TC path and intensity prediction method, equipment and medium based on AI and physical properties

Abstract

The present invention discloses a TC path and intensity prediction method, equipment and medium based on AI and physical properties, which aims to improve the short-time scale forecast accuracy of TC path and intensity. By integrating multiple factors with clear physical meanings, such as guiding airflow, ventilation flow, vorticity, ocean surface temperature, etc., these physical factors are subjected to correlation analysis and regression analysis, etc., to ensure that the input factors have a significant impact on the changes in TC path and intensity. A variety of AI models such as convolutional neural network (CNN), recurrent neural network (RNN), multi-layer perceptron (MLP), random forest (RF), extreme gradient boosting (XGBoost) and long short-term memory network (LSTM) are used to capture the complex nonlinear relationship between meteorological factors. In addition, by adopting a real-time feedback mechanism, new observation data is allowed to be dynamically input and model parameters are optimized, thereby improving the real-time nature of the prediction results. This method is particularly suitable for the northwest Pacific region and has broad application prospects, such as providing scientific support in meteorology, aviation, shipping and other fields.

Description

Technical Field

[0001] The present invention relates to the technical field of the intersection of weather forecasting and artificial intelligence (AI), and specifically to the prediction of the path and intensity of tropical cyclones. The present invention relates to a method, computer device, and readable storage medium for predicting the path and intensity of tropical cyclones (TCs) based on AI technology and meteorological and oceanic physical properties. By integrating AI technology and ocean-atmospheric physical processes, the accuracy, timeliness, and interpretability of tropical cyclone path and intensity predictions are improved. Background Art

[0002] Tropical cyclones (TCs) are a crucial and complex phenomenon in meteorology, capable of triggering extreme weather events such as strong winds, heavy rains, floods, and storm surges. Therefore, improving the accuracy and timeliness of TC track and intensity forecasts has always been a key issue in meteorological disaster forecasting. With the intensification of global climate change and the increasing frequency of extreme weather events, the accuracy and timeliness of TC forecasts have become increasingly important.

[0003] Traditional TC forecasts primarily rely on Numerical Weather Prediction (NWP) models, which simulate the evolution of meteorological systems by solving atmospheric dynamics and thermodynamics equations. However, while NWP models demonstrate high accuracy in long-term forecasts, they often face numerous challenges in short-term forecasts (e.g., 6-24 hours). First, while data assimilation technology is becoming increasingly mature, the quality and quantity of real-time observational data often fall short, hindering model performance. Second, NWP models require significant computing resources, especially at high resolution, which limits their application. Furthermore, the evolution of TCs is influenced by multiple meteorological factors, such as vertical wind shear, ocean surface temperature, and ambient humidity. Traditional models are ill-equipped to handle these complex nonlinear relationships, often leading to biased forecasts.

[0004] In recent years, the rapid development of artificial intelligence (AI) technology has opened up new possibilities for weather forecasting. Leveraging machine learning and deep learning algorithms, AI models can extract useful features from large-scale meteorological data and capture the complex relationships between meteorological factors. For example, advanced AI models such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs) have been widely used in forecasting TC tracks and intensities. These models' strengths lie in their powerful feature extraction capabilities and adaptability to complex patterns, enabling them to surpass traditional numerical models in certain situations. However, while AI models perform well in some situations, their forecasts often lack interpretability, particularly when multiple meteorological factors interact.

[0005] Most current research involving AI models fails to fully consider factors with obvious physical significance for TC paths and intensities, such as guiding air currents that significantly influence typhoon paths, ocean surface and subsurface temperatures, and mesoscale ocean eddy information that significantly influence typhoon intensity. These factors play a crucial role in influencing the movement and intensity of TCs, but existing AI models, because they focus more on data-driven algorithm optimization, often overlook the direct connection between these physical factors and tropical cyclone behavior. Incorporating these factors into AI model inputs is expected to significantly improve forecast accuracy and reliability. Secondly, although existing AI models have powerful feature extraction capabilities, they often lack a deep understanding of physical processes, which limits the model's generalization ability and performance in extreme situations.

[0006] In summary, how to effectively combine physical processes with artificial intelligence technology, fully consider the synergistic effects of multiple influencing factors, and construct an innovative TC path and intensity forecast model with high efficiency, high accuracy, and good interpretability, thereby improving the accuracy and timeliness of TC forecasts and providing more effective support for meteorological services, disaster management, and related decision-making, is a technical problem that needs to be urgently solved in the current field of TC path and intensity forecast research. Summary of the Invention

[0007] (1) Purpose of the invention

[0008] In response to the defects and shortcomings in existing tropical cyclone path and intensity forecasts, such as insufficient accuracy and timeliness, poor physical interpretability of AI models, and insufficient consideration of important meteorological physical factors, in order to solve at least one of the above-mentioned and other technical problems in the prior art, the present invention aims to provide a TC path and intensity prediction method, computer equipment and medium based on AI and physical characteristics. By identifying and integrating multiple influencing factors with clear physical significance, such as guiding airflow, ventilation flow, vorticity, ocean surface temperature and vertical wind shear, combined with AI models and introducing a real-time feedback mechanism to dynamically update the prediction data, efficient and accurate TC path and intensity predictions are achieved. This is particularly suitable for responding to climate change and extreme weather events, and not only significantly improves the accuracy, timeliness and interpretability of forecasts, but also provides reliable support for meteorological services, disaster management and related industries.

[0009] (2) Technical solution

[0010] In order to achieve the purpose of the invention and solve the technical problems, the present invention adopts the following technical solutions:

[0011] The first object of the present invention is to provide a TC track and intensity prediction method based on AI and physical properties, which is used to predict the track and intensity changes of tropical cyclones on a short time scale (6-24 hours). The prediction method, when implemented, includes at least the following steps:

[0012] SS1. Data Collection

[0013] At least five consecutive years of historical meteorological data must be obtained from multiple meteorological and oceanographic data sources, including at least TC Best Track data that records track and intensity information, Final Analysis Data (FNL) that provides six-hourly updated meteorological variables, and ocean reanalysis data generated by the Global Ocean Data Assimilation System (GODAS).

[0014] SS2. Data Preprocessing

[0015] Based on the historical meteorological and oceanographic data collected in step SS1, extract relevant data for each TC occurrence period within the target prediction area (e.g., the northwest Pacific region) and perform data quality control. This includes identifying and removing missing or anomalous data, standardizing and normalizing the data to ensure temporal and spatial consistency across different data sources, and then partitioning the processed dataset into training and validation sets. Ensure that the training and validation data are independent in terms of temporal (different years or different time periods) and spatial (different geographic regions) characteristics.

[0016] SS3. Selection and calculation of model output factors

[0017] Based on the historical TC Best Track data preprocessed in step SS2 and the characteristics of the target forecast area, multiple variables closely related to TC track and intensity changes are selected as model output factors. These model output factors include at least TC position changes (longitude and latitude changes) and intensity changes (minimum central pressure and maximum wind speed changes). Time series of these variables are calculated to extract characteristic changes and establish model output targets.

[0018] SS4. Selection and calculation of model input factors

[0019] Based on the physical mechanisms of TC motion and intensity changes, and in combination with the FNL atmospheric reanalysis data and GODAS ocean reanalysis data preprocessed in step SS2, multiple physical factors related to TC motion and intensity changes are selected and calculated. Correlations between these selected physical factors and TC paths and intensity changes are then analyzed. The correlation coefficients between each physical factor and the model output factors are calculated, and the model input factor set is ultimately determined based on these analysis results.

[0020] SS5. Construction and Training of TC Track and Intensity Prediction Model

[0021] Based on the model output factors and model input factors determined in steps SS3 and SS4, select at least one AI model from among convolutional neural network (CNN), recurrent neural network (RNN), multi-layer perceptron (MLP), long short-term memory network (LSTM), random forest (RF), and extreme gradient boosting (XGBoost) to build a TC path and intensity prediction model, and use the training set data in step SS2 to train the constructed TC path and intensity prediction model;

[0022] SS6. Verification and Optimization of TC Track and Intensity Prediction Models

[0023] Validate the TC track and intensity prediction model trained in step SS5 using the validation data from step SS2. Evaluate the model's performance at different forecast timescales (e.g., 6-hour, 12-hour, and 24-hour). The evaluation metrics include at least the track forecast error and the intensity forecast error. Based on the validation results, optimize the model by adjusting hyperparameters, adding regularization, and / or using cross-validation to improve the model's prediction accuracy and generalization ability.

[0024] SS7. Real-time forecast and dynamic updates

[0025] Establish a real-time feedback mechanism to dynamically input new pre-processed observational meteorological data into the model optimized in step SS6. Continuously adjust and optimize model parameters through online learning and / or regular retraining triggered in real time by new data changes. Regularly evaluate model performance and update the input factor set and model structure to improve the timeliness and accuracy of TC track and intensity forecasts.

[0026] SS8. Output of prediction results

[0027] Generate and output forecast results of TC path changes (longitude and latitude) and intensity changes (minimum air pressure and maximum wind speed) in the next 6 hours, 12 hours and 24 hours, calculate and output forecast errors, including but not limited to path forecast error and intensity forecast error, provide credibility assessment of forecast results, and provide decision support information for meteorological departments and disaster prevention and mitigation agencies.

[0028] The second invention object of the present invention is to provide a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, it implements the steps of the above-mentioned TC path and intensity prediction method based on AI and physical characteristics.

[0029] The third invention object of the present invention is to provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-mentioned TC path and intensity prediction method based on AI and physical characteristics.

[0030] (3) Technical effects

[0031] Compared with the existing technology, the TC path and intensity prediction method, device and medium based on AI and physical properties of the present invention have the following beneficial and significant technical effects:

[0032] (1) This invention significantly improves the accuracy of tropical cyclone (TC) path and intensity forecasts by integrating traditional meteorological physics with modern AI technology. By selecting physical factors related to TC movement and intensity changes, such as guiding airflow, ventilation flow, vertical wind shear, and sea surface temperature, and combining them with the powerful nonlinear fitting capabilities of AI models, the evolution of TCs can be captured more accurately. At the same time, the prediction method of the present invention is suitable for short-term forecasts (6 to 24 hours), greatly improving the response speed to sudden meteorological events and providing effective technical support for disaster prevention and mitigation.

[0033] (2) The core of this invention lies in identifying and integrating 35 influencing factors with clear physical meanings, such as guiding airflow, ventilation flow, vorticity, sea surface temperature, and vertical wind shear. Through correlation analysis and regression analysis, these factors were determined to have a significant impact on the changes in TC track and intensity. This approach not only enhances the interpretability of the model but also makes the prediction results more consistent with meteorological principles, thereby improving the accuracy and reliability of the prediction.

[0034] (3) This paper designs a variety of artificial intelligence models, including convolutional neural networks (CNNs), recurrent neural networks (RNNs), multi-layer perceptrons (MLPs), random forests (RFs), extreme gradient boosting (XGBoost), and long short-term memory networks (LSTMs). These models are designed to capture the complex nonlinear relationships between meteorological factors and provide accurate TC path and intensity forecasts. In addition, these models are trained using TC data from 2015 to 2020, and data from 2021 to 2022 are used for validation to ensure the effectiveness and adaptability of the models.

[0035] (4) This invention also introduces a real-time feedback mechanism that allows new observational data to be dynamically fed into the model, thereby continuously adjusting and optimizing the forecast results. This mechanism not only improves the real-time performance of TC forecasts but also enables online learning and regular retraining based on the latest data, ensuring that the model maintains a high level of forecast accuracy.

[0036] (5) The forecast model of the present invention not only outputs the TC path and intensity changes for the next 6, 12, and 24 hours, but also provides error assessment and reliability analysis of the path and intensity forecasts. This function helps meteorologists make decisions based on the reliability of the forecast results, thereby improving the usability and reliability of the forecast information.

[0037] (6) The TC forecast model of the present invention can significantly improve the accuracy and timeliness of path and intensity forecasts, providing an important basis for decision-making in meteorology, aviation, shipping, agriculture, and other fields. This technology has broad application prospects in the field of meteorological forecasting and can effectively enhance the ability to respond to extreme weather. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0039] Figure 1 FIG2 is a flowchart illustrating an implementation of the TC path and intensity prediction method based on AI and physical properties of the present invention;

[0040] Figure 2 Figure 2 shows the schematic diagram of variable factor statistics, where (A) is the schematic diagram of the calculation area for atmospheric background field variables, and (B) is the schematic diagram of the calculation area for ocean variables and geographic variables.

[0041] Figure 3 Figure 2 shows a comparison of the forecast errors of different models for all TC tracks and intensities in 2021-2022, where (a) shows the 6-24h forecast error for TC tracks and (b) shows the 6-24h forecast error for TC intensities.

[0042] Figure 4 The figure shows a comparison of the track forecast results of TC No. 10 in 2021 by various models, where (a) to (d) correspond to the track forecast results of Models 1 to 4, and (e) to (j) correspond to the track forecast results of Models 5 to 10.

[0043] Figure 5The figure shows a comparison of the intensity forecast results of various models for TC No. 10 in 2021, where (a) to (d) correspond to the intensity forecast results of Models 1 to 4, and (e) to (j) correspond to the intensity forecast results of Models 5 to 10.

[0044] Figure 6 The figure shows the path return results of the AI ​​model RF for all TCs from August to October 2021, where (a) to (l) correspond to the path return results of TC202109 to TC202120, respectively;

[0045] Figure 7 Shown is a schematic diagram of the intensity return results of the AI ​​model RF for all TCs from August to October 2021, where (a) to (l) correspond to the intensity return results of TC202109 to TC202120 respectively. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solutions and advantages of the implementation of the present invention clearer, the technical solutions in the embodiments of the present invention will be described in more detail below with reference to the drawings in the embodiments of the present invention. The described embodiments are part of the embodiments of the present invention, rather than all the embodiments. The embodiments described below with reference to the drawings are exemplary and are intended to be used to explain the present invention, and should not be understood as limiting the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0047] The present invention aims to provide a TC path and intensity prediction method, computer equipment and medium based on AI and physical properties. By identifying and integrating multiple influencing factors with clear physical meanings, such as guiding airflow, ventilation flow, vorticity, ocean surface temperature and vertical wind shear, combined with AI models and introducing a real-time feedback mechanism to dynamically update prediction data, efficient and accurate TC path and intensity prediction can be achieved. It is particularly suitable for responding to climate change and extreme weather events. It not only significantly improves the accuracy, timeliness and interpretability of forecasts, but also provides reliable support for meteorological services, disaster management and related industries.

[0048] As a specific example, Figure 1 As shown, the TC path and intensity prediction method based on AI and physical properties of the present invention is used to predict the path and intensity changes of tropical cyclones on a short time scale (6-24 hours). The implementation mainly includes the following steps:

[0049] At least five consecutive years of historical meteorological data were obtained from multiple meteorological and oceanographic data sources, including TC Best Track data that records track and intensity information, atmospheric reanalysis data (FNL) that provides meteorological variables updated every six hours, and ocean reanalysis data generated by the Global Ocean Data Assimilation System (GODAS).

[0050] The data and related information used in this paper are commonly used datasets in TC research. The TC BestTrack dataset can come from a variety of sources. The Japan Meteorological Agency (JMA) provides the most up-to-date data, so we used the JMA BestTrack dataset, which was downloaded directly from its website. FNL and GODAS data can be downloaded from the website of the China Meteorological Administration Information Center. All data were acquired from January 1, 2015, to December 31, 2022, using data from the period of TC occurrence in the northwestern Pacific. Some data may be missing.

[0051] Based on the historical meteorological and oceanographic data collected in step SS1, extract relevant data for each TC occurrence period in the target prediction area (e.g., the northwest Pacific region) and perform data quality control on it. This includes identifying and removing missing or anomalous data, standardizing and normalizing the data to ensure temporal and spatial consistency across different data sources, and then dividing the processed dataset into a training set and a validation set. Ensure that the training and validation data are independent in terms of temporal (different years or different periods) and spatial (different geographical regions) characteristics.

[0052] Preferably, in steps SS1 and SS2, information on the occurrence period of TCs in the Northwest Pacific Ocean between January 1, 2015, and December 31, 2022, is selected from multiple meteorological and oceanographic data sources. For each TC, the atmospheric and oceanic background field information in the reports is collected every six hours, with missing data removed. The TC occurrence period from January 1, 2015, to December 31, 2020, is used as training data, comprising 148 TCs and 2,660 statistical reports. All TCs from 2021 to 2011 are used as test data, comprising 47 TCs and 621 statistical reports.

[0053] Preferably, in step SS2, when performing data quality control, missing data can be addressed by using mean filling, interpolation, or deletion of records with missing values. Outliers can be identified and removed through statistical analysis using Z-scores or boxplots. Z-score standardization or Min-Max normalization can be used to process data of varying scales. Furthermore, the collected historical meteorological data can be time-synchronized to ensure consistent temporal resolution across all data sources (e.g., one data point every six hours). Furthermore, data quality control can also employ automated data cleaning methods based on machine learning, identifying potential anomalies through historical data learning and automatically performing data correction or filling to improve the efficiency and accuracy of data preprocessing.

[0054] Based on the historical TC Best Track data preprocessed in step SS2 and the characteristics of the target forecast area, multiple variables closely related to TC path and intensity changes are selected as model output factors. These model output factors include at least TC position changes (changes in longitude and latitude) and intensity changes (changes in minimum central pressure and maximum wind speed). Time series of these variables are calculated to extract characteristic changes and establish model output targets.

[0055] Preferably, in this step, the output factors of the model are the hourly position change delta(LON) and delta(LAT) of the TC, and the hourly intensity change delta(SLP) and delta(MWS) of the TC, calculated from the TC Best Track dataset. The intensity change of the TC is generally represented by the minimum pressure SLP at the center of the TC and the maximum wind speed MWS at the center of the TC. The hourly position and intensity changes of the TC are calculated using the formula X_n=X(t+n)-X(t), where X is LON, LAT, SLP, and MWS, t represents the current time, and n represents the hour. In the present invention, n is respectively set to 6h, 12h, and 24h. There are 12 output factors in total, as shown in OutputTarget in Table 1 below.

[0056] Based on the physical mechanisms of TC motion and intensity changes, and combined with the FNL atmospheric reanalysis data and GODAS ocean reanalysis data preprocessed in step SS2, multiple physical factors related to TC motion and intensity changes are selected and calculated. Then, correlation analysis is performed on the selected physical factors and TC path and intensity changes. The correlation coefficient between each physical factor and the model output factor is calculated, and the model input factor set is finally determined based on the analysis results.

[0057] Preferably, in step SS4, the multiple physical factors selected that are related to TC motion and intensity changes should at least include guiding airflows in the middle layer (500hPa-700hPa) and upper layer (300hPa-500hPa), Beta drift force and ventilation flow, airflow vorticity in the middle and upper layers, vertical wind shear at high and low altitudes, lower atmospheric temperature, lower atmospheric humidity, diagnostic CAPE value, sea surface temperature (SST), average temperature of the upper 200m of the ocean, and LANDMASK of the TC underlying surface, so as to comprehensively consider the key physical processes affecting TC motion and intensity changes.

[0058] Further preferably, in step SS4, the calculation of the Beta drift force and the ventilation flow at least includes:

[0059] SS41. Extract the U and V wind field data for the 500hPa, 600hPa, and 700hPa pressure layers within a 0-500km radius around the TC from the FNL atmospheric reanalysis data. U represents the x-direction wind speed component, and V represents the y-direction wind speed component. The U and V wind fields for these three extracted pressure layers are vertically averaged to obtain the average wind field representing the middle atmosphere.

[0060] SS42. Calculate the TC's moving speed using the vertically averaged U and V wind data using the TC's current position and its position over the previous six hours. Subtract this speed from the averaged U and V wind data to obtain the ambient wind field relative to the TC, eliminating errors caused by the TC's own movement.

[0061] SS43. The wind field within a 500km radius of the TC center is selected as the calculation area and decomposed into tangential wind and radial wind. The tangential wind represents the wind speed component along the TC rotation direction, and the radial wind represents the wind speed component perpendicular to the tangential direction.

[0062] SS44. Perform Fourier decomposition on the radial and tangential winds, extracting the first-order wave components that describe the dominant asymmetric components in the wind field. Then, use an inverse Fourier transform to convert these first-order wave components back into the U and V wind speed components in the Cartesian coordinate system.

[0063] SS45. The inversely calculated U and V wind speed components are averaged over the entire calculation area to obtain the magnitude and direction of the Beta drift force and ventilation flow. The Beta drift force represents the deviation of the TC relative to the ambient airflow, and the ventilation flow represents the influence of the asymmetric secondary airflow.

[0064] Preferably, in step SS4, when determining the final model input factors, the following steps are at least included:

[0065] First, the correlation coefficient between each physical factor and the model output factor is calculated, and the correlation coefficient is tested for significance. Based on the preset significance level (such as 1% or 5%), the correlation between each physical factor is evaluated to see whether it is statistically significant. Physical factors with statistically significant correlation are screened out, and factors that fail the significance level test are excluded.

[0066] Secondly, multiple linear regression was performed on the factors that passed the significance test to evaluate the relative contribution of each physical factor to the changes in TC paths and intensities, in order to further screen out the input factors with higher importance.

[0067] Afterwards, principal component analysis (PCA) was performed on the physical factors that passed the significance test and preliminary regression analysis. The PCA method was used to extract principal components whose cumulative explained variance reached a preset threshold (e.g., 95%) to reduce the data dimension and eliminate the remaining multicollinearity.

[0068] Finally, the results of correlation analysis, significance level test, regression analysis and principal component analysis are comprehensively considered to finally determine the input factor set of the model to ensure the prediction accuracy and stability of the model.

[0069] More specifically, there are many factors that affect TC movement and intensity. First, we need to rely on relevant theories of TC movement and change to select input factors:

[0070] (1) Steering Flow: From the perspective of vortex dynamics, the movement of TC mainly depends on the "steering effect" of large-scale circulation on TC. The large-scale environmental airflow around TC is called steering flow. The movement of TC has a clear relationship with the surrounding large-scale airflow. The average wind at an altitude of 500hPa-700hPa and a latitude radius of 5-7° from the center of TC can be defined as the steering flow of TC, which has the highest correlation with the movement direction of TC; in the Northern Hemisphere, TC always moves 10-20° to the left of the direction of the steering flow. Generally speaking, the low-level steering flow is not consistent with the direction of TC movement, mainly because the low-altitude weather system is greatly affected by the terrain and boundary layer. The influence of the steering flow in the middle and high layers on TC is relatively stable. Therefore, the present invention first calculates the steering flow of TC in the middle layer (500hPa-700hPa) and the steering flow in the high layer (300hPa-500hPa) as the primary input factors of the model.

[0071] (2) Beta drift and ventilation flow: Beta drift also significantly affects the movement of TCs, representing most of the deviation between TC motion and the environmental guiding airflow. Beta drift is mainly caused by beta vortex pairs, which are mainly composed of a wave in the tangential direction of the TC wind field and appear as a pair of counter-rotating vortices, also known as ventilation flow. Beta drift can be regarded as an asymmetric secondary guiding airflow induced by the interaction between cyclones and beta. Its initial distribution is the result of the dispersion of linear Rossby waves causing the symmetric circulation to deform. The maintenance and development of beta vortex pairs are mainly understood from the perspective of energetics. They can be strengthened by obtaining energy from the axisymmetric cyclonic circulation.

[0072] (3) Vorticity: From the perspective of physical equations, a TC can be considered as a vortex. The motion of a vortex obeys the vorticity equation, so vorticity is one of the factors that must be considered. Moreover, when a TC exhibits abnormal path and intensity, it is closely related to the influence of a nearby high-altitude cold vortex. This is why the vorticity factor must be considered. Therefore, this invention also uses the vorticity of the airflow in the middle and upper layers as an input factor.

[0073] (4) Other factors: In addition to the factors mentioned above that have a clear theoretical effect on TC paths, other factors also have a significant theoretical impact on TC intensity. For example, vertical wind shear at high and low altitudes can also cause TCs to spin, significantly affecting the intensity of TCs. Therefore, it is also one of the factors that have a very important impact on TCs. In addition, the temperature of the lower atmosphere, the humidity of the lower atmosphere, and the diagnostic CAPE value can indicate the potential for convective activity in a region. As a strong convective system, TCs are also sensitive to this. The ocean, as the main source of energy for TCs, continuously provides water vapor and heat to TCs to promote the development and movement of TCs. Therefore, SST is very important for the movement and intensity of TCs. At the same time, TCs can also cause significant sea surface cooling, inhibiting the development of TCs. Using the average temperature of the upper ocean can, to a certain extent, indicate the maximum sea surface cooling that TCs can cause. The effect of TCs on the ocean generally reaches about 200 meters below the ocean, so the average temperature of the upper 200 meters of the ocean is introduced. When a TC encounters land, the intensity of the TC will change rapidly due to the huge friction of the land. Therefore, the LANDMASK of the TC underlying surface is also an important factor affecting the TC intensity.

[0074] Based on the above considerations of the physical mechanisms, the following physical variables are directly extracted from atmospheric and oceanic reanalysis data and geographic data: U, V, T, VOR, hum2m, CAPE, and To. The subscript o represents ocean variables; those without the subscript o are assumed to represent atmospheric variables. Through simple calculations, the various physical factors described above can be obtained.

[0075] The guiding airflow is referred to as flow, with the x-component being u_flow and the y-component being v_flow. Vertical wind shear is abbreviated as vws, with the x-component being u_vws and the y-component being v_vws. Sea surface temperature is abbreviated as sst. Because TCs exist in the atmosphere, when calculating atmospheric background variables, variables within 500 km of the TC center are first removed. Only values ​​within 500-1000 km of the TC center are counted to prevent the TC's structural field from being mistakenly included in the background field. Since TCs only affect the ocean but do not reside within it, this step is not performed. Furthermore, the ocean's humidification and heating of TCs directly affect the TC center. Therefore, when calculating ocean variables, the SST values ​​within 500 km of the TC center are counted. Since the landmask's effect on a TC primarily affects the TC center, when calculating landmasks, the landmask within 500 km of the TC center is also counted.

[0076] The calculation diagram is as follows Figure 2 As shown: (A) in the figure is the calculation of atmospheric background field variables. The figure shows an annular area with an inner radius of R=0-5° (about 500km) and an outer radius of R=5-10° (about 500-1000km). During the calculation, only the values ​​within 500-1000km of the TC center are counted (i.e., the blue annular area) to avoid miscounting the structural field of the TC itself into the background field. The annular area is divided into four sectors (1, 2, 3, 4) for calculating the differences in variables in each direction. (B) in the figure is the calculation of ocean variables and geographic variables. The figure shows a circular area with a radius of R=0-5° (about 500km). During the calculation, the sea temperature variable values ​​and landmask within 500km of the TC center (i.e., the blue circular area) are counted. The circular area is also divided into four sectors (1, 2, 3, 4) for calculating the differences in variables in each direction. For both cases, Figure 2 The formulas for calculating variables are given in the table. Var-mean represents the average value of the four sectors, Var-xdiff calculates the difference in the x direction (east-west direction), and Var-ydiff calculates the difference in the y direction (north-south direction).

[0077] To calculate the mean field of a variable, the average value of the relevant variables within the statistical range of the TC center is used. This is hereinafter referred to as "mean" followed by the variable name. To calculate the gradient field of a variable, the right-side average minus the left-side average, or the upper-side average minus the lower-side average, of the relevant variables within the statistical range of the TC center is used. This is hereinafter referred to as "diff" followed by the variable name. Subscript x is used in the latitude direction, and subscript y is used in the longitude direction. Different vertical atmospheric layers also have different impacts on a TC. The subscript high represents the average value of 300-500hPa, mid represents the average value of 500-700hPa, and low represents the average value of 700-850hPa. Vertical wind shear is the difference between 200hPa and 850hPa.

[0078] The TC's current latitude is also considered an important deterministic internal force, so the current latitude and longitude are directly included as one of the model input variables in the input factor considerations. Furthermore, considering the existence of TC inertia, the TC's position and intensity change value (pre) and change trend value (trend) for the previous 6 hours are added. The calculation formula is as follows:

[0079] X_pre=X(t)-X(t-6h)

[0080] X_trend=[X(t)-X(t-6h)]-[X(t-6h)-X(t-12h)]

[0081] When calculating Xpre and Xtrend for MWS and SLP, the SLP value is used uniformly. This is because in the TCBest Track data, the quantitative value of MWS appears much later than the SLP value, resulting in sudden changes from 0 to high values, and vice versa, which is unrealistic. However, when the maximum wind speed is 0 m / s, the SLP value is already given, so it is used uniformly in the calculation.

[0082] All the variables and calculation methods mentioned above are used to calculate 35 influencing factors. The influencing factors are shown in Table 1 Input Factor. Note that the input factors in Table 1 are not arranged by correlation, but rather factors with similar physical relationships are arranged together to facilitate the use of subsequent AI models. Factors 1-6 represent upper atmospheric information, Factors 7-12 represent middle atmospheric information, Factors 13-15 represent lower atmospheric information, Factors 16-20 represent vertical or horizontal differences in the atmosphere, Factors 21-26 represent sea-air interface information, and Factors 27-35 represent the current and past 6-hour position and intensity information of the TC. Table 2 shows the meaning of each variable name and abbreviation. Table 3 shows the linear correlation coefficient between the model input factors and output factors.

[0083] Based on their physical relationships, these factors are divided into several categories: time series factors (Factors 27-35, hereafter referred to as "Time Only"); guiding airflow and secondary guiding airflow factors (Factors (1-3) & (7-12), hereafter referred to as "Flow"); eddy covariance factors (Factors (4-6) & (13-15), hereafter referred to as "Vor"); and the remaining factors. The use of all factors is referred to as "All Vars."

[0084] Table 1 Model input factors and output factors

[0085]

[0086] Table 2 Variable names and abbreviations

[0087]

[0088] Table 3 Linear correlation coefficients between model input factors and output factors

[0089]

[0090] **Indicates that the test passed the 1% significance level

[0091] *Indicates that the 5% significance level test was passed

[0092] No * indicates that the result did not pass the 5% significance level test.

[0093] Based on the model output and input factors determined in steps SS3 and SS4, select at least one AI model from among convolutional neural networks (CNN), recurrent neural networks (RNN), long short-term memory networks (LSTM), random forests (RF), and extreme gradient boosting (XGBoost) to construct a TC track and intensity prediction model. This model is then trained using the training data from step SS2 to capture the complex nonlinear relationships between meteorological and oceanographic factors.

[0094] Preferably, in the above step SS5, for the CNN model, a combination of multi-layer convolutional layers and pooling layers is adopted, each convolution layer is followed by batch normalization and ReLU activation function, and finally the prediction result is output through the fully connected layer; for the RNN and LSTM models, a bidirectional network structure suitable for time series data is designed to capture the front-end and back-end dependencies of the time series; for the RF and XGBoost models, the complexity and generalization ability of the model are balanced by optimizing the hyperparameters including at least the depth and number of trees.

[0095] Use the validation data from step SS2 to validate the TC track and intensity prediction model trained in step SS5. Evaluate the model's performance at different forecast timeframes (e.g., 6-hour, 12-hour, and 24-hour). The evaluation metrics include at least the track forecast error and the intensity forecast error. Based on the validation results, optimize the model by adjusting hyperparameters, adding regularization, and / or using cross-validation to improve the model's prediction accuracy and generalization ability.

[0096] Preferably, in steps SS5 and SS6 above, the prediction model training and optimization strategy includes: using batch normalization technology to improve the convergence speed and stability of the model; using dropout technology to prevent overfitting; applying a learning rate decay strategy to achieve more refined parameter adjustment in the later stage of training; and determining the optimal hyperparameter combination through k-fold cross validation.

[0097] More specifically, in steps SS5 and SS6 above, the present invention constructs four linear regression models (Models 1-4) to assess the relative contributions of different meteorological and physical factors to changes in tropical cyclone paths and intensities. These linear regression models capture the linear relationship between input factors and prediction targets, quantifying the impact of each input factor on path and intensity predictions, thereby providing a benchmark for input factor selection in AI models. Furthermore, Models 1-4, as baseline performance models, can quickly provide linear prediction results for path and intensity. These results provide an important reference for subsequent AI model performance evaluation, particularly when capturing complex nonlinear relationships. The improved performance of AI models over linear regression models can be clearly demonstrated through performance comparisons.

[0098] Models 1-4 are constructed using different combinations of input factors: Model 1 uses only the time series factor Time Only to evaluate the temporal evolution characteristics of path and intensity changes; Model 2 uses Time+Flow, and adds a guided airflow factor on the basis of Model 1 to evaluate the impact of atmospheric guided airflow on the TC path; Model 3 uses Time+Flow+Vor, and adds a vorticity factor on the basis of Model 2 to analyze the role of cyclone vorticity on path and intensity changes; Model 4 uses all factors All Vars to comprehensively analyze the joint influence of different physical processes on the TC path and intensity.

[0099] To further improve forecast accuracy, this paper also constructed six artificial intelligence models (Models 5-10). These models capture the complex nonlinear relationships between meteorological factors and are particularly suitable for scenarios where TC paths and intensities fluctuate dramatically in the short term. The specific models and their functions are as follows:

[0100] Model 5: Multi-Layer Perceptron (MLP)

[0101] The multilayer perceptron (MLP) is a classic feedforward neural network model consisting of an input layer, one or more hidden layers, and an output layer. It can learn complex nonlinear mapping relationships and is applicable to a wide range of regression and classification tasks. The core of the MLP model lies in its hidden layers. By using nonlinear activation functions (such as ReLU), the model can capture the nonlinear characteristics of the input data. The MLP model used in this invention contains multiple layers of fully connected layers (nn.Linear), each followed by a ReLU activation function (nn.ReLU). This structure helps the model capture the nonlinear characteristics of the input data. The model architecture includes a series of fully connected layers, with the number of neurons in each layer increasing and then decreasing, and finally outputting the required dimensions. Through this structure, the model can fit complex data distributions to a certain extent.

[0102] Model 6: Convolutional Neural Network (CNN)

[0103] A convolutional neural network (CNN) is a deep learning model specifically designed to process grid-structured data, such as images and sequence data. It has demonstrated remarkable capabilities in fields such as computer vision and signal processing. Key components of a CNN model include convolutional layers, activation functions, pooling layers, and fully connected layers. The CNN model implemented in this paper uses a one-dimensional convolutional layer (nn.Conv1d), suitable for processing one-dimensional sequence data. The model consists of three convolutional layers, each followed by a ReLU activation function (nn.ReLU) and a maximum pooling layer (nn.MaxPool1d). Convolutional layers capture local features, while pooling layers reduce the spatial dimensionality of features, improving computational efficiency. Finally, the model flattens the convolutional features and passes them to a fully connected layer (nn.Linear) for final prediction.

[0104] Model 7: Improved Recurrent Neural Network Model (RNN)

[0105] The improved recurrent neural network model uses long short-term memory (LSTM) to process sequence data. LSTM effectively captures long-term dependencies and is particularly useful for time series analysis tasks. The model in this paper uses a multi-layer bidirectional LSTM (nn.LSTM) to process the input sequence in both forward and reverse directions, thereby better utilizing contextual information. In addition, the model includes two fully connected layers (nn.Linear) to further process the LSTM output. During the forward propagation process, the model takes the output of the last time step of the sequence and passes it through the fully connected layers to make the final prediction.

[0106] Model 8: Improved bidirectional LSTM model

[0107] The improved bidirectional LSTM model also uses long short-term memory units, but processes the input sequence in both the forward and reverse directions. Bidirectional LSTM can capture information from both the beginning and end of the sequence, which is crucial for many natural language processing and time series forecasting tasks. The model employs a multi-layer bidirectional LSTM (nn.LSTM) and concatenates the forward and reverse outputs through two fully connected layers for final prediction. This structure allows the model to fully exploit contextual information in the sequence data.

[0108] Model 9: Extreme Gradient Boosting Regressor (XGBoost Regressor)

[0109] Extreme Gradient Boosting (XGBoost) is an efficient gradient boosting framework designed to quickly and accurately solve regression and classification problems. XGBoost gradually reduces the prediction error of training data by building multiple weak learners (typically decision trees). The XGBoost regressor used in this paper builds a model by setting specific parameters. XGBoost is suitable for processing large datasets, can handle missing values, and has good generalization capabilities. XGBoost supports parallel computing and can effectively utilize multi-core processors to accelerate the training process.

[0110] Model 10: Random Forest Regressor (RF)

[0111] Random forest is an ensemble learning method that improves prediction performance by building multiple decision trees and averaging their results. This reduces the risk of overfitting and achieves good results without requiring excessive parameter tuning. The random forest regressor used in this paper builds the model by setting the number of trees n_estimators = 500 and the random seed random_state = 42. Random forests are well suited for processing datasets with high-dimensional feature spaces and can handle missing values ​​and non-numeric data.

[0112] Figure 3 The following diagrams compare the forecast errors of different models for all TC tracks and intensities in 2021-2022, with (a) showing the 6-24 hour forecast error for TC tracks and (b) showing the 6-24 hour forecast error for TC intensity. Performance comparisons show that AI models demonstrate greater adaptability when handling complex nonlinear relationships. Model 10 (Random Forest Regressor) particularly excels when handling multiple combinations of physical factors, effectively reducing track and intensity forecast errors. In contrast, while CNN and MLP models can extract complex spatial and nonlinear features, they perform slightly worse than the Random Forest model in certain complex scenarios. Furthermore, Model 8 (Improved Bidirectional LSTM Model) performs particularly well in forecasting tracks and intensities over long timescales (e.g., 24 hours), making it particularly well-suited for scenarios with drastic TC intensity fluctuations.

[0113] Figure 4 and Figure 5 The results of the various models for the path and intensity forecast of TC No. 10 in 2021 are shown. It can be seen that the regression model has a significantly better effect after adding guiding airflow and ventilation flow to its path. Its intensity forecast results have also been significantly improved after adding factors related to intensity. This result is consistent with the Figure 3 The results are consistent with those in

[15] . For several AI models, there are some models with obvious deviations, such as CNN and MLP. For other models, the results are not intuitively different.

[0114] Figure 6 and Figure 7This figure shows the track and intensity forecasts for all TCs from August to October 2021, using the best-performing AI model, RF. As can be seen, this model performed well for most TCs, such as TCs 202109, 202110, 202112, 202116, and 202120. Its predicted tracks and intensities were consistent with the actual conditions. It also accurately predicted most TC track turns, with only a few significant errors, such as those for TCs 202111, 202114, and 202119. Some individual forecasts, such as TCs 202117 and 202118, showed significant deviations.

[0115] Establish a real-time feedback mechanism to dynamically input new pre-processed observational meteorological data into the model optimized in step SS6. Continuously adjust and optimize model parameters through online learning and / or regular retraining triggered in real time by new data changes. Regularly evaluate model performance and update the input factor set and model structure to improve the timeliness and accuracy of TC track and intensity forecasts.

[0116] Preferably, in the above step SS7, the implementation method of the real-time feedback mechanism includes: updating the real-time forecast at intervals of 6 hours, setting a sliding time window (such as the last 7 days) for online learning of the model, regularly (such as monthly) using accumulated new data to retrain the model to update the model parameters, and dynamically adjusting the model's prediction weight by comparing the deviation between the model prediction results and the actual observation results to maintain the model's adaptability to the latest meteorological patterns.

[0117] Generate and output forecast results of TC path changes (longitude and latitude) and intensity changes (minimum air pressure and maximum wind speed) in the next 6 hours, 12 hours and 24 hours, calculate and output forecast errors, including but not limited to path forecast error and intensity forecast error, provide credibility assessment of forecast results, and provide decision support information for meteorological departments and disaster prevention and mitigation agencies.

[0118] Preferably, in the above steps SS7 and SS8, when the prediction error exceeds the preset threshold, the model retraining process is triggered; it also includes an automatic identification and manual review mechanism for abnormal prediction results, and makes necessary adjustments to the model output in combination with the experience of meteorological experts, and generates forecast products in a standard format to ensure the reliability and practicality of the prediction results.

[0119] The above embodiments fully and effectively achieve the objectives of the present invention. Those skilled in the art will appreciate that the present invention includes, but is not limited to, the contents described in the accompanying drawings and the above specific embodiments. Although the present invention has been described with reference to the embodiments currently considered to be the most practical and preferred, it should be understood that the present invention is not limited to the disclosed embodiments, and any modifications that do not deviate from the functional and structural principles of the present invention are intended to be included within the scope of the claims.

Claims

1. A TC path and intensity prediction method based on AI and physical properties, characterized by: The prediction method comprises at least the following steps when implemented: SS1. Acquire at least five consecutive years of historical meteorological data from multiple meteorological and oceanographic data sources, including at least TC Best Track data, FNL atmospheric reanalysis data, and GODAS ocean reanalysis data; SS2. Preprocess the collected historical meteorological and oceanographic data and divide the preprocessed dataset into a training set and a validation set. SS3. Based on preprocessed historical TC Best Track data and the characteristics of the target forecast area, multiple variables closely related to TC track and intensity changes are selected as model output factors. These model output factors include at least TC position and intensity changes. Time series of these variables are calculated to extract characteristic changes and establish model output targets. SS4. Based on the physical mechanisms of TC motion and intensity changes, and in combination with preprocessed FNL atmospheric reanalysis data and GODAS ocean reanalysis data, select and calculate multiple physical factors related to TC motion and intensity changes. These factors should include at least mid- and upper-level guiding airflows, Beta drift and ventilation flows, mid- and upper-level airflow vorticity, vertical wind shear at high and low altitudes, lower-atmospheric temperature, lower-atmospheric humidity, diagnostic CAPE values, sea surface temperature (SST), average temperature in the upper 200 m of the ocean, and the LANDMASK of the TC underlying surface. Correlations between these selected physical factors and TC track and intensity changes should be analyzed. The correlation coefficients between each physical factor and the model output factors should be calculated one by one. The final model input factor set should be determined based on the analysis results. The calculation of Beta drift and ventilation flows should include at least the following substeps: The U and V wind field data of the 500hPa, 600hPa, and 700hPa pressure layers within the range of 0-500km around the TC were extracted from the FNL atmospheric reanalysis data, where U is the wind speed component in the x-direction and V is the wind speed component in the y-direction. The U and V wind fields of the three extracted pressure layers were vertically averaged to obtain the average wind field representing the middle atmosphere. After obtaining the vertically averaged U and V wind field data, the TC's moving speed is calculated using the TC's current position and its position in the previous 6 hours. This speed is then subtracted from the averaged U and V wind fields to obtain the ambient wind field relative to the TC. The wind field within a 500km radius of the TC center is selected as the calculation area and decomposed into tangential wind and radial wind. The tangential wind represents the wind speed component along the TC rotation direction, and the radial wind represents the wind speed component perpendicular to the tangential direction. Fourier decomposition is performed on the radial wind and tangential wind respectively to extract the first-order fluctuation component to describe the most important asymmetric component in the wind field. The first-order fluctuation component obtained after decomposition is then converted back into the U and V wind speed components in the Cartesian coordinate system through inverse Fourier transform. The back-calculated U and V wind speed components are averaged over the entire calculation area to obtain the magnitude and direction of the Beta drift force and ventilation flow. The Beta drift force represents the deviation of the TC relative to the atmospheric environment guiding airflow, and the ventilation flow represents the influence of the asymmetric secondary airflow. SS5. Based on the determined model output factors and model input factors, select at least one AI model to construct a TC path and intensity prediction model, and train the constructed TC path and intensity prediction model using the training set data. SS6. Use validation data to validate the trained TC path and intensity prediction model and optimize the model based on the validation results. SS7. Establish a real-time feedback mechanism to dynamically input newly observed meteorological data into the optimized model after preprocessing. This mechanism uses real-time triggered regular retraining to continuously adjust and optimize model parameters. Furthermore, it regularly evaluates model performance and updates the input factor set and model structure. SS8. Generate and output predictions of TC track and intensity changes over the next 6, 12, and 24 hours, calculate and output the prediction errors, and provide a credibility assessment of the prediction results.

2. The TC path and intensity prediction method based on AI and physical properties according to claim 1, characterized in that: In steps SS1 and SS2, information on the occurrence period of TCs in the northwestern Pacific Ocean between January 1, 2015, and December 31, 2022, was obtained from multiple meteorological and oceanographic data sources. For each TC case, the atmospheric and oceanic background field information in the reports was statistically analyzed every 6 hours, and missing data records were removed. The data from 2015 to 2020 were used for model training, and the data from 2021 to 2022 were used for model verification.

3. The TC path and intensity prediction method based on AI and physical properties according to claim 1, characterized in that: In step SS2, based on the collected historical meteorological and oceanographic data, relevant data for each TC occurrence period in the target prediction area are extracted and subjected to data quality control, including identification and elimination of missing or abnormal data, standardization and normalization of the data to ensure temporal and spatial consistency between different data sources, and when performing data quality control, mean filling, interpolation, or deletion of records with missing values ​​are used to handle missing parts in the data, outliers in the data are identified and removed through statistical analysis methods such as Z-score or box plots, Z-score standardization or Min-Max normalization is used to process data of different scales, and the collected historical meteorological data are time synchronized to ensure consistency in the temporal resolution of all data sources.

4. The TC path and intensity prediction method based on AI and physical properties according to claim 3, characterized in that: In step SS2, data quality control also uses an automated data cleaning method based on machine learning to identify potential abnormal data through historical data learning and automatically perform data correction or filling to improve the efficiency and accuracy of data preprocessing.

5. The TC path and intensity prediction method based on AI and physical properties according to claim 1, characterized in that: In step SS3, each output factor is calculated using X_n=X(t+n)-X(t) to calculate the hourly position and intensity changes of the TC, where X represents the longitude LON, latitude LAT, central minimum pressure SLP, and central maximum wind speed MWS, t represents the current time, and n is 6 hours, 12 hours, and 24 hours, respectively, forming a total of 12 model output factors.

6. The TC path and intensity prediction method based on AI and physical properties according to claim 1, characterized in that: In step SS4, when determining the final model input factors, at least the following steps are involved: First, the correlation coefficient between each physical factor and the model output factor is calculated, and the correlation coefficient is tested for significance level. Based on the preset significance level, the correlation between each physical factor is evaluated to see whether it is statistically significant. Physical factors with statistically significant correlation are screened out, and factors that fail the significance level test are excluded. Secondly, multiple linear regression was performed on the factors that passed the significance test to evaluate the relative contribution of each physical factor to the changes in TC paths and intensities, in order to further screen out the input factors with higher importance. Afterwards, principal component analysis was performed on the physical factors that passed the significance test and preliminary regression analysis. The principal components whose cumulative explained variance reached the preset threshold were extracted using the PCA method to reduce the data dimension and eliminate the remaining multicollinearity. Finally, the results of correlation analysis, significance level test, regression analysis and principal component analysis are comprehensively considered to finally determine the input factor set of the model to ensure the prediction accuracy and stability of the model.

7. The TC path and intensity prediction method based on AI and physical properties according to claim 1, characterized in that: In step SS5, when the AI ​​model selects the CNN model, a combination of multi-layer convolutional layers and pooling layers is used. Each convolution layer is followed by batch normalization and ReLU activation function, and finally the prediction result is output through the fully connected layer; when the AI ​​model selects the RNN or LSTM model, a bidirectional network structure suitable for time series data is designed to capture the front-end and back-end dependencies of the time series; when the AI ​​model selects the RF or XGBoost model, the complexity and generalization ability of the model are balanced by optimizing hyperparameters including the depth and number of trees.

8. The TC path and intensity prediction method based on AI and physical properties according to claim 1, characterized in that: In steps SS5 and SS6, the prediction model training and optimization strategies include: using batch normalization technology to improve the convergence speed and stability of the model; using dropout technology to prevent overfitting; applying a learning rate decay strategy to achieve more refined parameter adjustments in the later stages of training; and determining the optimal hyperparameter combination through k-fold cross-validation.

9. The TC path and intensity prediction method based on AI and physical properties according to claim 1, characterized in that: In step SS7, the implementation method of the real-time feedback mechanism includes: updating the real-time forecast at intervals of 6 hours, setting a sliding time window for online learning of the model, regularly using accumulated new data to retrain the model to update the model parameters, and dynamically adjusting the model's prediction weight by comparing the deviation between the model prediction results and the actual observation results to maintain the model's adaptability to the latest meteorological patterns.

10. The TC path and intensity prediction method based on AI and physical properties according to claim 1, characterized in that: In steps SS7 and SS8, when the prediction error exceeds the preset threshold, the model retraining process is triggered; it also includes an automatic identification and manual review mechanism for abnormal prediction results, and combines the experience of meteorological experts to make necessary adjustments to the model output and generate forecast products in a standard format to ensure the reliability and practicality of the prediction results.

11. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the TC path and intensity prediction method based on AI and physical properties are implemented as described in any one of claims 1 to 10.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the TC path and intensity prediction method based on AI and physical properties according to any one of claims 1 to 10 are implemented.

Images