Timing sequence generative model-based significant wave height space-time prediction method and system
Through the method based on the time-series generation model, ERA5 reanalysis data is used to predict the effective wave height of the ocean wave, which solves the problems of high computing costs and limited prediction capabilities in the prior art, and achieves more efficient and accurate wave height prediction.
Patent Information
- Application Number
- CN202411989902.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-31
- Publication Date
- 2025-04-29
AI Technical Summary
The prior art has problems such as high computational cost, strong dependence on initial and boundary conditions, limited prediction ability of data sparse data, and limited generalization ability and interpretability of model in the prediction of effective waves.
Using a time-series generative model method, the data is acquired by obtaining historical ERA5 and reanalyzing, encoding and dimensionality reduction processing is used to generate discrete variables, and input them together with the initial data to the DALSTM model, extracting and constructing feature matrices, and using ConvLSTMCell unit for prediction.
It effectively reduces the data dimension, improves the extraction and fusion of wave features, and improves the accuracy and efficiency of effective wave height prediction.
Smart Images
Figure CN119903476A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field related to wave height prediction, and in particular relates to a method and system for spatiotemporal prediction of effective wave height based on a time series generative model. Background Art
[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.
[0003] In the fields of oceanography and meteorology, accurate prediction of effective wave height is crucial for marine engineering, navigation and environmental monitoring. The effective wave height is a key parameter that describes the energy and size of waves, and has an important impact on marine operation safety, marine resource development, coastal disaster prevention and mitigation, etc. Traditional wave prediction methods mainly rely on physical models and statistical models. These methods can provide wave height predictions to a certain extent, but they have certain limitations.
[0004] Physical models, such as SWAN (Simulating WAves Nearshore) and WAVEWATCH III, simulate the generation, development and extinction of waves by analyzing physical equations. These models can provide relatively accurate wave height predictions, but they are computationally expensive and highly dependent on initial and boundary conditions, which limits their prediction capabilities in data-sparse areas.
[0005] Statistical models, such as neural networks and machine learning methods, predict significant wave heights by learning patterns from historical data. These methods excel in processing large amounts of data and capturing nonlinear relationships, but often require large amounts of labeled data and have limited generalization and interpretability.
[0006] With the development of remote sensing technology and numerical simulation technology, wave prediction methods based on reanalysis data have gradually attracted attention. Reanalysis datasets, such as ERA5 (the fifth generation of European Centre for Medium-Range Weather Forecasts reanalysis data), provide high-resolution global meteorological and oceanographic variables, providing a rich source of information for wave prediction. However, how to effectively extract useful information from these high-dimensional data to achieve more accurate predictions of significant wave heights is a problem that needs to be solved. Summary of the invention
[0007] In order to overcome the shortcomings of the above-mentioned prior art, the present invention provides a spatiotemporal prediction method and system for significant wave height based on a time series generative model, which effectively reduces the dimension of data through data dimensionality reduction, comprehensively utilizes discrete variables and initial data, realizes in-depth extraction and fusion of wave characteristics, and improves the accuracy and efficiency of subsequent significant wave height prediction.
[0008] In order to achieve the above object, the present invention adopts the following technical solution:
[0009] In a first aspect, the present invention provides a spatiotemporal prediction method for significant wave height based on a time series generative model, comprising:
[0010] Obtain historical ERA5 reanalysis data for the target area corresponding to the target prediction time;
[0011] The acquired historical ERA5 reanalysis data of the target area are input into the trained prediction model. When the prediction model processes the historical ERA5 reanalysis data of the target area, the key characteristic information of the wave field is encoded based on the historical ERA5 reanalysis data of the target area, and the dimension reduction processing is performed to generate discrete variables from the dimension-reduced data;
[0012] The generated discrete variables and the corresponding historical ERA5 reanalysis data are input into the DALSTM model of the prediction model to extract and construct a feature matrix;
[0013] Based on the feature matrix, the ConvLSTMCell unit of the prediction model is used to perform prediction to obtain the effective wave height corresponding to the target prediction time.
[0014] In a second aspect, the present invention provides a spatiotemporal prediction system for significant wave height based on a time series generative model, comprising:
[0015] An acquisition module is configured to: acquire historical ERA5 reanalysis data of a target area corresponding to a target prediction time;
[0016] A discrete module is configured to: input the acquired historical ERA5 reanalysis data of the target area into a trained prediction model, and when the prediction model processes the historical ERA5 reanalysis data of the target area, based on the historical ERA5 reanalysis data of the target area, encode key characteristic information of the wave field and perform dimensionality reduction processing, and generate discrete variables from the dimensionality reduced data;
[0017] An extraction and construction module is configured to: input the generated discrete variables and the corresponding historical ERA5 reanalysis data into the DALSTM model of the prediction model to extract and construct a feature matrix;
[0018] The prediction module is configured to: perform prediction based on the feature matrix using the ConvLSTMCell unit of the prediction model to obtain the effective wave height corresponding to the target prediction time.
[0019] In a third aspect, the present invention provides an electronic device comprising a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the method described in the first aspect is performed.
[0020] In a fourth aspect, the present invention provides a computer-readable storage medium for storing computer instructions, wherein when the computer instructions are executed by a processor, the method described in the first aspect is performed.
[0021] In a fifth aspect, the present invention provides a computer program product, comprising a computer program, which, when executed by a processor, implements the method described in the first aspect.
[0022] One or more of the above technical solutions have the following beneficial effects:
[0023] In the present invention, the historical ERA5 reanalysis data of the target area corresponding to the target prediction time are encoded and dimensionally reduced to generate discrete variables, which effectively reduces the dimension of the data while retaining key information; the generated discrete variables and the corresponding historical ERA5 reanalysis data of the target area are jointly input into the DALSTM model, which can comprehensively utilize the discrete variables and the initial data to achieve in-depth extraction and fusion of wave characteristics, thereby improving the accuracy and efficiency of subsequent effective wave height prediction.
[0024] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS
[0025] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.
[0026] Figure 1 This is a network structure diagram of the prediction model in Embodiment 1 of the present invention;
[0027] Figure 2 This is a structural diagram of the WaveVae module in the first embodiment of the present invention;
[0028] Figure 3 This is a structural diagram of the DALSTM model in Embodiment 1 of the present invention;
[0029] Figure 4 Schematic diagram of the DALSTM model processing process in Example 1 of the present invention. DETAILED DESCRIPTION
[0030] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.
[0031] It should be noted that the terms used herein are for describing specific embodiments only and are not intended to be limiting of exemplary embodiments according to the present invention.
[0032] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.
[0033] Embodiment 1
[0034] This embodiment discloses a spatiotemporal prediction method for significant wave height based on a time series generative model, including:
[0035] Step 1: Obtain the historical ERA5 reanalysis data of the target area corresponding to the target prediction time;
[0036] Step 2: input the acquired historical ERA5 reanalysis data of the target area into the trained prediction model. When the prediction model processes the historical ERA5 reanalysis data of the target area, the key characteristic information of the wave field is encoded based on the historical ERA5 reanalysis data of the target area, and the dimension reduction processing is performed to generate discrete variables from the dimension-reduced data;
[0037] The generated discrete variables and the corresponding historical ERA5 reanalysis data are input into the DALSTM model of the prediction model to extract and construct a feature matrix;
[0038] Based on the feature matrix, the ConvLSTMCell unit of the prediction model is used to perform prediction to obtain the effective wave height corresponding to the target prediction time.
[0039] This embodiment achieves efficient and accurate prediction of ocean wave significant wave height data by combining the ERA5 reanalysis data package and advanced machine learning technology.
[0040] The core of this embodiment lies in the encoding dimension reduction and DALSTM model of the WaveVae module, which work together to improve the efficiency of data compression and ensure the maximum retention of information integrity by performing double dimensionality reduction compression on spatiotemporal data. This method has broad application prospects in the fields of oceanographic research, wave energy utilization, and maritime safety, and can provide strong technical support for research and practice in related fields. With the implementation of the technical solution of this embodiment, it is expected to bring significant technological progress and economic benefits in the fields of marine environment monitoring, wave energy utilization, and maritime safety.
[0041] Combine the following Figure 1-Figure 3 The present embodiment is described in detail, specifically:
[0042] Step 1: Obtain the historical ERA5 reanalysis data for the target area corresponding to the target prediction time.
[0043] The ERA5 reanalysis datasets are all continuous time series with a temporal resolution of 1 h and a spatial resolution of 0.25°x0.25°, and are divided into training, validation, and test sets in a time length ratio of 3:1:1.
[0044] Specifically, it includes the significant wave height of ERA5 reanalysis data, the U component of sea surface wind of ERA5 reanalysis data, and the V component of sea surface wind of ERA5 reanalysis data.
[0045] Step 2: Input the acquired historical ERA5 reanalysis data of the target area into the trained prediction model. When the prediction model processes the historical ERA5 reanalysis data of the target area, it encodes the key feature information of the wave field and performs dimensionality reduction based on the historical ERA5 reanalysis data of the target area, and generates discrete variables from the reduced dimensionality data; the generated discrete variables and the corresponding historical ERA5 reanalysis data are input into the DALSTM model of the prediction model to extract and construct the feature matrix; based on the feature matrix, the ConvLSTMCell unit of the prediction model is used to perform prediction to obtain the effective wave height corresponding to the target prediction time.
[0046] The prediction model of this embodiment is the WavePrediction model, and the WavePrediction model architecture consists of two main modules: the WaveVae module and the DALSTM module. The WaveVae module is responsible for information compression of the initial field data, while the DALSTM module focuses on data feature extraction and time series prediction.
[0047] The input data for WavePredication model training is the historical ERA5 reanalysis data from time xt to time x, and the output prediction data is the effective wave height of the entire target area.
[0048] The training process of the WavePrediction model is divided into two stages: in the first stage, the WaveVae generative model is trained to learn the potential representation of the data; in the second stage, the DALSTM module is trained, which fuses the features of discrete variable data and performs time series prediction of effective wave height.
[0049] like Figure 2 As shown, the input data sets in the WaveVae module are the significant wave heights of the historical ERA5 reanalysis data, the components of the sea surface wind V and U of the ERA5 reanalysis data, and the output is the significant wave height of the ERA5 reanalysis data.
[0050] First, the WaveVae module is trained using the historical ERA5 reanalysis dataset. The WaveVae module first encodes and reduces the dimension of the significant wave height data, the component of the sea surface wind V, and the component of the sea surface wind U through the Conv3DEencode layer, and then generates the reduced-dimensional data into discrete variables through the codebook technology. Finally, 3DDecode is used to restore the discrete variables to the corresponding significant wave height data.
[0051] The DALSTM fusion module consists of two key components: CNNDecoder and data feature extraction fusion layer. These two components work together to achieve efficient processing and feature extraction of input data. Among them, the CNNDecoder component focuses on processing discrete variables in the data. Its role is to decode these discrete variables, that is, to convert them into a form that can be further analyzed. This step is necessary because discrete variables are usually contained in the original data, and if they are not properly decoded, their characteristics and information cannot be used by subsequent processing steps.
[0052] The data feature extraction fusion layer extracts the features of the input data through spatial attention and channel attention. The data feature extraction fusion layer intervenes to perform in-depth feature extraction on the decoded data. This layer is implemented through two attention mechanisms: spatial attention and channel attention. The spatial attention mechanism focuses on the spatial dimension of the data. It can identify the most important features in a specific area, thereby helping the model focus on the information that is most critical to the final task. The channel attention mechanism focuses on the channel dimension of the data. It can identify the most informative features in different channels, further enhancing the model's understanding of the data.
[0053] The input data of the DALSTM module is divided into two main parts: the first part is the initial field data of the waves, which provides the initial conditions of the wave state; the second part is the discrete variables generated by the WaveVae module after compressing the initial field of the waves. These discrete variables represent the compressed encoding of the initial field data, which aims to capture and encode the key characteristic information of the wave field. Through this structural design, the DALSTM module can comprehensively utilize the initial field data and the compressed discrete variables to achieve in-depth extraction and effective fusion of wave characteristics, thereby improving the accuracy and efficiency of wave height prediction. The specific execution process types of the model are as follows.
[0054] CNNDecoder is used to decode the discrete variables and generate the second feature F2, which contains all the required compressed information in the variable. The significant wave height data, the component of the sea surface wind V and the component data X of the sea surface wind U in EAR5 are normalized into the first feature F1, and the normalized input of the first feature F1 and the second feature F2 input features are superimposed in dimension by concatenating (connection) to form the third feature F3. This method enhances the expressive power of the network because it allows the network to retain more information at a deeper level. The third feature F3 then generates the fourth feature F4 through channel attention, and then the fourth feature F4 and the third feature F3 are multiplied by element multiplication to generate the fifth feature F5. This step helps the model to make more refined adjustments at the feature level. The fifth feature F5 generates the sixth feature F6 through spatial attention, and then uses element-wise multiplication to multiply the sixth feature F6 and the fifth feature F5 to generate the seventh feature F7. The goal of this step is to reduce information reduction and amplify the feature matrix features, so that the wave input data and the feature matrix can obtain global dimensional interactive features; the next step is to use the feature fusion layer to convolve the seventh feature F7 to further extract the feature information in the seventh feature F7, and compress the channel of the seventh feature F7 matrix to half of the original, and generate the input eighth feature F8, as shown Figure 4 shown.
[0055] The eighth feature F8 of the feature matrix output by the DALSTM model is input into the ConvLSTMCell unit. ConvLSTMCell extracts the features of spatiotemporal data by using convolution operations in input-to-state and state-to-state conversions instead of traditional fully connected operations. This method allows the model to capture local correlations in space and long-term dependencies in time. By stacking multiple ConvLSTMCell units, a coding prediction structure is formed, and an end-to-end training model is established to predict future significant wave heights.
[0056] This embodiment performs denormalization on the prediction data and uses the mean absolute error (MAE), mean square error (MSE), and root mean square error (RMSE) to score the prediction results.
[0057] This embodiment proposes a spatiotemporal prediction method for effective wave height based on a generated prediction model, which achieves efficient and accurate prediction of effective wave height data. Compared with traditional methods, this embodiment specifically uses WaveVae and DALSTM technology to perform double dimensionality reduction compression on spatiotemporal data, which not only improves the efficiency of data compression, but also ensures the maximum retention of information integrity. This innovation is not only groundbreaking in theory, but also shows significant practical value in practical applications, providing a more accurate prediction tool for fields such as marine engineering, marine navigation, and environmental monitoring.
[0058] This embodiment shows significant innovation in the field of spatiotemporal prediction of ocean significant wave height. Its WaveVae module and related training methods provide a novel technical solution through precise data processing and advanced algorithm optimization. This method not only improves the accuracy of prediction, but also optimizes the computational efficiency through the application of data dimension reduction and generative models. Therefore, the present invention has broad application prospects in the fields of oceanographic research, wave energy utilization, and navigation safety, and can provide strong technical support for research and practice in related fields. With the implementation of the technical solution of the present invention, it is expected to bring significant technological progress and economic benefits in the fields of marine environmental monitoring, wave energy utilization, and navigation safety. The innovation and practicality of the present invention not only provides a new perspective for the field of oceanographic research, but also opens up new possibilities for the development and utilization of marine resources, indicating that it will have a far-reaching impact in related fields.
[0059] Embodiment 2
[0060] The purpose of this embodiment is to provide a spatiotemporal prediction system for significant wave height based on a time series generative model, including:
[0061] An acquisition module is configured to: acquire historical ERA5 reanalysis data of a target area corresponding to a target prediction time;
[0062] A discrete module is configured to: input the acquired historical ERA5 reanalysis data of the target area into a trained prediction model, and when the prediction model processes the historical ERA5 reanalysis data of the target area, encode and reduce the dimension of the historical ERA5 reanalysis data of the target area, and generate discrete variables from the reduced dimension data;
[0063] An extraction and construction module is configured to: input the generated discrete variables and the corresponding historical ERA5 reanalysis data into the DALSTM model of the prediction model to extract and construct a feature matrix;
[0064] The prediction module is configured to: based on the feature matrix, use the ConvLSTMCell unit of the prediction model to make a prediction to obtain the effective wave height corresponding to the target prediction time
[0065] In further embodiments, there is also provided:
[0066] An electronic device includes a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the method described in Embodiment 1 is performed. For the sake of brevity, no further description is given here.
[0067] It should be understood that in this embodiment, the processor may be a central processing unit CPU, and the processor may also be other general-purpose processors, digital signal processors DSP, application-specific integrated circuits ASIC, off-the-shelf programmable gate arrays FPGA or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.
[0068] The memory may include a read-only memory and a random access memory, and provide instructions and data to the processor. A portion of the memory may also include a non-volatile random access memory. For example, the memory may also store information about the device type.
[0069] A computer-readable storage medium is used to store computer instructions. When the computer instructions are executed by a processor, the method described in embodiment 1 is completed.
[0070] The method in the first embodiment can be directly embodied as a hardware processor, or a combination of hardware and software modules in the processor. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware. To avoid repetition, it will not be described in detail here.
[0071] A computer program product includes a computer program, and when the computer program is executed by a processor, the method described in the first embodiment is implemented.
[0072] The present invention also provides at least one computer program product tangibly stored on a non-transitory computer-readable storage medium. The computer program product includes computer executable instructions, such as instructions included in a program module, which are executed in a device on a real or virtual processor of the target to perform the process / method as described above. Typically, program modules include routines, programs, libraries, objects, classes, components, data structures, etc. that perform specific tasks or implement specific abstract data types. In various embodiments, the functions of program modules can be combined or divided between program modules as needed. Machine executable instructions for program modules can be executed in local or distributed devices. In distributed devices, program modules can be located in local and remote storage media.
[0073] The computer program code for implementing the method of the present invention can be written in one or more programming languages. These computer program codes can be provided to the processor of a general-purpose computer, a special-purpose computer or other programmable data processing device, so that the program code, when executed by the computer or other programmable data processing device, causes the function / operation specified in the flow chart and / or block diagram to be implemented. The program code can be executed completely on a computer, partially on a computer, as an independent software package, partially on a computer and partially on a remote computer or completely on a remote computer or server.
[0074] In the context of the present invention, computer program codes or related data may be carried by any appropriate carrier to enable a device, apparatus or processor to perform the various processes and operations described above. Examples of carriers include signals, computer readable media, etc. Examples of signals may include electrical, optical, radio, acoustic or other forms of propagation signals, such as carrier waves, infrared signals, etc.
[0075] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with this embodiment can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.
[0076] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.
Claims
1. A spatiotemporal prediction method for significant wave height based on a time series generative model, characterized in that: include: Obtain historical ERA5 reanalysis data for the target area corresponding to the target prediction time; The acquired historical ERA5 reanalysis data of the target area are input into the trained prediction model. When the prediction model processes the historical ERA5 reanalysis data of the target area, the key characteristic information of the wave field is encoded based on the historical ERA5 reanalysis data of the target area, and the dimension reduction processing is performed to generate discrete variables from the dimension-reduced data; The generated discrete variables and the corresponding historical ERA5 reanalysis data are input into the DALSTM model of the prediction model to extract and construct a feature matrix; Based on the feature matrix, the ConvLSTMCell unit of the prediction model is used to perform prediction to obtain the effective wave height corresponding to the target prediction time.
2. The spatiotemporal prediction method of significant wave height based on a time series generative model according to claim 1, characterized in that: The historical ERA5 reanalysis data of the target area include significant wave height, U component of sea surface wind and V component of sea surface wind.
3. The spatiotemporal prediction method of significant wave height based on a time series generative model according to claim 1, characterized in that: The historical ERA5 reanalysis data of the target area are encoded and reduced in dimension, and the reduced-dimensional data are converted into discrete variables, specifically: The Conv3D Eencode layer is used to encode and reduce the dimension of the historical ERA5 reanalysis data of the target area; The codebook technology is used to generate the reduced-dimensional data into discrete variables.
4. The spatiotemporal prediction method of significant wave height based on a time series generative model according to claim 1, characterized in that: The generated discrete variables and the corresponding historical ERA5 reanalysis data are input into the DALSTM model of the forecasting model to extract and construct the feature matrix, specifically: Use the CNNDecoder layer to decode the generated discrete variables to obtain the decoded features; Normalizing the significant wave height data, the component of the sea surface wind V, and the component of the sea surface wind U corresponding to the discrete variables to obtain normalized features; The decoded features and normalized features are fused using the DALSTM module to obtain the feature matrix.
5. The spatiotemporal prediction method of significant wave height based on a time series generative model according to claim 1, characterized in that: The decoded features and normalized features are fused using the DALSTM module to obtain the feature matrix, which is: The decoded features and normalized features are superimposed on the dimension to form the third feature; The third feature is passed through a channel attention module to generate a fourth feature; Multiplying the fourth feature and the third feature by element-wise multiplication to obtain a fifth feature; The fifth feature is passed through a spatial attention module to generate a sixth feature; The sixth feature and the fifth feature are multiplied by element-wise multiplication to generate a seventh feature, and channel compression is performed on the seventh feature to obtain a feature matrix.
6. The spatiotemporal prediction method of significant wave height based on a time series generative model according to claim 1 or 4, characterized in that: The feature matrix is predicted by stacking multiple ConvLSTMCell units to obtain the effective wave height corresponding to the target prediction time.
7. A spatiotemporal prediction system for significant wave height based on a time series generative model, characterized in that: include: An acquisition module is configured to: acquire historical ERA5 reanalysis data of a target area corresponding to a target prediction time; A discrete module is configured to: input the acquired historical ERA5 reanalysis data of the target area into a trained prediction model, and when the prediction model processes the historical ERA5 reanalysis data of the target area, based on the historical ERA5 reanalysis data of the target area, encode key characteristic information of the wave field and perform dimensionality reduction processing, and generate discrete variables from the dimensionality reduced data; An extraction and construction module is configured to: input the generated discrete variables and the corresponding historical ERA5 reanalysis data into the DALSTM model of the prediction model to extract and construct a feature matrix; The prediction module is configured to: perform prediction based on the feature matrix using the ConvLSTMCell unit of the prediction model to obtain the effective wave height corresponding to the target prediction time.
8. An electronic device, characterized in that: The method comprises a memory and a processor and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the method according to any one of claims 1 to 6 is completed.
9. A computer-readable storage medium, characterized in that: Used to store computer instructions, which, when executed by a processor, complete the method described in any one of claims 1 to 6.
10. A computer program product, characterized in that The invention comprises a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 6.