Rainstorm magnitude calculation method in combination with objective weather typing and relative deviation fuzzy matrix

Abstract

The invention discloses a rainstorm magnitude calculation method in combination with objective weather typing and a relative deviation fuzzy matrix. The method comprises the following steps: carrying out objective typing on historical atmosphere reanalysis data; carrying out clustering analysis on the typing result and the mode forecast by adopting a clustering algorithm, and judging similar types; historical ground precipitation data are divided into different rainstorm magnitudes, and a relative deviation fuzzy matrix comprehensive evaluation technology is utilized to obtain rainstorm forecasting schemes of all the magnitudes; constructing a rainstorm potential calculation operator of each magnitude, calculating the rainstorm probability of each magnitude, and taking the rainstorm magnitude with the maximum probability as output; and adopting an exponential method to carry out empty elimination correction on the output rainstorm magnitude according to a mode forecast result. According to the method, scheme and weight distribution is carried out on main environment variable parameters influencing a weather system through the relative deviation fuzzy matrix, graded rainstorm potential forecasting is achieved in combination with a rainstorm potential calculation operator, and technical support is provided for quantitative and refined rainstorm magnitude early warning.

Description

Technical Field

[0001] The present invention relates to the technical field of meteorological data processing, and more particularly to a method for calculating heavy rain magnitude by combining objective weather typing and relative deviation fuzzy matrix. Background Art

[0002] In recent years, with the development of objective forecasting technologies such as meteorological element objective forecasting based on numerical model interpretation application and ensemble forecasting post-processing, and classification severe convective weather potential forecasting based on deep learning models, the forecasting ability of heavy rain, severe convective weather and other weather has been effectively improved.

[0003] Currently, there are also many forecasting services and products dedicated to such weather. For example, the heavy rain recurrence period forecasting product developed in recent years can achieve extreme precipitation forecasting for the next 1-10 days. However, its quantification and refinement degree far cannot meet the requirements of forecasting services. For example, in order to avoid being affected by heavy rain as much as possible and not affect normal operation, the power transmission and distribution network needs to adopt different measures for prevention according to different heavy rain magnitudes. However, the current mainstream forecasting products can only provide large-scale heavy rain forecasts, which can neither determine the heavy rain magnitude nor be refined to the layout area of the power transmission and distribution network. Therefore, it is necessary to perform secondary development on the basis of the current heavy rain forecasting products to meet the requirements of quantification and refinement. Especially for some areas with frequent heavy rain and high intensity, the demand for this research is more urgent. Summary of the Invention

[0004] The present invention aims to solve at least one of the technical problems existing in the related technologies to some extent.

[0005] The purpose of the present invention is to provide a method for calculating heavy rain magnitude by combining objective weather typing and relative deviation fuzzy matrix. Based on high-resolution multi-source real-time information and multi-scale, multi-member, long-history numerical model data, a heavy rain potential calculation operator and an objective algorithm framework are constructed, so as to calculate the heavy rain magnitude and meet the requirements of quantitative and refined early warning.

[0006] In order to achieve the above purpose, the present invention provides a method for calculating heavy rain magnitude by combining objective weather typing and relative deviation fuzzy matrix, including:

[0007] S100. Obtain historical atmospheric reanalysis data and historical surface precipitation data;

[0008] S200. Use the rotated T-mode objective typing method to objectively type the 500hPa height field of the historical atmospheric reanalysis data;

[0009] S300. Use the clustering algorithm to perform clustering analysis on the typing results in step S200 and the 500hPa geopotential height field of the model forecast respectively. Then, use the root mean square error to calculate the difference between the model forecast and each objective typing type, and the type with the smallest difference is determined as the similar type;

[0010] S400. Under the similar type determined in step S300, divide the historical surface precipitation data into different rainstorm magnitudes, and use the relative deviation fuzzy matrix comprehensive evaluation technology to obtain the rainstorm forecast schemes for each magnitude respectively;

[0011] S500. Based on the rainstorm forecast schemes for each magnitude, construct the rainstorm potential calculation operators for each magnitude, calculate the rainstorm probabilities for each magnitude, and take the rainstorm magnitude with the maximum probability as the output;

[0012] S600. Use the exponential method to perform blanking correction on the output rainstorm magnitude based on the model forecast result.

[0013] A further preferred technical solution of the present invention is that step S300 specifically includes:

[0014] S310. Use the K-means clustering algorithm to perform clustering analysis on the typing results in step S200, and obtain the clusters and geopotential heights in each category to represent different weather systems;

[0015] S320. Use the K-means clustering algorithm again to perform clustering analysis on the 500hPa geopotential height field of the model forecast;

[0016] S330. Use the root mean square error to calculate the difference between the model forecast and the clusters and geopotential heights of each objective typing type, and the type with the smallest difference is determined as the similar type.

[0017] Preferably, the calculation formula for the difference in step S330 is:

[0018]

[0019] Where S pct represents the comprehensive difference measurement value between the model forecast and the typing result; x and y represent the longitude and latitude of the cluster, z represents the geopotential height of the members within the cluster, n and m are the total number of clusters obtained by the K-means clustering algorithm for the typing result and the model forecast respectively; x 1i is the longitude coordinate of the i-th cluster in the typing result, x 2j is the longitude coordinate of the j-th cluster in the model forecast, y 1i is the latitude coordinate of the i-th cluster in the typing result, y 2j is the latitude coordinate of the j-th cluster in the model forecast, z 1i is the geopotential height of the members within the i-th cluster in the typing result, z 2jis the geopotential height of the members within the j-th cluster in the pattern forecast; w1, w2, and w3 represent weights, which are obtained by taking the maximum values of (x 1i -x 2j ) 2 , (y 1i -y 2j ) 2 and (z 1i -z 2j ) 2 respectively, and then through normalization processing.

[0020] Preferably, in step S400, the comprehensive evaluation technology of the relative deviation fuzzy matrix is used to obtain the heavy rain forecast schemes for each magnitude. The specific method is as follows:

[0021] S410. Let U = {u1, u2,..., u n} be the set of n schemes to be evaluated, and V = {v1, v2,..., v m} be the set of evaluation factors, including benefit-type indicators and cost-type indicators;

[0022] S420. Measure each scheme in U with each factor in V to obtain a technical scheme matrix A, which is expressed as:

[0023]

[0024] where a ij represents the index value of the j-th evaluation factor in the i-th scheme;

[0025] Assume the ideal scheme is V α , which is expressed as:

[0026]

[0027] where max{a ij} is the benefit-type indicator, and min{a ij} is the cost-type indicator;

[0028] S430. Establish a relative deviation fuzzy matrix:

[0029]

[0030] where

[0031] S440. Adopt the coefficient of variation method to determine the normalized weight coefficient w j of each evaluation index, which is expressed as:

[0032]

[0033] Among them, is the mean value of the jth convective variable;

[0034] S450. Establish a comprehensive evaluation model, expressed as:

[0035]

[0036] F i represents the priority of each type of plan. If F t < F s , it means that the tth plan is superior to the sth plan. Correspondingly, select the weight coefficient of the tth plan and the preset plan.

[0037] Preferably, in step S410, the benefit-type index represents the obvious difference of the convective parameter relative to the climate mean state during rainstorms of different intensities. The higher the benefit-type index, the better the effect. The calculation formula of the benefit-type index is:

[0038]

[0039] The cost-type index represents the stability of the convective parameter itself. The smaller the cost-type index, the better the effect. The calculation formula of the cost-type index is:

[0040]

[0041] In the formula, I kj is the parameter value of a certain convective parameter of the jth historical case in a rainstorm of magnitude k; is the 15-day sliding climate average value of the convective parameter corresponding to the jth historical case; is the average value of the convective parameter corresponding to a rainstorm of magnitude k at the moment when p convective weather events are approaching.

[0042] Preferably, in step S400, the historical surface precipitation data is divided into 4 rainstorm magnitudes, which are: precipitation of 20 - 50 mm / h is defined as rainstorm, precipitation of 50 - 80 mm / h is defined as heavy rainstorm, precipitation of 80 - 100 mm / h is defined as extremely heavy rainstorm, and precipitation > 100 mm / h is defined as extreme rainstorm.

[0043] Preferably, the rainstorm potential calculation operator constructed in step S500 is expressed as:

[0044]

[0045] Among them, k represents the magnitude of the rainstorm, m represents the month, i is the convective parameter, i = 1, 2, 3,..., 20; p km represents the probability of a rainstorm of magnitude k occurring in the mth month; F kmi is the optimal plan for a rainstorm of magnitude k in the mth month, wkmi is the corresponding weight;

[0046] Preferably, calculate the probability of rainstorms of each magnitude in the m-th month, and use the maximum probability as the discriminant basis for deterministic forecasting, which is expressed as:

[0047] S p = max{p km |k = 1, 2, 3, 4};

[0048] where S p is the rainstorm magnitude of the maximum probability output.

[0049] Preferably, the specific method of step S600 is as follows:

[0050] S610. Obtain the model forecast rainfall;

[0051] S620. Introduce the exponential method, and perform blanking by gradually reducing the probability of rainfall within a certain range above and below the upper and lower limits of the output rainstorm magnitude, which is expressed as:

[0052]

[0053] where x is the model forecast rainfall and u is the upper and lower limits of the graded rainfall.

[0054] Beneficial effects: The present invention considers the influence of different synoptic circulation situations on rainstorms, and through the relative deviation fuzzy matrix, assigns schemes and weights to the main environmental variable parameters affecting the weather system, and combines the rainstorm potential calculation operator to realize the graded rainstorm potential forecasting, providing technical support for quantitative and refined rainstorm magnitude early warning. Brief Description of the Drawings

[0055] Figure 1 is the flowchart of the rainstorm magnitude calculation method of the present invention;

[0056] Figure 2 are 4 types of synoptic circulations obtained by objective weather typing in the embodiment;

[0057] Figure 3 is the spatial distribution map of the forecast rainstorm magnitude in the embodiment. Detailed Embodiments

[0058] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions in the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the present invention. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention, and they should not be construed as limiting the present invention. Based on the embodiments in the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of the present invention. In the description of the present invention, it should be understood that the terms used are only for the purpose of description and should not be construed as indicating or implying relative importance.

[0059] The following will Figures 1-3 describe the heavy rain magnitude calculation method provided by the present invention that combines objective weather typing and relative deviation fuzzy matrix.

[0060] Embodiment: To solve the problems of quantitative and refined heavy rain magnitude early warning, this embodiment will mainly take the potential prediction of extreme rainfall intensity in summer in the Yangtze-Huaihe region as the research object. Based on high-resolution multi-source real-time information and multi-scale, multi-member, long-history numerical model data, etc., physical quantity statistical analysis is respectively adopted, and through the analysis of convective parameters characterizing extreme rainfall intensity events, an objective algorithm that can reflect extreme events is developed, including constructing a potential calculation operator and an objective algorithm framework, etc., to improve the prediction ability of precipitation extremes. An extreme heavy rain classification potential early warning with a resolution of up to the kilometer level is realized.

[0061] The heavy rain magnitude calculation method provided by this embodiment that combines objective weather typing and relative deviation fuzzy matrix, as Figure 1 shown, the process is as follows:

[0062] S100. Obtain historical atmospheric reanalysis data and historical surface precipitation data;

[0063] S200. Use the rotational T-mode objective typing method to objectively type the 500hPa height field of the historical atmospheric reanalysis data;

[0064] S300. Adopt a clustering algorithm to respectively perform clustering analysis on the typing results in step S200 and the model forecast 500hPa height field, and then use the root mean square error to calculate the difference between the model forecast and each objective typing type, and the type with the smallest difference is determined as the similar type;

[0065] S400. Under the similar type determined in step S300, divide the historical surface precipitation data into different heavy rain magnitudes, and use the relative deviation fuzzy matrix comprehensive evaluation technology to respectively obtain the heavy rain forecast schemes for each magnitude;

[0066] S500. Based on the rainstorm forecasting schemes for each magnitude level, construct the rainstorm potential calculation operators for each magnitude level, calculate the rainstorm probabilities for each magnitude level, and take the rainstorm magnitude with the maximum probability as the output.

[0067] S600. Adopt the exponential method to perform blanking correction on the output rainstorm magnitude using the model forecast results.

[0068] Taking the European Centre for Medium-Range Weather Forecasts (ECMWF) ERA5 reanalysis data and the numerical weather forecasts of the Jiangsu Meteorological Bureau as examples, each step is described in detail.

[0069] In S200, the 500 hPa geopotential height field in the historical atmospheric reanalysis data of the East Asian region (100 - 140°E, 15 - 55°N) from 2000 to 2022 is objectively classified using the rotated T-mode objective classification method. The types of geopotential height fields after classification are 4, as Figure 2 shown, and are named T1 - T4 respectively.

[0070] In S300, the numerical model forecast of the 500 hPa geopotential height field is used as the input, and the K-means clustering algorithm is used to perform clustering analysis on the numerical model forecast and the objective classification results, and the type with the closest Euclidean distance is selected as the classification result. The specific steps are as follows:

[0071] S310. Use the K-means clustering algorithm to perform clustering analysis on the classification results to obtain the clusters and geopotential heights in each category, which are used to represent weather systems such as the subtropical high, trough, and typhoon.

[0072] S320. Again, use the K-means clustering algorithm to perform clustering analysis on the 500 hPa height field of the model forecast to obtain the clusters and centroids.

[0073] S330. Calculate the Euclidean distance between the numerical model forecast and the subsequent basic model, and the type with the smallest difference is determined as the similar type.

[0074] Among them, the Euclidean distance calculation formula is as formula (1):

[0075]

[0076] Among them, S pct represents the comprehensive difference measurement value between the model forecast and the classification result; x and y represent the longitude and latitude of the cluster, z represents the geopotential height of the members within the cluster, n and m are the total number of clusters obtained by the K-means clustering algorithm for the classification result and the model forecast respectively; x 1i is the longitude coordinate of the i-th cluster in the classification result, x 2j is the longitude coordinate of the j-th cluster in the model forecast, y 1i is the latitude coordinate of the i-th cluster in the classification result, y 2jis the latitude coordinate of the j-th cluster in the model forecast, z 1i is the geopotential height of the members within the i-th cluster in the classification result, z 2j is the geopotential height of the members within the j-th cluster in the model forecast; w1, w2, and w3 represent weights, which are obtained by taking the maximum of the corresponding (x 1i -x 2j ) 2 , (y 1i -y 2j ) 2 and (z 1i -z 2j ) 2 respectively, and then obtained through normalization processing.

[0077] In step S400, under the determined similarity type, the corresponding surface precipitation data in 2023 is divided into four levels of heavy rain (20 - 50 mm / h), very heavy rain (50 - 80 mm / h), extremely heavy rain (80 - 100 mm / h), and extreme heavy rain (> 100 mm / h). The relative deviation fuzzy matrix comprehensive evaluation technology is used to obtain different heavy rain forecast schemes for different levels. The specific steps are as follows:

[0078] S410. Let U = {u1, u2,..., u n} be a set of n schemes to be evaluated, where here are 20 convective parameter variables, as shown in Table 1 specifically; V = {c1, v2,..., v m} is the evaluation factor set, including benefit-type indicators and cost-type indicators. The higher the benefit-type indicator, the better the effect, representing that the convective parameters should show obvious differences relative to the climatological mean state during heavy rain events of different intensities, which is expressed by formula (2) here; the smaller the cost-type indicator, the better, and it is necessary to reflect the stability of the convective parameters themselves, which is expressed by the standard deviation formula (3) of each convective parameter.

[0079]

[0080] In the formula, I kj is the parameter value of a certain convective parameter of the j-th historical case in the k-level heavy rain; is the 15-day sliding climatological mean of the convective parameter corresponding to the j-th historical case; is the mean value of the convective parameter corresponding to the k-level heavy rain at the approaching moment of p convective weather events.

[0081] Table 1 Set of physical quantities of convective parameters

[0082]

[0083] S420. Each scheme in U is measured by each factor in V to obtain a technical scheme matrix A, which is expressed as:

[0084]

[0085] where a ij represents the index value of the j-th evaluation factor in the i-th solution;

[0086] Assume the ideal solution is V α , which is expressed as:

[0087]

[0088] where max{a ij} is a benefit-type index, and min{a ij} is a cost-type index;

[0089] S430. Establish a relative deviation fuzzy matrix:

[0090]

[0091] where:

[0092]

[0093] S440. Use the coefficient of variation method to determine the normalized weight coefficient w j of each evaluation index, which is expressed as:

[0094]

[0095] where is the mean value of the j-th convective variable;

[0096] S450. Establish a comprehensive evaluation model, which is expressed as:

[0097]

[0098] F i represents the priority of each type of solution. If F t < F s , it means that the t-th solution is better than the s-th solution. Correspondingly, select the weight coefficient of the t-th solution and the preset solution.

[0099] Step S500 is to construct a calculation operator for the heavy rain potential of each magnitude.

[0100] The constructed heavy rain potential calculation operator is expressed as:

[0101]

[0102] In formula (10), k represents the magnitude of heavy rain, m represents the month, i is the convective parameter, where i = 1, 2, 3, …, 20; p km represents the probability of heavy rain of level k occurring in the m-th month; F kmi is the optimal solution for heavy rain of level k in the m-th month, and w kmi is the corresponding weight. In this way, the probability forecasts of short-term heavy rains at all levels are obtained. Since the current operational forecasts still require deterministic forecasts, S p = max{p km |k = 1, 2, 3, 4}, that is, the maximum probability, is used as the discriminant basis for deterministic forecasts.

[0103] Step S600 uses the exponential method to eliminate the blank by gradually reducing the probability of rainfall within a certain range of the upper and lower limits of classified heavy rain.

[0104] The potential forecast completely depends on the design and construction of convective parameters and actually reflects the properties and state conditions of the ambient atmosphere at a certain moment. Favorable environmental conditions do not necessarily result in heavy rain. Therefore, it is very necessary to take appropriate blank elimination measures. The rainfall forecast of the model itself is the amount of atmospheric water vapor condensation and fall calculated through parameterization according to the evolution of cloud microphysics. The model precipitation can be used to constrain the potential forecast to a certain extent. The exponential method (e) is introduced to eliminate the blank by gradually reducing the probability of rainfall within a certain range exceeding the upper and lower limits of the established classified heavy rain. The specific method is as follows:

[0105] S610. Obtain the model forecast rainfall;

[0106] S620. Introduce the exponential method to eliminate the blank by gradually reducing the probability of rainfall within a certain range exceeding the upper and lower limits of the output heavy rain magnitude, which is expressed as:

[0107]

[0108] where x is the model forecast rainfall and u is the upper and lower limits of classified rainfall.

[0109] For example, for rainfall with an intensity of 20 - 50 mm / h, the lower limit is 20 mm and the upper limit is 50 mm. When the model forecast rainfall is outside the upper and lower limits, formula (11) is used to constrain and reduce the rainfall forecast. After moderately constraining the original probability forecast with the model precipitation, the range of the probability forecast relative to the large value forecast is significantly reduced, and the false alarm range is significantly decreased.

[0110] After being processed by the above solution, the spatial distribution of the forecast heavy rain magnitude in the Jianghuai region is as Figure 3 shown, achieving quantitative and refined heavy rain early warning with a resolution of up to the kilometer level.

[0111] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for estimating rainstorm magnitude by combining objective weather classification and relative deviation fuzzy matrix, characterized in that: include: S100, obtain historical atmospheric reanalysis data and historical surface precipitation data; S200, using the rotating T mode objective classification method, objectively classify the 500hPa height field of historical atmospheric reanalysis data; S300, using a clustering algorithm to perform cluster analysis on the classification results of step S200 and the model forecast 500hPa height field, and then using the root mean square error to calculate the difference between the model forecast and each objective classification type, and the type with the smallest difference is determined as a similar type; S400, under the similarity type determined in step S300, the historical ground precipitation data are divided into different rainstorm levels, and the rainstorm forecast schemes of each level are obtained respectively by using the relative deviation fuzzy matrix comprehensive evaluation technology; S500, based on the rainstorm forecasting schemes of various magnitudes, constructing rainstorm potential calculation operators of various magnitudes, calculating the probability of rainstorms of various magnitudes, and taking the rainstorm magnitude with the maximum probability as output; S600, using the exponential method, the output rainstorm magnitude is corrected based on the model forecast results.

2. The method for calculating the rainstorm magnitude by combining objective weather classification and relative deviation fuzzy matrix according to claim 1 is characterized in that: Step S300 specifically includes: S310, using K-means clustering algorithm to perform cluster analysis on the classification results of step S200, and obtain clusters and potential heights in each category to represent different weather systems; S320, again using the K-means clustering algorithm to perform cluster analysis on the model forecast 500hPa height field; S330. Calculate the difference in cluster and potential height between the model prediction and each objective classification type using the root mean square error. The type with the smallest difference is determined as a similar type.

3. The method for calculating rainstorm magnitude by combining objective weather classification and relative deviation fuzzy matrix according to claim 1 is characterized in that: The calculation formula of the difference in step S330 is: Among them, S pct represents the comprehensive difference measure between the model prediction and the typing result; x and y represent the longitude and latitude of the cluster, z represents the potential height of the members in the cluster, n and m are the total number of clusters obtained by the typing result and the model prediction through the K-means clustering algorithm, respectively; x 1i is the longitude coordinate of the i-th cluster in the typing results, x 2j is the longitude coordinate of the jth cluster in the model forecast, y 1i is the latitude coordinate of the i-th cluster in the typing results, y 2j is the latitude coordinate of the jth cluster in the model forecast, z 1i is the potential height of the members in the i-th cluster in the typing results, z 2j is the potential height of the members in the jth cluster in the model forecast; w1, w2 and w3 represent weights, which are corresponding to (x 1i -x 2j ) 2 ,(y 1i -y 2j ) 2 and (z 1i -z 2j ) 2 Take the maximum value respectively and then normalize it.

4. The method for calculating rainstorm magnitude by combining objective weather classification and relative deviation fuzzy matrix according to claim 1, characterized in that: In step S400, the relative deviation fuzzy matrix comprehensive evaluation technology is used to obtain the rainstorm forecast schemes of various magnitudes respectively. The specific method is as follows: S410, let U = {u1, u2, ..., u n } is a set of n solutions to be evaluated, V = {v1,v2,…,v m } is a set of evaluation factors, including benefit-based indicators and cost-based indicators; S420, measure each solution in U with each factor in V to obtain a technical solution matrix A, expressed as: Among them, a ij represents the index value of the jth evaluation factor in the i-th scheme; Assuming the ideal solution is V α , expressed as: in, max{a ij } is a benefit indicator, min{a ij } is a cost indicator; S430, establish a relative deviation fuzzy matrix: in, S440, using the coefficient of variation method, determine the normalized weight coefficient w of each evaluation index j , expressed as: in, is the mean value of the jth convection variable; S450, establish a comprehensive evaluation model, expressed as: F i Indicates the priority of various solutions. If F t <F s , indicating that the t-th solution is better than the s-th solution. Correspondingly, the weight coefficient of the t-th solution and the pre-set solution are selected.

5. The method for calculating the rainstorm magnitude by combining objective weather classification and relative deviation fuzzy matrix according to claim 4 is characterized in that: In step S410, the benefit index represents the obvious difference of the convective parameters relative to the climate average state when heavy rains of different intensities occur. The higher the benefit index, the better the effect. The calculation formula of the benefit index is: The cost index represents the stability of the convection parameters themselves. The smaller the cost index, the better the effect. The cost index calculation formula is: In the formula, I kj is the parameter value of a convective parameter in the j-th historical case in the k-level rainstorm; is the 15-day sliding climate mean of the convective parameters corresponding to the jth historical case; It is the average value of the convective parameters corresponding to the k-level rainstorm at the time when the p-th convective weather event is about to occur.

6. The method for estimating rainstorm magnitude by combining objective weather classification and relative deviation fuzzy matrix according to claim 1, characterized in that: In step S400, the historical ground precipitation data is divided into four rainstorm levels, namely: precipitation of 20-50 mm / h is defined as rainstorm, precipitation of 50-80 mm / h is defined as heavy rainstorm, precipitation of 80-100 mm / h is defined as extremely heavy rainstorm, and precipitation >100 mm / h is defined as extreme rainstorm.

7. The method for estimating rainstorm magnitude by combining objective weather classification and relative deviation fuzzy matrix according to claim 1, characterized in that: The rainstorm potential calculation operator constructed in step S500 is expressed as: Where k represents the magnitude of the rainstorm, m represents the month, i is the convection parameter, i = 1, 2, 3, ..., 20; p km F represents the probability of a k-level rainstorm in the mth month; kmi is the optimal solution for k-level rainstorm in the mth month, w kmi The corresponding weight 。 8. The method for estimating rainstorm magnitude by combining objective weather classification and relative deviation fuzzy matrix according to claim 7 is characterized in that: Calculate the probability of heavy rain of various magnitudes in the mth month, and use the magnitude of heavy rain with the greatest probability as the basis for determining the deterministic forecast, expressed as: S p =max{p km |k=1,2,3,4}; Among them, S p The maximum probability of the output rainstorm magnitude.

9. The method for estimating rainstorm magnitude by combining objective weather classification and relative deviation fuzzy matrix according to claim 1, characterized in that: The specific method of step S600 is: S610, obtaining the model forecast rainfall; S620, introduce the exponential method to gradually reduce the probability of rainfall exceeding a certain range of the upper and lower limits of the output rainstorm level to eliminate the void, which is expressed as: Among them, x is the model forecast rainfall, and u is the upper and lower limits of the graded rainfall.

Images