A method for forecasting rainstorms and floods based on the similarity of rainstorm center of gravity migration characteristics

Abstract

This invention discloses a method for forecasting rainstorms and floods based on the similarity of rainstorm center of gravity migration characteristics, belonging to the field of flood forecasting and prediction technology. The method includes: constructing a rainstorm and flood event association scenario database; establishing a dynamic rainstorm structured representation matrix; representing rainstorm event data to form multiple rainstorm cluster units, extracting the center trace of the rainstorm center of gravity migration characteristics of each rainstorm cluster unit; calculating the spatial migration characteristics of the target rainstorm event's center of gravity, comparing it with the center trace of the rainstorm cluster units to determine the similarity between the target rainstorm event and each rainstorm cluster unit, and identifying the historical rainstorm event with the highest similarity; inputting the data into the rainstorm and flood event scenario database to obtain the corresponding flood event, and performing flood forecasting simulation. This invention considers the rainstorm center of gravity migration characteristics, directly linking the dynamic spatial evolution of rainstorms with flood response, thereby improving the accuracy of flood forecasting and prediction based on rainstorm and flood similarity knowledge.

Description

Technical Field

[0001] This invention relates to a method for forecasting rainstorms and floods based on the similarity of the center-of-gravity migration characteristics of rainstorms, belonging to the field of flood forecasting and prediction technology. Background Technology

[0002] In recent years, climate change has led to increasingly frequent floods, posing a serious threat to urban operations and the safety of people's lives and property. Therefore, rapid and accurate flood forecasting has become a crucial measure to reduce flood losses. Conducting studies on the similarity between rainstorms and floods has been proven to be an effective way to extend flood forecast periods. Currently, most flood forecasting methods based on the similarity of rainstorms and floods focus on rainfall magnitude and temporal characteristics. The spatial characteristics of rainfall are mostly expressed by calculating areal rainfall, but areal rainfall often fails to include information on the location of the rainstorm's center of gravity, resulting in low accuracy in similar rainstorm searches and consequently affecting the accuracy of the final flood forecast. The location of the rainstorm's center of gravity, especially its spatial migration characteristics, is closely related to the flood process. Therefore, it is urgent to construct a dynamic structured characterization index system for rainstorms that incorporates the migration characteristics of the rainstorm's center of gravity to improve the accuracy of similar rainstorm searches and thus enhance flood forecast accuracy. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to overcome the defects of the prior art and provide a rainstorm flood forecasting method and related equipment based on the similarity of rainstorm center of gravity migration characteristics. By considering the rainstorm center of gravity migration characteristics, the dynamic spatial evolution process of rainstorms is directly linked to flood response, thereby improving the accuracy of flood forecasting based on rainstorm flood similarity knowledge.

[0004] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0005] A method for forecasting rainstorm floods based on the similarity of rainstorm center of gravity migration characteristics includes the following steps:

[0006] Historical rainstorm and flood data for the study area were obtained. After dividing the rainstorm data into events, rainstorm event data was obtained. Then, the rainstorm event data and flood event data were correlated to build a rainstorm and flood event correlation scenario database.

[0007] A dynamic rainstorm structured characterization matrix is ​​established by selecting the spatial migration feature, magnitude feature, and temporal feature of the rainstorm centroid migration characteristics from the rainstorm data.

[0008] The dynamic rainstorm structured representation matrix is ​​used to represent rainstorm data. This includes first using a multivariate weighted similarity index to evaluate the similarity of the structured representation matrices of each rainstorm, then using a clustering analysis algorithm to cluster similar rainstorms to form multiple rainstorm clustering units, and finally extracting the center trace of the rainstorm centroid migration feature of each rainstorm clustering unit.

[0009] The centroid spatial migration characteristics of the target rainstorm event are calculated. The calculation results are compared with the center traces of the rainstorm cluster units to determine the similarity between the target rainstorm event and each rainstorm cluster unit, so that the target rainstorm event is assigned to a similar rainstorm cluster unit. Then, based on the similarity principle, a search is performed in the assigned rainstorm cluster unit to find the historical rainstorm event with the highest similarity.

[0010] The historical rainstorm events with the highest similarity are input into the rainstorm and flood event scenario database to obtain the corresponding flood events, and flood forecasting is performed based on the information of the corresponding flood events.

[0011] The process of dividing the rainstorm data into rainstorm sessions and then establishing a correlation between the rainstorm session data and the flood session data involves: selecting rainstorm session data from all stations that experienced precipitation within the corresponding time period in the rainstorm data based on the flood process time in the flood session data, and establishing a spatiotemporal correlation between the flood session data and the rainstorm session data.

[0012] The representation process of the centroid spatial migration feature is as follows:

[0013] The spatial distribution of precipitation in each time period is regarded as a mass distribution on a two-dimensional plane, and the centroid of the rainstorm is obtained by calculating the centroid of the precipitation field in that time period.

[0014] Arranging a series of rainstorm centers in chronological order yields a rainstorm center path sequence;

[0015] Based on the storm center of gravity path sequence, four key indicators were extracted from four dimensions: storm center of gravity path distance, storm center of gravity coverage area, storm center of gravity axis length, and storm center of gravity peak rainfall and peak time.

[0016] The centroid spatial migration feature matrix is ​​obtained based on four key indicators.

[0017] The formula for calculating the center of gravity of a rainstorm is:

[0018] (1),

[0019] In the formula: and The first The latitude and longitude coordinates of the center of gravity of the period of heavy rainfall. For the first The number of stations where precipitation occurred during the time period. It is a website In the Rainfall during the period and These are the sites Latitude and longitude coordinates.

[0020] The expression for the path sequence of the rainstorm's center of gravity is: ,in, Let t be the latitude and longitude coordinates of the center of gravity of the rainstorm in time period t, and T be the total number of rainstorm periods.

[0021] The distance of the center of gravity migration of the rainstorm is obtained by calculating the sum of the straight-line distances between the centers of gravity of rainstorms in consecutive time periods. The specific formula is as follows:

[0022] (2),

[0023] In the formula: This refers to the distance of the shift in the center of gravity of the rainstorm. and The first The latitude and longitude coordinates of the centroid of the periodic heavy rainfall, projected onto the plane with equal intervals. and The first The latitude and longitude coordinates of the centroid of the rainstorm during a given period are projected onto the plane with equal intervals, and T is the total number of rainstorm periods.

[0024] The coverage area of ​​the storm's centroid is obtained by calculating the area of ​​the minimum convex polygon formed by the centroids of storms across all time periods using the convex hull algorithm. The specific formula is as follows:

[0025] (3),

[0026] In the formula: The area covered by the center of the rainstorm. and The first The latitude and longitude coordinates of the ordered vertices of the convex hull, projected onto the plane using equal-distance projection. and The first The latitude and longitude coordinates of the ordered vertices of the convex hull are the horizontal and vertical coordinates of the plane under the equidistant projection, where m is the total number of vertices;

[0027] The centroidal length of a rainstorm is obtained by calculating the straight-line distance between the two rainstorm centroids with the greatest distance between them over all time periods. The specific formula is as follows:

[0028] (4),

[0029] In the formula: This indicates taking the maximum value in the sample sequence. Because the center of gravity of the rainstorm is long, and For the first The coordinates of the centroid of the rainfall during a given period, projected onto a plane with horizontal and vertical axes. and For the first The coordinates of the centroid of the rainstorm during a given period are represented by the horizontal and vertical coordinates of the plane under equidistant projection.

[0030] The specific form of the dynamic rainstorm structured characterization matrix is ​​as follows:

[0031] (5),

[0032] (6),

[0033] (7),

[0034] (8),

[0035] In the formula, This is a structured representation matrix for dynamic rainstorms. This reflects the characteristics of the shift in the center of gravity of the rainstorm. As a characteristic of magnitude, As a time feature, This refers to the distance of the shift in the center of gravity of the rainstorm. The area covered by the center of the rainstorm. Because the center of gravity of the rainstorm is long, This represents the peak rainfall at the center of the rainstorm. This is the peak of the rainstorm. Total rainfall, For the peak rainfall, For the duration of rainfall, This is the peak rainfall period.

[0036] The formula for calculating the similarity of the structured representation matrices of each rainstorm event using a multivariate weighted similarity index is as follows:

[0037] (9),

[0038] (10)

[0039] (11),

[0040] In the formula, For similarity, For the first Feature index vector of rainfall field For the first Feature index vector of rainfall field Let Euclidean distance be the characteristic index vectors of the two rainfall events. The angle between the characteristic index vectors of the two rainfall events. Total number of rainfall events, To be used for balance and The scale factor for differences in magnitude. The Euclidean norm of a vector. The weighting coefficient is between 0 and 1.

[0041] The formula for calculating the center trace is as follows:

[0042] (12)

[0043] (13)

[0044] (14)

[0045] (15)

[0046] In the formula: For the first The center trace features of each cluster unit Indicates the first The average value of the rainstorm centroid migration characteristics of each cluster unit. Indicates the first The average value of features at the order of magnitude of cluster units Indicates the first The average time characteristic of each cluster unit For the first The number of samples in each cluster unit. For the first A sample set of cluster units, For the first Characteristics of the center of gravity shift in rainfall events. For the first The magnitude characteristics of the rainfall event. For the first The temporal characteristics of the rainfall.

[0047] The calculation method for assigning the target rainstorm events to similar rainstorm cluster units is as follows:

[0048] (16)

[0049] In the formula: For the corresponding rainstorm cluster unit, To find the position function of the maximum value, The target of the rainstorm events and the first The similarity of the center traces of each cluster unit.

[0050] The beneficial effects of this invention are as follows: This invention provides a rainstorm and flood forecasting method based on the similarity of rainstorm center of gravity migration characteristics. By constructing a dynamic rainstorm structured characterization index system, it accurately quantifies key physical characteristics such as rainstorm center of gravity migration and intensity evolution. It also utilizes machine learning algorithms to achieve objective classification and fine matching of rainstorm events, directly linking the dynamic spatial evolution of rainstorms with flood response, thereby improving the accuracy of flood forecasting based on rainstorm-flood similarity knowledge. Furthermore, based on the physical assumption that similar rainfall patterns lead to similar flood responses, it performs forecast extrapolation by retrieving the most similar historical rainstorm-flood scenarios. The overall technical route has a clear physical mechanism and rigorous logic. It mainly relies on easily obtainable rainfall data, is highly adaptable to areas with data shortages, and the forecast results have good interpretability and operability, providing technical support for regional flood control work. Attached Figure Description

[0051] Figure 1 This is a flowchart illustrating a rainstorm flood forecasting method based on the similarity of rainstorm center of gravity migration characteristics according to the present invention.

[0052] Figure 2 A schematic diagram of the flood and target flood process lines corresponding to the mid-level similarity scenario 1 in this invention;

[0053] Figure 3 A schematic diagram of the flood and target flood process lines corresponding to the second similar scenario in this invention;

[0054] Figure 4 A schematic diagram of the flood and target flood process lines corresponding to the third similarity scenario in this invention;

[0055] Figure 5 This invention presents the visualization results of flood clustering groups based on multidimensional scale transformation technology. Detailed Implementation

[0056] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to illustrate the technical solution of the present invention more clearly, and should not be used to limit the scope of protection of the present invention.

[0057] Example 1

[0058] like Figure 1 As shown, this invention discloses a method for forecasting rainstorms and floods based on the similarity of rainstorm center of gravity migration characteristics, comprising the following steps:

[0059] Step 1: Obtain historical rainstorm data and flood event data for the study area. After dividing the rainstorm data into events, obtain rainstorm event data. Then, establish a correlation between the rainstorm event data and the flood event data to build a rainstorm and flood event correlation scenario database.

[0060] Step 2: Select the spatial migration feature, magnitude feature, and temporal feature of the center of gravity migration feature from the rainstorm data to establish a dynamic rainstorm structured representation matrix.

[0061] Step 3: Based on the dynamic rainstorm structured representation matrix, the rainstorm data is represented. First, the similarity of the structured representation matrices of each rainstorm is evaluated using a multivariate weighted similarity index. Then, a clustering analysis algorithm is used to cluster similar rainstorms to form multiple rainstorm clustering units. Finally, the center trace of the rainstorm centroid migration feature of each rainstorm clustering unit is extracted.

[0062] Step 4: Calculate the centroid spatial migration characteristics of the target rainstorm event, compare the calculation results with the center traces of the rainstorm cluster units to determine the similarity between the target rainstorm event and each rainstorm cluster unit, so that the target rainstorm event is assigned to a similar rainstorm cluster unit. Then, based on the similarity principle, search in the assigned rainstorm cluster unit to find the historical rainstorm event with the highest similarity.

[0063] Step 5: Input the historical rainstorm events with the highest similarity into the rainstorm and flood event scenario database to obtain the corresponding flood events, and perform flood forecast simulations based on the information of the corresponding flood events.

[0064] This invention constructs a dynamic structured characterization index system for rainstorms, accurately quantifying key physical characteristics such as the migration of rainstorm center of gravity and intensity evolution. It utilizes machine learning algorithms to achieve objective classification and precise matching of rainstorm events, directly linking the dynamic spatial evolution of rainstorms with flood response. This improves the accuracy of flood forecasting based on the similarity between rainstorms and floods. Furthermore, based on the physical assumption that similar rainfall patterns lead to similar flood responses, it performs forecast extrapolation by retrieving the most similar historical rainstorm-flood scenarios. The overall technical approach has a clear physical mechanism and rigorous logic, primarily relying on readily available rainfall data. It is highly adaptable to areas with data shortages, and the forecast results possess good interpretability and operability, providing technical support for regional flood control efforts.

[0065] Example 2

[0066] like Figure 1 As shown, this invention discloses a method for forecasting rainstorms and floods based on the similarity of rainstorm center of gravity migration characteristics, comprising the following steps:

[0067] Step 1: Obtain historical rainstorm data and flood event data for the study area. After dividing the rainstorm data into events, obtain rainstorm event data. Then, establish a correlation between the rainstorm event data and the flood event data. Specifically, based on the flood process time in the flood event data, select the rainstorm event data of all stations that experienced precipitation within the corresponding time period in the rainstorm data, establish the spatiotemporal correlation between the flood event data and the rainstorm event data, and finally build a rainstorm and flood event correlation scenario database.

[0068] Step 2: Select the spatial migration feature, magnitude feature, and temporal feature of the center of gravity migration feature from the rainstorm data to establish a dynamic rainstorm structured representation matrix.

[0069] The migration characteristics of the center of gravity of rainstorms are the core indicators for characterizing the spatiotemporal evolution of rainstorm systems. By quantifying geometric and physical parameters, the system reveals the movement trajectory, intensity changes, and structural characteristics of the main rain belt or the core area of ​​heavy precipitation.

[0070] The process of characterizing the spatial migration characteristics of the center of gravity is as follows: the spatial distribution of precipitation in each time period is regarded as a mass distribution on a two-dimensional plane, and the precipitation amount is the "mass". The center of gravity of the rainstorm is the "centroid" of the precipitation field in that time period. Its coordinates reflect the spatial concentration of precipitation intensity. The center of gravity of the rainstorm is obtained by calculating the center of gravity of the precipitation field in that time period.

[0071] The formula for calculating the center of gravity of a rainstorm is:

[0072] (1),

[0073] In the formula: and The first The latitude and longitude coordinates of the center of gravity of the period of heavy rainfall. For the first The number of stations where precipitation occurred during the time period. It is a website In the Rainfall during the period and These are the sites Latitude and longitude coordinates.

[0074] Arranging a series of rainstorm centroids in chronological order yields a centroid coordinate sequence, which is the rainstorm centroid path sequence.

[0075] The expression for the path sequence of the rainstorm's center of gravity is: ,in, Let t be the latitude and longitude coordinates of the center of gravity of the rainstorm in time period t, and T be the total number of rainstorm periods.

[0076] The distance of the center of gravity migration of the rainstorm is obtained by calculating the sum of the straight-line distances between the centers of gravity of rainstorms in consecutive time periods. The specific formula is as follows:

[0077] (2),

[0078] In the formula: This refers to the distance of the shift in the center of gravity of the rainstorm. and The first The latitude and longitude coordinates of the centroid of the periodic heavy rainfall, projected onto the plane with equal intervals. and The first The latitude and longitude coordinates of the centroid of the rainstorm during a given period are projected onto the plane with equal intervals, and T is the total number of rainstorm periods.

[0079] The coverage area of ​​the storm's centroid is obtained by calculating the area of ​​the minimum convex polygon formed by the centroids of storms across all time periods using the convex hull algorithm. The specific formula is as follows:

[0080] (3),

[0081] In the formula: The area covered by the center of the rainstorm. and The first The latitude and longitude coordinates of the ordered vertices of the convex hull, projected onto the plane using equal-distance projection. and The first The latitude and longitude coordinates of the ordered vertices of the convex hull are the horizontal and vertical coordinates of the plane under the equidistant projection, where m is the total number of vertices;

[0082] The centroidal length of a rainstorm is obtained by calculating the straight-line distance between the two rainstorm centroids with the greatest distance between them over all time periods. The specific formula is as follows:

[0083] (4),

[0084] In the formula: This indicates taking the maximum value in the sample sequence. Because the center of gravity of the rainstorm is long, and For the first The coordinates of the centroid of the rainfall during a given period, projected onto a plane with horizontal and vertical axes. and For the first The coordinates of the centroid of the rainstorm during a given period are represented by the horizontal and vertical coordinates of the plane under equidistant projection.

[0085] Based on the storm center path sequence, four key indicators were extracted from four dimensions: storm center path distance, storm center coverage, storm center axis length, and storm center rainfall peak and peak time.

[0086] Based on four key indicators, the centroid spatial migration feature matrix is ​​obtained. The specific form of the dynamic rainstorm structured characterization matrix is ​​as follows:

[0087] (5),

[0088] (6),

[0089] (7),

[0090] (8),

[0091] In the formula, This is a structured representation matrix for dynamic rainstorms. This reflects the characteristics of the shift in the center of gravity of the rainstorm. As a characteristic of magnitude, As a time feature, This refers to the distance of the shift in the center of gravity of the rainstorm. The area covered by the center of the rainstorm. Because the center of gravity of the rainstorm is long, This represents the peak rainfall at the center of the rainstorm. This is the peak of the rainstorm. Total rainfall, For the peak rainfall, For the duration of rainfall, This is the peak rainfall period.

[0092] Step 3: Based on the dynamic rainstorm structured representation matrix, the rainstorm data is represented. First, the similarity of the structured representation matrices of each rainstorm is evaluated using a multivariate weighted similarity index. Then, a clustering analysis algorithm is used to cluster similar rainstorms to form multiple rainstorm clustering units. Finally, the center trace of the rainstorm centroid migration feature of each rainstorm clustering unit is extracted.

[0093] The formula for calculating the similarity of the structured representation matrices of each rainstorm event using a multivariate weighted similarity index is as follows:

[0094] (9),

[0095] (10)

[0096] (11),

[0097] In the formula, For similarity, For the first Feature index vector of rainfall field No. Feature index vector of rainfall field Let Euclidean distance be the characteristic index vectors of the two rainfall events. The angle between the characteristic index vectors of the two rainfall events. Total number of rainfall events, As a scaling factor, used for balancing and Differences in magnitude The Euclidean norm of a vector. The weighting coefficient is between 0 and 1.

[0098] By associating and storing the statistical characteristics of the center trace, a traceable database of rainstorm center migration patterns is formed. The formula for calculating the center trace is as follows:

[0099] (12)

[0100] (13)

[0101] (14)

[0102] (15)

[0103] In the formula: For the first The center trace features of each cluster unit Indicates the first The average value of the rainstorm centroid migration characteristics of each cluster unit. Indicates the first The average value of features at the order of magnitude of cluster units Indicates the first The average time characteristic of each cluster unit For the first The number of samples in each cluster unit. For the first A sample set of cluster units, For the first Characteristics of the center of gravity shift in rainfall events. For the first The magnitude characteristics of the rainfall event. For the first The temporal characteristics of the rainfall.

[0104] Step 4: Calculate the centroid spatial migration characteristics of the target rainstorm events. Referring to formula (5), compare the calculated results with the center traces of the rainstorm cluster units to determine the similarity between the target rainstorm events and each rainstorm cluster unit. Referring to formulas (9) to (11), the target rainstorm events are assigned to similar rainstorm cluster units. The calculation method is as follows:

[0105] (16)

[0106] In the formula: For the corresponding rainstorm cluster unit, arg max is the function that takes the maximum value. The target of the rainstorm events and the first The similarity of the center trajectories of each cluster unit is then used to search within the corresponding rainstorm cluster unit based on the similarity principle to find the historical rainstorm events with the highest similarity.

[0107] Step 5: Input the historical rainstorm events with the highest similarity into the rainstorm and flood event scenario database to obtain the corresponding flood events, and perform flood forecast simulations based on the information of the corresponding flood events.

[0108] To verify the reliability of the dynamic rainstorm structured characterization matrix in step two for pointing to similar floods, a flood characteristic index matrix is ​​constructed, including total flood volume, duration, peak location, peak discharge, flood morphology, flood rise rate, and flood recession rate, etc., with the specific formulas as follows:

[0109] (17)

[0110] In the formula, Q is the flood characteristic index matrix. For the total flood volume, For the peak flow, For the duration of the flood, For the peak of the flood, For the rate of flood rise, , For the coefficient of variation and This is the skewness coefficient.

[0111] This invention constructs a flood characteristic index system according to Table 1, and uses the K-means clustering algorithm to directly cluster flood events, obtaining the flood cluster Flood_0; then, it obtains the flood cluster Flood_1 corresponding to the rainstorm events characterized in step two. Multidimensional scaling (DS) technology is used to map the eight indicators of the two flood clusters Flood_0 and Flood_1 into a two-dimensional space, visually displaying the distances between samples. Figure 5 It can be seen that the mapping results of the flood cluster Flood_1 and Flood_0 are very close, which also verifies the reliability of the method of the present invention.

[0112] Table 1 Flood Characteristic Indicators

[0113]

[0114] Example 3

[0115] This embodiment discloses a specific application based on Embodiment 2. Specifically, the historical precipitation data consists of rainstorm data from 250 upstream stations from 1960 to 2019, and the flood data consists of flood data from the main upstream control station, Yichang station, from 1960 to 2019. A total of three flood forecasts are performed, and the flood and target flood process lines corresponding to similar scenarios are shown below. Figure 2 , Figure 3 and Figure 4 As shown. This invention uses relative error (RE), Nash efficiency coefficient (NSE), and Kling-Gupta efficiency coefficient (KGE) to evaluate the similarity between floods corresponding to similar scenarios and the target flood. NSE and KGE are used to assess the goodness of fit between the two floods, while RE describes the systematic bias between the two floods. The calculation formula for the evaluation index is as follows:

[0116] (18)

[0117] (19)

[0118] (20)

[0119] (twenty one),

[0120] (twenty two),

[0121] (twenty three),

[0122] In the formula: The flood volume is typical of a flood; The flood volume selected for the clustering method; This represents the average flood volume of a typical flood. The average flood volume of the flood selected by the clustering method; Pearson correlation coefficient, which reflects the degree of linear correlation between floods in similar scenarios and the target flood; BR is the bias ratio, and RV is the relative rate of change; and , respectively, represent the standard deviations of floods corresponding to similar scenarios and the target flood; N is the total number of flood periods. The final calculation results are shown in Table 2.

[0123] Table 2 Calculation Results of Similarity Assessment Indicators

[0124]

[0125] In addition, the flood and target flood hydrographs corresponding to similar scenarios in each event are as follows: Figure 2 , Figure 3 and Figure 4 As shown. This invention analyzes the similarity assessment index calculation results (Table 2) and flood hydrographs of similar scenarios and target floods. Figure 2 , Figure 3 and Figure 4It can be found that the flood process corresponding to the historical similar rainstorms identified by the rainstorm flood search algorithm based on the similarity of the rainstorm center of gravity migration characteristics is closer to the actual flood process. All evaluation indicators show good performance, and the identification process is faster, which effectively improves the accuracy and efficiency of flood forecasting and can provide technical support for regional flood control work.

[0126] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for forecasting rainstorm floods based on the similarity of rainstorm center of gravity migration characteristics, characterized in that: Includes the following steps: Historical rainstorm and flood data for the study area were obtained. After dividing the rainstorm data into events, rainstorm event data was obtained. Then, the rainstorm event data and flood event data were correlated to build a rainstorm and flood event correlation scenario database. A dynamic rainstorm structured characterization matrix is ​​established by selecting the spatial migration feature, magnitude feature, and temporal feature of the rainstorm centroid migration characteristics from the rainstorm data. The dynamic rainstorm structured representation matrix is ​​used to represent rainstorm data. This includes first using a multivariate weighted similarity index to evaluate the similarity of the structured representation matrices of each rainstorm, then using a clustering analysis algorithm to cluster similar rainstorms to form multiple rainstorm clustering units, and finally extracting the center trace of the rainstorm centroid migration feature of each rainstorm clustering unit. The centroid spatial migration characteristics of the target rainstorm event are calculated. The calculation results are compared with the center traces of the rainstorm cluster units to determine the similarity between the target rainstorm event and each rainstorm cluster unit, so that the target rainstorm event is assigned to a similar rainstorm cluster unit. Then, based on the similarity principle, a search is performed in the assigned rainstorm cluster unit to find the historical rainstorm event with the highest similarity. The historical rainstorm events with the highest similarity are input into the rainstorm and flood event scenario database to obtain the corresponding flood events, and flood forecasting is performed based on the information of the corresponding flood events.

2. The rainstorm flood forecasting method based on the similarity of rainstorm center of gravity migration characteristics according to claim 1, characterized in that: The process of dividing the rainstorm data into rainstorm sessions and then establishing a correlation between the rainstorm session data and the flood session data involves: selecting rainstorm session data from all stations that experienced precipitation within the corresponding time period in the rainstorm data based on the flood process time in the flood session data, and establishing a spatiotemporal correlation between the flood session data and the rainstorm session data.

3. The rainstorm flood forecasting method based on the similarity of rainstorm center of gravity migration characteristics according to claim 1, characterized in that: The representation process of the centroid spatial migration feature is as follows: The spatial distribution of precipitation in each time period is regarded as a mass distribution on a two-dimensional plane, and the centroid of the rainstorm is obtained by calculating the centroid of the precipitation field in that time period. Arranging a series of rainstorm centers in chronological order yields a rainstorm center path sequence; Based on the storm center of gravity path sequence, four key indicators were extracted from four dimensions: storm center of gravity path distance, storm center of gravity coverage area, storm center of gravity axis length, and storm center of gravity peak rainfall and peak time. The centroid spatial migration feature matrix is ​​obtained based on four key indicators.

4. The rainstorm flood forecasting method based on the similarity of rainstorm center of gravity migration characteristics according to claim 3, characterized in that: The formula for calculating the center of gravity of a rainstorm is: （1）， In the formula: and The first The latitude and longitude coordinates of the center of gravity of the period of heavy rainfall. For the first The number of stations where precipitation occurred during the time period. It is a website In the Rainfall during the period and These are the sites Latitude and longitude coordinates.

5. The rainstorm flood forecasting method based on the similarity of rainstorm center of gravity migration characteristics according to claim 3, characterized in that: The expression for the path sequence of the rainstorm's center of gravity is: ,in, Let t be the latitude and longitude coordinates of the center of gravity of the rainstorm in time period t, and T be the total number of rainstorm periods.

6. The rainstorm flood forecasting method based on the similarity of rainstorm center of gravity migration characteristics according to claim 3, characterized in that: The distance of the center of gravity migration of the rainstorm is obtained by calculating the sum of the straight-line distances between the centers of gravity of rainstorms in consecutive time periods. The specific formula is as follows: （2）， In the formula: This refers to the distance of the shift in the center of gravity of the rainstorm. and The first The latitude and longitude coordinates of the centroid of the periodic heavy rainfall, projected onto the plane with equal intervals. and The first The latitude and longitude coordinates of the centroid of the rainstorm during a given period are projected onto the plane with equal intervals, and T is the total number of rainstorm periods. The coverage area of ​​the storm's centroid is obtained by calculating the area of ​​the minimum convex polygon formed by the centroids of storms across all time periods using the convex hull algorithm. The specific formula is as follows: （3）， In the formula: The area covered by the center of the rainstorm. and The first The latitude and longitude coordinates of the ordered vertices of the convex hull, projected onto the plane using equal-distance projection. and The first The latitude and longitude coordinates of the ordered vertices of the convex hull are the horizontal and vertical coordinates of the plane under the equidistant projection, where m is the total number of vertices; The centroidal length of a rainstorm is obtained by calculating the straight-line distance between the two rainstorm centroids with the greatest distance between them over all time periods. The specific formula is as follows: （4）， In the formula: This indicates taking the maximum value in the sample sequence. Because the center of gravity of the rainstorm is long, and For the first The coordinates of the centroid of the rainfall during a given period, projected onto a plane with horizontal and vertical axes. and For the first The coordinates of the centroid of the rainstorm during a given period are represented by the horizontal and vertical coordinates of the plane under equidistant projection.

7. The rainstorm flood forecasting method based on the similarity of rainstorm center of gravity migration characteristics according to claim 3, characterized in that: The specific form of the dynamic rainstorm structured characterization matrix is ​​as follows: （5）， （6）， （7）， （8）， In the formula, This is a structured representation matrix for dynamic rainstorms. This reflects the characteristics of the shift in the center of gravity of the rainstorm. As a characteristic of magnitude, As a time feature, This refers to the distance of the shift in the center of gravity of the rainstorm. The area covered by the center of the rainstorm. Because the center of gravity of the rainstorm is long, This represents the peak rainfall at the center of the rainstorm. This is the peak of the rainstorm. Total rainfall, For the peak rainfall, For the duration of rainfall, This is the peak rainfall period.

8. The rainstorm flood forecasting method based on the similarity of rainstorm center of gravity migration characteristics according to claim 1, characterized in that: The formula for calculating the similarity of the structured representation matrices of each rainstorm event using a multivariate weighted similarity index is as follows: （9）， （10）， （11）， In the formula, For similarity, For the first Feature index vector of rainfall field For the first Feature index vector of rainfall field Let Euclidean distance be the characteristic index vectors of the two rainfall events. The angle between the characteristic index vectors of the two rainfall events. Total number of rainfall events, To be used for balance and The scale factor for differences in magnitude. The Euclidean norm of a vector. The weighting coefficient is between 0 and 1.

9. The rainstorm flood forecasting method based on the similarity of rainstorm center of gravity migration characteristics according to claim 1, characterized in that: The formula for calculating the center trace is as follows: （12）， （13）， （14）， （15）， In the formula: For the first The center trace features of each cluster unit Indicates the first The average value of the rainstorm centroid migration characteristics of each cluster unit. Indicates the first The average value of features at the order of magnitude of cluster units Indicates the first The average time characteristic of each cluster unit For the first The number of samples in each cluster unit. For the first A sample set of cluster units, For the first Characteristics of the center of gravity shift in rainfall events. For the first The magnitude characteristics of the rainfall event. For the first The temporal characteristics of the rainfall.

10. The rainstorm flood forecasting method based on the similarity of rainstorm center of gravity migration characteristics according to claim 1, characterized in that: The calculation method for assigning the target rainstorm events to similar rainstorm cluster units is as follows: （16）， In the formula: For the corresponding rainstorm cluster unit, To find the position function of the maximum value, The target of the rainstorm events and the first The similarity of the center traces of each cluster unit.

Images