Tornado early warning method and system based on thermodynamic coupling
Through the tornado early warning method based on thermal coupling, the weight parameters are optimized using historical data and particle swarm algorithms, and tornado prediction combined with real-time meteorological data, the problem of low prediction lag and accuracy in the existing technology is solved, and higher prediction timeliness and accuracy is achieved.
Patent Information
- Application Number
- CN202510364230.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-26
- Publication Date
- 2025-07-08
AI Technical Summary
The existing technology has strong lag in tornado prediction and lacks deep-level mechanism analysis, resulting in low prediction accuracy.
Based on the thermally coupled tornado early warning method, an initial tornado prediction model is constructed by obtaining historical tornado sample data, and the weight parameters of thermal and dynamic contribution are optimized using particle swarm algorithms, and prediction is carried out in combination with real-time meteorological data.
It improves the timeliness and accuracy of tornado prediction, reduces the influence of human experience and subjective cognition, and enhances the real-time and accuracy of the model.
Smart Images
Figure CN120276072A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of meteorology, and particularly relates to a tornado warning method and system based on thermo - dynamic coupling. Background Art
[0002] A tornado is a highly destructive small - and - medium - scale meteorological disaster. Its instantaneous wind speed can reach 50 - 100 m / s, and even in extreme events exceeding EF5 level, the wind speed can reach more than 130 m / s, which poses great harm to society. For example, tornadoes can cause casualties; for another example, strong tornadoes can destroy houses, transmission towers, transportation facilities, and trigger secondary disasters such as fires and chemical leaks, resulting in losses of billions or even tens of billions. Therefore, accurately predicting tornadoes in advance helps to take corresponding measures in advance to ensure the personal safety of the public and reduce economic losses at the same time.
[0003] Currently, when predicting and warning tornadoes, the commonly used methods are based on numerical weather prediction models or helicity models. However, these methods mainly rely on macroscopic detection means such as radar and satellite data. They not only have strong hysteresis and are difficult to achieve timely prediction of tornadoes, but also lack accurate analysis of the deep - level mechanism of tornadoes, resulting in low prediction accuracy. Summary of the Invention
[0004] In view of this, aiming at the above - mentioned deficiencies, it is necessary to propose a tornado warning method and system based on thermo - dynamic coupling to improve the timeliness and accuracy of tornado prediction and warning.
[0005] In a first aspect, the present invention provides a tornado warning method based on thermo - dynamic coupling, including:
[0006] Pre - obtain historical tornado sample data; wherein, the historical tornado sample data includes: historical tornado intensity data, historical temperature data, historical wind speed data, historical thermal relaxation time data, and historical Coriolis parameter data;
[0007] Construct an initial tornado prediction model based on thermo - dynamic coupling;
[0008] Train the initial tornado prediction model using the historical tornado sample data to obtain a target tornado prediction model; wherein, when training the initial tornado prediction model, the particle swarm algorithm is used to optimize the weight parameters of the thermal contribution and dynamic contribution of the initial tornado prediction model;
[0009] Collect real - time meteorological data of the target area; wherein, the meteorological data includes temperature data, wind speed data, thermal relaxation time, and Coriolis parameter;
[0010] Input the real - time meteorological data into the target tornado prediction model and output the tornado prediction result;
[0011] Based on the tornado prediction result and a preset tornado warning threshold, a tornado warning is carried out.
[0012] Preferably, the initial tornado prediction model includes:
[0013] ψ = (ΔT / τ) α ×(ΔV / f) β
[0014] where ψ is the tornado prediction level, ΔT is the temperature gradient, τ is the thermal relaxation time, ΔV is the wind speed gradient, f is the Coriolis parameter, and α and β are the weight parameters of the thermal contribution and the dynamic contribution respectively, obtained through neural network learning.
[0015] Preferably, the training of the initial tornado prediction model using the historical tornado sample data includes:
[0016] Dividing the historical tornado sample data into a training data set and a test data set according to preset conditions;
[0017] Performing data preprocessing on the training data set to obtain a preprocessed training data set;
[0018] Based on the particle swarm optimization algorithm, using the preprocessed training data set to optimize the weight parameters α of the thermal contribution and β of the dynamic contribution;
[0019] Validating the tornado prediction model with optimized parameters using the test data set;
[0020] Determining the tornado prediction model composed of the weight parameters that meet the validation conditions as the target tornado prediction model.
[0021] Preferably, the data preprocessing of the training data set includes:
[0022] Using the exponential smoothing method to fill in the missing values of the historical thermal relaxation time data;
[0023] Using the interpolation method to fill in the missing values of the historical temperature data and the historical wind speed data;
[0024] Using the Z - score normalization method to normalize the historical temperature data, historical wind speed data, historical thermal relaxation time data, and historical Coriolis parameter data after filling in the missing values.
[0025] Preferably, the using the exponential smoothing method to fill in the missing values of the historical thermal relaxation time data includes:
[0026] Using the following formula to fill in the missing values of the historical thermal relaxation time data:
[0027] τ t = λτ t-1 + (1 - λ)τ obs
[0028] where τ t represents the filling value of the thermal relaxation time at time t, τ t-1 represents the thermal relaxation time of the previous time period, τ obs the most recent valid observed value of the thermal relaxation time, and λ is the smoothing coefficient.
[0029] Preferably, the interpolation method is used to fill the missing values of the historical temperature data and historical wind speed data, including:
[0030] Based on the following formula, the Kriging interpolation method is used to fill the missing values of the historical temperature data and historical wind speed data:
[0031]
[0032] where X(x) is the historical temperature data or historical wind speed data, ω i is the weight calculated from the semivariogram, and x i is the spatial coordinate of the neighboring site.
[0033] Preferably, the Z-score normalization method is used to normalize the historical temperature data, historical wind speed data, historical thermal relaxation time data, and historical Coriolis parameter data after filling the missing values:
[0034] The following calculation formula is used for normalization:
[0035]
[0036] where X norm is the result after normalizing the current data X to be normalized, μ X is the mean of the data to be normalized, and σ X is the standard deviation of the data to be normalized.
[0037] Preferably, based on the particle swarm algorithm, the weight parameters β of the thermal contribution α and dynamic contribution are optimized using the preprocessed training dataset, including:
[0038] Define particle parameters: The position of each particle is (α i , β i ), the velocity The population size is N, and the maximum number of iterations is T max ;
[0039] Calculate the fitness value of each particle in the t-th iteration;
[0040] Update the best position of the k-th particle using the initial position of the k-th particle, and update the best position of the population according to the fitness values of each particle;
[0041] Update the velocity and position of the k-th particle in the t-th iteration according to the best position of the particle and the best position of the population;
[0042] Repeat the above iterative steps until a preset stop condition is reached;
[0043] Among them, the optimized stop conditions include at least one of the following:
[0044] The iteration reaches a preset number of iterations;
[0045] The change rate of the fitness value is less than a preset threshold.
[0046] Preferably, updating the velocity and position of the k-th particle in the t-th iteration includes:
[0047] Update the particle using the following calculation formula:
[0048]
[0049] Among them, w is used to represent the inertia weight for balancing global and local search, c1 and c2 are respectively the individual and group learning factors, r1 and r2 are both random numbers, p best,i is the historical optimal position of particle i, g best is the global optimal position, represents the position of the i-th particle in the t-th iteration, represents the velocity of the i-th particle in the t-th iteration;
[0050] The calculation method of the fitness value is:
[0051]
[0052] Among them, F is the fitness value, ψ p is the predicted value of the tornado intensity, ψ t is the true value of the tornado intensity.
[0053] In a second aspect, the present invention provides a tornado warning system based on thermo - dynamic coupling, and the system includes: a model training module, a real - time data acquisition module, a prediction module, and a warning module;
[0054] The model training module is configured to pre-acquire historical tornado sample data; wherein, the historical tornado sample data includes: historical tornado intensity data, historical temperature data, historical wind speed data, historical heat relaxation time data, and historical Coriolis parameter data; construct an initial tornado prediction model based on thermo-dynamic coupling; use the historical tornado sample data to train the initial tornado prediction model to obtain a target tornado prediction model; wherein, when training the initial tornado prediction model, the particle swarm optimization algorithm is used to optimize the weight parameters of the thermal contribution and dynamic contribution of the initial tornado prediction model.
[0055] The real-time data acquisition module is configured to acquire real-time meteorological data of a target area; wherein, the meteorological data includes temperature data, wind speed data, heat relaxation time, and Coriolis parameter.
[0056] The prediction module is configured to input the real-time meteorological data into the target tornado prediction model and output a tornado prediction result.
[0057] The warning module is configured to perform a tornado warning according to the tornado prediction result and a preset tornado warning threshold.
[0058] As can be seen from the above technical solutions, in the tornado warning method and system based on thermo-dynamic coupling provided by the present invention, when performing a tornado warning, first, historical tornado sample data is pre-acquired, and an initial tornado prediction model constructed by using the historical tornado sample data is trained based on the particle swarm optimization algorithm to obtain a target tornado prediction model. Then, real-time meteorological data of the target area is acquired and input into the target tornado prediction model, and then a prediction result of the tornado intensity can be output. Furthermore, according to the tornado prediction result and a preset tornado warning threshold, a tornado warning can be performed. Thus, it can be seen that this solution can use the tornado prediction model to predict the tornado intensity according to the real-time acquired meteorological data, which has stronger real-time performance. Moreover, the traditional method relying on radar and satellite data is also affected by human experience and subjective cognition, while this solution performs prediction based on a model trained with historical data, and the influence of human experience and subjective cognition is smaller, so it also has higher prediction accuracy. In addition, the tornado prediction model constructed by this solution is based on thermo-dynamic coupling, fully considering the coupled synergistic effect of thermodynamic factors and atmospheric dynamic factors, and can also greatly improve the accuracy of tornado intensity prediction. Description of the Drawings
[0059] Figure 1 It is a flowchart of a tornado warning method based on thermo-dynamic coupling provided by an embodiment of the present invention.
[0060] Figure 2 It is a schematic diagram of a tornado warning system based on thermo-dynamic coupling provided by an embodiment of the present invention. Detailed Implementation Manner
[0061] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the accompanying drawings required for use in the embodiments. Obviously, the accompanying drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can also be obtained based on these drawings.
[0062] As Figure 1 shown, the present invention provides a tornado warning method based on thermo - dynamic coupling, and the method may include the following steps:
[0063] Step 101: Pre - obtain historical tornado sample data; wherein, the historical tornado sample data includes: historical tornado intensity data, historical temperature data, historical wind speed data, historical thermal relaxation time data, and historical Coriolis parameter data;
[0064] Step 102: Construct an initial tornado prediction model based on thermo - dynamic coupling;
[0065] Step 103: Use the historical tornado sample data to train the initial tornado prediction model to obtain a target tornado prediction model; wherein, when training the initial tornado prediction model, the particle swarm algorithm is used to optimize the weight parameters of the thermal contribution and dynamic contribution of the initial tornado prediction model;
[0066] Step 104: Collect real - time meteorological data of the target area; wherein, the meteorological data includes temperature data, wind speed data, thermal relaxation time, and Coriolis parameter;
[0067] Step 105: Input the real - time meteorological data into the target tornado prediction model and output the tornado prediction result;
[0068] Step 106: Perform tornado warning according to the tornado prediction result and a preset tornado warning threshold.
[0069] In this embodiment, the tornado prediction model can predict the tornado intensity according to the real - time collected meteorological data, which has stronger real - time performance. Moreover, the traditional method relying on radar and satellite data is also affected by human experience and subjective cognition. However, this solution predicts based on a model trained with historical data, with less influence of human experience and subjective cognition, and thus has higher prediction accuracy. In addition, the tornado prediction model constructed in this solution is based on thermo - dynamic coupling, fully considering the coupling and synergistic effects of thermodynamic factors and atmospheric dynamic factors, and can also greatly improve the accuracy of tornado intensity prediction.
[0070] In addition, this solution considers optimizing the weight parameters of the thermal contribution and dynamic contribution of the initial tornado prediction model through the particle swarm algorithm, which can enable the model to more accurately reflect the actual contributions of thermodynamic and kinetic parameters to the tornado intensity. Moreover, based on the particle swarm algorithm, through the swarm intelligence mechanism, the weight combinations of thermodynamic and kinetic parameters are efficiently and robustly explored, enabling the model to capture complex nonlinear relationships, possess high precision and strong generalization ability. Its advantages and purposes can closely revolve around the actual needs of tornado prediction, providing reliable technical support for meteorological disaster warning.
[0071] For step 101, historical tornado sample data is obtained in advance.
[0072] In this step, considering obtaining historical tornado sample data in advance, it can specifically include: historical tornado intensity data, historical temperature data, historical wind speed data, historical thermal relaxation time data, and historical Coriolis parameter data; for historical sample data, it can be considered to be obtained through meteorological stations in southern provinces. For example, the historical tornado sample data can include the data of 37 tornado events from the Guangdong Meteorological Observatory from 2019 to 2022. Of course, in some embodiments, the historical tornado sample data can also include data from other countries and regions, such as meteorological data integrated through historical tornado event databases such as NOAA Storm Events and real-time satellite remote sensing data.
[0073] Step 102: Construct an initial tornado prediction model based on thermo-dynamic coupling.
[0074] In this step, considering constructing the initial model based on the thermo-dynamic coupling method, the constructed initial tornado prediction model can be:
[0075] ψ = (ΔT / τ) α ×(ΔV / f) β
[0076] where ψ is the tornado prediction level, ΔT is the temperature gradient, τ is the thermal relaxation time, ΔV is the wind speed gradient, f is the Coriolis parameter, and α and β are the weight parameters of the thermal contribution and dynamic contribution respectively, obtained through neural network learning.
[0077] In this embodiment, when ΔT / τ decreases, the thermal relaxation time increases, and the energy accumulation time prolongs, suppressing short-term strong convection and enabling higher prediction accuracy for persistent strong tornadoes. When the ΔV / f ratio increases, such as in low-latitude regions where f is small, it can promote the rapid enhancement of cyclonic vorticity and can be used to improve the prediction accuracy of tornado rotation speed. Therefore, this embodiment can deeply consider the formation mechanism of tornadoes by quantifying the non-linear interaction of thermal relaxation time, temperature gradient, wind speed gradient, and Coriolis parameter, thereby contributing to improving the prediction accuracy of tornadoes.
[0078] Step 103: Train the initial tornado prediction model using historical tornado sample data to obtain the target tornado prediction model.
[0079] In this step, when training the initial tornado prediction model, it is considered to use the particle swarm algorithm to optimize the weight parameters of the thermal contribution and dynamic contribution of the initial tornado prediction model. Since particle swarm optimization shares information through particle swarm cooperation, it can combine individual optimal and global optimal to guide the search direction and avoid falling into local optima. In the tornado prediction model, the interaction relationship between the thermodynamic term and the dynamic term is highly non-linear. The traditional gradient descent method is easily interfered by local optima, while the particle swarm algorithm can comprehensively explore the search space. Moreover, particle swarm optimization only depends on the fitness function and does not require the gradient information of the objective function. In meteorological models, the physical meanings of α and β are complex, and the objective function may be non-differentiable or the derivative is difficult to analyze. The particle swarm algorithm can bypass this limitation.
[0080] Specifically, when training the initial tornado prediction model using historical tornado sample data, it can be achieved in the following way:
[0081] Step S31: Divide the historical tornado sample data into a training data set and a test data set according to preset conditions.
[0082] In this step, for example, it can be considered to divide the historical tornado sample data according to a ratio of 70% and 30%. Take 70% of the historical tornado sample data as the training data set, and the remaining 30% as the test data set.
[0083] Step S32: Perform data preprocessing on the training data set to obtain the preprocessed training data set.
[0084] In this step, when performing data preprocessing, it can be considered:
[0085] (1) Use the exponential smoothing method to fill in the missing values of the historical thermal relaxation time data;
[0086] For example, use the following formula to fill in the missing values of the historical thermal relaxation time data:
[0087] τ t = λτ t-1 + (1 - λ)τ obs
[0088] where τ t represents the filling value of the thermal relaxation time at time t, τ t-1 represents the thermal relaxation time of the previous time period, τ obs is the observed value of the most recent effective thermal relaxation time, and λ is the smoothing coefficient.
[0089] (2) Use the interpolation method to fill in the missing values of the historical temperature data and historical wind speed data;
[0090] For example, based on the following formula, use the Kriging interpolation method to fill in the missing values of the historical temperature data and historical wind speed data:
[0091]
[0092] where X(x) is the historical temperature data or historical wind speed data, ω i is the weight calculated from the semivariogram, and x i is the spatial coordinate of the neighboring station.
[0093] (3) Use the Z-score standardization method to standardize the historical temperature data, historical wind speed data, historical thermal relaxation time data, and historical Coriolis parameter data after filling in the missing values.
[0094] For example, use the following calculation formula for standardization:
[0095]
[0096] where X norm is the result after standardizing the current data X to be standardized, μ X is the mean of the data to be standardized, and σ X is the standard deviation of the data to be standardized.
[0097] Step S33: Based on the particle swarm algorithm, use the preprocessed training dataset to optimize the thermal contribution α and the weight parameter β of the dynamic contribution;
[0098] In this step, it can be specifically implemented in the following way:
[0099] Define the particle parameters: The position of each particle is (α i , β i ), the velocity The population size is N, and the maximum number of iterations is T max ;
[0100] Calculate the fitness value of each particle in the t-th iteration;
[0101] Update the best position of the k-th particle using the initial position of the k-th particle, and update the best position of the population according to the fitness values of each particle;
[0102] Update the velocity and position of the k-th particle in the t-th iteration according to the best position of the particle and the best position of the population;
[0103] Repeat the above iteration steps until a preset stop condition is reached;
[0104] Among them, the optimization stop conditions include at least one of the following:
[0105] The iteration reaches the preset number of iterations;
[0106] The change rate of the fitness value is less than the preset threshold.
[0107] Specifically, a parameter range can be given, α ∈ [0.1, 3.0], β ∈ [0.1, 2.0], N = 50, T max = 200. When calculating the fitness value, it can be calculated through the formula where F is the fitness value, ψ p is the predicted value of the tornado intensity, and ψ t is the true value of the tornado intensity.
[0108] After calculating the fitness values of each particle, consider updating the best position of the k-th particle and the best position of the population. In the initial stage, the best position of the particle is updated through the initial position of the particle, that is, the best position of each particle is updated according to the values in the particle position. The best position of the population can be updated through the fitness value. For example, a fitness threshold can be set. After calculating the fitness value of the particle, if the fitness value is greater than the preset fitness threshold, the value of the particle is updated to the best position of the population, otherwise no update is made.
[0109] Furthermore, when updating the velocity and position of the k-th particle in the t-th iteration, the particle can be specifically updated using the following calculation formula:
[0110]
[0111] where w is used to represent the inertia weight for balancing global and local search, c1 and c2 are the individual and group learning factors respectively, r1 and r2 are both random numbers, p best,i is the historical optimal position of particle i, and g best is the global optimal position, represents the position of the i-th particle in the t-th iteration, Characterizes the speed of the ith particle at the tth iteration.
[0112] In this step, w can be set to 0.7, c1 and c2 can both be set to 1.5. The preset threshold in the optimized stop condition can be 10 -5 .
[0113] Step S34: using the test data set to verify the tornado prediction model after optimizing the parameters;
[0114] In this step, consider using the test data set to calculate the mean square error to verify the model. For example, the mean square error can be calculated using the following formula:
[0115]
[0116] Among them, N val is the number of test sample sets, ψ pred,i is the tornado intensity predicted by the model, ψ true,i The true intensity of the tornado.
[0117] Step S35: Determine the tornado prediction model composed of weight parameters that meet the verification conditions as the target tornado prediction model.
[0118] In this step, the error of the validation set is used to determine the reliability of the model. For example, if the error of the validation set does not decrease after 10 consecutive rounds of training, the training can be terminated to obtain the final model. For example, the final training result is ψ = (ΔT / τ) 0.6 ×(ΔV / f) 0.4 .
[0119] Step 104: Collecting real-time meteorological data of the target area; wherein the meteorological data includes temperature data, wind speed data, thermal relaxation time and Coriolis parameters;
[0120] Step 105: input the real-time meteorological data into the target tornado prediction model, and output the tornado prediction result;
[0121] Step 106: Issue a tornado warning based on the tornado prediction result and the preset tornado warning threshold.
[0122] In this step, it is used to determine whether a tornado warning is needed based on the prediction results. For example, when the prediction result output by the tornado prediction model is greater than 3.7, consider issuing a warning through the warning device to prompt personnel to take corresponding preventive measures.
[0123] like Figure 2As shown in the figure, the present invention also provides a tornado warning system based on thermo - dynamic coupling. The system includes: a model training module 201, a real - time data acquisition module 202, a prediction module 203, and a warning module 204;
[0124] The model training module 201 is configured to pre - obtain historical tornado sample data; where the historical tornado sample data includes: historical tornado intensity data, historical temperature data, historical wind speed data, historical thermal relaxation time data, and historical Coriolis parameter data; construct an initial tornado prediction model based on thermo - dynamic coupling; and use the historical tornado sample data to train the initial tornado prediction model to obtain a target tornado prediction model; where, when training the initial tornado prediction model, the particle swarm algorithm is used to optimize the weight parameters of the thermal contribution and dynamic contribution of the initial tornado prediction model.
[0125] The real - time data acquisition module 202 is configured to collect real - time meteorological data of the target area; where the meteorological data includes temperature data, wind speed data, thermal relaxation time, and Coriolis parameter.
[0126] The prediction module 203 is configured to input the real - time meteorological data into the target tornado prediction model and output a tornado prediction result.
[0127] The warning module 204 is configured to perform tornado warning according to the tornado prediction result and a preset tornado warning threshold.
[0128] In this embodiment, the system can be an embedded meteorological warning terminal, and its chip uses an FPGA dedicated computing chip to improve the operation rate of the CPU. For example, the terminal can also be linked with the smart grid, and when the tornado intensity reaches a certain value, the power supply in high - risk areas can be automatically cut off to reduce the risk of secondary disasters. At the same time, the warning information can also be pushed to the emergency management department through 4G or 5G networks.
[0129] This specification also provides a computer - readable storage medium, on which a computer program is stored. When the computer program is executed on a computer, the computer is made to execute the method in any one of the embodiments in the specification.
[0130] This specification also provides a computing device, including a memory and a processor. The memory stores executable code, and when the processor executes the executable code, the method in any one of the embodiments in the specification is implemented.
[0131] Regarding the information interaction, execution process, etc. among the various units within the above - mentioned device, since it is based on the same concept as the method embodiment of this specification, the specific content can be referred to the description in the method embodiment of this specification, and will not be elaborated here.
[0132] The modules or units in the device according to the embodiments of the present invention can be combined, divided, and deleted according to actual needs. The above-disclosed content is only the preferred embodiments of the present invention, and of course, it cannot be used to limit the scope of the rights of the present invention. Those of ordinary skill in the art can understand all or part of the processes of implementing the above embodiments, and the equivalent changes made according to the claims of the present invention still fall within the scope covered by the present invention.
Claims
1. A tornado warning method based on thermo-mechanical coupling, characterized in that Including: Pre-acquire historical tornado sample data; wherein, the historical tornado sample data includes: historical tornado intensity data, historical temperature data, historical wind speed data, historical thermal relaxation time data, and historical Coriolis parameter data; Construct an initial tornado prediction model based on thermo-hydrodynamic coupling; Train the initial tornado prediction model using the historical tornado sample data to obtain a target tornado prediction model; wherein, when training the initial tornado prediction model, the particle swarm optimization algorithm is used to optimize the weight parameters of the thermal contribution and dynamic contribution of the initial tornado prediction model; Collect real-time meteorological data of the target area; wherein, the meteorological data includes temperature data, wind speed data, thermal relaxation time, and Coriolis parameter; Input the real-time meteorological data into the target tornado prediction model and output the tornado prediction result; Perform tornado warning according to the tornado prediction result and a preset tornado warning threshold; 2. The tornado warning method based on thermo-fluid coupling according to claim 1, characterized in that The initial tornado prediction model includes: ψ = (ΔT / τ) α × (ΔV / f) β Wherein, ψ is the tornado prediction level, ΔT is the temperature gradient, τ is the thermal relaxation time, ΔV is the wind speed gradient, f is the Coriolis parameter, and α and β are respectively the weight parameters of the thermal contribution and dynamic contribution, which are obtained through neural network learning.
3. The tornado warning method based on thermo-mechanical coupling according to claim 2, wherein The training of the initial tornado prediction model using the historical tornado sample data includes: Divide the historical tornado sample data into a training data set and a test data set according to preset conditions; Perform data preprocessing on the training data set to obtain a preprocessed training data set; Based on the particle swarm optimization algorithm, use the preprocessed training data set to optimize the weight parameters α of the thermal contribution and β of the dynamic contribution; Use the test data set to verify the tornado prediction model with optimized parameters; Determine the tornado prediction model composed of weight parameters that meet the verification conditions as the target tornado prediction model.
4. The tornado warning method based on thermo-fluid coupling according to claim 3, characterized in that Performing data preprocessing on the training data set includes: Adopt the exponential smoothing method to fill in the missing values of the historical thermal relaxation time data; Adopt the interpolation method to fill in the missing values of the historical temperature data and historical wind speed data; Adopt the Z-score standardization method to standardize the historical temperature data, historical wind speed data, historical thermal relaxation time data, and historical Coriolis parameter data after missing value filling; 5. The tornado warning method based on thermo-mechanical coupling according to claim 4, wherein, The adoption of the exponential smoothing method to fill in the missing values of the historical thermal relaxation time data includes: Use the following formula to fill in the missing values of the historical thermal relaxation time data: τ t = λτ t-1 + (1 - λ)τ obs Among them, τ t represents the filling value of the thermal relaxation time at time t, τ t-1 represents the thermal relaxation time of the previous time period, τ obs is the most recent valid observed value of the thermal relaxation time, and λ is the smoothing coefficient.
6. The tornado warning method based on thermo - kinematic coupling according to claim 4, wherein The adoption of the interpolation method to fill in the missing values of the historical temperature data and historical wind speed data includes: Based on the following formula, adopt the Kriging interpolation method to fill in the missing values of the historical temperature data and historical wind speed data; Among them, X(x) is historical temperature data or historical wind speed data, ω i is the weight calculated by the semivariogram, and x i is the spatial coordinate of the neighboring site.
7. The tornado warning method based on thermo-mechanical coupling according to claim 4, characterized in that, The adoption of the Z-score standardization method to standardize the historical temperature data, historical wind speed data, historical thermal relaxation time data, and historical Coriolis parameter data after missing value filling: Adopt the following calculation formula for standardization processing: Among them, X norm is the result after standardizing the currently to-be-standardized data X, μ X is the mean of the to-be-standardized data, and σ X is the standard deviation of the to-be-standardized data.
8. The tornado warning method based on thermo - kinematic coupling according to claim 3, wherein, The optimization of the weight parameters α of the thermal contribution and β of the dynamic contribution using the preprocessed training data set based on the particle swarm optimization algorithm includes: Define particle parameters: The position of each particle is (α i , β i ), and the velocity The population size is N, and the maximum number of iterations is T max ; Calculate the fitness value of each particle in the t-th iteration; Update the best position of the k-th particle using the initial position of the k-th particle, and update the best position of the population according to the fitness values of each particle; Update the velocity and position of the k-th particle in the t-th iteration according to the best position of the particle and the best position of the population; Repeat the above iteration steps until a preset stopping condition is reached; Among them, the optimized stopping conditions include at least one of the following: The iteration reaches a preset number of iterations; The change rate of the fitness value is less than a preset threshold.
9. The tornado warning method based on thermo-fluid coupling according to claim 8, wherein, The update of the velocity and position of the k-th particle in the t-th iteration includes: Update the particle using the following calculation formula: Among them, w is used to represent the inertia weight for balancing global and local searches, c1 and c2 are the individual and swarm learning factors respectively, r1 and r2 are both random numbers, and p best,i is the historical best position of particle i, and g best is the global best position, represents the position of the i-th particle at the t-th iteration, and represents the velocity of the i-th particle at the t-th iteration; The calculation method of the fitness value is: Among them, F is the fitness value, and ψ p is the predicted value of the tornado intensity, and ψ t is the true value of the tornado intensity.
10. A tornado warning system based on thermo-mechanical coupling, characterized in that, The system includes: a model training module, a real-time data acquisition module, a prediction module, and a warning module; The model training module is configured to pre-obtain historical tornado sample data; among them, the historical tornado sample data includes: historical tornado intensity data, historical temperature data, historical wind speed data, historical thermal relaxation time data, and historical Coriolis parameter data; construct an initial tornado prediction model based on thermo-dynamic coupling; use the historical tornado sample data to train the initial tornado prediction model to obtain a target tornado prediction model; among them, when training the initial tornado prediction model, the particle swarm algorithm is used to optimize the weight parameters of the thermal contribution and dynamic contribution of the initial tornado prediction model; The real-time data acquisition module is configured to collect real-time meteorological data of the target area; among them, the meteorological data includes temperature data, wind speed data, thermal relaxation time, and Coriolis parameter; The prediction module is configured to input the real-time meteorological data into the target tornado prediction model and output a tornado prediction result; The warning module is configured to perform tornado warning according to the tornado prediction result and a preset tornado warning threshold.