Data-intensive rain type intelligent identification and urban rainfall scenario dataset construction method

Abstract

The application provides a data-intensive rain type intelligent identification and urban rainfall scenario dataset construction method. The main steps of the construction method are: 1) urban market rainfall sample dataset generation; 2) urban rain type intelligent classification and rain type dataset generation; 3) urban each rainfall station standard rain type identification; 4) scene design rainfall process derivation; and 5) urban rainfall scenario dataset generation. The urban rainfall scenario dataset constructed by the application includes historical scene rainfall and each rain type design rainfall with different guarantee rates, covers possible urban rainfall processes, and can provide rainfall scenarios for digital twin city flood control system construction.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of urban design rain type derivation, and particularly relates to a data-intensive rain type intelligent identification and urban rainfall scenario data set construction method, that is, a method for obtaining an urban rainfall scenario data set through the steps of intelligent classification of urban match rainfall rain type, definition of rain type standard characteristic value, standard rain type identification, and derivation of match design rainfall process. BACKGROUND

[0002] With the rapid development of hydrological automatic monitoring and remote transmission technology, urban rainfall monitoring station network construction has also been vigorously developed. Intensive rainfall data, which is characterized by high density of observation sites and high time resolution, is gradually accumulating in some cities. Intensive rainfall data has higher spatial resolution, more accurate rainfall estimation, and more comprehensive time series than traditional data, which can help improve the understanding and prediction of regional rainfall characteristics.

[0003] Rainfall scenario data is a key basis for predicting the possibility and intensity of floods, and is also the data foundation for the construction of digital twin city flood control system. By using rainfall scenario data, the flood inundation range and water level change under different rainfall scenarios can be simulated, the effectiveness of existing flood control facilities can be evaluated, the planning and construction of flood control and water conservancy facilities can be optimized, intelligent decision-making and emergency response can be carried out, the flood control response capability of the city can be improved, and the flood risk degree of different regions can be evaluated to provide scientific basis for urban planning and development.

[0004] The application carries out intelligent classification of urban match rainfall rain type, definition of rain type standard characteristic value, standard rain type identification, and derivation of match design rainfall process. The constructed urban rainfall scenario data set includes not only historical match rainfall but also design rainfall processes of each rain type with different guarantee rates, covers possible rainfall processes in the city, and can provide rainfall scenarios for urban flood control system construction, serve as an important data source for predicting flood possibility and intensity, optimizing flood control and water conservancy facilities, providing intelligent decision-making for flood control, and supporting emergency response for disaster reduction, and help improve the flood control and flood resistance capabilities of the city. SUMMARY

[0005] The application proposes a data-intensive rain type intelligent identification and urban rainfall scenario data set construction method. The constructed urban rainfall scenario data set includes not only historical match rainfall but also design rainfall processes of each rain type with different guarantee rates, and can provide rainfall scenarios for urban flood control system construction.

[0006] To achieve the above-mentioned purpose, the application adopts the following technical solutions:

[0007] The data-intensive rain type intelligent identification and urban rainfall scenario data set construction method comprises the following steps:

[0008] Step 1, urban market rainfall sample data set generation: based on the automatic monitoring data of urban rainfall stations, the rainfall event is automatically extracted and manually audited, and the urban rainfall data series of each rainfall station are obtained; the rainfall station two-dimensional rainfall sample data set is generated by using the rainfall data, and the urban three-dimensional rainfall sample data set is formed by each rainfall station;

[0009] Step 2, intelligent recognition of urban rain type and generation of rain type data set: based on the two-dimensional rainfall sample set of each rainfall station in the city, the self-organizing map algorithm (Self-Organizing Map, SOM) is used for unsupervised intelligent classification, the rain type division is completed, and the rain type data set of each rainfall station in the city is generated;

[0010] Step 3, standard rain type recognition of each rainfall station in the city: the duration of each rainfall station, the peak duration, the maximum rainfall in the period, the average rainfall in the period, the total rainfall of the scene, the frequency analysis is carried out to obtain the rain type characteristic value of different guarantee rate, and the 50% guarantee rate rain type characteristic value is taken as the standard rain type characteristic value, and the history similar rainfall process of the standard rain type characteristic value is identified as the standard rain type;

[0011] Step 4, scene design rainfall process derivation: based on the standard rain type scene rainfall process, the same ratio amplification method is used to derive the scene rainfall process for the maximum rainfall in the period, and the scene rainfall process of each rainfall station under different guarantee rates is obtained;

[0012] Step 5, generation of urban rainfall scenario data set: the scene rainfall monitoring data extracted from the historical monitoring data of the rainfall station and the rainfall process data set of different guarantee rates of each rain type together constitute the urban rainfall scenario data set.

[0013] The technical scheme of the present application has the following advantages:

[0014] (1) Because the intensive monitoring data usually has higher spatial and temporal resolution, the amount of data generated is larger, and the processing and management of large-scale data require powerful computing and storage capacity, which increases the complexity of data processing, and the traditional method is difficult to abstract its characteristic rule, and the intelligent clustering method is more suitable for data-intensive scene rainfall type classification and rain type feature extraction;

[0015] (2) The 50% guarantee rate is defined as the rain type standard characteristic value, and the history similar rainfall process is identified as the standard rainfall process of each rain type, so the standard rain type is selected from the measured rainfall process rather than artificially constructed, which ensures the objectivity of the standard rain type;

[0016] (3) The urban rainfall scenario data set includes not only the historical scene rainfall, but also the design scene rainfall process of each rain type under different guarantee rates, which covers the possible rainfall scenarios of the city and can provide rainfall scenarios for the construction of urban flood control system. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 Rainfall processes of type I classified in 5 classes.

[0018] Figure 2 Rainfall processes of type II classified in 5 classes.

[0019] Figure 3 Rainfall processes of type III classified in 5 classes.

[0020] Figure 4 Rainfall processes of type IV classified in 5 classes.

[0021] Figure 5 Rainfall processes of type V classified in 5 classes.

[0022] Figure 6 Rainfall processes of type I classified in 4 classes.

[0023] Figure 7 Rainfall processes of type II classified in 4 classes.

[0024] Figure 8 Rainfall processes of type III classified in 4 classes.

[0025] Figure 9 Rainfall processes of type IV classified in 4 classes.

[0026] Figure 10 Rainfall processes of type I classified in 3 classes.

[0027] Figure 11 Rainfall processes of type II classified in 3 classes.

[0028] Figure 12 Rainfall processes of type III classified in 3 classes.

[0029] Figure 13 Histogram of rainfall processes of type I.

[0030] Figure 14 Histogram of rainfall processes of type II.

[0031] Figure 15 Histogram of rainfall processes of type III.

[0032] Figure 16 Histogram of rainfall processes of type IV. DETAILED DESCRIPTION

[0033] The application is further described by the following examples, which do not limit the application in any way, and any modification or change made by those skilled in the art without departing from the technical solutions of the application shall fall within the scope of the claims of the application.

[0034] The application will be further described in detail below with reference to the accompanying drawings.

[0035] The application is further described by the following examples, which do not limit the application in any way, and any modification or change made by those skilled in the art without departing from the technical solutions of the application shall fall within the scope of the claims of the application.

[0036] 1. The data-intensive rain type intelligent recognition and digital twin city flood control rainfall scenario dataset construction method according to claim 1, wherein the method for generating the city event rainfall sample set in step 1 comprises:

[0037] Step 1.1, collection, processing and saving of rainfall data: collecting continuous rainfall monitoring data of the city rainfall station FH station from 2019 to 2020, and processing the annual rainfall data into continuous rainfall data with an isochronous period Δt = 1h;

[0038] Step 1.2, automatic extraction of event rainfall: sequentially reading the continuous rainfall time series of the FH station in June to September in the database, and dividing it into independent event rainfall; regarding the rainfall process with a period rainfall greater than or equal to 2mm and a continuous rainfall duration greater than 3h as effective event rainfall, that is, p' = 2mm, ΔT = 3h, and selecting the effective event rainfall;

[0039] Step 1.3, manual review of event rainfall: regarding two event rainfalls with an interval time less than 2h as the same event rainfall, that is, Δt' = 2h, and setting the interval period rainfall as 2mm / h, that is, equal to the effective period rainfall p', for example, combining the automatically extracted event rainfalls of 2019 / 8 / 16 11:00-2019 / 8 / 17 1:00 and 2019 / 8 / 17 4:00-2019 / 8 / 18 8:00 into one event rainfall of 2019 / 8 / 16 11:00-2019 / 8 / 18 8:00, as shown in Table 1:

[0040] Table 1: Event rainfall merging

[0041]

[0042]

[0043] Step 1.4, after the automatic extraction of event rainfall and the manual review of event rainfall of the FH station, 33 rainfalls are obtained, and the longest event rainfall has 46 periods, therefore, the FH station generates a two-dimensional sample dataset of 33*46 event rainfalls Similarly, each rainfall station generates a two-dimensional field rainfall sample data set for each rainfall station according to the same steps and methods as the FH station; all rainfall stations in the city form a three-dimensional sample data set of the city's rainfall, which is omitted here;

[0044] 2. The data-intensive rain type intelligent identification and urban rainfall scenario data set construction method according to claim 1, characterized in that the method of step 2 rain type intelligent identification and rain type data set generation comprises:

[0045] Step 2.1, based on the two-dimensional sample data set of the field rainfall process of the rainfall station k Classify the α field rainfall types using the SOM algorithm, initially set to 5 categories, i.e. α = 5, and form the field rainfall data set of each rain type Wherein, the rain type serial number i = I, II, III, IV, V, and the field rainfall process of each rain type is shown in Figures 1-5 Similarly, the field rainfall process of each rain type is divided into 4 categories, as shown in Figures 6-9 The field rainfall process of each rain type is divided into 3 categories, as shown in Figures 10-12 ;

[0046] Step 2.2, visually evaluate the clustering effect of different classification numbers, and the field rainfall process of the same rain type has similarity, and the field rainfall process of different rain types has obvious difference, and the comprehensive evaluation of 4 categories is more reasonable;

[0047] Step 2.3, repeat the above steps for each rainfall station in the city to obtain the field rainfall data sample set of each rain type for each rainfall station in the city

[0048] 3. The data-intensive rain type intelligent identification and urban rainfall scenario data set construction method according to claim 1, characterized in that the step 3 standard rain type identification method comprises:

[0049] Step 3.1, based on the field rainfall data set of the FH station rain type I M FH,1 , N FH,1 represent the number of field rainfall of the FH station rain type I as 6 and the maximum number of field rainfall periods as 36; the rainfall duration, peak occurrence duration, period maximum rainfall, period average rainfall, and field total rainfall of each field rainfall are counted to form a rain type I statistical characteristic sample set l = TT, PPT, PP, MP, TP, respectively representing the duration, peak occurrence duration, period maximum rainfall, period average rainfall, and field total rainfall of the field rainfall, as shown in Table 2; similarly, the characteristic values of the field rainfall of each rain type are shown in Tables 3-5;

[0050] Table 2 Rain type I field rainfall series characteristic values

[0051]

[0052] Table 3 Rain type II series of each field rainfall characteristic value

[0053]

[0054]

[0055] Table 4 Rain type III series of each field rainfall characteristic value

[0056]

[0057] Table 5 Rain type IV series of each field rainfall characteristic value

[0058]

[0059] Step 3.2, FH station-based rain type I field rainfall statistical characteristic data set The frequency calculation is carried out by using the box plot method, and the rain type characteristic values exceeding the 5%, 25%, 50%, 75% and 95% guarantee rate are obtained The rain type characteristic value at the 50% guarantee rate is taken as the standard rain type characteristic value The results are shown in Table 6; similarly, the other rain types are shown in Tables 7-9;

[0060] Table 6 Rain type I field rainfall characteristic guarantee rate characteristic value

[0061]

[0062]

[0063] Table 7 Rain type II field rainfall characteristic guarantee rate characteristic value

[0064]

[0065] Table 8 Rain type III field rainfall characteristic guarantee rate characteristic value

[0066]

[0067] Table 9 Rain type IV field rainfall characteristic guarantee rate characteristic value

[0068]

[0069] Step 3.3, based on sample set The dynamic time warping method is used to identify the standard rain type characteristic value The historical similar rain type characteristic value is taken as the standard rain type The rain type I standard rain type process is shown in Figure 13; Similarly, the standard rain type process of other rain types is shown in Figures 14-16 ;

[0070] 4. The data-intensive rain type intelligent identification and digital twin city flood control rainfall scenario dataset construction method according to claim 1, characterized in that the step 4 design rainfall process derivation method comprises:

[0071] Step 4.1, based on the standard rain type Calculate the maximum rainfall of each guarantee rate p period The ratio β of the maximum rainfall of the standard rain type period Using the same ratio amplification method to obtain the corresponding field rainfall process of the given design guarantee rate Calculate the characteristic values corresponding to the characteristic guarantee rates of 5%, 25%, 75%, and 95%, and the calculation results are shown in Table 10; Similarly, the field rainfall process of any guarantee rate p can be calculated, which is omitted here;

[0072] Table 10 Rain type I characteristic guarantee rate field rainfall process

[0073]

[0074] Step 4.2, FH station rainfall repeats the above steps to obtain the design rain type dataset of each rain type The calculation results are shown in Tables 11-13; Repeat the above steps for each rainfall station in the city to obtain the design rain type dataset of the city

[0075] Table 11 Rain type II characteristic guarantee rate field rainfall process

[0076]

[0077] Table 12 Rain type III characteristic guarantee rate field rainfall process

[0078]

[0079] Table 13 Rain type IV characteristic guarantee rate field rainfall process

[0080]

[0081]

[0082] 5. The data-intensive rain type intelligent identification and city rainfall scenario dataset construction method according to claim 1, characterized in that the step 5 city flood control rainfall scenario dataset generation comprises:

[0083] Step 5.1, 33 rainfall processes are measured by FH station; similarly, the rainfall sample data set of each rainfall station in the city is extracted As historical rainfall data;

[0084] Step 5.2, the rainfall process of each rain type feature guarantee rate of FH station is shown in Tables 10-13; similarly, the design rain type data of different guarantee rates of each rainfall station in the city is calculated; the design rain type data set of the city is shown in Table 14 As possible rainfall scenario data set;

[0085] Step 5.3, the rainfall data of each rain type of FH station is shown in Table 15 And The city rainfall scenario database is constructed as a data source.

[0086] In summary, the application provides a data-intensive rain type intelligent identification and city rainfall scenario data set construction method, and the present application is only an example for the method described in the patent, and with the continuous accumulation of monitoring data, the rationality and reliability of rain type classification and standard rain type calculation results will be improved when the sample capacity is sufficient.

[0087] The preferred embodiments of the application disclosed above are only used to help explain the application. The preferred embodiments do not describe all the details, nor limit the application to the specific embodiments described. Obviously, according to the content of the specification, many modifications and changes can be made. The specification selects and describes these embodiments in order to better explain the principles and practical applications of the application, so that those skilled in the art can well understand and utilize the application. The application is limited by the claims and their entire scope and equivalents.

Claims

1. A data-intensive rain type intelligent recognition and urban rainfall scenario data set construction method, characterized in that, Includes the following steps: Step 1: Generation of sample dataset of rainfall events in the city: Based on the automatic monitoring data of urban rain gauges, rainfall events are automatically extracted and manually reviewed to obtain a series of rainfall data from each rain gauge station in the city; Two-dimensional rainfall sample datasets from rain gauge stations are generated using rainfall data from each rainfall event, and three-dimensional rainfall sample datasets from each rain gauge station are formed for the city. Step 2, Intelligent Identification of Urban Rainfall Patterns and Generation of Rainfall Pattern Datasets: Based on the two-dimensional rainfall sample set of each rain gauge station in the city, the Self-Organizing Map (SOM) algorithm is used for unsupervised intelligent classification to complete the rainfall pattern division and generate the rainfall pattern dataset of each rain gauge station in the city. Step 3, Standard Rainfall Pattern Identification at Each Rain Gauge Station in the City: Statistically analyze the rainfall duration, peak duration, maximum rainfall, average rainfall, and total rainfall for each rainfall pattern at each rain gauge station to form a statistical feature sample set of rainfall patterns. Perform frequency analysis to obtain rainfall pattern feature values ​​with different guarantee rates. Use the rainfall pattern feature value with a 50% guarantee rate as the standard rainfall pattern feature value, and then identify historical similar rainfall processes of the standard rainfall pattern feature value as the standard rainfall pattern. Step 4, Derivation of the designed rainfall process for each rainfall event: Based on the standard rainfall pattern, the rainfall process for each event is derived using the same scaling-up method for the maximum rainfall amount in the time period, and the rainfall process for each rain gauge with different guarantee rates is obtained; Step 5, Generation of Urban Rainfall Scenario Dataset: The urban rainfall scenario dataset is composed of rainfall monitoring data extracted from historical monitoring data of rain gauges and various rainfall process datasets with different guarantee rates.

2. The data-intensive rain type intelligent identification and urban rainfall scenario data set construction method according to claim 1, characterized in that, The method for generating the urban market secondary rainfall sample set in step 1 includes: Step 1.1, Collection, processing and storage of rainfall data: Collect automatic monitoring data from various rain gauge stations in the city, process the rainfall monitoring data into equal-time Δt rainfall data and store it in the database. The size of the time period Δt can be determined according to the city's flood control requirements or the city's runoff conditions. Step 1.2, Automatic extraction of rainfall events: The continuous rainfall time series from June to September of each rain gauge station in the rainfall monitoring database are read sequentially. Rainfall events where the rainfall during a given period is greater than the effective rainfall amount p' and the duration of continuous rainfall is greater than the effective rainfall duration ΔT are considered as valid rainfall events. Valid rainfall events for each rain gauge station are selected. The two thresholds p' and ΔT for valid rainfall events can be set according to flood control requirements and the actual runoff conditions of the city. Step 1.3, Manual review of rainfall events: The valid rainfall events extracted automatically are manually reviewed. Rainfall events with an interval time of less than Δt' are considered as the same rainfall event, and the rainfall amount during the interval is set as p'. Δt' can be determined according to the specific situation. Step 1.4, generating a two-dimensional sample data set of the field rainfall process based on the effective field rainfall process data of each rain gauge, and each rain gauge forms a three-dimensional sample data set of the urban field rainfall process; taking rain gauge k as an example, field rainfall serial number m = 1, 2, …, M, M is the number of field rainfall of rain gauge k; the number of field rainfall period n = 1, 2, …, N, N is the maximum number of field rainfall periods, and the rainfall period is less than N, and the rainfall amount of the non-rainfall period is 0; generating a two-dimensional field rainfall sample data set of M rows and N columns of field rainfall As shown in Table 1, the rainfall amount of rain gauge k field 2, period 4 is 3 mm, and period 7 has no rainfall; finally, a three-dimensional urban field rainfall process sample data set of K*M*N is formed wherein, the serial number of the rain gauge k = 1, 2, …, K, and K is the number of rain gauges; Table 1 Field of rain gauge k's two-dimensional sample set of rainfall 3. The data-intensive rain type intelligent identification and urban rainfall scenario data set construction method according to claim 1, characterized in that, The method for intelligent rain pattern recognition and rain pattern dataset generation in step 2 includes: Step 2.1, two-dimensional sample data set of the event rainfall process based on the rain gauge station k Step 2.2, set the initial number of categories α, and divide the event rainfall into α categories by using the SOM algorithm to form the event rainfall data set of each type of rain pattern wherein the rain pattern serial number i = 1, 2, …, α, and the rain pattern event rainfall data sample set corresponding to the α types of rain patterns of the rain gauge station k is generated Step 2.2: Compare the rainfall events of different rain types and evaluate the clustering effect of different numbers of categories. The evaluation principles are: 1) Rainfall events of the same category have good similarity; 2) Rainfall events of different categories have significant differences; Select appropriate categories to classify rain types. Step 2.

3. Repeat the above steps for each rain gauge in the city to obtain the sample set of rainfall data for each rain type at each rain gauge in the city 4. The data-intensive rain type intelligent identification and urban rainfall scenario data set construction method according to claim 1, characterized in that, The standard rain pattern identification method in step 3 includes: Step 3.1, the field rainfall data set of the rain pattern i based on the rain gauge station k M k,i , N k,i respectively represent the number of field rainfall and the number of maximum field rainfall period of the rain pattern i of the rain gauge station k; the rainfall duration, the peak duration, the maximum rainfall of the period, the average rainfall of the period, and the total rainfall of the field are counted for each field rainfall sample to form the rainfall characteristic sample set of the rain pattern i l = TT, PPT, PP, MP, TP, respectively represent the duration of the field rainfall, the peak duration, the maximum rainfall of the period, the average rainfall of the period, and the total rainfall of the field Step 3.2, the match rainfall characteristic data set of the rain type i based on the rain station k Frequency analysis is performed on each characteristic index to calculate the rain type characteristic value of each guarantee rate p The rain type characteristic value of 50% guarantee rate is taken as the standard rain type characteristic value Step 3.3, based on the sample set The standard rain pattern characteristic value is identified by using dynamic time warping method The historical similar rain pattern characteristic value corresponding to the rainfall process as the standard rain pattern 5. The data-intensive rain type intelligent identification and urban rainfall scenario data set construction method according to claim 1, characterized in that, Step 4, the design of rainfall events, includes: Step 4.1, based on the standard rain pattern Calculate the period maximum rainfall for each guarantee rate p The ratio β of the period maximum rainfall to the standard rain pattern The corresponding field rainfall process for a given design guarantee rate p is obtained using the same ratio β amplification method Step 4.2, design storm data set for each rain gauge k Repeat the above steps to obtain the design storm data set for each rain pattern Repeat the above steps to obtain the design storm data set for each rain pattern 6. The data-intensive rain type intelligent identification and urban rainfall scenario data set construction method according to claim 1, characterized in that, Step 5, generating the urban rainfall scenario dataset, includes: Step 5.1 involves extracting a dataset of urban rainfall samples based on measured rainfall data from urban rain gauges. As historical rainfall data; Step 5.2 involves generating the derived urban design rainfall pattern dataset. As a dataset of possible rainfall scenarios; Step 5.3, will and A database of urban rainfall scenarios was constructed using rainfall events as data sources.

Images