerical simulation methods based on physical models, such as spectral models and nonlinear shallow water equations. While these models are theoretically sound, they are highly dependent on marine environmental inputs and have certain limitations in terms of real-time performance, adaptability, and modeling of complex sea conditions.

[0003] As an improvement, machine learning techniques have been applied to ocean wave prediction in recent years. For example, models such as Artificial Neural Networks (ANNs) and Support Vector Regression (SVR) learn wave evolution patterns from large amounts of historical data, enabling rapid prediction of future wave states without complex physical modeling. However, current conventional techniques mostly rely on direct modeling based on time-domain signals, neglecting the frequency-domain structural characteristics of wave signals. Waves are nonlinear processes formed by the superposition of multiple frequency energies. Especially under complex sea conditions, the wave behavior of different frequency bands differs significantly. Direct modeling often results in low prediction accuracy, poor stability, and short effective duration due to frequency domain aliasing, limiting its practical application. Furthermore, conventional machine learning models are prone to overfitting, local optima, and rapid error accumulation when facing high-frequency disturbances and non-stationary wave signals. Therefore, in summary, current wave prediction does not fully utilize frequency domain features to directly model wave time-domain signals, leading to frequency domain aliasing, poor prediction accuracy and stability; the models may suffer from overfitting and error accumulation, resulting in relatively weak generalization ability. Summary of the Invention

[0004] This invention provides a short-term wave prediction method and system based on support vector machines. The purpose is to solve the problems in current wave prediction, which do not fully utilize frequency domain features to directly model wave time domain signals, resulting in frequency domain aliasing, poor prediction accuracy and stability, and the possibility of overfitting and error accumulation in the model, as well as relatively weak generalization ability.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: This invention provides a short-term wave prediction method based on support vector machines, comprising the following steps: The wave surface rise time series under measured or experimental conditions is collected, and preprocessing operations such as missing value imputation, trend term removal, baseline normalization and noise filtering are performed in sequence to obtain the cleaned wave rise time series signal sequence. The wave rise time series signal after cleaning is subjected to spectrum conversion processing to screen the main frequency band, and a bandpass filter adapted to the main frequency band is designed. The main frequency band is divided into multiple sub-frequency bands, and the signals of each sub-frequency band are reconstructed by inverse transformation to obtain multiple sub-frequency band wave rise time series. For each sub-frequency band wave rise time series, an independent least squares support vector regression model is constructed and trained to obtain multiple trained least squares support vector regression models corresponding to each sub-frequency band; time series prediction is carried out independently for the trained least squares support vector regression models of each sub-frequency band to obtain the predicted time series corresponding to each sub-frequency band. The predicted time series corresponding to all sub-bands are linearly superimposed, and a preset window function is used to smooth the edges of each sub-band. After time series reconstruction, the final predicted wave rise sequence is obtained, realizing short-term wave prediction based on support vector machine.

[0006] In some implementations, noise filtering is achieved using smoothing filtering, such as median filtering or wavelet denoising; baseline normalization processing includes normalizing the wave height signal of the wave surface rise time series to a preset numerical range.

[0007] In some implementations, the spectrum conversion is achieved through Fast Fourier Transform (FFT) to convert the time-domain wave signal into a frequency-domain spectral signal; the main frequency band is selected based on the wave spectrum energy distribution characteristics.

[0008] In some implementations, the bandpass filter is a Butterworth infinite impulse response filter, and the filter order is set to a preset reasonable order range to control the fluctuation in the passband to be within a preset minimum range; the main frequency band is divided into a preset number of sub-bands, and the width of each sub-band is consistent and within a reasonable range.

[0009] In some implementations, an autoregressive structure is used to design the input vector when constructing the least squares support vector regression model. The input vector consists of the sub-band wave rise values ​​at the current time and consecutive historical time points. The expression for the input vector is as follows: ; in, This is the model input vector at the current moment. This represents the current sub-band wave rise value. The number of historical wave height values ​​contained in the input vector.

[0010] In some implementations, the regression function expression of the least squares support vector regression model is as follows: ; in, The wave rise prediction value output by the model. This represents the total number of samples during the model training phase. For Lagrange multipliers, For the model kernel function, Let be the input vector of the sample to be predicted. For the first The input vector of each training sample, This is the model bias term.

[0011] Furthermore, the Gaussian kernel function is chosen, and its expression is: ; in, For any input vector, Let be the Euclidean distance between the two input vectors. This is the kernel width parameter.

[0012] In some implementations, the parameters of the least squares support vector regression model include regularization parameters and kernel width parameters, both of which are selected from a preset reasonable range of values; the parameter tuning process is achieved through cross-validation to select the optimal combination of hyperparameters.

[0013] In some implementations, the training phase of the least squares support vector regression model uses the least squares method to transform the model solution process into solving a system of linear equations. When using the trained least squares support vector regression model to carry out time series prediction, a multi-step iterative prediction mode is adopted, using the predicted value output by the model at the previous time step as the input of the model at the next time step to achieve continuous wave short-term prediction within a preset time period.

[0014] This invention also provides a short-term wave prediction system based on support vector machines (SVMs) to implement the aforementioned short-term wave prediction method based on SVMs. The system includes a data acquisition and preprocessing module, a data inverse transformation and reconstruction module, a time series prediction module, and a wave rise sequence module; wherein: The data acquisition and preprocessing module is used to: acquire wave surface rise time series under measured or experimental conditions, and sequentially perform preprocessing operations such as missing value imputation, trend term removal, baseline normalization and noise filtering to obtain the cleaned wave rise time series signal sequence. The data inverse transformation and reconstruction module is used to: perform spectral transformation processing on the cleaned wave rise time series signal sequence, filter the main frequency band, design a bandpass filter adapted to the main frequency band, divide the main frequency band into multiple sub-frequency bands, and then perform inverse transformation and reconstruction on the signals of each sub-frequency band to obtain multiple sub-frequency band wave rise time series. The time series prediction module is used to: construct and train an independent least squares support vector regression model for each sub-frequency band wave rise time series, thereby obtaining multiple trained least squares support vector regression models corresponding to each sub-frequency band; and independently perform time series prediction on the trained least squares support vector regression models for each sub-frequency band to obtain the predicted time series corresponding to each sub-frequency band. The wave rise sequence module is used to: linearly superimpose the predicted time series corresponding to all sub-frequency bands, and simultaneously use a preset window function to smooth the edges of each sub-frequency band. After completing the time series reconstruction, the final predicted wave rise sequence is obtained, realizing short-term wave prediction based on support vector machine.

[0015] Compared with existing technologies, the wave short-term prediction method and system based on support vector machines of the present invention have the following advantages: This invention effectively removes missing information, trend interference, and noise components from the original data through preprocessing of the wave surface rise time series, ensuring the regularity and reliability of the input data. Through spectrum conversion, main frequency band selection, and frequency division processing, the complex multi-frequency superimposed wave signal is decomposed into multiple independent sub-frequency band time series signals, which is relatively consistent with the essential characteristics of wave nonlinear evolution. This invention can, to a certain extent, avoid the frequency domain aliasing problem that easily occurs when directly modeling the full-band time domain signal. For each sub-frequency band, an independent least-squares support vector regression model is constructed and trained, enabling each model to specifically learn the wave fluctuation patterns of the corresponding frequency band. Compared to processing the full-band signal with a single model, this effectively improves the adaptability to wave characteristics of different frequency bands. Furthermore, this invention achieves accurate restoration of the wave time series signal through linear superposition of sub-frequency band prediction results and smooth reconstruction using window functions. It does not rely on complex ocean physical models, ensuring the real-time performance of short-term wave prediction and enhancing adaptability to dynamic and complex sea conditions, thus effectively improving the accuracy and stability of short-term wave prediction to a certain extent. Attached Figure Description

[0016] The accompanying drawings are provided to further understand the invention and constitute a part of this invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0017] Figure 1 This is a flowchart illustrating a short-term wave prediction method based on support vector machines according to the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0019] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0020] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0021] It should be noted that the apparatus and methods disclosed in the embodiments herein can also be implemented in other ways. The apparatus embodiments described above are merely illustrative; for example, the flowcharts and block diagrams in the accompanying drawings show the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments herein. In this regard, each block in a flowchart or block diagram may represent a module, program, or part of code containing one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system to perform the specified function or action, or can be implemented using a combination of dedicated hardware and computer instructions.

[0022] In addition, the functional modules in the various embodiments of this article can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0023] How can we propose a machine learning wave prediction method that combines wave frequency domain information and has strong generalization ability, so as to effectively improve the accuracy and effective duration of short-term prediction and thus adapt to dynamic and complex sea conditions?

[0024] Based on this, such as Figure 1 As shown, this invention provides a short-term wave prediction method based on support vector machines, comprising the following steps: The wave surface rise time series under measured or experimental conditions is collected, and preprocessing operations such as missing value imputation, trend term removal, baseline normalization and noise filtering are performed in sequence to obtain the cleaned wave rise time series signal sequence. The wave rise time series signal after cleaning is subjected to spectrum conversion processing to screen the main frequency band, and a bandpass filter adapted to the main frequency band is designed. The main frequency band is divided into multiple sub-frequency bands, and the signals of each sub-frequency band are reconstructed by inverse transformation to obtain multiple sub-frequency band wave rise time series. For each sub-frequency band wave rise time series, an independent least squares support vector regression model is constructed and trained to obtain multiple trained least squares support vector regression models corresponding to each sub-frequency band; time series prediction is carried out independently for the trained least squares support vector regression models of each sub-frequency band to obtain the predicted time series corresponding to each sub-frequency band. The predicted time series corresponding to all sub-bands are linearly superimposed, and a preset window function is used to smooth the edges of each sub-band. After time series reconstruction, the final predicted wave rise sequence is obtained, realizing short-term wave prediction based on support vector machine.

[0025] This invention integrates frequency domain decomposition technology with least squares support vector machine (LS-SVM). By decomposing wave signals into multiple frequency bands for independent modeling and then reconstructing them in time series, it can improve the predictive performance and engineering adaptability of the model to a certain extent.

[0026] Specifically, the noise filtering in this invention employs a smoothing filter method, using median filtering or wavelet denoising; baseline normalization processing includes normalizing the wave height signal of the wave surface rise time series to a preset numerical range; the smoothing filter method used can specifically filter out high-frequency clutter, and the normalization processing can unify the data amplitude range, further improving the purity of the preprocessed data. The spectrum conversion in this invention is achieved through Fast Fourier Transform (FFT), which can efficiently complete the conversion from the time domain to the frequency domain, and selecting the main frequency band based on energy distribution can accurately focus on effective wave information.

[0027] Furthermore, the bandpass filter of this invention uses a Butterworth infinite-length unit impulse response filter, with the filter order set within a preset reasonable range to control the ripple within the passband to a preset minimum range. The main frequency band is divided into a preset number of sub-bands, with each sub-band having a consistent width within a reasonable range. The Butterworth filter ensures the accuracy and smoothness of the frequency band segmentation, avoids introducing additional errors during the frequency division process, and provides a reliable decomposition signal for sub-band modeling.

[0028] Furthermore, this invention employs an autoregressive structure to design the input vector when constructing the least squares support vector regression model. The input vector consists of the wave rise values ​​of the sub-frequency bands at the current time and consecutive historical time points. This autoregressive input vector design fully utilizes historical time-series data, enhancing the model's ability to capture the evolution patterns of wave time series. The expression for the input vector is as follows: ; in, This is the model input vector at the current moment. This represents the current sub-band wave rise value. The number of historical wave height values ​​contained in the input vector.

[0029] The regression function expression for the least squares support vector regression model is as follows: ; in, The wave rise prediction value output by the model. This represents the total number of samples during the model training phase. For Lagrange multipliers, For the model kernel function, Let be the input vector of the sample to be predicted. For the first The input vector of each training sample, This is the model bias term.

[0030] The regression function uses the Gaussian kernel function, whose expression is as follows: ; in, For any input vector, Let be the Euclidean distance between the two input vectors. The kernel width parameter is used. By employing a Gaussian kernel function, the nonlinear characteristics of wave signals can be accurately adapted, improving the model's fitting effect and generalization ability.

[0031] In some operating conditions, this invention requires optimization of the least squares support vector regression model and improvement of the prediction execution process. By using a reasonable parameter value range and cross-validation tuning method, the model's fitting and generalization capabilities can be effectively balanced, reducing the risk of overfitting and underfitting. Furthermore, the least squares method is used in the training phase of the least squares support vector regression model to simplify the solution process and improve the model training efficiency. The multi-step iterative prediction mode can realize wave prediction for continuous duration, meeting the application needs of short-term prediction in practical engineering.

[0032] This invention also provides a short-term wave prediction system based on support vector machines, including a data acquisition and preprocessing module, a data inverse transformation and reconstruction module, a time series prediction module, and a wave rise sequence module; wherein: The data acquisition and preprocessing module is used to: acquire wave surface rise time series under measured or experimental conditions, and sequentially perform preprocessing operations such as missing value imputation, trend term removal, baseline normalization and noise filtering to obtain the cleaned wave rise time series signal sequence. The data inverse transformation and reconstruction module is used to: perform spectral transformation processing on the cleaned wave rise time series signal sequence, filter the main frequency band, design a bandpass filter adapted to the main frequency band, divide the main frequency band into multiple sub-frequency bands, and then perform inverse transformation and reconstruction on the signals of each sub-frequency band to obtain multiple sub-frequency band wave rise time series. The time series prediction module is used to: construct and train an independent least squares support vector regression model for each sub-frequency band wave rise time series, thereby obtaining multiple trained least squares support vector regression models corresponding to each sub-frequency band; and independently perform time series prediction on the trained least squares support vector regression models for each sub-frequency band to obtain the predicted time series corresponding to each sub-frequency band. The wave rise sequence module is used to: linearly superimpose the predicted time series corresponding to all sub-frequency bands, and simultaneously use a preset window function to smooth the edges of each sub-frequency band. After completing the time series reconstruction, the final predicted wave rise sequence is obtained, realizing short-term wave prediction based on support vector machine.

[0033] The present invention will be further described in detail below through specific embodiments.

[0034] This invention discloses a short-term wave prediction method based on support vector machines, which is carried out according to the following steps: Wave surface rise time series obtained under actual measurement or experimental conditions, with a sampling frequency of not less than 100Hz; Missing values ​​were filled in and trend terms were removed from the data; the wave height signal was normalized to the range of [-1,1] or [0,1]; smoothing filtering (median filtering or wavelet denoising) was used to remove high-frequency noise; and the cleaned wave rise time series signal was obtained. The cleaned wave rise time series signal was converted into a frequency domain spectrum using Fast Fourier Transform (FFT). Used to analyze regions of concentrated energy; Based on experience and spectral energy distribution, only frequency components in the range of 0.1Hz to 2.0Hz are retained (to filter out low-frequency background drift and high-frequency noise), and the main frequency band is selected. A Chebyshev type II IIR filter is used, with an order of 4 to 6, and the passband ripple is controlled to be less than 0.1dB, to design a bandpass filter. The main frequency band is divided into N sub-bands (N=5), each with a width of 0.3 to 0.5Hz.

[0035] Perform IFFT on each sub-band signal to obtain the sub-time series. ; The set of sub-band wave rise time series is as follows: ; For each sub-band Construct an independent least squares support vector regression model (LS-SVM) for time series forecasting: Using an autoregressive structure, the past d wave height values ​​are used as the input vector. : ; Output y t =η(t+1); The core function of the model uses the following regression function: ; Among them, kernel function Linear kernels can be used ; Gaussian kernel (RBF): ; The parameter γ (regularization) controls the balance between fitting and generalization ability, and σ is the kernel width parameter.

[0036] As an optimization, the following parameters are tuned.

[0037] γ ∈ [10 - ², 10³], σ ∈ [0.01, 5.0]; The optimal combination of hyperparameters is selected through k-fold cross-validation; The least squares method is used instead of the SVM's insensitive loss function ε to improve the solution efficiency (the solution is a system of linear equations). During the testing phase, the model can iteratively use the previous step's prediction as the next input to achieve T. p Continuous prediction in steps (e.g., 6 seconds); Output the predicted time series for each frequency band. ; The predicted time series of each sub-band are linearly superimposed: ; Note the transition band issue caused by the overlap of filter frequency responses between different frequency bands. Window functions (such as the Hanning window) can be used to smooth band edges and reduce spectral leakage. Output the final predicted wave rise sequence. .

[0038] In summary, this invention provides a short-term wave prediction method based on support vector machines (SVMs). It ensures the regularity and reliability of the input wave time-series data. By introducing spectrum transformation and frequency division processing, it decomposes complex nonlinear wave signals into multiple sub-frequency band sequences, improving the frequency domain aliasing problem in full-band time-domain modeling. This invention constructs independent least-squares support vector regression models for each sub-frequency band, specifically learning the wave patterns of different frequency bands and improving feature adaptability. Signal reconstruction is completed through linear superposition and window function smoothing to restore the wave time series. It does not rely excessively on complex ocean physical models, possesses a certain degree of reliability, and can adapt to dynamically complex sea conditions.

[0039] Finally, it should be noted that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Anyone skilled in the art can readily implement the present invention according to the description and above. Any modifications, alterations, or equivalent variations made using the technical content disclosed above are equivalent embodiments of the present invention. Furthermore, any modifications, alterations, or variations made to the above embodiments based on the essential technology of the present invention are still within the protection scope of the present invention.

Claims

1. A short-term wave prediction method based on support vector machines, characterized in that, Includes the following steps: The wave surface rise time series under measured or experimental conditions is collected, and preprocessing operations such as missing value imputation, trend term removal, baseline normalization and noise filtering are performed in sequence to obtain the cleaned wave rise time series signal sequence. The wave rise time series signal after cleaning is subjected to spectrum conversion processing to screen the main frequency band, and a bandpass filter adapted to the main frequency band is designed. The main frequency band is divided into multiple sub-frequency bands, and the signals of each sub-frequency band are reconstructed by inverse transformation to obtain multiple sub-frequency band wave rise time series. For each sub-frequency band wave rise time series, an independent least squares support vector regression model is constructed and trained to obtain multiple trained least squares support vector regression models corresponding to each sub-frequency band; time series prediction is carried out independently for the trained least squares support vector regression models of each sub-frequency band to obtain the predicted time series corresponding to each sub-frequency band. The predicted time series corresponding to all sub-bands are linearly superimposed, and a preset window function is used to smooth the edges of each sub-band. After time series reconstruction, the final predicted wave rise sequence is obtained, realizing short-term wave prediction based on support vector machine.

2. The wave short-term prediction method based on support vector machine according to claim 1, characterized in that, Noise filtering is achieved using a smoothing filter method, employing median filtering or wavelet denoising; baseline normalization processing includes normalizing the wave height signal of the wave surface rise time series to a preset numerical range.

3. The short-term wave prediction method based on support vector machine according to claim 1, characterized in that, Spectrum conversion is achieved through Fast Fourier Transform, which is used to convert time-domain wave signals into frequency-domain spectral signals; the main frequency band is selected based on the energy distribution characteristics of the wave spectrum.

4. The short-term wave prediction method based on support vector machine according to claim 1, characterized in that, The bandpass filter is a Butterworth infinite impulse response filter, and the filter order is set to a preset reasonable order range to control the fluctuation in the passband to be within a preset minimum range. The main frequency band is divided into a preset number of sub-bands, and the width of each sub-band is consistent and within a reasonable range.

5. The short-term wave prediction method based on support vector machine according to claim 1, characterized in that, When constructing the least squares support vector regression model, an autoregressive structure is used to design the input vector. The input vector consists of the wave rise values ​​of the sub-frequency bands at the current time and the historical consecutive time points. The expression of the input vector is as follows: ； in, This is the model input vector at the current moment. This represents the current sub-band wave rise value. The number of historical wave height values ​​contained in the input vector.

6. The short-term wave prediction method based on support vector machine according to claim 1, characterized in that, The regression function expression for the least squares support vector regression model is as follows: ； in, The wave rise prediction value output by the model. This represents the total number of samples during the model training phase. For Lagrange multipliers, For the model kernel function, Let be the input vector of the sample to be predicted. For the first The input vector of each training sample, This is the model bias term.

7. The short-term wave prediction method based on support vector machine according to claim 6, characterized in that, The kernel function chosen is the Gaussian kernel function, whose expression is: ； in, For any input vector, Let be the Euclidean distance between the two input vectors. This is the kernel width parameter.

8. The short-term wave prediction method based on support vector machine according to claim 1, characterized in that, The parameters of the least squares support vector regression model include regularization parameters and kernel width parameters, both of which are selected from a preset reasonable range of values. The parameter tuning process is achieved through cross-validation to select the optimal combination of hyperparameters.

9. The short-term wave prediction method based on support vector machine according to claim 1, characterized in that, During the training phase of the least squares support vector regression model, the least squares method is used to transform the model solution process into solving a system of linear equations. When using the trained least squares support vector regression model for time series prediction, a multi-step iterative prediction mode is adopted, using the predicted value output by the model at the previous time step as the input of the model at the next time step to achieve continuous wave short-term prediction within a preset time period.

10. A short-term wave prediction system based on support vector machines, used to implement the short-term wave prediction method based on support vector machines as described in any one of claims 1-9, characterized in that, It includes a data acquisition and preprocessing module, a data inverse transformation and reconstruction module, a time series prediction module, and a wave rise sequence module; among which: The data acquisition and preprocessing module is used to: acquire wave surface rise time series under measured or experimental conditions, and sequentially perform preprocessing operations such as missing value imputation, trend term removal, baseline normalization and noise filtering to obtain the cleaned wave rise time series signal sequence. The data inverse transformation and reconstruction module is used to: perform spectral transformation processing on the cleaned wave rise time series signal sequence, filter the main frequency band, design a bandpass filter adapted to the main frequency band, divide the main frequency band into multiple sub-frequency bands, and then perform inverse transformation and reconstruction on the signals of each sub-frequency band to obtain multiple sub-frequency band wave rise time series. The time series prediction module is used to: construct and train an independent least squares support vector regression model for each sub-frequency band wave rise time series, thereby obtaining multiple trained least squares support vector regression models corresponding to each sub-frequency band; and independently perform time series prediction on the trained least squares support vector regression models for each sub-frequency band to obtain the predicted time series corresponding to each sub-frequency band. The wave rise sequence module is used to: linearly superimpose the predicted time series corresponding to all sub-frequency bands, and simultaneously use a preset window function to smooth the edges of each sub-frequency band. After completing the time series reconstruction, the final predicted wave rise sequence is obtained, realizing short-term wave prediction based on support vector machine.

Images