Typhoon prediction method and device, electronic equipment and storage medium
By clustering and classifying historical typhoon data and combining it with the correction of the multilayer sensing model, the problem of insufficient accuracy of existing typhoon prediction methods has been solved, and accurate prediction of typhoon paths has been achieved.
Patent Information
- Application Number
- CN202511068461.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-31
- Publication Date
- 2025-11-14
AI Technical Summary
Existing typhoon forecasting methods mainly rely on meteorological monitoring data, lacking effective forecasting methods, resulting in insufficient accuracy.
By acquiring various types of historical typhoon data, clustering and random forest classification algorithms are used for data processing. Combined with multilayer sensing models and dynamic time adjustment algorithms, the model is corrected to improve prediction accuracy.
It has enabled accurate prediction of typhoon paths, improving the accuracy and reliability of typhoon forecasting.
Smart Images
Figure CN120951153A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of information technology, and in particular to a typhoon forecasting method, apparatus, electronic device, and storage medium. Background Technology
[0002] Typhoons are highly destructive natural disasters, and their formation, path evolution, and intensity changes are influenced by a combination of meteorological factors (such as temperature, air pressure, monsoons, and ocean currents). Accurate prediction of typhoon classification, path, and attenuation characteristics is crucial for disaster prevention and mitigation. However, current typhoon forecasting still relies primarily on meteorological monitoring data and lacks effective forecasting methods. Summary of the Invention
[0003] The purpose of this application is to provide a typhoon forecasting method, apparatus, electronic device, and storage medium to solve the problem of how to perform typhoon forecasting. The specific technical solution is as follows:
[0004] A first aspect of this application provides a typhoon forecasting method, including:
[0005] Obtain various types of historical typhoon data;
[0006] The historical typhoon data is clustered using a preset clustering algorithm to obtain multiple types of first cluster data; the multiple types of first cluster data are then classified and predicted using a preset random forest classification algorithm to obtain multiple types of second cluster data.
[0007] By using a preset multi-layer perception model, the position at the next time step is calculated based on the position information at multiple times in the second clustering data of various types, and the predicted path information is obtained. The predicted path information includes the position information at multiple times. By using a preset dynamic time adjustment algorithm, the loss of the preset multi-layer perception model is calculated based on the predicted path information and the actual path information. Based on the loss of the preset multi-layer perception model, the preset multi-layer perception model is corrected to obtain the corrected multi-layer perception model.
[0008] Real-time typhoon data is acquired, and the output path information is calculated based on the modified multilayer sensing model.
[0009] In one possible implementation, the historical typhoon data is clustered using a preset clustering algorithm to obtain multiple types of first cluster data, including:
[0010] By using a preset interpolation method, missing values are calculated for the various types of historical typhoon data to obtain continuous typhoon data.
[0011] By using a preset K-means clustering algorithm, the continuous typhoon data is clustered to obtain multiple types of first cluster data.
[0012] In one possible implementation, the step of calculating the position at the next time step based on the position information at multiple times in the various types of second clustering data using a preset multilayer sensing model to obtain predicted path information includes:
[0013] Calculate the Coriolis force at the next moment based on the historical typhoon data.
[0014] The Coriolis force and the location information at multiple times in the second cluster data of the various types are input into a preset multilayer sensing model, so that the preset multilayer sensing model performs weighted summation calculation through weight matrix and bias vector, and outputs the predicted path information.
[0015] In one possible implementation, the step of calculating the loss of a preset multilayer perception model based on predicted path information and actual path information using a preset dynamic time adjustment algorithm includes:
[0016] Based on the predicted path information and the actual path information, using the formula:
[0017]
[0018] Calculate the loss of the preset multilayer perception model; where A is the actual path information, A = (a1, a2, ..., a...). i …a n B represents the predicted path information, B = (b1, b2, ..., bb2). j …b n );d(a i b j ) represents a i and b j The distance between them; a1, a2…a i …a n These represent the actual positions at time 1, time 2, ..., time i, ..., time n, respectively; b1, b2, ..., bn, respectively. j …b n These represent the predicted positions at time 1, time 2, ..., time j, ..., time n, respectively.
[0019] In one possible implementation, after classifying and predicting the various types of first cluster data using a preset random forest classification algorithm to obtain various types of second cluster data, the method further includes:
[0020] Based on the various types of second clustering data, an exponential decay model of wind speed over time, an exponential decay model of precipitation over time, and a power function decay model of precipitation with distance from the typhoon center were created.
[0021] Acquire real-time typhoon data, and calculate and output wind speed information and precipitation information based on the exponential decay model of wind speed over time, the exponential decay model of precipitation over time, and the power function decay model of precipitation with distance from the typhoon center.
[0022] A second aspect of this application provides a typhoon forecasting device, comprising:
[0023] The data acquisition module is used to acquire various types of historical typhoon data;
[0024] The data clustering module is used to cluster the historical typhoon data using a preset clustering algorithm to obtain multiple types of first cluster data; and to classify and predict the multiple types of first cluster data using a preset random forest classification algorithm to obtain multiple types of second cluster data.
[0025] The model creation module is used to calculate the position at the next time step using a preset multilayer sensing model based on the position information at multiple time steps in the second clustering data of various types, thereby obtaining predicted path information, wherein the predicted path information includes the position information at multiple time steps; calculate the loss of the preset multilayer sensing model using a preset dynamic time adjustment algorithm based on the predicted path information and the actual path information; and correct the preset multilayer sensing model based on the loss of the preset multilayer sensing model to obtain a corrected multilayer sensing model.
[0026] The path prediction module is used to acquire real-time typhoon data and calculate the output path information based on the modified multilayer sensing model.
[0027] In one possible implementation, the data clustering module is specifically used to calculate missing values in the various types of historical typhoon data using a preset interpolation method to obtain continuous typhoon data; and to cluster the continuous typhoon data using a preset K-means clustering algorithm to obtain multiple types of first cluster data.
[0028] In one possible implementation, the model creation module is specifically used to calculate the Coriolis force corresponding to the next moment based on the historical typhoon data; input the Coriolis force and the location information of multiple moments in the second cluster data of multiple types into a preset multilayer sensing model, so that the preset multilayer sensing model performs weighted summation calculation through weight matrix and bias vector, and outputs the predicted path information.
[0029] In one possible implementation, the model creation module is specifically used to, based on the predicted path information and the actual path information, use the formula:
[0030]
[0031] Calculate the loss of the preset multilayer perception model; where A is the actual path information, A = (a1, a2, ..., a...). i …a n B represents the predicted path information, B = (b1, b2, ..., bb2). j …b n );d(a i b j ) represents a i and b j The distance between them; a1, a2…a i …a n These represent the actual positions at time 1, time 2, ..., time i, ..., time n, respectively; b1, b2, ..., bn, respectively. j …b n These represent the predicted positions at time 1, time 2, ..., time j, ..., time n, respectively.
[0032] In one possible implementation, the device further includes:
[0033] The attenuation prediction module is used to create an exponential decay model of wind speed over time, an exponential decay model of precipitation over time, and a power function decay model of precipitation with distance from the typhoon center based on the various types of second clustering data; acquire real-time typhoon data, and calculate and output wind speed information and precipitation information based on the exponential decay models of wind speed over time, precipitation over time, and precipitation with distance from the typhoon center.
[0034] Another aspect of the embodiments of this application also provides an electronic device, including:
[0035] Memory, used to store computer programs;
[0036] The processor, when executing the program stored in memory, implements any of the above typhoon forecasting methods.
[0037] In another aspect of the embodiments of this application, a computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements any of the typhoon forecasting methods described above.
[0038] In another aspect of the embodiments of this application, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to execute any of the typhoon forecasting methods described above.
[0039] Beneficial effects of the embodiments in this application:
[0040] This application provides a typhoon forecasting method, apparatus, electronic device, and storage medium. The method includes: acquiring various types of historical typhoon data; clustering the historical typhoon data using a preset clustering algorithm to obtain various types of first cluster data; classifying and predicting the various types of first cluster data using a preset random forest classification algorithm to obtain various types of second cluster data; calculating the position at the next time moment using a preset multi-layer sensing model based on the position information at multiple times in the various types of second cluster data to obtain predicted path information, wherein the predicted path information includes the position information at multiple times; calculating the loss of the preset multi-layer sensing model using a preset dynamic time adjustment algorithm based on the predicted path information and the actual path information; correcting the preset multi-layer sensing model based on the loss of the preset multi-layer sensing model to obtain a corrected multi-layer sensing model; acquiring real-time typhoon data and calculating output path information based on the corrected multi-layer sensing model. The solution in this application embodiment can use a clustering algorithm to cluster historical typhoon data to obtain multiple types of typhoon data. Then, using these multiple types of typhoon data and a preset multi-layer perception model, the path can be predicted. The preset multi-layer perception model can be corrected based on the predicted path, thereby predicting the typhoon path using real-time typhoon data and obtaining output path information.
[0041] Of course, implementing any product or method of this application does not necessarily require achieving all of the advantages described above at the same time. Attached Figure Description
[0042] To more clearly illustrate the technical solutions in the embodiments of this application 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 only some embodiments of this application. For those skilled in the art, other embodiments can be obtained based on these drawings.
[0043] Figure 1 A flowchart illustrating a typhoon forecasting method provided in an embodiment of this application;
[0044] Figure 2 A flowchart illustrating the clustering of historical typhoon data provided in this application embodiment;
[0045] Figure 3 A flowchart illustrating the predicted path information provided in an embodiment of this application;
[0046] Figure 4A schematic diagram of a typhoon forecasting device provided in an embodiment of this application;
[0047] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation
[0048] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art based on this application are within the scope of protection of this application.
[0049] A first aspect of the embodiments of this application provides a typhoon forecasting method, see [link to previous document]. Figure 1 , Figure 1 A flowchart illustrating a typhoon forecasting method provided in this application embodiment includes:
[0050] Step S11: Obtain various types of historical typhoon data;
[0051] Step S12: The historical typhoon data is clustered using a preset clustering algorithm to obtain multiple types of first cluster data; the multiple types of first cluster data are classified and predicted using a preset random forest classification algorithm to obtain multiple types of second cluster data.
[0052] Step S13: Using a preset multilayer sensing model, the position at the next time step is calculated based on the position information at multiple times in the second clustering data of various types, to obtain predicted path information, wherein the predicted path information includes the position information at multiple times; using a preset dynamic time adjustment algorithm, the loss of the preset multilayer sensing model is calculated based on the predicted path information and the actual path information; based on the loss of the preset multilayer sensing model, the preset multilayer sensing model is corrected to obtain a corrected multilayer sensing model.
[0053] Step S14: Obtain real-time typhoon data and calculate the output path information based on the corrected multi-layer sensing model.
[0054] Corresponding to step S11 above, various types of historical typhoon data are acquired, including typhoon characteristic data such as wind speed, air pressure, intensity level, latitude and longitude, and seasonal information, as well as environmental meteorological parameters such as temperature, ocean currents, and wind fields, to construct a comprehensive typhoon characteristic dataset. In one example, the various types of historical typhoon data provided in this application embodiment can primarily come from authoritative meteorological databases, containing detailed typhoon path information and meteorological environmental data. The typhoon path data comes from the Tropical Cyclone Best Path Database, an internationally recognized standard source of typhoon data, providing global tropical cyclone best path data since 1945. Each record can include key parameters such as typhoon number, timestamp accurate to the hour, latitude and longitude coordinates, maximum sustained wind speed, and minimum central pressure. The meteorological environmental data comes from comprehensive observation data from NOAA (National Oceanic and Atmospheric Administration) and the China Meteorological Administration, including sea surface temperature, wind field data, and pressure field elements, with a time resolution of 6 hours and a spatial resolution of 0.25 degrees × 0.25 degrees. In one example, factors influencing typhoon track prediction include: Temperature: Typhoons typically form and move over warm ocean areas, and due to the ocean's thermal energy supply, their paths often tend towards tropical and subtropical regions. Air Pressure: Typhoons move in low-pressure areas and are influenced by surrounding high-pressure systems (such as the subtropical high), so their paths usually move along troughs. Ocean Currents and Wind Fields: Ocean currents affect the distribution of sea surface temperature, and wind fields control the outer circulation of the typhoon, determining its direction of movement. Therefore, when acquiring historical typhoon data, information on temperature, air pressure, ocean currents, and wind fields can be obtained.
[0055] Corresponding to step S12 above, when clustering the historical typhoon data using a preset clustering algorithm, corresponding feature vectors can be constructed based on the historical typhoon data, and clustering can be performed by calculating the distance between feature vectors. For example, a distance threshold can be set, and vectors with a distance less than the threshold can be clustered into one class. The random forest classification algorithm is a classifier that uses multiple trees to train and predict samples. It includes multiple decision trees, and its output class is determined by the mode of the classes output by individual trees. It can be used to handle classification and regression problems, as well as dimensionality reduction problems. When classifying and predicting the various types of first clustered data using the preset random forest classification algorithm, the various types of first clustered data can be input into the preset random forest classification algorithm to calculate the various types of second clustered data.
[0056] Corresponding to step S13 above, when calculating the position of the next moment based on the position information of multiple moments in the various types of second cluster data using a preset multi-layer sensing model, a nonlinear function can be used for calculation. In one example, this nonlinear function can be fitted using historical typhoon data, thereby obtaining the corresponding positions of multiple next moments using the position information of multiple moments in the various types of second cluster data and the nonlinear function. The predicted path information includes the position information of multiple next moments. In one example, let the position of the typhoon at time t be (x... t ,y t The predicted position of the next time step is (x) t+1 ,y t+1 );
[0057] (x t+1 ,y t+1 )=f(T,P,C,W)+∈ (1)
[0058] Where f() is a nonlinear function, ∈ is a random error term, and T, P, C, W represent various types of second cluster data.
[0059] When calculating the loss of a preset multilayer perception model using a preset dynamic time adjustment algorithm based on predicted and actual path information, the distance between the predicted and actual path information can be calculated and used as the loss of the preset multilayer perception model. This loss can then be used to correct the preset multilayer perception model. It should be noted that in actual use, after correcting the preset multilayer perception model based on its loss, the corrected model can be determined to meet preset requirements, such as whether its loss is less than a preset loss threshold. If the requirements are met, the corrected multilayer perception model is obtained and output; otherwise, the process returns to the step of calculating the loss of the preset multilayer perception model using the preset dynamic time adjustment algorithm based on predicted and actual path information, and training continues until the loss is less than the preset loss threshold.
[0060] Corresponding to step S14 above, when real-time typhoon data is acquired and the output path information is calculated based on the modified multilayer sensing model, real-time data of various types of typhoons can be used to predict the path based on the modified multilayer sensing model and obtain the output path information.
[0061] As can be seen, the method of this application embodiment can cluster historical typhoon data using a clustering algorithm to obtain multiple types of typhoon data. Then, using these multiple types of typhoon data and a preset multi-layer perception model, the path can be predicted. The preset multi-layer perception model can be corrected based on the predicted path, thereby predicting the typhoon path using real-time typhoon data and obtaining output path information.
[0062] In one possible implementation, the historical typhoon data is clustered using a preset clustering algorithm to obtain multiple types of first cluster data, see [link to relevant documentation]. Figure 2 ,include:
[0063] Step S21: Using a preset interpolation method, calculate the missing values of the various types of historical typhoon data to obtain continuous typhoon data;
[0064] Step S22: The continuous typhoon data is clustered using a preset K-means clustering algorithm to obtain multiple types of first cluster data.
[0065] When calculating missing values for various types of historical typhoon data using pre-defined interpolation methods, a spatiotemporal joint interpolation method can be employed. For example, linear interpolation can be used to fill in consecutive missing values in the time dimension; inverse distance weighted interpolation can be used in the spatial dimension, such as considering the observations from the five nearest surrounding stations for weighted calculation. In practical application, data sequences with more than three consecutive missing time points can be removed after quality checks. For another example, textual missing values can be filled through data association, such as obtaining missing typhoon name information by associating typhoon numbers with historical databases, and fields that cannot be filled in can be uniformly marked, such as "NA". In practical application, after calculating missing values, feature extraction and selection can also be performed. Multidimensional features are extracted from the original data; typhoon dynamic features can include 6-hour wind speed change rate, movement path curvature, etc.; environmental field features include derived variables such as vertical wind shear and ocean heat content. During feature selection, the Pearson correlation coefficient between each feature and the target variable can be calculated, retaining features with an absolute correlation coefficient greater than 0.3. Then, the correlation between features is analyzed visually using a heatmap, eliminating highly collinear features. All selected features are Z-score standardized to ensure uniformity of dimensions.
[0066] In this embodiment, the continuous typhoon data is clustered using a preset K-means clustering algorithm. K-means clustering can be used to construct feature vectors based on the characteristics of the typhoons, dividing them into several classes. Clustering can then be performed by maximizing the between-group variance or minimizing the within-group variance. Specifically, this can be achieved using the formula:
[0067]
[0068] Perform clustering, where J represents the clustering result, μ k Let C represent the mean of multiple eigenvectors, where x represents an eigenvector, and C represents the mean of multiple eigenvectors. k Let K represent the set of feature vectors, where K is the total number of feature vector sets.
[0069] After clustering using formula (2), predictive analysis can be performed using the random forest algorithm. The classification result of each tree can be used to determine the final predicted category through a majority vote.
[0070]
[0071] Where T is the total number of trees, y T The prediction result for the T-th tree is given, and mode() represents the random forest algorithm. This indicates the prediction result.
[0072] In this embodiment, historical typhoon data can distinguish between summer typhoons and autumn typhoons. The difference between the feature mean vectors of summer and autumn typhoons can be calculated and quantified using Euclidean distance. Specifically, the mean vectors of summer and autumn typhoons can be set as X. 夏 and X 秋 Then the difference D can be expressed as:
[0073] D=∥∥X 夏 -X 秋 ∥∥ (4)
[0074] In one possible implementation, the step involves using a preset multi-layer sensing model to calculate the position at the next time step based on the position information at multiple times within the various types of second clustering data, thereby obtaining predicted path information. (See [link to relevant documentation]). Figure 3 ,include:
[0075] Step S31: Calculate the Coriolis force corresponding to the next moment based on the historical typhoon data;
[0076] Step S32: Input the Coriolis force and the location information at multiple times in the second cluster data of the various types into a preset multilayer sensing model, so that the preset multilayer sensing model performs weighted summation calculation through weight matrix and bias vector, and outputs the predicted path information.
[0077] In this embodiment of the application, when calculating the Coriolis force corresponding to the next moment based on the historical typhoon data, the Coriolis force is a deflection force caused by the Earth's rotation. In the Northern Hemisphere, the typhoon path is deflected to the right, while in the Southern Hemisphere, it is deflected to the left. The formula for calculating the Coriolis force is:
[0078] F c =2mωvsin (φ) (5)
[0079] Where m represents the typhoon's air mass, ω is the Earth's rotational angular velocity, v is the typhoon's speed, φ is the latitude, and F... c This indicates the Coriolis force.
[0080] In this embodiment, when the predicted path information is obtained by weighting and calculating using a weight matrix and bias vector based on the Coriolis force and the positions at multiple times in the various types of second clustering data, a preset multilayer perceptron model can be used for calculation. Specifically, the preset multilayer perceptron model can consist of an input layer, one or more hidden layers, and an output layer, with each layer consisting of multiple neurons. Each neuron is connected to all neurons in the previous layer and calculates the output based on the input. The input layer is the input feature vector. The hidden layers can have multiple layers, typically using activation functions to introduce nonlinearity. The output layer generates the prediction result. In the preset multilayer perceptron model, forward propagation can calculate the output of each neuron. Assuming the number of neurons in the l-th layer is n... l The output h of the l-th layer neuron (l) It can be represented as:
[0081] z (l) =W (l) h (l-1) +b (l) (6)
[0082] h( l) =f(z) (l) (7)
[0083] Among them: W (l) b is the weight matrix of the l-th layer. (l) h is the bias vector of the l-th layer. (l-1) This is the output of the (l-1)th layer. f() is the activation function, and the weighted sum introduces nonlinearity through the activation function.
[0084] In one possible implementation, the step of calculating the loss of a preset multilayer perception model based on predicted path information and actual path information using a preset dynamic time adjustment algorithm includes: calculating the loss of the preset multilayer perception model based on predicted path information and actual path information using the formula:
[0085]
[0086] Calculate the loss of the preset multilayer perception model; where A is the actual path information, A = (a1, a2, ..., a...). i …a n B represents the predicted path information, B = (b1, b2, ..., bb2). j …b n );d(a i b j ) represents a i and b j The distance between them; a1, a2…a i …a n These represent the actual positions at time 1, time 2, ..., time i, ..., time n, respectively; b1, b2, ..., bn, respectively. j …b n These represent the predicted positions at time 1, time 2, ..., time j, ..., time n, respectively.
[0087] In one possible implementation, after classifying and predicting the various types of first cluster data using a preset random forest classification algorithm to obtain various types of second cluster data, the method further includes: creating an exponential decay model of wind speed over time, an exponential decay model of precipitation over time, and a power function decay model of precipitation with respect to the distance from the typhoon center based on the various types of second cluster data; acquiring real-time typhoon data, and calculating and outputting wind speed information and precipitation information based on the exponential decay model of wind speed over time, the exponential decay model of precipitation over time, and the power function decay model of precipitation with respect to the distance from the typhoon center.
[0088] In this embodiment of the application, based on the various types of second clustering data, an exponential decay model of wind speed over time, an exponential decay model of precipitation over time, and a power function decay model of precipitation with distance from the typhoon center are created.
[0089] After a typhoon makes landfall, its wind speed decreases rapidly over time, which can usually be described using an exponential decay model.
[0090] V(t) = V0e -αt (11)
[0091] Where V0 is the initial wind speed, α is the wind speed attenuation coefficient, t is time, and V(t) is the real-time wind speed.
[0092] Precipitation also decreases over time, which can be described by an exponential model:
[0093] R(t)=R0e -βt (12)
[0094] Where R0 is the initial precipitation, β is the precipitation attenuation coefficient, t is time, and R(t) is the real-time precipitation.
[0095] Precipitation R decreases as the distance from the typhoon center d increases, and can be described by a power function:
[0096]
[0097] Where d0 is the reference distance, γ is the precipitation attenuation coefficient, d is the real-time distance, R0() is the power function, and R(d) is the precipitation at the corresponding distance d.
[0098] To illustrate the solutions of the embodiments of this application, the following description is provided in conjunction with a specific example:
[0099] In one example, the data comes from the China Meteorological Administration's Typhoon Network and the China Meteorological Administration. First, the typhoon data from the founding of the People's Republic of China to 2023 are processed, and some of the data is shown in Table 1.
[0100] Table 1. Results of preprocessed data
[0101] serial number Start time End time longitude latitude strength grade wind speed air pressure movement speed 194901 1949011308 1949012508 134.8 9.6 3 7 15 1005 6.3617340 194901 1949011308 1949012508 134.5 10.8 2 8 20 1002 4.11833061 194901 1949011308 1949012508 134.5 11.6 5 10 25 1000 0.2289916 194901 1949011308 1949012508 134.6 12.2 5 11 30 998 2.6225834 202317 2023121708 2023121820 126.6 7.7 2 8 18 1000 38 202317 2023121708 2023121820 126.5 7.5 2 8 18 1000 33 202317 2023121708 2023121820 126.2 7.3 3 7 16 1002 4.2122561 202317 2023121708 2023121820 125.8 7.4 3 7 16 1002 7.7836689
[0102] Further rules were established to classify data from June to August as summer typhoons and data from September to November as autumn typhoons. The mathematical expression is as follows:
[0103]
[0104] Among them, typhoon intensity and level are significantly positively correlated, jointly representing disaster intensity; wind speed, as a core indicator, is strongly positively correlated with both, while air pressure shows a significant negative correlation, with the low-pressure core area confirming the physical characteristics of typhoons; the direction of movement and speed have interactive effects, influenced by both the typhoon's internal dynamics and environmental conditions. The correlation mechanism of each parameter conforms to the multi-factor coupling characteristics of the typhoon dynamic system.
[0105] Furthermore, heatmaps can be created based on Spearman's rank correlation coefficient. Heatmaps can reveal the statistical relationships between variables such as "units (mm / cm²)," "numerical values," "time," and "temperature." "Units" and "numerical values" show a perfect positive correlation, while "time" shows a strong negative correlation with both, with both decreasing significantly over time. In contrast, "temperature" has a weaker correlation with other variables. Further, typhoon data can be processed using K-means. Different types of typhoons exhibit significant differences in wind speed. By segmenting typhoon data using wind speed as a single variable, characteristics of typhoons of different intensities can be identified. In actual typhoon observations, typhoons occur more frequently within certain wind speed ranges, making it easier for typhoons to fall within these ranges. In this embodiment, functional principal component analysis can be performed on the latitude and longitude trajectories of multiple typhoons at consecutive time points to extract the main changing features of typhoon paths. These features can then be used as input to construct a multilayer perceptron neural network model.
[0106] Furthermore, the difference in characteristic mean vectors between summer and autumn typhoons was calculated, and the difference was quantified by Euclidean distance. Then, the typhoon paths, cloud maps, and provinces they passed through in 2024 were predicted, as shown in Table 2.
[0107] Table 2. Provinces Pathed by Typhoons
[0108]
[0109] The prediction of longitude and latitude of the data was conducted, and the results were compared. Due to the relatively backward technology during the Republic of China era, the records of typhoons may have been inaccurate, which significantly affected model training and resulted in poor typhoon path prediction. Therefore, the Coriolis force was introduced to correct the longitude and latitude to ensure the accuracy of typhoon path prediction. The DTW (Dynamic Time Warping) algorithm was used to compare the predicted paths, and Table 3 shows the typhoon center locations.
[0110] Table 1. Location of Typhoon Center
[0111] time Typhoon center location (longitude / latitude) 14:00 on the 13th 14.8 / 140.5 14:00 on the 14th 21.2 / 138.9 14:00 on the 15th 24.6 / 134 14:00 on the 16th 28.2 / 129.2 14:00 on the 17th 30.5 / 124.6
[0112] Substituting the data into (8), the DTW distance between the predicted path and the actual path is 139348.20. From September 16 to September 18, 2024, the central wind speed of Typhoon Bebinca showed obvious fluctuations, indicating that the typhoon's intensity was unstable during this period, and the central wind speed gradually weakened over time. Combining the precipitation change map, it can be found that when the wind speed is high in certain periods, the precipitation is also relatively high. This indicates that strong winds may bring more water vapor transport, thus leading to increased precipitation. Changes in wind speed also affect precipitation. Strong winds can transport more water vapor to the typhoon-affected area, thereby increasing the intensity and range of precipitation. However, changes in wind speed are not always completely synchronized with precipitation. It is speculated that precipitation is also regulated by other factors, such as atmospheric stability and water vapor content. The typhoon moves between longitudes 116 and 123. As longitude changes, the typhoon's latitude also changes, thus forming a specific typhoon path. The typhoon's movement path is relatively smooth, without significant turns or fluctuations, indicating that the typhoon's movement is relatively stable. Based on its changes in longitude and latitude, the overall direction of movement is northwestward.
[0113] A second aspect of this application provides a typhoon forecasting device, see [link to relevant documentation]. Figure 4 ,include:
[0114] Data acquisition module 401 is used to acquire various types of historical typhoon data;
[0115] The data clustering module 402 is used to cluster the historical typhoon data using a preset clustering algorithm to obtain multiple types of first cluster data; and to classify and predict the multiple types of first cluster data using a preset random forest classification algorithm to obtain multiple types of second cluster data.
[0116] The model creation module 403 is used to calculate the position at the next time step using a preset multilayer sensing model based on the position information at multiple time steps in the second clustering data of the various types, thereby obtaining predicted path information, wherein the predicted path information includes the position information at multiple time steps; calculate the loss of the preset multilayer sensing model using a preset dynamic time adjustment algorithm based on the predicted path information and the actual path information; and correct the preset multilayer sensing model based on the loss of the preset multilayer sensing model to obtain a corrected multilayer sensing model.
[0117] The path prediction module 404 is used to acquire real-time typhoon data and calculate the output path information based on the modified multilayer sensing model.
[0118] In one possible implementation, the data clustering module is specifically used to calculate missing values in the various types of historical typhoon data using a preset interpolation method to obtain continuous typhoon data; and to cluster the continuous typhoon data using a preset K-means clustering algorithm to obtain multiple types of first cluster data.
[0119] In one possible implementation, the model creation module is specifically used to calculate the Coriolis force corresponding to the next moment based on the historical typhoon data; input the Coriolis force and the location information of multiple moments in the second cluster data of multiple types into a preset multilayer sensing model, so that the preset multilayer sensing model performs weighted summation calculation through weight matrix and bias vector, and outputs the predicted path information.
[0120] In one possible implementation, the model creation module is specifically used to, based on the predicted path information and the actual path information, use the formula:
[0121]
[0122] Calculate the loss of the preset multilayer perception model; where A is the actual path information, A = (a1, a2, ..., a...). i …a n B represents the predicted path information, B = (b1, b2, ..., bb2). j …b n );d(a i b j ) represents a i and b j The distance between them; a1, a2…a i …a n These represent the actual positions at time 1, time 2, ..., time i, ..., time n, respectively; b1, b2, ..., bn, respectively. j …b n These represent the predicted positions at time 1, time 2, ..., time j, ..., time n, respectively.
[0123] In one possible implementation, the device further includes:
[0124] The attenuation prediction module is used to create an exponential decay model of wind speed over time, an exponential decay model of precipitation over time, and a power function decay model of precipitation with distance from the typhoon center based on the various types of second clustering data; acquire real-time typhoon data, and calculate and output wind speed information and precipitation information based on the exponential decay models of wind speed over time, precipitation over time, and precipitation with distance from the typhoon center.
[0125] As can be seen, the apparatus of this application embodiment can cluster historical typhoon data using a clustering algorithm to obtain multiple types of typhoon data. Then, using these multiple types of typhoon data and a preset multi-layer perception model, the path can be predicted. The preset multi-layer perception model can be corrected based on the predicted path, thereby predicting the typhoon path using real-time typhoon data and obtaining output path information.
[0126] This application also provides an electronic device, such as... Figure 5 As shown, it includes:
[0127] Memory 501 is used to store computer programs;
[0128] When processor 502 executes the program stored in memory 501, it performs the following steps:
[0129] Obtain various types of historical typhoon data;
[0130] The historical typhoon data is clustered using a preset clustering algorithm to obtain multiple types of first cluster data; the multiple types of first cluster data are then classified and predicted using a preset random forest classification algorithm to obtain multiple types of second cluster data.
[0131] By using a preset multi-layer perception model, the position at the next time step is calculated based on the position information at multiple times in the second clustering data of various types, and the predicted path information is obtained. The predicted path information includes the position information at multiple times. By using a preset dynamic time adjustment algorithm, the loss of the preset multi-layer perception model is calculated based on the predicted path information and the actual path information. Based on the loss of the preset multi-layer perception model, the preset multi-layer perception model is corrected to obtain the corrected multi-layer perception model.
[0132] Real-time typhoon data is acquired, and the output path information is calculated based on the modified multilayer sensing model.
[0133] The communication bus mentioned in the above electronic devices can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc. For ease of illustration, only one thick line is used to represent it in the diagram, but this does not mean that there is only one bus or one type of bus.
[0134] The communication interface is used for communication between the aforementioned electronic devices and other devices.
[0135] The memory may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.
[0136] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.
[0137] In another embodiment provided in this application, a computer-readable storage medium is also provided, which stores a computer program that, when executed by a processor, implements the steps of any of the typhoon prediction methods described above.
[0138] In another embodiment provided in this application, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to execute any of the typhoon forecasting methods described above.
[0139] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a solid-state drive (SSD), etc.
[0140] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, 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. Without further limitations, 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.
[0141] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on its differences from other embodiments. In particular, the embodiments for apparatus, electronic devices, and storage media are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.
[0142] The above description is merely a preferred embodiment of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application are included within the scope of protection of this application.
Claims
1. A typhoon forecasting method, characterized in that, include: Obtain various types of historical typhoon data; By using a preset clustering algorithm, the historical typhoon data is clustered to obtain multiple types of first cluster data; By using a preset random forest classification algorithm, the various types of first cluster data are classified and predicted to obtain various types of second cluster data; By using a preset multi-layer perception model, the position at the next time step is calculated based on the position information at multiple times in the second clustering data of various types, and the predicted path information is obtained. The predicted path information includes the position information at multiple times. By using a preset dynamic time adjustment algorithm, the loss of the preset multi-layer perception model is calculated based on the predicted path information and the actual path information. Based on the loss of the preset multi-layer perception model, the preset multi-layer perception model is corrected to obtain the corrected multi-layer perception model. Real-time typhoon data is acquired, and the output path information is calculated based on the modified multilayer sensing model.
2. The method according to claim 1, characterized in that, The historical typhoon data is clustered using a preset clustering algorithm to obtain multiple types of first-cluster data, including: By using a preset interpolation method, missing values are calculated for the various types of historical typhoon data to obtain continuous typhoon data. By using a preset K-means clustering algorithm, the continuous typhoon data is clustered to obtain multiple types of first cluster data.
3. The method according to claim 1, characterized in that, The step involves using a preset multi-layer sensing model to calculate the position at the next time step based on the position information at multiple times in the various types of second clustering data, thereby obtaining predicted path information, including: Calculate the Coriolis force at the next moment based on the historical typhoon data. The Coriolis force and the location information at multiple times in the second cluster data of the various types are input into a preset multilayer sensing model, so that the preset multilayer sensing model performs weighted summation calculation through weight matrix and bias vector, and outputs the predicted path information.
4. The method according to claim 1, characterized in that, The step of calculating the loss of a preset multilayer perception model using a preset dynamic time adjustment algorithm based on predicted path information and actual path information includes: Based on the predicted path information and the actual path information, using the formula: Calculate the loss of the preset multilayer perception model; where A is the actual path information, A = (a1, a2, ..., a...). i …a n B represents the predicted path information, B = (b1, b2, ..., bb2). j …b n );d(a i b j ) represents a i and b j The distance between them; a1, a2…a i …a n These represent the actual positions at time 1, time 2, ..., time i, ..., time n, respectively; b1, b2, ..., bn, respectively. j …b n These represent the predicted positions at time 1, time 2, ..., time j, ..., time n, respectively.
5. The method according to claim 1, characterized in that, After classifying and predicting the various types of first cluster data using a preset random forest classification algorithm to obtain various types of second cluster data, the method further includes: Based on the various types of second clustering data, an exponential decay model of wind speed over time, an exponential decay model of precipitation over time, and a power function decay model of precipitation with distance from the typhoon center were created. Acquire real-time typhoon data, and calculate and output wind speed information and precipitation information based on the exponential decay model of wind speed over time, the exponential decay model of precipitation over time, and the power function decay model of precipitation with distance from the typhoon center.
6. A typhoon forecasting device, characterized in that, include: The data acquisition module is used to acquire various types of historical typhoon data; The data clustering module is used to cluster the historical typhoon data using a preset clustering algorithm to obtain multiple types of first cluster data; By using a preset random forest classification algorithm, the various types of first cluster data are classified and predicted to obtain various types of second cluster data; The model creation module is used to calculate the position at the next time step based on the position information at multiple times in the second clustering data of the various types of data through a preset multi-layer perception model, thereby obtaining the predicted path information; to calculate the loss of the preset multi-layer perception model based on the predicted path information and the actual path information through a preset dynamic time adjustment algorithm; and to correct the preset multi-layer perception model based on the loss of the preset multi-layer perception model, thereby obtaining the corrected multi-layer perception model. The path prediction module is used to acquire real-time typhoon data and calculate the output path information based on the modified multilayer sensing model.
7. The apparatus according to claim 6, characterized in that, The data clustering module is specifically used to calculate missing values in the various types of historical typhoon data using a preset interpolation method to obtain continuous typhoon data; and to cluster the continuous typhoon data using a preset K-means clustering algorithm to obtain multiple types of first cluster data.
8. The apparatus according to claim 6, characterized in that, The model creation module is specifically used to calculate the Coriolis force corresponding to the next moment based on the historical typhoon data; input the Coriolis force and the location information of multiple moments in the second cluster data of various types into a preset multilayer sensing model, so that the preset multilayer sensing model performs weighted summation calculation through weight matrix and bias vector, and outputs the predicted path information.
9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, when executing a program stored in memory, implements the method described in any one of claims 1-5.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium contains a computer program that, when executed by a processor, implements the method described in any one of claims 1-5.