Rainstorm flood model library intelligent online generation method and system

Abstract

The application discloses a heavy rain flood mode library intelligent online generation method and system, and the method comprises the following steps: importing data, extracting the station name and position information from the station information file, reading the corresponding station detection data file, reading the time series data into the database for classified storage, and managing the existing stations; automatically filling the monitoring data in the stations in the same basin, filling the discontinuous rainfall monitoring data into the time series data with the same granularity as the flow monitoring data, and aligning; based on the optimized FLUSS algorithm, the time series flow data is segmented, the start and end time of the flood is segmented, and the start and end time is automatically adjusted; extracting the segmented time series data, calculating the multivariate flood characteristic value, realizing the conversion from scene data to feature mode; and distinguishing and viewing the management of the basins. The application enhances the accuracy and rationality of the flood scene division, and provides strong support for the construction and optimization of the flood mode library.

Description

Technical Field

[0001] This invention belongs to the field of flood field management, specifically relating to a method and system for intelligent online generation of rainstorm and flood pattern databases. Background Technology

[0002] Currently, flood pattern management and identification primarily rely on manual operation and traditional data processing methods. Administrators must manually organize monitoring data from various hydrological stations, perform data processing and time-series data alignment, and handle missing values. Subsequently, traditional offline event segmentation algorithms are used to segment flood events, and the event data is stored locally. However, this method has several key problems. Limitations of the segmentation algorithm: Traditional time-series segmentation algorithms, such as FLUSS, may fail to accurately identify and segment flood events when processing time-series data with smooth state characteristics, especially on both sides of flood events. This can lead to problems in finding state segmentation points, affecting the accuracy of segmentation. The cumbersome manual adjustment: The segmented event results require dedicated output to check the rationality of the segmentation. If it is unreasonable, the administrator must manually adjust the start and end times and re-segment the data, a process that is both time-consuming and error-prone. Limitations of feature calculation: During feature calculation, there is a risk of feature calculation bias due to individual data anomalies. Current methods cannot effectively identify and adjust these biases. Lack of online integration and visualization: Currently, there is no integrated online system for flood event segmentation and characteristic calculation that enables visual viewing and adjustment of flood patterns. All of these methods limit the efficiency and accuracy of flood management. Summary of the Invention

[0003] Purpose of the invention: This invention provides an intelligent online generation method and system for a rainstorm and flood pattern library, aiming to solve the problems of complex and non-visualized generation steps and complicated manual adjustment processes in the existing flood pattern generation and management process.

[0004] Technical solution: The present invention provides an intelligent online generation method for a rainstorm and flood pattern library, comprising the following steps:

[0005] (1) Import data, extract station name and location information from station information file and read the corresponding station detection data file, read the time series data into the database, classify and save it, and manage the existing stations;

[0006] (2) Automatically fill in the monitoring data of stations in the same watershed, fill the discontinuous rainfall monitoring data into time series data with the same granularity as the flow monitoring data and align them; the aligned rainfall and flow data are used for subsequent field division work.

[0007] (3) The time series flow data is segmented based on the optimized FLUSS algorithm to separate the start and end times of the flood, and the start and end times are automatically adjusted; the optimized FLUSS algorithm introduces a similar state crossing monitoring mechanism into the FLUSS algorithm to realize the segmentation of long-term time series data.

[0008] (4) Extract the segmented time series data, calculate the multivariate flood characteristic values, and realize the transformation of scene data into feature patterns;

[0009] (5) Management is carried out by watershed.

[0010] Furthermore, the implementation process of step (1) is as follows:

[0011] (11) New watershed information: Import new watershed data, which includes the watershed coverage area, administrative division, and geographical environment description;

[0012] (12) Import station information: Read the station information document and import the station information contained in the newly created watershed. This information includes station identification information, station type, geographical location and time span of detection data.

[0013] (13) Import station data: Read the station monitoring data document and import the station's historical monitoring data, which includes rainfall-related data and flow-related data.

[0014] Furthermore, the implementation process of step (2) is as follows:

[0015] (21) Read time series data: Read the monitoring data from all stations in the same watershed, divide them into rainfall data and flow data, find the maximum time span in the data, and perform interpolation preprocessing on the missing value part in the test data according to the maximum time span;

[0016] (22) Select time span: Based on the existing station data coverage time, manually adjust the time range used for flood field division and cut off the redundant data;

[0017] (23) Filling time series data: Use quadratic interpolation to fill the rainfall data into continuous time series data;

[0018] (24) Aligning Rainfall and Flood Data: Align and organize the complete rainfall and flow data according to the start and end times of alignment at a uniform time interval, in the format of {monitoring time, flow value, rainfall value 1, rainfall value 2...}, where the rainfall value is the measurement value of different rain gauge stations.

[0019] Furthermore, the process of segmenting time-series traffic data based on the optimized FLUSS algorithm in step (3) is as follows:

[0020] Divide the time-series traffic data into multiple subsequences of length length;

[0021] Within a specified limit, calculate and compare the Euclidean distance between adjacent subsequences to assess their similarity.

[0022] For each subsequence, determine the subsequence that is most similar to it, and record this similarity relationship as an arc;

[0023] Count the number of these arcs spanning the domain at each time point to form a new sequence;

[0024] The new sequence is compared with the expected number of cross-domain arcs when randomly distributed under ideal conditions to obtain a ratio sequence;

[0025] In the ratio sequence, find the minimum value. The time point corresponding to the minimum value is the dividing point, which is used to divide the time series data into flood state and normal state.

[0026] Store the start and end times of the flood state in a flood array.

[0027] Furthermore, the automatic adjustment of the start and end times described in step (3) is implemented as follows:

[0028] Compile rainfall monitoring data within the flood timeframe. If the flood start and end times coincide with the rainfall time, adjust the flood start point forward according to the rainfall start time; if the flood end time coincides with the rainfall time, adjust the flood end point backward according to the rainfall end time.

[0029] Furthermore, the implementation process of step (4) is as follows:

[0030] (41) Data reading: For each station, read the measurement data within the time period of the flood field, as well as the station's own information;

[0031] (42) Flow feature extraction: Based on the flow time series of the station, calculate the flow-related features as follows: {flood peak pattern, flood duration, peak flow, peak water level, time when the flood reaches the peak flow, total flow of the flood, how long after the flood starts to reach the peak, duration after reaching the peak, duration after reaching the peak, water level sequence};

[0032] (43) Grid rainfall calculation: Using the latitude and longitude information of the rain gauge, the precipitation data is extended to the entire watershed through interpolation methods to obtain the rainfall distribution sequence of each grid unit in the watershed;

[0033] (44) Rainfall feature extraction: Based on the grid rainfall sequence, calculate the features related to rainfall, as follows: {rainfall sequence, total rainfall, hourly grid rainfall peak, rainfall trend, hourly rainfall sum, hourly rainfall center, rainstorm center point trajectory, hourly rainfall peak index, total rainfall for each grid point, maximum total rainfall and index for each grid point};

[0034] (45) Feature merging or addition: Compare the calculated flow and rainfall features with the existing flood and rainfall features in the model library, and select to merge or add them;

[0035] (46) Pattern Recognition and Fusion: The calculated flood and flow characteristics are combined into new flood models. Each model represents a specific flood development process, described by a series of flow and rainfall characteristics. The models are matched with existing flood models in the model library using the Spatiotemporal Attribute Similarity Measurement (STDTW) algorithm. The following operations are performed based on the matching results:

[0036] If the similarity between a new feature and a pattern in the library reaches a preset standard, it is classified as an instance of that pattern, and the new feature is integrated into that pattern to optimize and improve the pattern.

[0037] If the similarity between a new feature and all patterns in the database does not meet the preset standard, it is considered to represent a new rainfall-flood pattern and is added to the pattern database as a new pattern for future analysis and prediction.

[0038] Similarity pattern fusion standard: If the similarity between the comparison pattern and the matched pattern ω = 1 / 1 + dist is greater than N / N+1, then the two patterns are considered similar and are weighted and merged; where dist is the DTW distance and N is the number of current feature patterns.

[0039] Furthermore, the feature merging described in step (45) is performed using a weighted average based on weighting factors, as shown in the following formula:

[0040] T x =(T' x *n+T) / n+1

[0041] Where T is the calculated feature, T' x Similar features in the pattern library, where n is the weighting factor.

[0042] Furthermore, the implementation process of step (5) is as follows:

[0043] Select the watershed and topological image information and read the watershed ID; query the list of flood event IDs that occurred within the watershed based on the obtained ID; obtain the start and end times of the flood event and the spatiotemporal characteristic sequence of the event based on the flood event ID, and display the time series and characteristic patterns of the flood event through a graphical interface.

[0044] The present invention discloses an intelligent online generation system for a rainstorm and flood pattern database, comprising a data import module, a rainstorm-flood alignment module, a field segmentation module, a pattern generation module, and a historical flood pattern management module; wherein:

[0045] Data import module: Reads data from uploaded files and saves it to the database;

[0046] Rainfall and flood alignment module: Time-calibrates imported rainfall and flow data to provide time-series data for segmentation in the event segmentation module;

[0047] The flood season segmentation module: segments the start and end times of the flood and automatically adjusts them; it segments the time-series flow data of the aligned data based on the optimized FLUSS algorithm, segments the start and end times of the flood, and automatically adjusts the start and end times; the optimized FLUSS algorithm introduces a similar state crossing monitoring mechanism into the FLUSS algorithm.

[0048] The pattern generation module extracts the segmented time-series data, calculates multivariate flood characteristic values, and then fuses or adds to the existing characteristic patterns in the comparison pattern library.

[0049] Historical Flood Pattern Management Module: Allows for viewing and managing generated flood patterns by watershed, and enables deletion and modification operations.

[0050] Beneficial Effects: Compared with existing technologies, the present invention offers the following advantages: By introducing a univariate time-series segmentation algorithm and online intelligent generation technology, the present invention achieves automation and intelligence in flood event management, greatly improving the efficiency and accuracy of flood event generation; the present invention employs an optimized low-cost time-series segmentation algorithm, avoiding complex pre-training or full-scale fine-tuning processes, reducing system implementation costs and operational complexity; through automated data import and alignment, intelligent event segmentation and feature calculation, the system can quickly and accurately generate flood event information, effectively solving the problems existing in existing technologies; this efficient flood event management capability makes the system highly practical in the fields of flood control management and decision support; the present invention implements strict constraints and verifications in data processing and feature calculation, avoiding errors caused by data anomalies or algorithm deviations, ensuring that the system output is more accurate and reliable; in summary, the present invention not only improves user trust in the system and reduces risks caused by erroneous information, but also fills the gap in the field of intelligent flood model management, possessing extremely high practical value and commercial potential; in terms of model fusion, the present invention achieves effective integration of new and old flood models, a process that not only enriches the flood model library but also improves the accuracy of model matching and the reliability of flood prediction. Attached Figure Description

[0051] Figure 1 It is a framework diagram for the intelligent online generation of rainstorm and flood pattern database;

[0052] Figure 2 This is an operational diagram of the rainwater alignment module proposed in this invention;

[0053] Figure 3 This is a flowchart of the session division module proposed in this invention;

[0054] Figure 4 This is a flowchart of the pattern generation module proposed in this invention. Detailed Implementation

[0055] The present invention will now be described in further detail with reference to the accompanying drawings.

[0056] like Figure 1 As shown, this invention provides a method for intelligent online generation of a rainstorm and flood pattern library, comprising the following steps:

[0057] Step 1: Use the data import module to import data, extract station names, locations, and other information from the station information file, and read the corresponding station detection data file to import the time-series data into the database for categorized storage. Existing stations can also be managed. The relationship between flow stations and rainfall stations in a unified watershed is often one-to-many, providing external data support for field production.

[0058] Create new watershed information: Import new watershed data, which includes the watershed's coverage area, administrative division, and geographical description. This mainly includes the watershed's latitude and longitude range, name, ID, and administrative division.

[0059] Import Station Information: This function reads and imports station information from a newly created watershed. This information includes station identification, station type, geographical location, and the time span of the data collected. Key details include station ID, whether it is a hydrological station, whether it is a rainfall station, the station's latitude and longitude, and the watershed it belongs to. A watershed often contains multiple stations.

[0060] Import station data: Read the station monitoring data document and import the station's historical monitoring data, which includes rainfall-related data and flow-related data. The time-series data format is [station ID, measurement time, measurement value].

[0061] Step 2: As Figure 2 As shown, the monitoring data from stations within the same watershed are automatically populated, filling discontinuous rainfall monitoring data with time-series data of the same granularity as the flow monitoring data and aligning them. This aligned rainfall and flow data is used for subsequent event segmentation. The rainwater-flow alignment module performs time calibration on the imported rainfall and flow data, providing time-series data for event segmentation. For example, it aligns rainfall and flow data from May 1, 1989 to August 3, 1989.

[0062] Read time series data: Read the monitoring data from all stations in the same watershed, divide them into rainfall data and flow data, find the maximum time span in the data, and perform interpolation preprocessing on the missing value part of the test data according to the maximum time span.

[0063] Select time span: Based on the existing station data coverage time, manually adjust the time range used for flood event division and cut off the redundant data.

[0064] Filling time series data: Use quadratic interpolation to fill rainfall data into continuous time series data.

[0065] Aligning Rainfall and Flood Data: Complete rainfall and flow data are aligned and organized according to the start and end times of alignment at uniform time intervals, in a format such as {monitoring time, flow value, rainfall value 1, rainfall value 2...}, where the rainfall value is the measurement value from different rain gauge stations.

[0066] Save monitoring data: Save the interpolated station monitoring data to the database at hourly intervals.

[0067] Step 3: As Figure 3As shown, the optimized FLUSS algorithm is used to segment the time-series flow data, segmenting the start and end times of the flood, and automatically adjusting the start and end times; by introducing a similar state crossing monitoring mechanism, the long-term time-series data can be segmented.

[0068] Read the time-series flow data, apply the optimized FLUSS algorithm for state segmentation, dividing the time-series data into flood state and normal state, and store the start and end time nodes of the flood state in a flood array. Details are as follows:

[0069] 1) Divide the time-series traffic data into multiple subsequences of length length.

[0070] 2) Within the specified limit, calculate and compare the Euclidean distance between adjacent subsequences to assess their similarity.

[0071] 3) For each subsequence, determine the subsequence most similar to it and record this similarity relationship as an arc.

[0072] 4) Count the number of these arcs across the domain at each time point to form a new sequence.

[0073] 5) Compare this sequence with the expected number of cross-domain arcs when randomly distributed under ideal conditions to obtain a ratio sequence.

[0074] 6) In the ratio sequence, find the minimum value. The time point corresponding to this value is the dividing point, which is used to divide the time series data into flood state and normal state.

[0075] 7) Store the start and end time nodes of the flood status in a flood array with the data type [[Flood 1 start time, flood 1 end time], [Flood 2 start time, flood 2 end time], [Flood 3 start time, flood 3 end time], [Flood 4 start time, flood 4 end time]...] for subsequent analysis.

[0076] Align rainfall data: Organize rainfall monitoring data within the flood timeframe. If the flood start and end times coincide with the rainfall time, adjust the flood start time backward according to the rainfall start time; if the flood end time coincides with the rainfall time, adjust the flood end time backward according to the rainfall end time.

[0077] The front-end page displays the data of the divided flood events and provides the function of manually dragging the timeline so that users can make precise adjustments and corrections.

[0078] Step 4: As Figure 4 As shown, the segmented time-series data is extracted, multivariate flood feature values ​​are calculated, and the scene data is transformed into feature patterns.

[0079] Data Reading: First, the latitude and longitude boundaries of the target watershed are determined, and a list of IDs for all stations within the watershed is obtained. Then, for each station ID, its detailed information is retrieved and loaded, including the station type (flow monitoring station or rainfall monitoring station) and its specific latitude and longitude coordinates. Finally, based on the flood start and end times, flow and rainfall data for these stations within the corresponding time range are extracted and read from the flood event segmentation module.

[0080] Flow feature extraction: Based on the flow time series of the stations, calculate the flow-related features. Specifically, these include: {peak flow pattern, flood duration, peak flow, peak water level, time when the flood reaches the peak flow, total flow of the flood, how long after the flood starts to reach the peak, duration after reaching the peak, and water level sequence}.

[0081] Grid-based rainfall calculation: Using the latitude and longitude information of rain gauges, the precipitation data is extended to the entire watershed through interpolation methods, thereby obtaining the rainfall distribution sequence of each grid unit within the watershed.

[0082] Rainfall feature extraction: Based on the gridded rainfall sequence, calculate rainfall-related features. Specifically, these include {rainfall sequence, total rainfall, hourly grid rainfall peak, rainfall trend, hourly rainfall sum, hourly rainfall center, storm center trajectory, hourly rainfall peak index, total rainfall for each grid point, maximum total rainfall for each grid point and its index}.

[0083] Feature merging: The calculated flow and rainfall characteristics are compared with existing flood and flow characteristics in the model library, and then merged or added as needed. During merging, a weighted average is used based on weighting factors. The calculation formula is as follows: T x =(T' x *n+T) / n+1; n=n+1. Where T is the calculated feature, T' x Similar features in the pattern library, where n is the weight factor, and the default weight factor for merged features is 1.

[0084] Pattern recognition and fusion: The calculated flood and discharge features are combined into new flood models. Each model may represent a specific flood development process, described by a series of discharge and rainfall features. These models are then matched with existing flood models in the model library using the Spatiotemporal Attribute Similarity Measurement (STDTW) algorithm. Based on the matching results, the following operations are performed:

[0085] If the similarity between a new feature and a pattern in the library reaches a preset standard, it is classified as an instance of that pattern, and the new feature is integrated into that pattern to optimize and improve the pattern.

[0086] If the similarity between a new feature and any of the patterns in the database does not meet the preset standard, it is considered to represent a new rainfall-flood pattern and is added to the pattern database as a new pattern for future analysis and prediction.

[0087] Similarity pattern fusion standard: If the similarity between the comparison pattern and the matched pattern ω=1 / 1+dist is greater than N / N+1, then the two patterns are considered similar patterns and are weighted and merged, where dist is the DTW distance and N is the number of current feature patterns.

[0088] The STDTW algorithm is formally expressed as follows: U = {the set of all characteristic factors of the flood}, X = {S(i)} k |0 <i<=N,N=|U|},Y=R k , F=(X,Y), D={F=(X,Y)}; where, S(i) k Let S(1) represent the set of feature factor patterns for the k-th historical flood. Let R represent the flow pattern for the k-th historical flood. There is a mapping relationship F between the feature factor set X and the flow pattern Y. The flood pattern is defined as: F(X,Y). The pattern library is represented by the symbol D, and its formal expression is as follows: D={(X,Y)}.

[0089] Pattern saving: Each flood pattern and newly added flow and rainfall characteristics are saved to the database for subsequent analysis and use.

[0090] Step 5: Use the historical flood pattern management module to view and manage the generated flood event patterns by watershed.

[0091] Select a watershed, view the topology image information and read the watershed ID; obtain the flood list: query the list of flood event IDs that occurred within the watershed based on the obtained ID; display the flood process: obtain the start and end times of the flood event based on the flood event ID, as well as the spatiotemporal characteristic sequence of the event, and display the time series and characteristic patterns of the flood event through a graphical interface.

[0092] This invention also proposes an intelligent online generation system for a rainstorm and flood pattern database, comprising a data import module, a rainstorm-flood alignment module, a flood event segmentation module, a pattern generation module, and a historical flood pattern management module; wherein:

[0093] Data import module: Reads data from uploaded files and saves it to the database;

[0094] Rainfall and flood alignment module: Time-calibrates the imported rainfall and flow data to provide time-series data for segmentation in the event segmentation module;

[0095] The flood season segmentation module: segments the start and end times of the flood and automatically adjusts them; it segments the time-series flow data of the aligned data based on the optimized FLUSS algorithm, segments the start and end times of the flood, and automatically adjusts the start and end times; the optimized FLUSS algorithm introduces a similar state crossing monitoring mechanism into the FLUSS algorithm.

[0096] The pattern generation module extracts the segmented time-series data, calculates multivariate flood characteristic values, and then fuses or adds to the existing characteristic patterns in the comparison pattern library.

[0097] Historical Flood Pattern Management Module: Allows for viewing and managing generated flood patterns by watershed, and enables operations such as deletion and modification.

Claims

1. A method for intelligent online generation of a storm flood pattern library, characterized in that, Includes the following steps: (1) Import data, extract station name and location information from station information file and read the corresponding station detection data file, read the time series data into the database, classify and save it, and manage the existing stations; (2) Automatically fill in the monitoring data of stations in the same watershed, fill discontinuous rainfall monitoring data into time series data with the same granularity as flow monitoring data and align them; Aligning rainfall and flow data is used for subsequent field segmentation. (3) The time series flow data is segmented based on the optimized FLUSS algorithm to separate the start and end times of the flood, and the start and end times are automatically adjusted; the optimized FLUSS algorithm introduces a similar state crossing monitoring mechanism into the FLUSS algorithm to realize the segmentation of long-term time series data. (4) Extract the segmented time series data, calculate the multivariate flood characteristic values, and realize the transformation of scene data into feature patterns; (5) Management should be differentiated and conducted based on watershed; The process of segmenting time-series traffic data based on the optimized FLUSS algorithm in step (3) is as follows: Divide the time-series traffic data into multiple subsequences of length length; Within a specified limit, calculate and compare the Euclidean distance between adjacent subsequences to assess their similarity. For each subsequence, determine the subsequence that is most similar to it, and record this similarity relationship as an arc; Count the number of these arcs spanning the domain at each time point to form a new sequence; The new sequence is compared with the expected number of cross-domain arcs when randomly distributed under ideal conditions to obtain a ratio sequence; In the ratio sequence, find the minimum value. The time point corresponding to the minimum value is the dividing point, which is used to divide the time series data into flood state and normal state. Store the start and end time points of the flood state in a flood array; The implementation process of step (4) is as follows: (41) Data reading: For each station, read the measurement data within the time period of the flood event, as well as the station's own information; (42) Flow feature extraction: Based on the flow time series of the station, calculate the flow-related features, specifically as follows: {flood peak pattern, flood duration, peak flow, peak water level, time when the flood reaches the peak flow, total flow of the flood, how long after the flood starts to reach the peak, duration after reaching the peak, duration after reaching the peak, water level sequence}; (43) Grid rainfall calculation: Using the latitude and longitude information of rain gauge stations, the precipitation data is extended to the entire watershed through interpolation methods, thereby obtaining the rainfall distribution sequence of each grid unit in the watershed; (44) Rainfall feature extraction: Based on the grid rainfall sequence, calculate the features related to rainfall, as follows: {rainfall sequence, total rainfall, hourly grid rainfall peak, rainfall trend, hourly rainfall sum, hourly rainfall center, rainstorm center point trajectory, hourly rainfall peak index, total rainfall for each grid point, maximum total rainfall and index for each grid point}; (45) Feature merging or addition: Compare the calculated flow and rainfall features with the existing flood and rainfall features in the model library, and select to merge or add them; (46) Pattern recognition and fusion: The calculated flood and flow characteristics are combined into new flood models. Each model represents a specific flood development process, described by a series of flow and rainfall characteristics. The models are matched with existing flood models in the model library using the spatiotemporal similarity measurement algorithm STDTW. The following operations are performed based on the matching results: If the similarity between a new feature and a pattern in the library reaches a preset standard, it is classified as an instance of that pattern, and the new feature is integrated into that pattern to optimize and improve the pattern. If the similarity between a new feature and all patterns in the database does not meet the preset standard, it is considered to represent a new rainfall-flood pattern and is added to the pattern database as a new pattern for future analysis and prediction. Similarity pattern fusion standard: If the similarity between the comparison pattern and the matched pattern ω=1 / 1+dist is greater than N / N+1, then the two patterns are considered similar and are weighted and merged; where dist is the DTW distance and N is the number of current feature patterns.

2. The intelligent online generation method of storm flood pattern library according to claim 1, characterized in that, The implementation process of step (1) is as follows: (11) New watershed information: Import new watershed data, which includes the watershed coverage area, administrative division, and geographical environment description; (12) Import station information: Read the station information document and import the station information contained in the newly created watershed. This information includes station identification information, station type, geographical location and time span of detection data; (13) Import station data: Read the station monitoring data document and import the station's historical monitoring data, which includes rainfall-related data and flow-related data.

3. The intelligent online generation method of storm flood pattern library according to claim 1, characterized in that, The implementation process of step (2) is as follows: (21) Read time series data: Read the monitoring data from all stations in the same watershed, divide them into rainfall data and flow data, find the maximum time span in the data, and perform interpolation preprocessing on the missing value part in the test data according to the maximum time span; (22) Select time span: Based on the existing station data coverage time, manually adjust the time range used for flood field division and cut off the redundant data; (23) Filling time series data: Use quadratic interpolation to fill the rainfall data into continuous time series data; (24) Aligning Rainfall and Flood Data: The complete rainfall and flow data are aligned and organized according to the start and end times of alignment at a uniform time interval, in the following format: {monitoring time, flow value, rainfall value 1, rainfall value 2...}, where the rainfall value is the measurement value of different rain gauge stations.

4. The intelligent online generation method for a rainstorm and flood pattern library according to claim 1, characterized in that, The automatic adjustment of start and end times described in step (3) is implemented as follows: Compile rainfall monitoring data within the flood timeframe. If the flood start and end times coincide with the rainfall time, adjust the flood start point forward according to the rainfall start time; if the flood end time coincides with the rainfall time, adjust the flood end point backward according to the rainfall end time.

5. The intelligent online generation method for a rainstorm and flood pattern library according to claim 1, characterized in that, The feature merging described in step (45) is performed using a weighted average based on the weighting factors, as shown in the following formula: T x = (T' x *n+T) / n+1 where T is the computed feature, T x similar features in the pattern library, and n is a weighting factor.

6. The intelligent online generation method for a rainstorm and flood pattern library according to claim 1, characterized in that, The implementation process of step (5) is as follows: Select the watershed and topological image information and read the watershed ID; query the list of flood event IDs that occurred within the watershed based on the obtained ID; obtain the start and end times of the flood event and the spatiotemporal characteristic sequence of the event based on the flood event ID, and display the time series and characteristic patterns of the flood event through a graphical interface.

7. A smart online generation system for a rainstorm and flood pattern library using the method described in any one of claims 1 to 6, characterized in that, It includes a data import module, a rainwater and flood alignment module, a flood event segmentation module, a model generation module, and a historical flood model management module; among which: Data import module: Reads data from uploaded files and saves it to the database; The rainwater alignment module performs time calibration on the imported rainfall and flow data, providing time-series data for segmentation by the event segmentation module; The flood season segmentation module divides the flood start and end times and automatically adjusts them; based on the optimized FLUSS algorithm, the aligned data is segmented into time-series flow data to divide the flood start and end times and automatically adjust the start and end times; the optimized FLUSS algorithm introduces a similar state crossing monitoring mechanism into the FLUSS algorithm. The pattern generation module extracts the segmented time-series data, calculates multivariate flood characteristic values, and then fuses or adds to the existing characteristic patterns in the comparison pattern library. The historical flood pattern management module allows for viewing and managing flood event patterns by watershed, and enables deletion and modification operations.

Images