A method for classification and type identification of rainstorm floods in extremely large basins

Through classification methods and fuzzy identification theory based on rainfall and flood-related indicators, the problem of identifying heavy rainfall flood types in the super-large basin is solved, and the efficiency of flood resource utilization and the accuracy of reservoir scheduling is improved.

CN118797418BActive Publication Date: 2024-11-15DALIAN UNIV OF TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411276908.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-12
Publication Date
2024-11-15
Estimated Expiration
2044-09-12

AI Technical Summary

Technical Problem

The prior art is difficult to effectively classify and identify the types of heavy rainfall floods in extra-large basins, especially in real-time scheduling, which leads to low efficiency in flood resource utilization, and the existing methods have insufficient classification accuracy for flood processes affected by complex geographical areas and multiple factors.

Method used

The process flood classification method based on rainfall and flood-related indicators is adopted, and the proportion of sub-flood total and sub-storage total are classified. Combined with system clustering method and fuzzy identification theory, decision-making rules are generated to achieve dynamic identification of flood types. and real-time scheduling decisions.

Benefits of technology

Accurate and early identification of the types of heavy rainfall floods in the extreme basin has been achieved, the decision-making basis for reservoir scheduling has been improved, the efficiency of flood resource utilization has been enhanced, and real-time and high accuracy has been achieved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118797418B_ABST
    Figure CN118797418B_ABST
Patent Text Reader

Abstract

The present invention provides a method for classifying and identifying rainstorm floods in a particularly large river basin, and belongs to the technical field of flood classification and identification. The identification method first determines and calculates rainfall and flood-related indicators for classification; secondly, classifies floods based on flood classification indicators and analyzes the morphology, rate, and intensity characteristics of various flood processes; thirdly, classifies floods based on rainfall indicators and verifies the feasibility and accuracy of classification using rainfall information; finally, generates decision rules based on rainfall information according to the flood classification results and dynamically identifies flood types in three stages. The present invention uses the proportion of the total amount of secondary floods and the proportion of the total amount of secondary rainfall for classification, and uses forecast rainfall and ground rain for dynamic identification of flood types. Therefore, the identification method can segmentally identify types during the occurrence and development of floods, has the advantages of real-time performance, can make judgments in advance, and assist in early decision-making.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of flood classification and identification, and relates to a method for classifying and identifying flood events by utilizing the spatial composition of floods, and in particular to a method for classifying and identifying rainstorm floods in a particularly large river basin. Background Art

[0002] As a complex natural phenomenon, floods are both deterministic and random. The frequency, intensity and impact range of floods may change with climate change and human activities. The development process and impact range of different types of floods are different (Turkington T., Breinl K., Ettema J., et al. A new flood type classification method for use in climate change impact studies [J]. Weather and Climate Extremes, 2016, 14 (C), 1-16.), which affects the reservoir dispatching mode and restricts the efficiency of flood resource utilization. Classifying flood types according to certain principles and identifying and classifying them in real time during the regulation process is an effective way to understand the laws of basin floods and alleviate the contradiction between flood control and benefit development. It can effectively improve the potential of flood resource utilization while ensuring that flood control standards are not lowered.

[0003] At present, flood classification research is mainly concentrated in small and medium-sized river basins. More research is based on the characteristic information after the flood, such as flood volume, flood peak, flood duration, etc., based on the natural classification standards between the common points and differences of flood events, to divide the flood process into multiple types, and then apply it to classification forecasting to improve the forecast accuracy. Most of these studies take the entire flood process of small and medium-sized river basins as the research object, and divide it by similarity based on process indicators. There are few studies on flood classification in large river basins and real-time identification of flood types during flood development. Large river basins cover a vast geographical area, with large basin area, rich terrain types, multiple rainstorm weather systems, and complex main and tributary river systems. Floods are affected by multiple complex factors such as terrain characteristics, reservoir scheduling, and runoff generation and convergence mechanisms. These factors interact with each other, making the formation and development of floods highly complex, and making the flow process of main and tributary river systems at the same outlet section significantly different. Flood classification only relies on flood process indicators, and the similarity between floods is low, which will produce a large number of discrete points. However, the overall pattern of flood composition at the same outlet section is similar. Classification based on the differences in flood composition of each main and tributary river can effectively divide flood types and distinguish the main water areas.

[0004] There are three main methods for flood classification, namely, cause analysis, intelligent optimization and cluster analysis. The cause analysis method starts from the physical causes of floods, studies its related laws, and classifies floods with similar characteristics into one category. It can decompose complex problems into smaller parts and has the advantage of strong intuitiveness. However, cause analysis needs to rely on the experience and judgment of experts and has a certain degree of subjectivity. Intelligent optimization algorithms include artificial neural network algorithms, artificial bee colony algorithms, hybrid frog leaping algorithms, etc., which solve the robustness problem of the flood classification process and have high efficiency and global search capabilities (Chen Chunmei, Wang Zhengyuan, Qu Na. A review of artificial intelligence optimization algorithms [J]. University, 2020, (26): 71-72.). They are often used by scholars to optimize algorithms to improve classification accuracy, but it is necessary to select appropriate intelligent optimization algorithms according to the characteristics of specific problems. If the selection is inappropriate, it may lead to poor results. Cluster analysis is based on statistical theory, applies ordinary mathematical methods or fuzzy mathematical methods, and uses the characteristics of the data itself for analysis. It can find the structure and pattern in the data for clustering without prior knowledge. The clustering algorithm is simple, easy to understand, and highly operable. The algorithm is sensitive to initial parameters and data outliers, which can be improved through algorithm optimization and data review. Liang Yusheng et al. (Liang Yusheng, Yang Xiaofang, Hu Dan. Research and Application of Fuzzy Cluster Analysis in Hydrological Forecasting [J]. Mathematical Learning and Research, 2018 (19), 149.) used the ISODATA clustering algorithm to determine the number of clusters in advance when using the weighted fuzzy C-means clustering algorithm for flood classification in the Cao'e River Basin. Therefore, clustering algorithms have gradually become an effective mainstream method in the field of flood classification. Dinesh (Dinesh S.. Fuzzy classification of simulated droughts and floods of water bodies [J]. Journal of Applied Sciences, 2007, 7 (18): 2610-2616.) classified historical flood processes in multiple basins through fuzzy clustering, proving the good applicability of this method in different basins. At present, flood classification mainly relies on system clustering, k-means clustering and fuzzy clustering.

[0005] Flood identification is to retrieve historical flood processes or flood types similar to the current flood by obtaining relevant rainfall, flood and other information, combined with the existing long-term historical flood data in the basin. Current flood identification research focuses on using characteristic information after the flood occurs for identification, and there is less research on how to identify flood types in advance. Heavy rain floods refer to flood phenomena caused by heavy rainfall in a short period of time. The flood formation process is closely related to the formation and development of high-intensity rainfall. Therefore, the main cause of the flood (i.e., the amount of heavy rainfall) is used as a priori identification indicator to identify the type of flood that is about to occur in advance. It can predict the relevant characteristics of future floods based on the information provided by similar floods in history, or refer to the treatment plans of similar floods in history to build treatment plans for future floods.

[0006] The following is an explanation of the research fields of biometric smart card modules and data prediction:

[0007] In the research field of biometric smart card modules, existing modules usually combine contact and contactless communication interfaces. They are not only used for identity authentication, but also support multiple functions such as financial transactions and access control. The design realizes the integration of multiple smart card functions and can meet the needs of multiple applications. With the advancement of technology and the emphasis on user experience, biometric modules are increasingly integrated into smart cards. At the same time, the thickness and volume of the modules are designed to be smaller, which reduces the size and complexity of the modules, improves the convenience of installation, and ensures user convenience. With the development of society and the growing demand for electronic payment and identity authentication, the market demand for biometric smart cards will continue to rise, and research in this direction will develop towards higher integration, stronger security and better user experience.

[0008] In the field of data prediction, especially construction data prediction, machine learning algorithms (such as support vector machines, neural networks, etc.) are widely used in data prediction. These algorithms can handle complex nonlinear relationships and extract useful information from large amounts of historical data. Many prediction models can now receive and process new data in real time, thereby dynamically adjusting the prediction results and improving the timeliness and accuracy of the prediction. In order to further improve the accuracy of predictions, researchers have explored combining multiple algorithms to form hybrid models for use. This type of research can improve the generalization ability of the model while optimizing model parameters. It is worth noting that data quality directly affects the performance of the prediction model. Therefore, data preprocessing (such as denoising, filling missing values, standardization, etc.) is particularly important in the entire prediction process. Summary of the invention

[0009] In view of the shortcomings of the prior art, the present invention provides a method for classifying and dynamically identifying typical floods. Based on the composition characteristics of flood events, the method constructs flood classification methods with rainfall and flood as classification indicators, verifies the possibility and accuracy of using rainfall information to identify flood types, and finally uses rainfall information to dynamically identify flood types in real-time scheduling operations. The classification and identification method fully considers the problem of large differences in the composition of flood areas, can determine flood types in advance, and provide a decision-making basis for reservoir scheduling decisions.

[0010] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0011] A method for classifying and identifying rainstorm floods in a particularly large river basin. The identification method first determines and calculates rainfall and flood-related indicators for classification; secondly, classifies floods based on flood classification indicators and analyzes the morphology, rate, and intensity characteristics of various flood processes; thirdly, classifies floods based on rainfall indicators and verifies the feasibility and accuracy of using rainfall information for classification; finally, generates decision rules based on rainfall information based on flood classification results and dynamically identifies flood types in three stages. The schematic diagram of flood classification and identification process is attached. Figure 1 As shown. The following steps are included:

[0012] In the first step, the rainfall and flood-related indicators used in the classification were determined and calculated;

[0013] In view of the large area, many tributaries and intervals, and complex water flow process of the super-large basin, the flood composition is used as the classification index for classification. The classification index described in the present invention includes two types: flood classification index and rainfall classification index.

[0014] The proportion of total secondary flood volume is used as the flood classification index. The calculation of the flood classification index is shown in formula (1):

[0015] ; (1)

[0016] in, It is The basin in The proportion of floods in the total flood volume, It is Sub-basin, It is a flood; It is The sub-basin The amount of flood water flowing into the control section during the flood; It is The total flood volume at the flood outlet section.

[0017] The proportion of total rainfall is used as the rainfall classification index. The calculation of the rainfall classification index is shown in formula (2):

[0018] ; (2)

[0019] in, It is The partition is in The proportion of total rainfall in flood events, It is Partitions; It is The partition is in Total rainfall depth of flooding; It is The drainage area of ​​the subdivision.

[0020] Among them, the classification process using the two indicators of the proportion of secondary floods and the proportion of secondary rainfall is independent of each other.

[0021] The second step is to classify floods based on the proportion of total secondary floods;

[0022] Step 2.1) The present invention uses flood events as samples, adopts a systematic clustering method, and uses the principle of the shortest distance between samples to classify the flood events into types;

[0023] First, in step 2.1), it involves determining the number of classifications to clarify the optimal number of classifications. The present invention uses the AIC (Akaike Information) criterion to determine how many types of floods should be divided into most reasonably. The AIC index calculation is shown in formula (3):

[0024] ; (3)

[0025] Among them, AIC is an index for judging the optimal number of classifications, and the minimum AIC value corresponds to the optimal number of classifications; is the number of types obtained through systematic clustering; is the maximum value of the log-likelihood function of the classification result; L is the likelihood function.

[0026] Furthermore, the process of classifying flood types based on the system clustering method in step 2.1) is as follows:

[0027] (2.1.1) The Euclidean distance is used to calculate the distance between each flood event. The calculation is shown in formula (4):

[0028] d jj ' = [ ∑ k = 1 b ( x j k − x j ' k ) 2 ] 1 / 2 , ( j = 1 , 2 , . . . c , j ' = 1 , 2 , ⋯ , c ) ; (4)

[0029] in, It is , the Euclidean distance between two floods; It is The first flood indicators; It is The first flood indicators; is the total number of flood classification indicators.

[0030] The distance matrix is ​​obtained by calculating the Euclidean distance between flood events. :

[0031] ; (5)

[0032] (2.1.2) The distance between various types of floods is calculated using the class average distance. The calculation is shown in formula (6) and formula (7):

[0033] ; (6)

[0034] Among them, the definition It is Flood-like It is Flood-like; is the average of the two types of flood distances; and They are and Class Number of flood samples in ; Representation Class The samples in Representation Class Samples in; Definition For any type of flood, It is a new flood type after the two types of floods merge; and Merge into When any type of flood and The distance is:

[0035] ; (7)

[0036] in, yes and distance; yes and distance; Is a class The number of flood samples in .

[0037] (2.1.3) According to the distance between any two floods in the entire basin calculated by formula (5), the two floods with the smallest distance are merged into one type of flood; according to the distance between different types of floods calculated by formula (6), the two types of floods with the closest distance are merged into a new flood type. ; Then calculate any type of flood according to the public notice (7) With new flood types Then return to step (2.1.2) and repeat this process until all floods are merged into the number of types corresponding to the best AIC index.

[0038] Step 2.2) According to the flood classification index (the proportion of total flood volume) determined in the first step and the classification method in step 2.1), each flood is classified into four types of floods. The four types of floods are specifically the Mintuojiang River and Jialingjiang River encounter type, the Wujiang River and the main stream encounter type, the Jialing River type, and the Jialing River and Cunsan interval encounter type.

[0039] Step 2.3) Based on the flood classification results obtained in step 2.2), the characteristics of the four types of floods are analyzed. Different types of floods differ in morphology, flow process, etc. Therefore, the morphological index, rate index, and intensity index of each type of flood are selected to analyze the characteristics of each type of flood.

[0040] Furthermore, the specific indicators and their calculations included in the morphological index, rate index, and strength index in step 2.3) are as follows:

[0041] The morphological indicators include peak time skewness, flow variation coefficient, and high pulse duration ratio, which are calculated as shown in formula (8)-formula (10).

[0042] ; (8)

[0043] in, is the peak time skewness, It is the time of flood. It is the total duration of the flood. It reflects the relative time of the flood peak in each flood. The smaller the value, the earlier the flood peak occurs.

[0044] CV = σ [ Q ( t ) ] Q av ; (9)

[0045] in, is the coefficient of variation of flow rate, It is a flood process. σ [ Q ( t ) ] This is the standard deviation of the flood process, is the average flow rate of the flood process. The larger the value, the steeper and sharper the flood process.

[0046] ; (10)

[0047] in, is the high pulse duration ratio, It is the duration of the flood process with flow value greater than 75% of the peak flow value. The larger the value, the wider the flood peak shape and the longer the high water volume flood process lasts.

[0048] The rate indicators include flood rate and flood fall rate, which are calculated as shown in formula (11) and formula (12):

[0049] ; (11)

[0050] in, is the flood rate, and are the secondary flood peak flow and the rising flow, It's the time of flood. It is the time when the flood peak appears. The larger the value, the faster the flood rises, and it is easy to form a catastrophic flood in a short period of time.

[0051] ; (12)

[0052] in, is the flood rate, This flood terminated the traffic. It is the time when the flood ends.

[0053] The intensity characteristic indicators include total flood volume and peak flood flow.

[0054] For each type of flood, three morphological indices, two rate indices, and two intensity indices are obtained through formulas (8)-(12).

[0055] Step 2.4) Compare the seven indicators of each type of flood obtained in step 2.3) with the characteristic description of historical floods. If the typical floods of different types are consistent with the historical characteristic descriptions, the classification is considered reasonable and the classification result is accurate. If the typical floods of different types are inconsistent with the historical characteristic descriptions, the classification is considered inconsistent with the actual situation and the flood classification result is not suitable for subsequent flood type identification work. Return to the first step to adjust the flood classification indicators, classification methods, etc. and re-divide the flood types.

[0056] The third step is to classify floods based on the proportion of total rainfall;

[0057] Step 3.1) The present invention uses the rainfall events corresponding to the flood events as samples, adopts the system clustering method, and uses the principle of the shortest distance between samples to classify the rainfall events into types; the process of classifying the flood events into types based on the system clustering method in step 3.1) is consistent with step 2.1);

[0058] Step 3.2) According to the rainfall classification index (percentage of total rainfall) determined in the first step and the classification method in step 3.1), the rainfall corresponding to each flood is classified into four types of floods. The four types of floods are specifically the Mintuojiang River and Jialingjiang River encounter type, the Wujiang River and main stream encounter type, the Jialing River type, and the Jialing River and Cunsan interval encounter type.

[0059] The fourth step is to verify the feasibility of the flood classification results in the second and third steps;

[0060] If the accurate classification result based on the proportion of secondary floods in the second step overlaps with the classification result based on the proportion of secondary rainfall in the third step by 85% or more, the classification based on rainfall indicators (proportion of secondary rainfall total amount) in the third step is considered feasible, and the flood type identification work in the fifth step should be continued; otherwise, it is considered that the proportion of secondary rainfall total amount cannot be used as a discriminant indicator for flood classification and identification, and the indicator should be reselected, the classification should be reclassified, and the feasibility verification should be reconstructed.

[0061] Step 5: Determine the method used for flood type identification;

[0062] When identifying flood types, there are many indicators and the boundaries between different types of floods are unclear. Fuzzy logic provides an effective tool for dealing with uncertainty and fuzziness, which is suitable for dealing with data with different types of characteristics in flood events. The present invention adopts the rough set theory in fuzzy identification, mines the implicit relationship between classification indicators and flood types, determines the decision rules for identifying flood types, and identifies flood types in three stages.

[0063] The process of applying rough set theory to identify flood types is as follows:

[0064] Step 5.1) Attribute selection and parsimony;

[0065] The attribute reduction method of rough set theory is used to calculate the attribute importance of rainfall classification indicators and remove redundant attributes. The core attributes of rainfall classification indicators are obtained, and the most decisive proportion of total rainfall for different flood types is extracted. The greater the importance, the stronger the ability of the total rainfall proportion for classification.

[0066] Step 5.2) Determine the lower approximation and the upper approximation;

[0067] Based on the proportion of the most decisive rainfall total extracted in step 5.1), the lower approximation and upper approximation of each flood type are calculated to determine the typical characteristics and boundaries of each flood type. The lower approximation refers to the minimum set of samples that can be classified as a certain flood type; the upper approximation refers to the maximum set after excluding all samples that cannot be classified as a certain flood type.

[0068] Step 5.3) Determine the decision rule;

[0069] Based on the lower and upper approximations determined in step 5.2), a decision rule for identifying flood types is derived.

[0070] Step 5.4) Application and implementation;

[0071] Based on the decision rules determined in step 5.3), the flood type corresponding to the actual flood process is identified in three stages.

[0072] Further, the three stages of flood identification described in step 5.4) are as follows:

[0073] Stage 1: No rainfall has occurred yet, and flood type identification is fully based on the forecast rainfall information;

[0074] Phase 2: As rainfall continues to develop, since there are no less than two main rainfall areas for each type of flood, when there are no more than two sub-basins with high-intensity rainfall areas in the early stage, the flood type is identified by combining the measured rainfall and subsequent rainfall forecast information (the forecast product used in this patent is ECMWF);

[0075] Stage 3: High-intensity rainfall has occurred in various regions, and the rainfall intensity has decreased. At this time, the flood is identified by combining the measured rainfall and forecast rainfall information. The time interval of this node varies for different floods.

[0076] Step 6: Dynamic identification of flood types based on rainfall information;

[0077] According to the decision rules for flood type identification determined in the fifth step, combined with the forecast rainfall information (the forecast information used in this patent is the forecast rainfall data of the European Centre for Medium-Range Weather Forecasts (ECMWF)) and the falling rain information, the flood type to which the flood process belongs is dynamically identified.

[0078] Compared with the prior art, the present invention has the following advantages and effects:

[0079] The present invention provides a classification method based on composition, determines the decision rules for identification, and proposes a method for dynamically identifying flood types using rainfall information. According to the above process, the method uses the proportion of the total amount of secondary floods and the proportion of the total amount of secondary rainfall for classification, and uses the forecast rainfall and the rainfall to dynamically identify the flood type. Therefore, the identification method can segmentally identify the type during the occurrence and development of the flood, has the advantage of real-time, can make judgments in advance, and assist in early decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] Figure 1 This is a schematic diagram of the flood classification and identification process;

[0081] Figure 2 This is a comparison chart of the classification results based on rainfall composition and the classification results based on flood composition;

[0082] Figure 3 This is a diagram showing the dynamic identification results of four types of floods in different stages; Figure 3 (a) is the result map of the identification of the encounter flood in the Mintuo River and Jialing River; Figure 3 (b) is the result diagram of the identification of the type of flood encountered in the Wujiang River and its mainstream. DETAILED DESCRIPTION

[0083] The present invention is further described below in conjunction with specific embodiments.

[0084] The present invention takes the upper and middle reaches of the Hun Yangtze River as an example, with a time range of 1981, 1982, 1998, 1999 and 2014-2021, with a total of 43 floods (n=43). The specific implementation method is described in detail in combination with the technical scheme and the accompanying drawings, which specifically includes the following steps:

[0085] In the first step, the rainfall and flood-related indicators used in the classification were determined and calculated;

[0086] The present invention is based on the flood information of the event, and determines the flood index used for classification as the proportion of the total flood volume. , where m=7, n=43; the rainfall index used to determine the classification is the proportion of the total rainfall , where t=15.

[0087] The second step is to classify floods based on the proportion of total secondary floods;

[0088] Based on the system clustering method and the determined flood classification index (the proportion of the total flood volume in the 7 sub-basins), the 43 floods were classified into different types. The optimal number of categories was determined using the AIC criterion. The different numbers of flood types and the corresponding AIC index are shown in Table 1. The number of types corresponding to the minimum AIC value is taken as the optimal number of categories, that is, 4 categories.

[0089] Table 1. Number of different flood types and AIC index.

[0090]

[0091] The four types of floods are specifically the type encountered by the Min-Tuo River and Jialing River (type 1), the type encountered by the Wujiang River and the main stream (type 2), the Jialing River (type 3), and the type encountered by the Jialing River and the Cunsan interval (type 4). The results of the flood type classification and the proportion of the total amount of secondary floods in the basin corresponding to the four flood types are shown in Table 2.

[0092] Table 2. Average proportion of total secondary floods in the basin for each type of flood.

[0093]

[0094] Then, based on the flood classification results, the characteristic indicators of the four types of floods, such as morphology, rate, and intensity, are calculated. Among them, the morphological indicators include the peak time skewness , Flow rate variation coefficient , High pulse duration ratio , the rate index includes flood rate , flood rate The intensity characteristic indicators include the total amount of flood and peak flow The calculation results of characteristic indicators corresponding to various types of floods are shown in Table 3.

[0095] Table 3. Mean values ​​of characteristic indicators of various types of floods.

[0096]

[0097] The third step is to classify floods based on the proportion of total rainfall;

[0098] Based on the system clustering method and the determined rainfall classification index (the proportion of total rainfall in 15 sub-areas), the rainfall events corresponding to the 43 floods were classified into four types of floods. The four types of floods are specifically the Mintuojiang River and Jialingjiang River encounter type, the Wujiang River and main stream encounter type, the Jialing River type, and the Jialing River and Cunsan interval encounter type.

[0099] The fourth step is to verify the feasibility of the flood classification results in the second and third steps;

[0100] Compare the classification results based on the proportion of secondary floods in the second step with the classification results based on the proportion of secondary rainfall in the third step. If the overlap of the two classification results reaches 85% or above, it is considered that the classification based on rainfall indicators (proportion of secondary rainfall) in the third step is feasible, and the flood type identification work in the fifth step should be continued.

[0101] The results are as follows Figure 2 As shown, the results show that the consistency between the classification results based on the proportion of the total amount of secondary floods and the classification results based on the proportion of the total amount of secondary rainfall in the third step is as high as 91%, which means that the classification based on the proportion of the total amount of secondary rainfall is feasible.

[0102] Step 5: Determine the method used for flood type identification;

[0103] The rough set theory in fuzzy identification is used to explore the implicit relationship between classification indicators and flood types, determine the decision rules for identifying flood types, and identify flood types in three stages. The order of attribute importance of the proportion of total rainfall for decision rules is shown in Table (4); taking the encounter type of Mintuojiang River and Jialingjiang River and the encounter type of Wujiang River and main stream as examples, the decision rules used for flood identification are shown in Table (5)-Table (6).

[0104] Table 4. Importance of the proportion of total rainfall in each sub-region.

[0105]

[0106] The importance of the proportion of the total amount of rainfall in the upper reaches, middle reaches and lower reaches of the Jinsha River is 0, which is a redundant attribute in type identification, indicating that the ability to identify flood types has not decreased after removing these three indicators.

[0107] Table 5. Encounter type identification rules for the Min-Tuo River and Jialing River

[0108]

[0109] Table 6. Encounter type identification rules for Wujiang River and mainstream sections

[0110]

[0111] Step 6: Dynamic identification of flood types based on rainfall information

[0112] According to the decision rules used in the fifth step, the proportion of total rainfall is used to dynamically identify various types of floods in three stages. The results of the staged identification of various types of floods are shown in Figure 2. Figure 3 shown.

[0113] from Figure 3 As can be seen from (a), for the encounter flood process of the Min-Tuo River and Jialing River, the recognition accuracy in stage 1 and stage 2 is as high as 90.91%, and the recognition accuracy in stage 3 can reach 100%; Figure 3As can be seen in (b), for the encounter-type flood process in the Wujiang River and its main stream, the recognition accuracy in stage 2 is low, but it recovers in stage 3, so it does not affect the overall recognition effect. Based on the flood type recognition results, it can be concluded that as the flood process progresses, the cumulative rainfall from stage 1 to stage 3 increases, the proportion of measured rainfall increases, and the recognition results tend to be more accurate.

[0114] The above-described embodiments merely express the implementation methods of the present invention, but they cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that for those skilled in the art, several modifications and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention.

Claims

1. A method for classifying and identifying rainstorm floods in a particularly large river basin, characterized in that: The identification method first determines and calculates rainfall and flood-related indicators for classification; secondly, classifies flood events based on flood classification indicators and analyzes the morphology, rate, and intensity characteristics of various flood processes; thirdly, classifies flood events based on rainfall indicators and verifies the feasibility and accuracy of classification using rainfall information; finally, generates decision rules based on rainfall information according to flood event classification results and dynamically identifies flood types in three stages; the method includes the following steps: In the first step, the rainfall and flood-related indicators used in the classification were determined and calculated; Based on the characteristics of the extra-large basin, the flood composition is used as a classification index for classification; the classification index includes two types of flood classification index and rainfall classification index, and the classification process of the two indicators is independent of each other; among them, the proportion of the total amount of secondary floods is used as the flood classification index, and the proportion of the total amount of secondary rainfall is used as the rainfall classification index; The second step is to classify floods based on the proportion of total secondary floods; Step 2.1: Taking flood events as samples, the systematic clustering method is used to classify flood events according to the principle of the shortest distance between samples. The Akaike information AIC criterion is used to determine the optimal number of flood classifications, as shown in formula (3): AIC=2l-2ln(L) (3) Among them, AIC is the index for judging the optimal number of classifications, and the minimum AIC value corresponds to the optimal number of classifications; l is the number of types obtained by systematic clustering; ln(L) is the maximum value of the log-likelihood function of the classification result; L is the likelihood function; Step 2.2: Based on the proportion of the total flood volume and the classification method in step 2.1, each flood is classified into four types of floods; Step 2.3: Based on the flood classification results obtained in step 2.2, the characteristics of the four types of floods are analyzed, and the morphological index, rate index, and intensity index of each type of flood are selected to analyze the characteristics of each type of flood; Step 2.4: Compare the seven indicators of each type of flood obtained in step 2.3 with the characteristic description of historical floods. If the typical floods of different types are consistent with the historical characteristic descriptions, the classification is considered reasonable and the classification result is accurate. If the typical floods of different types are inconsistent with the historical characteristic descriptions, the classification is considered inconsistent with the actual situation and the flood classification result is not suitable for subsequent flood type identification work. Return to the first step, adjust the flood classification indicators and classification methods, and reclassify the flood types. The third step is to classify floods based on the proportion of total rainfall; Step 3.1: Taking the rainfall events corresponding to the flood events as samples, the systematic clustering method is used to classify the rainfall events according to the principle of the shortest distance between samples; the process of classifying the flood events based on the systematic clustering method in step 3.1 is the same as that in step 2.1; Step 3.2: Based on the proportion of total rainfall and the classification method in step 3.1, the rainfall corresponding to each flood is classified into four types of floods; The fourth step is to verify the feasibility of the flood classification results in the second and third steps; If the accurate classification result based on the proportion of the total amount of secondary floods obtained in the second step overlaps with the classification result based on the proportion of the total amount of secondary rainfall obtained in the third step by 85% or more, the classification based on the proportion of the total amount of secondary rainfall in the third step is considered feasible, and the flood type identification work in the fifth step should be continued; otherwise, it is considered that the proportion of the total amount of secondary rainfall cannot be used as a discriminant indicator for flood classification and identification, and the indicator should be re-selected, re-classified, and the feasibility verification should be re-performed; Step 5: Determine the method used for flood type identification; The rough set theory in fuzzy identification is used to explore the implicit relationship between classification indicators and flood types, determine the decision rules for identifying flood types, and identify flood types in three stages. Step 6: Dynamic identification of flood types based on rainfall information; According to the decision rules for flood type identification determined in the fifth step, combined with the forecast rainfall information and the rainfall information, the flood type to which the flood process belongs is dynamically identified.

2. The method for classifying and identifying rainstorm floods in a particularly large river basin according to claim 1 is characterized in that: In the first step: the calculation of the proportion of the total amount of secondary flood is as shown in formula (1): Among them, Q ij is the proportion of the total flood volume of the ith basin in the jth flood, i is the ith sub-basin, and j is the jth flood; W ij is the flood volume of the ith sub-basin flowing into the control section in the jth flood; W j is the total flood volume at the outlet section of the jth flood; The calculation of the proportion of the total rainfall is shown in formula (2): Among them, p aj is the proportion of the total rainfall of the a-th sub-district in the j-th flood, a is the a-th sub-district; R aj is the total rainfall depth of the a-th sub-district in the j-th flood; S a is the drainage area of ​​the ath partition.

3. The method for classifying and identifying rainstorm floods in a particularly large river basin according to claim 1 is characterized in that: The process of classifying flood types based on the system clustering method in step 2.1 is as follows: Step 2.1.1: Use Euclidean distance to calculate the distance between each flood event; the calculation is shown in formula (4): Among them, d jj' is the Euclidean distance between the jth and j'th floods; x jk is the kth indicator of the jth flood; x j'k is the kth index of the j'th flood; b is the total number of flood classification indicators; The distance matrix D is obtained by calculating the Euclidean distance between flood events: D=(d jj' ) c×c (5) Step 2.1.2: Use the class average distance to calculate the distance between various types of floods; the calculation is shown in formula (6) and formula (7): In which, define G p is a flood of type p, G q It is a flood of type q; is the average of the two types of flood distances; x p and x q They are class G p and class G q The number of flood samples in X P Represents class G p The sample in X q Represents class G q Samples in; define G r For any type of flood, G c is the new flood type after the two types of floods merge; when G p and G q Merge into G c When any type of flood G r With G c The distance is: Among them, D rp It's G r and G p Distance; D rq It's G r and G q The distance of x c It is class G c Number of flood samples in ; Step 2.1.3: According to the distance between any two floods in the entire basin calculated by formula (5), the two floods with the smallest distance are merged into one type of flood; according to the distance between different types of floods calculated by formula (6), the two types of floods with the closest distance are merged to form a new flood type G. c ; Then calculate any type of flood G according to the formula (7) r With the new flood type G c Then return to step 2.1.2 and repeat this process until all floods are merged into the number of types corresponding to the best AIC index.

4. The method for classifying and identifying rainstorm floods in a particularly large river basin according to claim 1 is characterized in that: In the step 2.2, the four types of floods are specifically the type encountered by the Minjiang and Tuojiang rivers, the type encountered by the Wujiang river and the main stream, the Jialingjiang river, and the type encountered by the Jialingjiang river and the Cunsan section.

5. The method for classifying and identifying rainstorm floods in a particularly large river basin according to claim 1 is characterized in that: The specific indicators and their calculations included in the morphological index, rate index, and strength index in step 2.3 are as follows: The morphological indicators include peak time skewness, flow variation coefficient, and high pulse duration ratio, which are calculated as shown in formula (8)-formula (10): Where TS is the peak time skewness, T R is the duration of flood, T is the total duration of flood; TS reflects the relative time of flood peak in each flood. The smaller the value, the earlier the flood peak appears. Where CV is the coefficient of variation of flow, Q(t) is the flood process, σ[Q(t)] is the standard deviation of the flood process, Q av It is the average flow of the flood process. The larger the CV value is, the steeper and thinner the flood process is. Among them, HDR is the high pulse duration ratio, T H It is the duration of the flood process when the flow value is greater than 75% of the peak flow value. The larger the HDR value is, the wider the flood peak shape is and the longer the high water volume flood process lasts. The rate indicators include flood rate and flood fall rate, which are calculated as shown in formula (11) and formula (12): Where RQ is the flood rate, Q m Q0 and Q1 are the peak flow and the rising flow of the secondary flood respectively, T0 is the time of the flood rising, T m It is the time when the flood peak appears; the larger the RQ value is, the faster the flood rises, and it is easy to form a disastrous flood in a short time; Where DQ is the flood rate, Q end This is the flood termination flow, T end It is the time when the flood ends; The intensity characteristic indicators include total flood volume and peak flood flow; For each type of flood, three morphological indices, two rate indices, and two intensity indices are obtained through formulas (8)-(12).

6. The method for classifying and identifying rainstorm floods in a particularly large river basin according to claim 1 is characterized in that: In the step 3.2, the four types of floods are specifically the type encountered by the Minjiang and Tuojiang rivers, the type encountered by the Wujiang river and the main stream, the Jialingjiang river, and the type encountered by the Jialingjiang river and the Cunsan section.

7. The method for classifying and identifying rainstorm floods in a particularly large river basin according to claim 1, characterized in that: In the fifth step, the process of identifying flood types using rough set theory is as follows: Step 5.1: Attribute selection and parsimony Calculate the attribute importance of rainfall classification indicators to remove redundant attributes, obtain the core attributes of rainfall classification indicators, and extract the most decisive proportion of total rainfall for different flood types. The greater the importance, the stronger the ability of the total rainfall proportion for classification. Step 5.2: Determine the lower and upper approximations Based on the proportion of the most decisive rainfall totals extracted in step 5.1, the lower approximation and upper approximation of each flood type are calculated to determine the typical characteristics and boundaries of each flood type; the lower approximation refers to the minimum set of samples that may be classified as a certain flood type; the upper approximation refers to the maximum set after excluding all samples that cannot be classified as a certain flood type; Step 5.3: Determine the decision rule Based on the lower and upper approximations determined in step 5.2, derive the decision rule for identifying the flood type; Step 5.4: Application and implementation Based on the decision rules determined in step 5.3, the flood type corresponding to the actual flood process is identified in three stages.

8. A method for classifying and identifying rainstorm floods in a particularly large river basin according to claim 7, characterized in that: In step 5.4, the three stages of flood identification are as follows: Stage 1: No rainfall has occurred yet, and flood type identification is fully based on the forecast rainfall information; Phase 2: As rainfall continues to develop, since there are no less than two main rainfall areas for each type of flood, when there are no more than two sub-basins with high-intensity rainfall areas in the early stage, the flood type is identified by combining the measured rainfall and subsequent rainfall forecast information; Stage 3: High-intensity rainfall has occurred in various regions, and the rainfall intensity has decreased. At this time, the flood is identified by combining the measured rainfall and forecast rainfall information. The time intervals of the nodes of different flood events are different.