Typhoon full-path simulation method based on local reinforcement of target area

Abstract

The invention discloses a typhoon full-path simulation method based on target area local reinforcement, and aims to solve the problem that an existing full-path model focuses on a large-scale area and neglects the modeling precision of an infrastructure system with small-scale space distribution. The typhoon full-path simulation method comprises the following steps: 1, collecting historical typhoon trajectory data; 2, determining a target center point and a target area; 3, constructing a typhoon full-path model; 4, calculating the distribution weight of the typhoon sample; 5, re-training the typhoon intensity model by using the weighted typhoon samples, and determining an objective function through typhoon intensity predicted values correspondingly output by each group of typhoon samples; 6, optimizing the objective function by using a genetic algorithm, and determining parameters in the distribution weight; 7, retraining the typhoon intensity model XGBP; and 8, carrying out typhoon full-path simulation. According to the invention, typhoon full-path simulation based on local reinforcement of the target area is realized, and higher precision can be achieved in the target area where the infrastructure system is located.

Description

Technical Field

[0001] The present invention belongs to the field of typhoon disaster simulation for disaster prevention and mitigation, and specifically relates to a typhoon full-path simulation method based on local strengthening of the target area to improve the accuracy of typhoon simulation. Background Art

[0002] The landing of a typhoon will cause serious social, economic and infrastructure damage to coastal areas, especially the loss of infrastructure functions in spatial distribution, which may bring huge adverse effects. Typhoon disaster modeling is an important part of risk analysis and assessment based on numerical simulation, and its accuracy is crucial for evaluating disaster risks and enhancing the resilience of coastal areas.

[0003] Existing typhoon simulation models include local path models and full-path models, but both of these methods have limitations when applied to infrastructure systems with spatial distribution. The local path model is suitable for simulating typhoons of a single structure, but not for widely distributed infrastructure systems; while the full-path model has advantages in the risk assessment of infrastructure systems with large-area distribution, but needs to be improved in terms of modeling accuracy. Summary of the Invention

[0004] The purpose of the present invention is to solve the problem that the existing full-path model focuses on large-scale regions and ignores the modeling accuracy of small-scale spatially distributed infrastructure systems, and to provide a typhoon full-path simulation method based on local strengthening of the target area to improve the accuracy of the typhoon model in the infrastructure system distribution area for more accurate risk analysis.

[0005] The typhoon full-path simulation method based on local strengthening of the target area of the present invention is implemented according to the following steps:

[0006] Step 1: Collect historical typhoon track data, where the typhoon track data includes information such as the longitude, latitude, central pressure and sea surface temperature of the typhoon;

[0007] Step 2: Obtain the spatial location of the target infrastructure system through the Geographic Information System (GIS), extract the central point location coordinates of the infrastructure system as the target central point, and use the area with a radius of 250 km centered on the target central point as the target area;

[0008] Step 3: Build a typhoon full-path model based on the historical typhoon track data in Step 1. The typhoon full-path model includes a starting point model, a typhoon path model, a typhoon intensity model and a typhoon decay model;

[0009] The process of building the typhoon full-path model is as follows:

[0010] Step 3.1: Build a starting point model;

[0011] Step 3.2: Based on the starting point, construct a typhoon path model through the XGBoost model. The typhoon path model consists of the typhoon path model XGB for longitude features λ and the typhoon path model XGB for latitude features ψ to predict the latitude ψ (°) and longitude λ (°) of the typhoon center at the (i + 1)-th moment. The typhoon path model for longitude features is shown in Equation (2), and the typhoon path model for latitude features is shown in Equation (3);

[0012] λ i+1 = XGB λ (ψ i-3 , λ i-3 , ψ i-2 , λ i-2 , ψ i-1 , λ i-1 , ψ i , λ i , t i+1 ) + ε λ (2)

[0013] ψ i+1 = XGB ψ (ψ i-3 , λ i-3 , ψ i-2 , λ i-2 , ψ i-1 , λ i-1 , ψ i , λ i , t i+1 ) + ε ψ (3)

[0014] where ψ i-3 , ψ i-2 , ψ i-1 , ψ i and ψ i+1 represent the latitudes of the typhoon positions at the (i - 3)-th, (i - 2)-th, (i - 1)-th, i-th, and (i + 1)-th moments respectively, λ i-3 , λ i-2 , λ i-1 , λ i and λ i+1 represent the longitudes of the typhoon positions at the (i - 3)-th, (i - 2)-th, (i - 1)-th, i-th, and (i + 1)-th moments respectively, t i+1 represents the total movement time of the typhoon at the (i + 1)-th moment, in hours; ε λ and ε ψ represent the longitude model error term and the latitude model error term respectively;

[0015] Step 3.3: Construct a typhoon intensity (central pressure) model XGB through the XGBoost model P, the typhoon intensity is the central pressure of the typhoon, and the typhoon intensity model is shown in formula (6);

[0016] P i+1 = XGB P (ψ i-2 , λ i-2 , SST i-2 , areapass i-2 , ψ i-1 , λ i-1 , SST i-1 , areapass i-1 , ψ i , λ i , SST i , areapass i , ψ i+1 , λ i+1 , SST i+1 , t i+1 , areapass i+1 ) + ε P (6)

[0017] Where P i+1 represents the predicted typhoon intensity value at the (i + 1)-th moment; ψ i-2 , ψ i-1 , ψ i , ψ i+1 represent the latitudes of the typhoon positions at the (i - 2)-th, (i - 1)-th, i-th, and (i + 1)-th moments of the typhoon sample respectively; λ i-2 , λ i-1 , λ i , λ i+1 represent the longitudes of the typhoon positions at the (i - 2)-th, (i - 1)-th, i-th, and (i + 1)-th moments of the typhoon sample respectively; SST i-2 , SST i-1 , SST i , SST i+1 represent the sea surface temperatures at the (i - 2)-th, (i - 1)-th, i-th, and (i + 1)-th moments of the typhoon sample respectively; t i+1 represents the typhoon migration time (hours) at the (i + 1)-th time point; ε P represents the intensity model error term; areapass i-2 , areapass i-1 , areapass i , areapass i+1 is a binary variable. If the typhoon sample path passes through the target area described in step 2, the binary variable value is 1; if the typhoon sample path does not pass through the target area described in step 2, the binary variable value is 0;

[0018] Step 3.4. Construct a typhoon decay model;

[0019] Step 4: Calculate the typhoon intensity model XGB P The distance d between the central coordinates (ψ i , λ i )(as shown in formula (6)) of the j-th typhoon sample at the i-th time and the coordinates of the target center point j,s , and calculate the allocation weight ω of the typhoon sample according to formula (7) j,s to enhance the local accuracy;

[0020]

[0021] where a and b are optimization variables (to be solved);

[0022] Step 5: Weight each group of typhoon samples used to train the typhoon intensity model XGB P using formula (7), and use the weighted typhoon samples to retrain the typhoon intensity model XGB P to obtain the typhoon intensity prediction value corresponding to the output of each group of typhoon samples. Compare the typhoon intensity prediction value with the true typhoon intensity value to determine the objective function F(X) in formula (8):

[0023]

[0024] where N represents the number of historical typhoons, X represents the parameter group (a, b), d i , s represents the distance from the target center point to the nearest point on the i-th historical typhoon track, and P i,predict (X) is the predicted intensity of the typhoon point closest to the target center point in the i-th typhoon sample output by the retrained typhoon intensity model XGB P , and P i,actual represents the actual intensity of the typhoon point closest to the target center point in the i-th typhoon sample;

[0025] Step 6: Use the genetic algorithm to optimize the objective function F(X) to determine the optimal model parameters a and b;

[0026] Step 7: Obtain the allocation weight of each historical typhoon sample through the optimal model parameters a and b, and retrain the typhoon intensity model XGB according to the allocation weight of each historical typhoon sample P to obtain an improved typhoon intensity model;

[0027] Step 8: Simulate the complete typhoon full path through the starting point model, typhoon path model, improved typhoon intensity model, and typhoon attenuation model to complete the typhoon full path simulation method based on local enhancement in the target area.

[0028] The present invention realizes the simulation of the full typhoon path based on local enhancement of the target area, and can achieve higher accuracy in the target area where the infrastructure system is located, that is, the intensity parameters of a large number of simulated typhoons are more consistent with the intensity parameters of historical typhoons in the target area, which is of great significance for improving the typhoon disaster modeling accuracy of the target infrastructure system.

[0029] The present invention integrates the respective advantages of the full-path model and the local-path model. It can not only artificially simulate and generate a large number of typhoon samples in a wide area, but also achieve higher accuracy in the target infrastructure area, which is of great significance for the risk assessment of infrastructure under typhoon disasters. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 It is a schematic diagram of the key steps for constructing the full typhoon path model with local enhancement in the target area of the present invention;

[0031] Figure 2 It is a schematic diagram of the distribution of the infrastructure system in the embodiment;

[0032] Figure 3 It is a comparison chart of the central pressure of typhoons simulated between the full-path model with local enhancement in the target area and the unimproved full-path model at 47 coastal stations in the embodiment;

[0033] Figure 4 It is a schematic diagram of the distribution comparison of the intensity parameters obtained from the full-path model with local enhancement in the target area and the unimproved full-path model at the central position of the target infrastructure system in the embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0034] DETAILED DESCRIPTION OF THE INVENTION 1: The typhoon full-path simulation method based on local enhancement in the target area of the present embodiment is realized according to the following steps:

[0035] Step 1: Collect historical typhoon track data, where the typhoon track data includes the longitude, latitude, central pressure, and sea surface temperature information of the typhoon;

[0036] Step 2: Obtain the spatial location of the target infrastructure system through the Geographic Information System (GIS), extract the center point location coordinates of the infrastructure system as the target center point, and use the area with a radius of 250 km centered on the target center point as the target area;

[0037] Step 3: Construct a typhoon full-path model based on the historical typhoon track data in Step 1. The typhoon full-path model includes a starting point model, a typhoon path model, a typhoon intensity model, and a typhoon attenuation model;

[0038] The process of constructing the typhoon full-path model is as follows:

[0039] Step 3.1: Construct the starting point model;

[0040] Step 3.2: Based on the starting point, construct the typhoon path model through the XGBoost model. The typhoon path model consists of the typhoon path model XGB of longitude features λ and the typhoon path model XGB of latitude features ψ to predict the latitude ψ (°) and longitude λ (°) of the typhoon center at the (i + 1)-th moment. The typhoon path model of longitude features is shown in Formula (2), and the typhoon path model of latitude features is shown in Formula (3);

[0041] λ i+1 = XGB λ (ψ i-3 , λ i-3 , ψ i-2 , λ i-2 , ψ i-1 , λ i-1 , ψ i , λ i , t i+1 ) + ε λ (2)

[0042] ψ i+1 = XGB ψ (ψ i-3 , λ i-3 , ψ i-2 , λ i-2 , ψ i-1 , λ i-1 , ψ i , λ i , t i+1 ) + ε ψ (3)

[0043] where ψ i-3 , ψ i-2 , ψ i-1 , ψ i and ψ i+1 represent the latitudes of the typhoon positions at the (i - 3)-th, (i - 2)-th, (i - 1)-th, i-th, and (i + 1)-th moments respectively, and λ i-3 , λ i-2 , λ i-1 , λ i and λ i+1 represent the longitudes of the typhoon positions at the (i - 3)-th, (i - 2)-th, (i - 1)-th, i-th, and (i + 1)-th moments respectively. t i+1 represents the total movement time of the typhoon at the (i + 1)-th moment, in hours; ε λ and ε ψ represent the longitude model error term and the latitude model error term respectively;

[0044] Step 3.3: Construct the typhoon intensity (central pressure) model XGB using the XGBoost model P The typhoon intensity is the central pressure of the typhoon. The typhoon intensity model is shown in Equation (6);

[0045] P i+1 = XGB P (ψ i-2 , λ i-2 , SST i-2 , areapass i-2 , ψ i-1 , λ i-1 , SST i-1 , areapass i-1 , ψ i , λ i , SST i , areapass i , ψ i+1 , λ i+1 , SST i+1 , t i+1 , areapass i+1 ) + ε P (6)

[0046] where P i+1 represents the predicted typhoon intensity value at the (i + 1)-th moment; ψ i-2 , ψ i-1 , ψ i , ψ i+1 represent the latitudes of the typhoon positions at the (i - 2)-th, (i - 1)-th, i-th, and (i + 1)-th moments of the typhoon sample, respectively; λ i-2 , λ i-1 , λ i , λ i+1 represent the longitudes of the typhoon positions at the (i - 2)-th, (i - 1)-th, i-th, and (i + 1)-th moments of the typhoon sample, respectively; SST i-2 , SST i-1 , SST i , SST i+1 represent the sea surface temperatures at the (i - 2)-th, (i - 1)-th, i-th, and (i + 1)-th moments of the typhoon sample, respectively; t i+1 represents the typhoon migration time (hours) at the (i + 1)-th time point; ε P represents the intensity model error term; areapass i-2 , areapass i-1 , areapass i , areapass i+1is a binary variable. If the path of the typhoon sample passes through the target area described in Step 2, the value of the binary variable is 1; if the path of the typhoon sample does not pass through the target area described in Step 2, the value of the binary variable is 0.

[0047] Step 3.4: Construct a typhoon attenuation model;

[0048] Step Four: Calculate the typhoon intensity model XGB P The distance d between the central coordinates (ψ i , λ i )(as shown in Equation (6)) of the j-th group of typhoon samples at the i-th time and the coordinates of the target center point j,s , and calculate the allocation weight ω of the typhoon sample according to Equation (7) j,s to enhance the local accuracy;

[0049]

[0050] where a and b are optimization variables (to be solved);

[0051] Step Five: Based on Equation (7), weight each group of typhoon samples used to train the typhoon intensity model XGB P , and use the weighted typhoon samples to retrain the typhoon intensity model XGB P , so as to obtain the typhoon intensity prediction value corresponding to the output of each group of typhoon samples. Compare the typhoon intensity prediction value with the true typhoon intensity value to determine the objective function F(X) in Equation (8):

[0052]

[0053] where N represents the number of historical typhoons, X represents the parameter group (a, b), d i , s represents the distance from the target center point to the nearest point on the i-th historical typhoon track, and P i,predict (X) is the predicted intensity of the typhoon point closest to the target center point in the i-th typhoon sample output by the retrained typhoon intensity model XGB P , and P i,actual represents the actual intensity of the typhoon point closest to the target center point in the i-th typhoon sample;

[0054] Step Six: Use the genetic algorithm to optimize the objective function F(X) to determine the optimal model parameters a and b;

[0055] Step Seven: Through the optimal model parameters a and b, obtain the allocation weight of each historical typhoon sample, and retrain the typhoon intensity model XGB P based on the allocation weight of each historical typhoon sample to obtain an improved typhoon intensity model;

[0056] Step 8: Simulate the complete typhoon full path through the starting point model, typhoon path model, improved typhoon intensity model, and typhoon decay model, thereby completing the typhoon full path simulation method based on local strengthening in the target area.

[0057] Specific Embodiment 2: The difference between this embodiment and Specific Embodiment 1 is that in Step 1, historical typhoon track data is collected from the CMA typhoon best track database.

[0058] Specific Embodiment 3: The difference between this embodiment and Specific Embodiment 1 or 2 is that in Step 1, the number of historical typhoon track data is greater than 2000.

[0059] Specific Embodiment 4: The difference between this embodiment and any one of Specific Embodiments 1 to 3 is that in Step 1, typhoon data with the starting point on land is removed from the historical typhoon track data.

[0060] Specific Embodiment 5: The difference between this embodiment and any one of Specific Embodiments 1 to 4 is that the process of constructing the starting point model in Step 3.1 is as follows:

[0061] According to the historical typhoon track data collected in Step 1, obtain the starting point data of each typhoon track data, and then perform typhoon starting point modeling by sampling from the starting point data to obtain the starting point model.

[0062] Specific Embodiment 6: The difference between this embodiment and Specific Embodiment 5 is that the starting point data includes the longitude, latitude, time, and central pressure of the starting point.

[0063] Specific Embodiment 7: The difference between this embodiment and any one of Specific Embodiments 1 to 6 is that in Step 3.4, the process of constructing the typhoon decay model XGB end is as follows:

[0064] Construct the typhoon decay model XGB through the XGBoost model end , and the typhoon decay model is shown in formula (5);

[0065] end i+1 = XGB end (ψ i , λ i , P i , SST i , t i ) (5)

[0066] where the latitude ψ i and longitude λ i respectively represent the latitude and longitude of the typhoon center at the i-th moment, P i represents the central pressure of the typhoon at the i-th moment, and SST iDenote the sea surface temperature at the i-th moment, t i Denote the total time (hours) of typhoon movement; when end i+1 = 0, the typhoon will continue to develop, when end i+1 = 1, the typhoon ends.

[0067] Specific implementation method eight: The difference between this implementation method and one of the first to seventh specific implementation methods is that the genetic algorithm in step six includes selection, crossover, and mutation operations.

[0068] Specific implementation method nine: The difference between this implementation method and the eighth specific implementation method is that in step six, when the objective function F(X) reaches the minimum value, the optimal model parameters a and b are determined.

[0069] Specific implementation method ten: The difference between this implementation method and one of the first to ninth specific implementation methods is that the initial population size of the genetic algorithm in step six is set to 100.

[0070] Specific implementation method eleven: The difference between this implementation method and the tenth specific implementation method is that the crossover rate of the genetic algorithm is set to 0.5 and the mutation rate is set to 0.2.

[0071] Example: The typhoon full-path simulation method based on local strengthening of the target area is implemented according to the following steps:

[0072] Step 1. Collect all historical typhoon track data in the Northwest Pacific region during the period 1949 - 2022, including information such as longitude, latitude, central pressure, sea surface temperature, etc. The specific process of step 1 is as follows:

[0073] Step 1.1. Extract all historical typhoon track data through the CMA typhoon best track database, and obtain the longitude, latitude, central pressure, maximum wind speed radius, and occurrence time (year, month, day, hour) of all typhoons;

[0074] Step 1.2. Obtain the monthly sea surface temperature situation in the Northwest Pacific through the HadISST database, and interpolate it spatially to each historical typhoon path point, then the sea surface temperature information of all typhoon data points in step 1.1 can be obtained;

[0075] Step 1.3. Preprocess the historical typhoon data, remove the typhoons whose starting points are on land, and finally obtain 2430 typhoon numbers and 71076 typhoon points.

[0076] Step 2. Obtain the spatial location of the target infrastructure system through the Geographic Information System GIS, and draw a spatial distribution map as Figure 3As shown, the central point position coordinates of the infrastructure system are extracted as the target central point. A region with a radius of 250 km centered at the target central point is used as the target area. In this embodiment, the position of the target central point s is 116°E, 23.5°N.

[0077] Step 3: Based on the collected historical typhoon disaster data, construct a typhoon full-path model, including a starting point model, a typhoon path model, a typhoon intensity model, and a typhoon attenuation model. The specific process of Step 3 is as follows:

[0078] Step 3.1: According to the preprocessed historical typhoon track data obtained in Step 1, obtain the starting point information (longitude, latitude, time, central pressure, etc.) of each typhoon path (track), and then model the typhoon starting point by sampling from the starting point data. A total of several starting points are simulated. In this embodiment, a total of 328,169 typhoons are simulated, totaling 10,000 years;

[0079] Step 3.2: Use the following formula to perform data normalization processing on all the obtained typhoon data:

[0080]

[0081] In the formula, X represents the variable to be normalized, including longitude, latitude, time, central pressure, etc., min(X) represents the minimum value in X, max(X) represents the maximum value in X, and X minmax represents the variable after normalization processing;

[0082] Step 3.3: Based on the starting point, construct a typhoon path model through the XGBoost model. The typhoon path model consists of the typhoon path model XGB λ of longitude features and the typhoon path model XGB ψ of latitude features. The typhoon path model of longitude features is shown in formula (2), and the typhoon path model of latitude features is shown in formula (3);

[0083] λ i+1 = XGB λ (ψ i-3 , λ i-3 , ψ i-2 , λ i-2 , ψ i-1 , λ i-1 , ψ i , λ i , t i+1 ) + ε λ (2)

[0084] ψ i-1 = XGB ψ (ψ i-3, λ i-3 , ψ i-2 , λ i-2 , ψ i-1 , λ i-1 , ψ i , λ i , t i+1 ) + ε ψ (3)

[0085] Among them, XGB ψ and XGB λ respectively represent the typhoon path models based on the XGBoost model, used to predict the latitude ψ (°) and longitude λ (°) of the next state (at the (i + 1)-th moment) of the typhoon center; ψ i-3 , ψ i-2 , ψ i-1 , ψ i and ψ i+1 respectively represent the latitudes of the typhoon positions at the (i - 3)-th, (i - 2)-th, (i - 1)-th, i-th, and (i + 1)-th moments, while λ i-3 , λ i-2 , λ i-1 , λ i and λ i+1 respectively represent the longitudes of the typhoon positions at the (i - 3)-th, (i - 2)-th, (i - 1)-th, i-th, and (i + 1)-th moments, t i+1 represents the total movement time of the typhoon at the (i + 1)-th moment, in hours; ε λ and ε ψ respectively represent the model error terms of longitude and latitude, fitted based on the deviation between the predicted value and the true value of historical data, and modeled as a normal distribution;

[0086] Step 3.4: Based on the simulated typhoon path, establish a typhoon intensity model and a decay model through the XGBoost model. The typhoon intensity model before local strengthening in the target area is given as follows:

[0087] P i+1 = XGB P (ψ i-2 , λ i-2 , SST i-2 , ψ i-1 , λ i-1 , SST i-1 , ψ i , λ i , SST i , ψ i+1 , λ i+1 , SST i+1 , t i+1 ) + ε P

[0088] In the formula, P i+1represents the intensity (central pressure) value at the (i + 1)-th predicted moment, XGB P indicates that this is an intensity (central pressure) prediction model based on the XGBoost machine learning algorithm. ψ i-2 , ψ i-1 , ψ i , ψ i+1 respectively represent the latitudes of the typhoon positions at the (i - 2)-th, (i - 1)-th, i-th, and (i + 1)-th moments. λ i-2 , λ i-1 , λ i , λ i+1 respectively represent the longitudes of the typhoon positions at the (i - 2)-th, (i - 1)-th, i-th, and (i + 1)-th moments. SST i-2 , SST i-1 , SST i , SST i+1 respectively represent the sea surface temperatures at the (i - 2)-th, (i - 1)-th, i-th, and (i + 1)-th moments. As an environmental variable, the sea surface temperature has an important impact on the typhoon intensity. t i+1 represents the typhoon's travel time (hours) at the (i + 1)-th time point, ε P represents the model error term, which is fitted based on the deviation between the predicted value and the true value of historical data and is modeled as a normal distribution;

[0089] Step 3.5: The termination state of the typhoon is judged by the typhoon decay model as shown in the following formula:

[0090] end i+1 = XGB end (ψ i , λ i , P i , SST i , t i )

[0091] In the formula, XGB end represents the typhoon decay model based on the XGBoost model; the latitude ψ i and the longitude λ i represent the latitude and longitude of the typhoon center at the i-th moment; P i represents the central pressure at the i-th moment; SST i represents the sea surface temperature at the i-th moment; t i represents the total travel time (hours) of the typhoon; when end i+1 = 0, the typhoon will continue to develop; when end i+1 = 1, the typhoon ends.

[0092] Step 4. Based on the following locally enhanced typhoon intensity model formula, replace the previous intensity model (Step 3.4) to perform local enhancement of the typhoon intensity model:

[0093] P i+1 = XGB P (ψ i-2 , λ i-2 , SST i-2 , areapass i-2 , ψ i-1 , λ i-1 , SST i-1 , areapass i-1 , ψ i , λ i , SST i , areapass i , ψ i+1 , λ i+1 , SST i+1 , t i+1 , areapass i+1 ) + ε P

[0094] In the formula, P i+1 represents the intensity (central pressure) value at the predicted (i + 1)-th moment, and XGB P represents an intensity (central pressure) prediction model based on the XGBoost machine learning algorithm. ψ i-2 , ψ i-1 , ψ i , ψ i+1 represent the latitudes of the typhoon positions at the (i - 2)-th, (i - 1)-th, i-th, and (i + 1)-th moments, respectively. λ i-2 , λ i-1 , λ i , λ i+1 represent the longitudes of the typhoon positions at the (i - 2)-th, (i - 1)-th, i-th, and (i + 1)-th moments, respectively. SST i-2 , SST i-1 , SST i , SST i+1 represent the sea surface temperatures at the (i - 2)-th, (i - 1)-th, i-th, and (i + 1)-th moments, respectively. t i+1 represents the typhoon travel time (hours) at the (i + 1)-th time point, and ε P represents the model error term, which is obtained by fitting the deviation between the predicted value and the true value based on historical data and is modeled as a normal distribution. areapass i-2 , areapass i-1 , areapass i , areapass i+1is a binary variable. If the path of the typhoon sample passes through the target area described in Step 2, the value of the binary variable is 1; if the path of the typhoon sample does not pass through the target area described in Step 2, the value of the binary variable is 0.

[0095] Step 5: Calculate the typhoon intensity model XGB P The distance d between the central coordinates (ψ i , λ i )(as shown in Formula (6)) of the j-th group of typhoon samples at the i-th time and the coordinates of the target center point j,s , and calculate the allocation weight ω of the typhoon sample according to Formula (7) j,s to enhance the local accuracy;

[0096]

[0097] where a and b are optimization variables to be solved and are solved by the genetic algorithm.

[0098] Step 6: Weight each group of typhoon samples used to train the typhoon intensity model XGB P in Formula (7), and use the weighted typhoon samples to retrain the typhoon intensity model XGB P to obtain the typhoon intensity prediction value corresponding to each group of typhoon samples. Compare the typhoon intensity prediction value with the true typhoon intensity value to determine the objective function F(X) in Formula (8):

[0099]

[0100] where N represents the number of historical typhoon samples, d i , s represents the distance from the target site s to the nearest point on the i-th historical typhoon track, and P i,predict (X) represents the allocation weight of each historical sample calculated by input parameters a and b, and according to Formulas (6) and (7), input the typhoon intensity model XGB P to predict the predicted intensity of the typhoon point closest to the target site s in the i-th typhoon sample, and P i,actual represents the actual intensity of the typhoon point closest to the target center point in the i-th typhoon sample.

[0101] Step 7: Use the genetic algorithm to optimize the objective function F(X) to determine the optimal model parameters a and b. In this embodiment, the operating parameters of the genetic algorithm are shown in Table 1, and the optimal parameters a and b are determined to be 0.045 and 0.382 respectively.

[0102] Table 1 Genetic algorithm parameters

[0103] Genetic algorithm parameters Value Population size 100 Maximum number of generations 40 Crossover rate 0.5 Mutation rate 0.2

[0104] Step 8: Obtain the distribution weights of each historical typhoon sample using the optimal model parameters a and b, and retrain the typhoon intensity model XGB based on the distribution weights of each historical typhoon sample. P to obtain an improved typhoon intensity model;

[0105] Step 9: Simulate the complete typhoon full path using the starting point model, typhoon path model, improved typhoon intensity model, and typhoon decay model to obtain a large number of typhoons for subsequent typhoon scenario analysis.

[0106] In this embodiment, the typhoons simulated by the improved model and the unimproved model (both are 328,169) are compared at 47 coastal stations (including the target station). The average central pressure of the typhoon points closest to the typhoons simulated by the two models within 500 km of the 47 stations is calculated and compared with the historical typhoons, as Figure 3 shown. It can be seen that although the effect of the full path model with local strengthening in the target area is not good at stations far from the target station where the infrastructure is located in this embodiment, very obvious improvement effects are achieved at and near the target station where the infrastructure is located, which is very close to the historical central pressure value, and it is undoubtedly very beneficial for the damage analysis of the infrastructure system under typhoon disasters. Figure 4 The distribution comparison of the central pressures of the typhoons simulated by the two models and the historical typhoons at the closest points to the stations within the range of the infrastructure system is given. It can be found that the improvement effect of the present invention is very obvious. Within the range of infrastructure distribution, the typhoons simulated by the typhoon full path model with local strengthening in the target area are very close to the historical typhoons in terms of the key parameter of central pressure. The improvement of the simulation accuracy of typhoons has great significance for the damage analysis and resilience assessment of the infrastructure system under typhoon disasters.

Claims

1. A typhoon full-path simulation method based on local enhancement of the target area, characterized in that The typhoon full-path simulation method based on local enhancement of the target area is implemented according to the following steps: Step 1: Collect historical typhoon track data, which includes the longitude, latitude, central pressure, and sea surface temperature information of typhoons; Step 2: Obtain the spatial location of the target infrastructure system through Geographic Information System (GIS), extract the central point location coordinates of the infrastructure system as the target central point, and use the area with a radius of 250 km centered on the target central point as the target area; Step 3: Build a typhoon full-path model based on the historical typhoon track data in Step 1. The typhoon full-path model includes a starting point model, a typhoon path model, a typhoon intensity model, and a typhoon decay model; The process of building the typhoon full-path model is as follows: Step 3.1: Build a starting point model; Step 3.2: Based on the starting point, construct a typhoon track model through the XGBoost model. The typhoon track model consists of the typhoon track model XGB of longitude features λ and the typhoon track model XGB of latitude features ψ to predict the latitude ψ and longitude λ of the typhoon center at the (i + 1)-th moment. The typhoon track model of longitude features is shown in Equation (2), and the typhoon track model of latitude features is shown in Equation (3). λ i+1 = XGB λ (ψ i-3 ,λ i-3 ,ψ i-2 ,λ i-2 ,ψ i-1 ,λ i-1 ,ψ i ,λ i ,t i+1 ) + ε λ (2) ψ i+1 = XGB ψ (ψ i-3 , λ i-3 , ψ i-1 , λ i-2 , ψ i-1 , λ i-1 , ψ i , λ i , t i+1 ) + ε φ (3) where ψ i-3 , ψ i-2 , ψ i-1 , ψ i and ψ i+1 represent the latitudes of the typhoon positions at the (i - 3)th, (i - 2)th, (i - 1)th, ith, and (i + 1)th moments respectively, λ i-3 , λ i-2 , λ i-1 , λ i and λ i+1 represent the longitudes of the typhoon positions at the (i - 3)th, (i - 2)th, (i - 1)th, ith, and (i + 1)th moments respectively, t i+1 represents the total moving time of the typhoon at the (i + 1)th moment, in hours; ε λ and ε ψ represent the longitude model error term and the latitude model error term, respectively; Step 3.

3. Construct the typhoon intensity model XGB using the XGBoost model P , where the typhoon intensity is the central pressure of the typhoon, and the typhoon intensity model is shown in formula (6); P i+1 = XGB P (ψ i-2 , λ i-2 , SST i-2 , areapass i-2 , ψ i-1 , λ i-1 , SST i-1 , areapass i-1 , ψ i , λ i , SST i , areapass i , ψ i+1 , λ i+1 , SST i+1 , t i+1 , areapass i+1 ) + ε P (6) where P i+1 represents the typhoon intensity value at the predicted (i + 1)-th moment; ψ i-2 , ψ i-1 , ψ i , ψ i+1 represent the latitudes of the typhoon positions at the (i - 2)-th, (i - 1)-th, i-th, and (i + 1)-th moments of the typhoon sample respectively; λ i-2 , λ i-1 , λ i , λ i+1 represent the longitudes of the typhoon positions at the (i - 2)-th, (i - 1)-th, i-th, and (i + 1)-th moments of the typhoon sample respectively; SST i-2 , SST i-1 , SST i , SST i+1 represent the sea surface temperatures at the (i - 2)-th, (i - 1)-th, i-th, and (i + 1)-th moments of the typhoon sample respectively; t i+1 represents the typhoon migration time at the (i + 1)-th time point; ε P represents the intensity model error term; areapass i-2 , areapass i-1 , areapass i , areapass i+1 is a binary variable. If the path of the typhoon sample passes through the target area described in step two, the binary variable value is 1; if the path of the typhoon sample does not pass through the target area described in step two, the binary variable value is 0; Step 3.4: Build a typhoon decay model; Step 4. Calculate the typhoon intensity model XGB P The distance d i between the central coordinates (ψ i , λ j,s ) of the j-th typhoon sample at the i-th time in P and the coordinates of the target center point is calculated, and the allocation weight ω j,s of the typhoon sample is calculated according to formula (7) to enhance the local accuracy; where a and b are optimization variables; Step 5: Based on formula (7), weight each group of typhoon samples used for training typhoon intensity model XGB P in the model, and use the weighted typhoon samples to retrain the typhoon intensity model XGB P to obtain the typhoon intensity prediction values corresponding to the output of each group of typhoon samples. Compare the typhoon intensity prediction values with the true typhoon intensity values to determine the objective function F(X) as shown in formula (8): Wherein, N represents the number of historical typhoons, X represents the parameter group (a, b), and d i,s represents the distance from the target center point to the nearest point on the i-th historical typhoon track, and P i,predict (X) is the typhoon intensity model XGB retrained P The predicted intensity of the typhoon point closest to the target center point in the i-th typhoon sample output, and P i,actual represents the actual intensity of the typhoon point closest to the target center point in the i-th typhoon sample; Step 6: Use the genetic algorithm to optimize the objective function F(X) to determine the optimal model parameters a and b; Step 7: Obtain the allocation weights of each historical typhoon sample using the optimal model parameters a and b, and retrain the typhoon intensity model XGB based on the allocation weights of each historical typhoon sample P to obtain an improved typhoon intensity model; Step 8: Simulate the complete typhoon full-path through the starting point model, the typhoon path model, the improved typhoon intensity model, and the typhoon decay model, thus completing the typhoon full-path simulation method based on local enhancement of the target area.

2. The typhoon full-path simulation method based on local enhancement of the target area according to claim 1, wherein In Step 1, collect historical typhoon track data from the CMA typhoon best track database.

3. The typhoon full-path simulation method based on local enhancement of the target area according to claim 1, wherein The number of historical typhoon track data in Step 1 is greater than 2000.

4. The typhoon full-path simulation method based on local enhancement of the target area according to claim 1, characterized in that In Step 1, remove the typhoon data with starting points on land from the historical typhoon track data.

5. The typhoon full-path simulation method based on local enhancement of the target area according to claim 1, characterized in that The process of building the starting point model in Step 3.1 is as follows: According to the historical typhoon track data collected in Step 1, obtain the starting point data of each typhoon track data, and then model the typhoon starting point by sampling from the starting point data to obtain the starting point model.

6. The typhoon full-path simulation method based on local enhancement of the target area according to claim 5, characterized in that The starting point data includes the longitude, latitude, time, and central pressure of the starting point.

7. The typhoon full-path simulation method based on local enhancement of the target area according to claim 1, characterized in that The process of constructing the typhoon attenuation model XGB in Step 3.4 end is as follows: Construct the typhoon attenuation model XGB through the XGBoost model end , and the typhoon attenuation model is shown in formula (5); end i+1 = XGB cnd (ψ i , λ i , P i , SST i , t i ) (5) where the latitude ψ i and the longitude λ i represent the latitude and longitude of the typhoon center at the i-th moment respectively, P i represents the central pressure of the typhoon at the i-th moment, SST i represents the sea surface temperature at the i-th moment, t i represents the total time of typhoon movement; when end i+1 = 0, the typhoon will continue to develop, when end i+1 = 1, the typhoon ends.

8. The typhoon full-path simulation method based on local enhancement of the target area according to claim 1, wherein In Step 6, the genetic algorithm includes selection, crossover, and mutation operations.

9. The typhoon full-path simulation method based on local enhancement of the target area according to claim 1, characterized in that In Step 6, the initial population size of the genetic algorithm is set to 100.

10. The typhoon full-path simulation method based on local enhancement of the target area according to claim 9, characterized in that The crossover rate of the genetic algorithm is set to 0.5, and the mutation rate is set to 0.2.

Images