Typhoon wind speed prediction method and system based on deep-pred framework
The typhoon wind speed prediction method built using the Deep-Pred framework solves the problems of computational resources and data dependence in existing technologies, achieving efficient typhoon wind speed prediction and improving prediction accuracy.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- WUHAN UNIV
- Filing Date
- 2024-01-17
- Publication Date
- 2026-05-19
AI Technical Summary
Existing typhoon wind speed forecasting methods rely on a large amount of computing resources and time, and their accuracy is highly dependent on the quality of the initial data, making real-time forecasting difficult.
The Deep-Pred framework is used to collect typhoon wind speed data, identify influencing factors, construct an environmental field of influencing factors, divide the dataset, and build a Deep-Pred framework containing multiple modules for model training and validation. The training results are used to adjust the model parameters to obtain a typhoon wind speed prediction model.
It can accurately extract the spatiotemporal features of multidimensional typhoon information, improving the prediction accuracy of typhoon intensity change trends and maximum wind speed, which is superior to existing models.
Smart Images
Figure CN118035670B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of data prediction technology, and in particular to a typhoon wind speed prediction method and system based on the Deep-Pred framework. Background Technology
[0002] Typhoon forecasting is a very important branch of extreme weather forecasting.
[0003] Currently, models for predicting typhoon intensity can be broadly categorized into three types: dynamic models, statistical models, and deep learning models. Dynamic models primarily utilize the physical and dynamic characteristics of typhoons to construct mathematical models, further predicting their intensity and trajectory. They exhibit high accuracy in predicting typhoon trajectories. However, a drawback of dynamic models is their heavy reliance on substantial computational resources and time. Furthermore, their accuracy is highly dependent on the quality of the initial data, which may hinder real-time prediction capabilities. Statistical model-based prediction methods mainly establish robust statistical models through the analysis and comparison of historical data. Common techniques used in this approach include regression analysis and Bayesian statistics. This method facilitates the analysis and comparison of historical data, leading to highly reliable predictions. However, compared to other methods, the prediction accuracy of statistical models is relatively low. Deep learning models possess the ability to capture sequential information from historical data and have been successfully applied to various prediction tasks. This method utilizes deep neural network models to extract relevant factors from historical data and predict future typhoon trends, achieving significant progress in predicting typhoon intensity.
[0004] Currently, typhoon intensity is influenced by a variety of factors, making prediction quite difficult. The most direct indicator of typhoon intensity is wind speed; therefore, a new method for predicting typhoon wind speed is needed. Summary of the Invention
[0005] This invention provides a typhoon wind speed prediction method and system based on the Deep-Pred framework, which addresses the shortcomings of existing typhoon wind speed prediction technologies.
[0006] In a first aspect, the present invention provides a typhoon wind speed prediction method based on the Deep-Pred framework, comprising:
[0007] Collect typhoon wind speed data for a preset year and time period to determine the factors affecting typhoon wind speed;
[0008] Based on the typhoon wind speed influencing factor data, an influencing factor environmental field is constructed, and a new typhoon dataset is formed by the influencing factor environmental field and the typhoon wind speed data.
[0009] The new typhoon dataset is divided into training set, validation set and test set;
[0010] Construct a Deep-Pred framework, use the training set and the validation set to train and validate the model of the Deep-Pred framework, adjust the model parameters according to the training results, and obtain a typhoon wind speed prediction model.
[0011] Input the test set and the typhoon wind speed data for the time period to be predicted into the typhoon wind speed prediction model to obtain the typhoon wind speed prediction results.
[0012] According to the present invention, a typhoon wind speed prediction method based on the Deep-Pred framework is provided, which determines the data of typhoon wind speed influencing factors including potential vorticity, relative vorticity, vertical wind shear, divergence, air temperature, relative humidity, wind u component, wind v component, vertical velocity, and sea surface temperature.
[0013] According to the present invention, a typhoon wind speed prediction method based on the Deep-Pred framework is provided, which constructs an environmental field of influencing factors based on the typhoon wind speed influencing factor data, including:
[0014] The data on typhoon wind speed influencing factors are divided with the typhoon's location as the center to obtain the environmental field of influencing factors.
[0015] According to the typhoon wind speed prediction method based on the Deep-Pred framework provided by the present invention, the new typhoon dataset is divided into a training set, a validation set, and a test set, including:
[0016] The new typhoon dataset is normalized, and a sliding window with four time steps is set for sliding processing. The dataset is then divided into the training set, validation set, and test set according to a preset partitioning ratio.
[0017] According to the present invention, a typhoon wind speed prediction method based on the Deep-Pred framework is provided, wherein the Deep-Pred framework is constructed, including:
[0018] The system comprises a data compression module, a spatiotemporal attention module, a ConvGPU neural unit, an Encoder-Decoder module, a convolutional attention mechanism, a feature enhancement module, and a prediction module, wherein:
[0019] The spatiotemporal attention module includes a location attention layer, a spatial attention layer, and a temporal attention layer;
[0020] The Encoder-Decoder module includes an encoder and a decoder.
[0021] According to the present invention, a typhoon wind speed prediction method based on the Deep-Pred framework is provided, which trains and validates the model of the Deep-Pred framework using the training set and the validation set, adjusts the model parameters according to the training results, and obtains a typhoon wind speed prediction model, including:
[0022] The model performance is evaluated using the mean absolute error (MAE). The Adam optimizer is used, and the preset learning rate and preset number of iterations are set to determine the final model parameters and output the typhoon wind speed prediction model.
[0023] Secondly, the present invention also provides a typhoon wind speed prediction system based on the Deep-Pred framework, comprising:
[0024] The data acquisition module is used to collect typhoon wind speed data for a preset year and time period, and to determine the data of factors affecting typhoon wind speed.
[0025] The construction module is used to construct an environmental field of influencing factors based on the typhoon wind speed influencing factor data, and to form a new typhoon dataset with the environmental field of influencing factors and the typhoon wind speed data.
[0026] The partitioning module is used to partition the new typhoon dataset to obtain a training set, a validation set, and a test set.
[0027] The training module is used to build the Deep-Pred framework, train and validate the model of the Deep-Pred framework using the training set and the validation set, adjust the model parameters according to the training results, and obtain the typhoon wind speed prediction model.
[0028] The prediction module is used to input the test set and the typhoon wind speed data for the time period to be predicted into the typhoon wind speed prediction model to obtain the typhoon wind speed prediction result.
[0029] Thirdly, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the typhoon wind speed prediction method based on the Deep-Pred framework as described above.
[0030] Fourthly, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the typhoon wind speed prediction method based on the Deep-Pred framework as described above.
[0031] Fifthly, the present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the typhoon wind speed prediction method based on the Deep-Pred framework as described above.
[0032] The typhoon wind speed prediction method and system based on the Deep-Pred framework provided by this invention can accurately extract the spatiotemporal features of multi-dimensional typhoon information by using the Deep-Pred framework for typhoon wind speed prediction. Compared with existing prediction models, it can better predict the trend of typhoon intensity change and maximum wind speed. Attached Figure Description
[0033] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.
[0034] Figure 1 This is one of the flowcharts of the typhoon wind speed prediction method based on the Deep-Pred framework provided by the present invention;
[0035] Figure 2 This is the second flowchart of the typhoon wind speed prediction method based on the Deep-Pred framework provided by this invention.
[0036] Figure 3 These are the prediction results and actual wind speed maps provided by the Deep-Pred framework of this invention;
[0037] Figure 4 This is a schematic diagram of the newly created data compression module provided by the present invention;
[0038] Figure 5 This is a structural diagram of the newly created spatiotemporal attention mechanism module provided by the present invention;
[0039] Figure 6 This is a diagram of the newly created ConvGPU neural unit and Encoder-Decoder structure provided by the present invention.
[0040] Figure 7 This is a diagram of the new convolutional attention mechanism network structure provided by the present invention;
[0041] Figure 8 This is a schematic diagram of the multi-head convolutional attention mechanism provided by the present invention;
[0042] Figure 9 This is a structural diagram of the Deep-Pred framework provided by the present invention;
[0043] Figure 10 This is a comparison chart of the 6-24h prediction results of the Deep-Pred framework provided by this invention with the results of other methods;
[0044] Figure 11This is a schematic diagram of the typhoon wind speed prediction system based on the Deep-Pred framework provided by the present invention;
[0045] Figure 12 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation
[0046] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.
[0047] Current technologies are limited by a variety of influencing factors, making typhoon intensity prediction difficult. Previous studies have shown that traditional prediction methods rely on a large amount of computing resources and time, and their accuracy is highly dependent on the quality of the initial data, which will hinder real-time prediction capabilities.
[0048] Figure 1 This is one of the flowcharts illustrating the typhoon wind speed prediction method based on the Deep-Pred framework provided in this embodiment of the invention, such as... Figure 1 As shown, it includes:
[0049] Step 100: Collect typhoon wind speed data for a preset year and time period to determine the data on factors affecting typhoon wind speed;
[0050] Step 200: Construct an environmental field of influencing factors based on the typhoon wind speed influencing factor data, and form a new typhoon dataset with the environmental field of influencing factors and the typhoon wind speed data;
[0051] Step 300: Divide the new typhoon dataset into training set, validation set and test set;
[0052] Step 400: Construct the Deep-Pred framework, use the training set and the validation set to train and validate the model of the Deep-Pred framework, adjust the model parameters according to the training results, and obtain the typhoon wind speed prediction model;
[0053] Step 500: Input the test set and the typhoon wind speed data for the time period to be predicted into the typhoon wind speed prediction model to obtain the typhoon wind speed prediction result.
[0054] The purpose of this invention is to consider the multiple influencing factors of typhoon intensity and to establish a new typhoon intensity prediction framework, Deep-Pred, which extracts features from multidimensional environmental variables to accurately predict typhoon wind speed. The results show that the prediction results of the Deep-Pred framework for 6-24 hours are better than those of the US National Hurricane Center (NHC), China Meteorological Administration (CMA), Joint Typhoon Warning Center (JTWC), LSTM, FFNN, TC-Pred framework, etc.
[0055] Specifically, such as Figure 2 As shown, this embodiment of the invention collects typhoon wind speed data and typhoon wind speed influencing factor data for a specified year and time period, processes the typhoon wind speed influencing factor data, divides the dataset, constructs a Deep-Pred framework containing multiple modules and trains the model, trains and optimizes the model parameters to obtain a typhoon wind speed prediction model, and finally inputs the test set and the actual typhoon wind speed data for the time period to be predicted into the model to obtain the typhoon wind speed prediction result.
[0056] like Figure 3 The diagram shows a comparison between the predicted wind speeds and the actual wind speeds for several historically significant typhoons using embodiments of the present invention.
[0057] This invention uses the Deep-Pred framework for typhoon wind speed prediction, which can accurately extract the spatiotemporal features of multidimensional typhoon information. Compared with existing prediction models, it can better predict the trend of typhoon intensity changes and maximum wind speed.
[0058] Based on the above embodiments, the data on factors affecting typhoon wind speed include potential vorticity, relative vorticity, vertical wind shear, divergence, air temperature, relative humidity, wind U-component, wind V-component, vertical velocity, and sea surface temperature.
[0059] Specifically, the main influencing factors determined in this embodiment of the invention include potential vorticity (600 hPa), relative vorticity (925 hPa), vertical wind shear (vector), divergence (925 hPa), air temperature (50 / 150 / 300 / 350 / 550 / 750 / 950 hPa), relative humidity (50 / 150 / 350 / 550 / 600 / 750 / 950 hPa), wind u component (50 / 150 / 350 / 550 / 750 / 950 hPa), wind v component (50 / 150 / 200 / 350 / 550 / 750 / 950), vertical velocity (200 / 300 / 400 / 500 / 600 / 700 hPa), and sea surface temperature.
[0060] The collected typhoon wind speed data covers typhoon wind speeds from 1979 to 2021.
[0061] Based on the above embodiments, an environmental field of influencing factors is constructed based on the typhoon wind speed influencing factor data, including:
[0062] The data on typhoon wind speed influencing factors are divided with the typhoon's location as the center to obtain the environmental field of influencing factors.
[0063] The new typhoon dataset is divided into a training set, a validation set, and a test set, including:
[0064] The new typhoon dataset is normalized, and a sliding window with four time steps is set for sliding processing. The dataset is then divided into the training set, validation set, and test set according to a preset partitioning ratio.
[0065] Specifically, the collected data on typhoon wind speed influencing factors are divided into environmental fields centered on the typhoon location, and integrated with the typhoon wind speed data into a new dataset.
[0066] Since different influencing factors have different ranges and time scales, it is necessary to normalize the data to make the typhoon intensity prediction model easier to train and achieve reasonable prediction results. The obtained data is normalized as follows.
[0067] X(i)=(0.1*(max-x(i))+0.9*(x(i)-min)) / (max-min).
[0068] The normalized data was divided into training, validation, and test sets in a 7:1:2 ratio.
[0069] Based on the above embodiments, the Deep-Pred framework is constructed, including:
[0070] The system comprises a data compression module, a spatiotemporal attention module, a ConvGPU neural unit, an Encoder-Decoder module, a convolutional attention mechanism, a feature enhancement module, and a prediction module, wherein:
[0071] The spatiotemporal attention module includes a location attention layer, a spatial attention layer, and a temporal attention layer;
[0072] The Encoder-Decoder module includes an encoder and a decoder.
[0073] Specifically, the embodiment of the present invention constructs the Deep-Pred framework, including:
[0074] To unify the dimensions of the experimental data, a data compression module was then used to compress the data volume, specifically as follows: Figure 4As shown, an environmental field is constructed using the location of the typhoon as input to the model. Then, a convolutional neural network is used to capture spatial information and compress the data to generate a new dataset. Finally, a fusion block module is used to generate the final representation.
[0075] Specifically, the experimental data is standardized in dimensions, and then the data volume is compressed using a data compression module, as follows: Figure 4 As shown, an environmental field is constructed using the location of the typhoon as input to the model. Then, a convolutional neural network is used to capture spatial information and compress the data to generate a new dataset. Finally, a fusion module is used to generate the final representation.
[0076] like Figure 5 As shown, the final representation of the data is input into the spatiotemporal attention mechanism to extract temporal and spatial information. The spatiotemporal attention mechanism (ST-Attention Module) consists of three sub-layers: a positional attention layer, a spatial attention layer, and a temporal attention layer. The positional attention layer learns the relationships between features by capturing the spatial dependencies between feature map locations. The spatial attention layer is introduced to enhance the representation of key regions. The temporal attention mechanism can capture the similarity between feature maps of corresponding time channels.
[0077] like Figure 6 As shown, the hidden state of the ConvGRU from the previous time step is used to highlight key features through SAL (Synchronization Algorithm), and then the hidden state is input into subsequent ConvGRU units to enhance its ability to capture the spatiotemporal relationships between various factors. The Encoder-Decoder model performs well in capturing long-term dependencies between sequences and achieving more accurate multi-step predictions, making it particularly suitable for typhoon intensity prediction. It consists of an encoder and a decoder. The encoder receives the sequence as input, learns deeper features and relationships, and generates an output sequence. The initial state and storage units from the encoder are passed to the decoder, which unconditionally takes the sequence as input to generate its output sequence.
[0078] like Figure 7 As shown, this embodiment introduces a novel feature enhancement layer, which is based on a combination of convolutional transformers and residual networks. This layer can identify relevant information and focus on information from different subspaces of different feature spaces.
[0079] like Figure 8 As shown, a multi-head convolutional attention mechanism is used. Its main function is to map the input sequence to different representation spaces, and then calculate the weighted average of the vectors in these spaces to obtain the final vector representation.
[0080] The final composition is as follows Figure 9 The Deep-Pred framework structure diagram.
[0081] Based on the above embodiments, the Deep-Pred framework model is trained and validated using the training set and the validation set. The model parameters are adjusted according to the training results to obtain a typhoon wind speed prediction model, including:
[0082] The model performance is evaluated using the mean absolute error (MAE). The Adam optimizer is used, and the preset learning rate and preset number of iterations are set to determine the final model parameters and output the typhoon wind speed prediction model.
[0083] Specifically, in this embodiment of the invention, the model is trained by assuming model parameters under the Deep-Pred framework, and the parameters are adjusted according to the training results to determine the final model parameters.
[0084] Calculated using MAE:
[0085] y t For the true value, This is a predicted value;
[0086] The model parameters are continuously corrected, and the model parameters are determined when the predicted MAE value is minimized: the optimizer is Adam optimizer, the learning rate is Lr = 0.00001, the number of iterations is Epoch = 20, and the final model parameters are determined.
[0087] Based on the above embodiments, as a preferred embodiment of the typhoon wind speed prediction method provided in this embodiment, it further includes:
[0088] The prediction results of the embodiments of the present invention are compared with the results of other existing methods.
[0089] Optionally, prediction results from the National Hurricane Center (NHC), China Meteorological Administration (CMA), Joint Typhoon Warning Center (JTWC), LSTM, FFNN, and TC-Pred frameworks are compared with those output by the Deep-Pred framework. Here, the models based on different neural units in the TC-Pred framework and the Deep-Pred framework are named TC and De, respectively, such as TCConvLSTM and DeConvConvLSTM.
[0090] The Deep-Pred framework can accurately extract the spatiotemporal features of multidimensional typhoon information. Experimental results show that, for example... Figure 10As shown, the typhoon wind speed predicted based on the Deep-Pred framework has a significantly lower forecast error than that based on the Typhoon Forecasting Service of the U.S. National Hurricane Center. Preliminary results are as follows: The MAE (Maximum Aspect Ratio) of the 6-24h typhoon intensity predicted by the DeConvGPU model under the Deep-Pred framework are 1.56 m / s, 2.44 m / s, 3.24 m / s, and 3.88 m / s, respectively. This accuracy is 2.50%, 3.56%, 1.22%, and 3.00% higher than the second-best performing DeConvLSTM network, 4.88%, 3.17%, 3.28%, and 3.72% higher than the basic unit CovGRU, and 11.86%, 4.17%, 4.56%, and 6.50% higher than the output results under the TC-Pred framework. This model can predict the trend of typhoon intensity changes and maximum wind speed relatively well.
[0091] The typhoon wind speed prediction system based on the Deep-Pred framework provided by this invention will be described below. The typhoon wind speed prediction system based on the Deep-Pred framework described below can be referred to in correspondence with the typhoon wind speed prediction method based on the Deep-Pred framework described above.
[0092] Figure 11 This is a schematic diagram of the typhoon wind speed prediction system based on the Deep-Pred framework provided in an embodiment of the present invention, as shown below. Figure 11 As shown, it includes: a data acquisition module 1101, a construction module 1102, a partitioning module 1103, a training module 1104, and a prediction module 1105, wherein:
[0093] The data acquisition module 1101 is used to collect typhoon wind speed data for a preset year and time period to determine the data of factors affecting typhoon wind speed; the construction module 1102 is used to construct an environmental field of influencing factors based on the data of factors affecting typhoon wind speed, and to form a new typhoon dataset with the environmental field of influencing factors and the typhoon wind speed data; the partitioning module 1103 is used to partition the new typhoon dataset to obtain a training set, a validation set and a test set; the training module 1104 is used to construct a Deep-Pred framework, use the training set and the validation set to train the model of the Deep-Pred framework, adjust the model parameters according to the training results, and obtain a typhoon wind speed prediction model; the prediction module 1105 is used to input the test set and the typhoon wind speed data of the time period to be predicted into the typhoon wind speed prediction model to obtain the typhoon wind speed prediction result.
[0094] Figure 12 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 12As shown, the electronic device may include: a processor 1210, a communications interface 1220, a memory 1230, and a communications bus 1240, wherein the processor 1210, the communications interface 1220, and the memory 1230 communicate with each other through the communications bus 1240. The processor 1210 can call logic instructions in the memory 1230 to execute a typhoon wind speed prediction method based on the Deep-Pred framework. This method includes: collecting typhoon wind speed data for a preset year and time period to determine typhoon wind speed influencing factor data; constructing an influencing factor environmental field based on the typhoon wind speed influencing factor data, and forming a new typhoon dataset with the influencing factor environmental field and the typhoon wind speed data; dividing the new typhoon dataset to obtain a training set, a validation set, and a test set; constructing the Deep-Pred framework, using the training set and the validation set to train and validate the model of the Deep-Pred framework, adjusting the model parameters according to the training results to obtain a typhoon wind speed prediction model; and inputting the test set and typhoon wind speed data for the time period to be predicted into the typhoon wind speed prediction model to obtain typhoon wind speed prediction results.
[0095] Furthermore, the logical instructions in the aforementioned memory 1230 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0096] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the typhoon wind speed prediction method based on the Deep-Pred framework provided by the above methods. The method includes: collecting typhoon wind speed data for a preset year and time period, and determining typhoon wind speed influencing factor data; constructing an influencing factor environmental field based on the typhoon wind speed influencing factor data, and forming a new typhoon dataset with the influencing factor environmental field and the typhoon wind speed data; dividing the new typhoon dataset to obtain a training set, a validation set, and a test set; constructing a Deep-Pred framework, using the training set and the validation set to train and validate the model of the Deep-Pred framework, adjusting the model parameters according to the training results, and obtaining a typhoon wind speed prediction model; inputting the test set and typhoon wind speed data for the time period to be predicted into the typhoon wind speed prediction model to obtain typhoon wind speed prediction results.
[0097] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the typhoon wind speed prediction method based on the Deep-Pred framework provided by the above methods. The method includes: collecting typhoon wind speed data for a preset year and time period, and determining typhoon wind speed influencing factor data; constructing an influencing factor environmental field based on the typhoon wind speed influencing factor data, and forming a new typhoon dataset with the influencing factor environmental field and the typhoon wind speed data; dividing the new typhoon dataset to obtain a training set, a validation set, and a test set; constructing a Deep-Pred framework, using the training set and the validation set to train and validate the model of the Deep-Pred framework, adjusting the model parameters according to the training results, and obtaining a typhoon wind speed prediction model; inputting the test set and typhoon wind speed data for the time period to be predicted into the typhoon wind speed prediction model to obtain typhoon wind speed prediction results.
[0098] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.
[0099] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.
[0100] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A typhoon wind speed prediction method based on the Deep-Pred framework, characterized in that, include: Collect typhoon wind speed data for a preset year and time period to determine the factors affecting typhoon wind speed; Based on the typhoon wind speed influencing factor data, an influencing factor environmental field is constructed, and a new typhoon dataset is formed by the influencing factor environmental field and the typhoon wind speed data. The new typhoon dataset is divided into training set, validation set and test set; Construct a Deep-Pred framework, use the training set and the validation set to train and validate the model of the Deep-Pred framework, adjust the model parameters according to the training results, and obtain a typhoon wind speed prediction model. Input the typhoon wind speed data of the test set and the time period to be predicted into the typhoon wind speed prediction model to obtain the typhoon wind speed prediction results; Building the Deep-Pred framework includes: The system comprises a data compression module, a spatiotemporal attention module, a ConvGPU neural unit, an Encoder-Decoder module, a convolutional attention mechanism, a feature enhancement module, and a prediction module, wherein: The spatiotemporal attention module includes a positional attention layer, a spatial attention layer, and a temporal attention layer. The positional attention layer learns the relationships between features by capturing the spatial dependencies between feature map positions. The spatial attention layer enhances the representation of key regions. The temporal attention mechanism captures the similarity between feature maps of corresponding time channels. The Encoder-Decoder module includes an encoder and a decoder. The hidden state of the ConvGRU in the previous time step is used to highlight key features through SAL. Then, the hidden state is input into the subsequent ConvGRU unit to enhance its ability to capture the spatiotemporal relationship between various factors. The encoder receives the sequence as input, learns deeper features and relationships, and generates an output sequence. The initial state and storage unit from the encoder are transmitted to the decoder. The decoder unconditionally takes the sequence as input to generate the decoder's output sequence. The feature enhancement module is based on a combination of convolutional transformer and residual network to identify relevant information and pay attention to information from different subspaces of different feature spaces. It adopts a multi-head convolutional attention mechanism to map the input sequence to different representation spaces, calculates the weighted average of vectors in different representation spaces, and obtains the final vector representation.
2. The typhoon wind speed prediction method based on the Deep-Pred framework according to claim 1, characterized in that, The data for determining the factors affecting typhoon wind speed include potential vorticity, vorticity, vertical wind shear, divergence, air temperature, humidity, wind U-component, wind V-component, vertical velocity, and sea surface temperature.
3. The typhoon wind speed prediction method based on the Deep-Pred framework according to claim 1, characterized in that, Based on the data on typhoon wind speed influencing factors, an environmental field of influencing factors is constructed, including: The data on typhoon wind speed influencing factors are divided with the typhoon's location as the center to obtain the environmental field of influencing factors.
4. The typhoon wind speed prediction method based on the Deep-Pred framework according to claim 1, characterized in that, The new typhoon dataset is divided into a training set, a validation set, and a test set, including: The new typhoon dataset is normalized, and a sliding window with four time steps is set for sliding processing. The dataset is then divided into the training set, validation set, and test set according to a preset partitioning ratio.
5. The typhoon wind speed prediction method based on the Deep-Pred framework according to claim 1, characterized in that, The Deep-Pred framework model is trained and validated using the training set and the validation set. The model parameters are adjusted based on the training results to obtain a typhoon wind speed prediction model, including: The model performance is evaluated using the mean absolute error (MAE). The Adam optimizer is used, and the preset learning rate and preset number of iterations are set to determine the final model parameters and output the typhoon wind speed prediction model.
6. A typhoon wind speed prediction system based on the Deep-Pred framework, comprising the typhoon wind speed prediction method based on the Deep-Pred framework as described in any one of claims 1 to 5, characterized in that, include: The data acquisition module is used to collect typhoon wind speed data for a preset year and time period, and to determine the data of factors affecting typhoon wind speed. The construction module is used to construct an environmental field of influencing factors based on the typhoon wind speed influencing factor data, and to form a new typhoon dataset with the environmental field of influencing factors and the typhoon wind speed data. The partitioning module is used to partition the new typhoon dataset to obtain a training set, a validation set, and a test set. The training module is used to build the Deep-Pred framework, train and validate the model of the Deep-Pred framework using the training set and the validation set, adjust the model parameters according to the training results, and obtain the typhoon wind speed prediction model. The prediction module is used to input the test set and the typhoon wind speed data for the time period to be predicted into the typhoon wind speed prediction model to obtain the typhoon wind speed prediction result.
7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the typhoon wind speed prediction method based on the Deep-Pred framework as described in any one of claims 1 to 5.
8. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the typhoon wind speed prediction method based on the Deep-Pred framework as described in any one of claims 1 to 5.
9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the typhoon wind speed prediction method based on the Deep-Pred framework as described in any one of claims 1 to 5.