A flood event classification method and system based on flood characteristics

Abstract

The application discloses a flood scene division method and system based on flood characteristics and belongs to the technical field of flood control. The existing scheme needs to set a flood peak threshold and other limit conditions artificially. Due to the subjectivity and non-uniformity of the standard, the flood scene extraction process is unreasonable. The flood scene division method based on flood characteristics constructs a flood characteristic simulation model, an event screening model, an event merging model and a scene division model, performs frequency analysis on flow data, obtains a flood peak threshold and is used for representing flood characteristics. According to the flood peak threshold, a plurality of flow data points higher than the flood peak threshold are screened out, and then the flow data points are marked as events with potential flood risks, so that the flood scene can be accurately screened, the flood peak threshold and other limit conditions need not be set artificially, the subjectivity and non-uniformity of the threshold set artificially are effectively avoided, and therefore the identification efficiency of flood events can be effectively improved.

Description

TECHNICAL FIELD

[0001] The application relates to a flood episode division method and system based on flood characteristics, and belongs to the technical field of flood control. BACKGROUND

[0002] A Chinese document (Li K, Liu KX, Liang LL. Automatic identification method of flood episode based on function properties Hydrology, 2021, 41(5): 26-31) proposes a method for automatically identifying flood episodes from continuous hydrological processes. The method is based on the concavity and convexity of the function and the limiting conditions of the flood peak threshold, the flood peak distance threshold, the start and end flow threshold, and the rain peak threshold to determine the flood event; and further analyzes the influence of the start and end flow threshold and the rain peak threshold on the identification result, so as to obtain the flood episode.

[0003] The above scheme needs to set the flood peak threshold and other limiting conditions artificially, and there is no public setting basis. The flood peak threshold is given artificially, and due to the subjectivity and different standards, the selection of the flood peak threshold often lacks consistency, and the complex relationship between flood and rainfall cannot be fully considered, which may lead to unreasonable flood episode extraction process; and in different basins, the size of the threshold often changes greatly, so the artificial setting of the flood peak threshold will have a large error, thereby affecting the accurate division of the flood episode.

[0004] The information disclosed in this background technology is only used to understand the background of the inventive concept, so it can include information that does not constitute prior art. SUMMARY

[0005] In view of the above problems or one of the above problems, the purpose of the present application is to provide a flood episode division method and system based on flood characteristics, which can accurately screen flood episodes without artificially setting flood peak threshold and other limiting conditions, effectively avoiding the subjectivity and different standards of artificial threshold setting, making the selection standard of the flood peak threshold tend to consistency, and thus effectively improving the identification efficiency of the flood event.

[0006] To solve the above problems or one of the above problems, the second object of the present application is to provide a flood episode division method and system based on flood characteristics, which can accurately screen flood episodes by calculating the flood peak threshold and flood interval and adjusting the relevant parameter threshold according to the basin characteristics and flood characteristics, and is not limited to identifying single-peak, multi-peak and other flood scenes; the start time and end time of each flood interval are fully considered in terms of rainfall and flow process to ensure that the parameter calibration requirements are met, thereby effectively improving the parameter calibration accuracy and work efficiency of flood forecasting, so that the flood episode can be accurately identified and thus widely applied in flood forecasting systems.

[0007] To achieve one of the above objects, the first technical solution of the present application is:

[0008] A flood episode division method based on flood characteristics, comprising the following steps:

[0009] Step one, using a pre-constructed flood characteristic simulation model, obtaining the flow data and rainfall data of a certain basin, and performing frequency analysis on the flow data to obtain the frequency distribution of the flow data; according to the frequency distribution, the flood peak threshold is calculated to represent the flood characteristics;

[0010] Step two, based on a pre-constructed event screening model, according to the flood peak threshold, screening a plurality of flow data points higher than the flood peak threshold from the flow data, and then marking the flow data points as events with potential flood risk;

[0011] Step three, using a pre-constructed event merging model, finding the maximum flow peak value and its occurrence time of each event, and obtaining the peak interval time of adjacent events; then according to the peak interval time, processing a plurality of events to generate flood events;

[0012] Step four, using a pre-constructed episode division model, obtaining the rainfall start and end times according to the rainfall data; and adjusting the time boundary of the flood events according to the rainfall start and end times and the flow data, to determine the flood episode.

[0013] The application can accurately screen flood scenes without setting a flood peak threshold and other limiting conditions, effectively avoiding the subjectivity and inconsistent standards of manually setting a threshold, making the selection criteria of the flood peak threshold consistent, thereby effectively improving the identification efficiency of flood events, and the scheme is scientific, reasonable and feasible.

[0014] As a preferred technical measure:

[0015] Step one, the method for obtaining the flow data and rainfall data of a certain basin by using the pre-constructed flood feature simulation model is as follows:

[0016] Collect the original flow data sequence, original rainfall data sequence and corresponding time sequence of the monitoring stations in a certain basin;

[0017] Detect the original flow data sequence and the original rainfall data sequence for mutation value respectively to obtain mutation data;

[0018] Replace the mutation data by using linear interpolation method, and complete the missing values in the original flow data sequence and the original rainfall data sequence to obtain a standard flow sequence and a standard rainfall sequence;

[0019] Based on the time sequence, the standard flow sequence and the standard rainfall sequence are processed for time consistency to obtain time stamps with unified format;

[0020] Resample the standard flow sequence and the standard rainfall sequence according to a predetermined time step to obtain flow data and rainfall data, so that the flow data and the rainfall data have the same time frequency.

[0021] As a preferred technical measure:

[0022] The method for detecting the original flow data sequence and the original rainfall data sequence for mutation value respectively to obtain mutation data is as follows:

[0023] Step 11, according to the historical flood data, the duration of the historical flood scenes is counted to obtain the interval range selected by the window algorithm, the window values in the interval range are cross-verified, and finally the optimal window value is obtained;

[0024] Step 12, obtain a raw flow data sequence and a raw rainfall data sequence, which respectively include a plurality of data points;

[0025] Step 13, according to the window value, intercept the data points within the window length to obtain window-in data; take the median of the window-in data to obtain a median; then calculate the residual of the window-in data and the median to obtain the absolute deviation of all flow data points from the median;

[0026] Step 14, based on the absolute deviation and the confidence interval of the abnormal value, a sensitive coefficient is calculated;

[0027] Step 15, based on the sensitive coefficient and the absolute deviation, it is judged whether each data point exceeds the upper limit of the abnormal value or the lower limit of the abnormal value; when the data point does not exceed the upper limit of the abnormal value or the lower limit of the abnormal value, the data point is taken as normal data;

[0028] When the data point exceeds the upper limit of the abnormal value or the lower limit of the abnormal value, the data point is taken as mutation data.

[0029] As a preferred technical measure:

[0030] Based on the absolute deviation and the confidence interval of the abnormal value, the sensitive coefficient is calculated as follows:

[0031] Set the sensitive coefficient value range and the calculation step;

[0032] Based on the calculation step, a plurality of sensitive values are generated;

[0033] According to the plurality of sensitive values and the absolute deviation, a plurality of abnormal threshold values are calculated;

[0034] Based on the plurality of abnormal threshold values, a data quantile diagram is drawn to obtain the confidence interval of each abnormal threshold value, and confidence data is obtained; according to the confidence data, a sensitive value is selected as a sensitive coefficient.

[0035] As a preferred technical measure:

[0036] Step 1, using a pre-constructed flood feature simulation model, frequency analysis is performed on the flow data to obtain a frequency distribution of the flow data; according to the frequency distribution, the flood peak threshold is calculated as follows:

[0037] The flow data is subjected to frequency analysis, and the flow points in the flow data are arranged in ascending order to obtain a frequency distribution;

[0038] According to the frequency distribution, the cumulative frequency of each flow point is calculated to obtain frequency data;

[0039] Based on the frequency data and the flood characteristics, a cumulative threshold is determined;

[0040] The flow point corresponding to the cumulative threshold is taken as the flood peak threshold.

[0041] As a preferred technical measure:

[0042] Step two, based on the pre-constructed event screening model, according to the flood peak threshold, several flow data points higher than the flood peak threshold are screened out from the flow data, and then the flow data points are marked as the method of events with potential flood risk as follows:

[0043] Step 21, according to the flood peak threshold, several flow data points higher than the flood peak threshold are screened out from the flow data;

[0044] Step 22, sort the screened several flow data points and mark their indexes;

[0045] Step 23, the flow data points with continuous indexes are marked as an event, and the interval set including the flow process data of each event is obtained;

[0046] Step 24, according to the objective function and constraint condition, the interval set is processed for multiple iterations to obtain the optimal interval set;

[0047] Step 25, based on the optimal interval set, the event with potential flood risk is determined.

[0048] As a preferred technical measure:

[0049] Step 24, according to the objective function and constraint condition, the interval set is processed for multiple iterations to obtain the optimal interval set as follows:

[0050] Firstly, based on the interval number m of the interval set, a set list with a length of is created, and the values in the set list are 0 or 1; each value in the set list represents whether the adjacent two intervals are merged, when the value is "0", it means keeping the current position and the interval is not merged; when the value is "1", it means merging the left and right intervals at this position;

[0051] Then, a group of 0 or 1 sequences is randomly generated to assign values to the set list to form a candidate merging scheme;

[0052] Based on the characteristics of flood, the objective function and constraint condition are constructed, the objective function is the deviation degree of the maximum flow value of a certain interval and the 99th percentile of the interval set; the constraint conditions include the length of the merged single interval, the index threshold and the flow difference threshold;

[0053] According to the objective function and constraint condition, the candidate merging scheme is iterated and screened to obtain the optimal binary sequence;

[0054] Map the binary sequence back to the set list to get the final merged interval list;

[0055] De-duplicate the final merged interval list to make all intervals unique and convert each interval to tuple form to generate the optimal interval set.

[0056] As a preferred technical measure:

[0057] Step three, using the pre-constructed event merging model, find the maximum flow peak value of each event and its occurrence time, and get the peak interval time of adjacent events; then according to the peak interval time, process several events to generate the flood event method as follows:

[0058] Obtain the basic data of a certain basin; the basic data includes the area of the basin, the benchmark minimum interval time and the area unit conversion coefficient;

[0059] Build a time compression formula to quantify and standardize the growth rate of the time interval between flood events;

[0060] Substitute the area of the basin and the area unit conversion coefficient into the time compression formula to get the flood interval time increment;

[0061] Add the flood interval time increment and the benchmark minimum interval time to get the independence test threshold;

[0062] Obtain a plurality of events with potential flood risk, and find the maximum flow peak value of each event and its occurrence time;

[0063] According to the occurrence time of the maximum flow peak value of each event, get the peak interval time of adjacent events;

[0064] Compare the peak interval time with the independence test threshold. If the peak interval time between two adjacent events is less than the independence test threshold, merge the two events into one event;

[0065] The merged event is the flood event.

[0066] As a preferred technical measure:

[0067] Step four, using the pre-constructed game division model, according to the rainfall data, obtain the rainfall start and end time; and according to the rainfall start and end time and the flow data, adjust the time boundary of the flood event to determine the flood game method as follows:

[0068] Step 41, according to the rainfall start and end time of the rainfall data, find from the starting position of each flood event until N consecutive zero rainfall values are found to determine the new starting position;

[0069] Step 42, according to the flow data, starting from the end position of each flood event, find backward until finding the continuous decreasing flow value to determine the new end position;

[0070] Step 43, according to the new start position and the new end position, determine the time boundary of the flood event;

[0071] Step 44, optimize the time boundary of the flood event, so that the time interval of each flood event does not overlap;

[0072] Step 45, the optimized flood event is used as the flood episode.

[0073] To achieve one of the above purposes, the second technical solution of the application is:

[0074] A flood episode division system based on flood characteristics, comprising:

[0075] One or more processing units;

[0076] Storage device for storing one or more programs;

[0077] When the one or more programs are executed by the one or more processing units, the one or more processing units implement the above-mentioned flood episode division method based on flood characteristics.

[0078] Compared with the prior art, the application has the following beneficial effects:

[0079] The application constructs a flood characteristic simulation model, an event screening model, an event merging model and an episode division model, performs frequency analysis on the flow data to obtain the frequency distribution of the flow data, calculates the flood peak threshold according to the frequency distribution, uses the flood peak threshold to represent the flood characteristics, screens a plurality of flow data points higher than the flood peak threshold from the flow data according to the flood peak threshold, and then marks the flow data points as events with potential flood risk, processes a plurality of events according to the peak interval time to generate flood events, and finally adjusts the time boundary of the flood events according to the rainfall start and end time and the flow data, so that the flood episode can be accurately screened without setting the flood peak threshold and other limiting conditions artificially, the subjectivity and inconsistent standards of the threshold set artificially are effectively avoided, the selection standard of the flood peak threshold tends to be consistent, and the identification efficiency of the flood event is effectively improved.

[0080] Further, the application can accurately screen flood scenes by calculating the flood peak threshold and flood interval and adjusting the relevant parameter threshold according to the basin characteristics and flood characteristics, and is not limited to identifying single-peak, multi-peak and other flood scenes; the starting time and ending time of each flood interval are fully considered in terms of rainfall and flow process, so as to ensure that the parameter rating requirements are met, effectively improve the parameter rating accuracy and work efficiency of flood forecasting, and thus the flood scenes can be accurately identified, and the application can be widely applied to flood forecasting systems. BRIEF DESCRIPTION OF DRAWINGS

[0081] Figure 1 is a flow chart of the flood scene division method based on flood characteristics of the application;

[0082] Figure 2 is a flood scene division effect diagram of a single-peak flood of a certain basin by applying the application;

[0083] Figure 3 is a flood scene division effect diagram of a multi-peak flood of a certain basin by applying the application. DETAILED DESCRIPTION

[0084] In order for those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should be within the scope of protection of the present application. The present application covers any alternative, modification, equivalent method and scheme within the spirit and scope of the present application defined by the claims.

[0085] As shown in Figure 1 , the first specific embodiment of the flood scene division method based on flood characteristics of the application is as follows:

[0086] A flood scene division method based on flood characteristics comprises the following steps:

[0087] Step 1: using a pre-constructed flood characteristic simulation model, obtaining the flow data and rainfall data of a certain basin, and performing frequency analysis on the flow data to obtain the frequency distribution of the flow data; according to the frequency distribution, calculating the flood peak threshold for representing the flood characteristics;

[0088] Step 2: based on a pre-constructed event screening model, according to the flood peak threshold, screening a plurality of flow data points higher than the flood peak threshold from the flow data, and then marking the flow data points as events with potential flood risk;

[0089] Step three, using the pre-constructed event merging model, find the maximum flow peak value of each event and its occurrence time, and get the peak interval time of adjacent events; then according to the peak interval time, process several events to generate flood events;

[0090] Step four, using the pre-constructed game division model, according to the rainfall data, obtain the rainfall start and end time; and according to the rainfall start and end time and the flow data, adjust the time boundary of the flood event, so as to determine the flood game.

[0091] The second specific embodiment of the flood game division method based on flood characteristics provided by the application is as follows:

[0092] A flood game division method based on flood characteristics comprises the following contents:

[0093] Collect the original flow data and original rainfall data of a certain basin; and pre-process the original flow data and original rainfall data to obtain flow data and rainfall data with time consistency;

[0094] Perform frequency analysis on the flow data to obtain the frequency distribution of the flow data; according to the frequency distribution, calculate the flood peak threshold value for representing the flood characteristics; according to the flood peak threshold value, screen several flow data points higher than the flow threshold value from the flow data, and then mark the flow data points as events with potential flood risk;

[0095] Find the maximum flow peak value of each event and its occurrence time, and get the peak interval time of adjacent events; when the peak interval time of two events is less than a preset time threshold value, merge the two events into one event to obtain a flood event;

[0096] According to the original rainfall data, obtain the rainfall start and end time; and according to the rainfall start and end time, adjust the time boundary of the flood event, so that the start and end time of the flood event matches the rainfall start and end time, to generate a flood game.

[0097] The third specific embodiment of the flood game division method based on flood characteristics provided by the application is as follows:

[0098] A flood game division method based on flood characteristics comprises the following steps:

[0099] S1, collect data and pre-process the data, which comprises the following contents:

[0100] Firstly, collect the flow and rainfall data of a specific basin, pre-process the collected data, and use a linear interpolation algorithm to process missing values and abnormal values (for example, negative values or unreasonable high values) to perfect the data set. Finally, ensure that all data are consistent in time, and unify the time stamp format and data frequency.

[0101] S2, construct a flood feature simulation model and an event screening model to preliminarily screen flood events, which includes the following contents:

[0102] After completing the data preprocessing, the flood feature simulation model is used to analyze the frequency distribution of the flow sequence data by frequency analysis of the flow data, and according to the frequency distribution, the 80% percentile is used as the flow threshold (flood peak threshold). Then, the event screening model is used to screen out the flow data points higher than the flow threshold, and mark these data points as events with potential flood risk.

[0103] S3, construct an event merging model to test the independence of events, which includes the following contents:

[0104] Find the maximum flow peak value of each event and its occurrence time, and compare the peak interval time of adjacent events. Combine the continuous and threshold-exceeding data points into a single flood event. If the interval between two adjacent events is less than the preset time threshold, the two events will be merged into one event to avoid over-segmentation of events caused by small data recording gaps or small-scale fluctuations.

[0105] S4, construct a field division model to adjust the boundary of the flood event, which includes the following contents:

[0106] Combine the rainfall data with the flow data to adjust the time boundary of the identified flood event. According to the start and end of rainfall and the change of flow, re-evaluate and adjust the start and end time of the flood event to ensure that the time limit of the event matches the actual rainfall and flow change.

[0107] The present application can accurately screen flood events by calculating the flood peak threshold and flood interval. The present application is not limited to identifying single-peak floods or multi-peak floods, and can adjust the flood parameter threshold according to the characteristics of the flood. The start time and end time of each flood interval are fully considered in terms of rainfall and runoff process to ensure that the parameter rate setting requirements are met, effectively improving the parameter rate setting accuracy and work efficiency of flood forecasting, thereby improving the efficiency of flood event identification, so that the flood event can be accurately identified, and the present application can be widely applied in flood forecasting systems.

[0108] In this embodiment, the specific method of step S1 is as follows:

[0109] First, collect the flow and rainfall data sequences of the monitoring stations in a specific watershed. Among them, the rainfall monitoring station provides a historical rainfall data sequence , the flow monitoring station provides a flow data sequence , and the corresponding time sequence .

[0110] Then, all data sequences are preprocessed, and the rainfall and flow data of each monitoring site are respectively subjected to outlier detection. Since the rainfall and flow changes of natural processes should be continuous, the outliers are defined as the abrupt values appearing in the data sequences, and the identification method is the moving window algorithm. The specific steps are as follows:

[0111] Step 1, set the window size, according to the length of the flood time in the historical flood data, the length of the window can be confirmed, the length of the historical flood scene is counted, the interval range of the window selection is obtained, the cross-checking is carried out according to the window value in the interval range, and finally the optimal window value is 5, that is, the previous and subsequent 5 data points are selected.

[0112] Step 2, calculate the sliding window statistics, intercept the flow data in the window length n, take the median of the data in the window, obtain the median MID, calculate the residual of the data in the window and the median, and obtain the median of the difference between all data points and the median, that is, the median absolute deviation MAD, the specific calculation formula is as follows:

[0113] When n is an odd number, the calculation formula of the median MID is as follows:

[0114]

[0115] When n is an even number, the calculation formula of the median MID is as follows:

[0116]

[0117] Wherein, is the rainfall or flow data at the i-th moment, and n is the window length.

[0118] Further, the calculation formula of the median absolute deviation MAD is as follows:

[0119]

[0120] Wherein, is the median calculation function.

[0121] Step 4, detect outliers, judge whether the current data is beyond the upper and lower limits of the change outliers, and the expression is as follows:

[0122]

[0123] Wherein, is the sensitive coefficient, is the current data.

[0124] The data point at a certain moment can be excluded from the smooth state to suddenly increase or decrease greatly, and the data trend presents a sharp condition.

[0125] In this embodiment, the sensitivity coefficient The value confirmation method is as follows:

[0126] The sensitivity analysis is performed on the value, and the changes of the detection result and the subsequent model performance are observed when the value changes within a certain range, and finally a value is selected which makes the model prediction performance optimal.

[0127] The value range is set, the value range is 4 to 6, and a certain step (optionally 0.5) is used for traversal.

[0128] The anomaly detection is performed on each value, the anomaly threshold value is calculated using each value, and the optimal value is selected according to the effect, and the calculation formula is as follows:

[0129]

[0130] The data Q-Q graph is drawn, and it is observed which value makes the confidence interval 95%, so as to maximize the distinction of abnormal values.

[0131] For the abnormal values and missing values in the data sequence of each monitoring site, the linear interpolation method is used for completion. For the data value missing at the time point , the known data point at the previous adjacent time and the known data point at the next adjacent time are used to calculate, and the calculation formula is as follows:

[0132]

[0133] In this embodiment, the timestamps of all monitoring sites are converted into a unified format, which is “YYYY (year) -MM (month) -DD (day) HH (hour): MM (minute): SS (second)”, to realize the time consistency processing of data.

[0134] In this embodiment, the method for unifying the data frequency is as follows:

[0135] According to the predetermined time step (such as every hour or every day), the rainfall and flow data of all monitoring sites are resampled to ensure that each data sequence has the same time frequency. For the original data frequency higher than the target frequency, the average value within the time window is used for down-sampling, and the specific calculation formula is:​​​

[0136]

[0137] wherein, is the down-sampled data, is the original data point within the time window, and N is the number of data points within the time window.

[0138] For the original data whose frequency is lower than the target frequency, linear interpolation is used for up-sampling, and the newly added data points are completed by the following formula :

[0139]

[0140] wherein, and are the times of adjacent known data points, and are the corresponding values, is the time point that needs to be interpolated.

[0141] Before the screening of flood events, strict data quality inspection and data cleaning and completion work must be carried out. The data processing method of the present application can effectively reduce the burden of manual operation, improve work efficiency, and also enhance the accuracy of event division and the reliability of event extraction.

[0142] In the present embodiment, the specific method of step S2 is as follows:

[0143] Firstly, the frequency analysis is carried out on the flow data, the data is arranged in ascending order, and then the cumulative frequency of each flow value is calculated , and the calculation formula is as follows:

[0144]

[0145] wherein, represents the order position of the current data point in the sequence, is the total number of data points.

[0146] The purpose of this step is to identify the flow interval with flood characteristics and exclude the influence of base flow. For this purpose, the minimum flow method Tennant is used to determine the percentage of base flow. Generally, the average 20% base flow standard is considered reasonable, and the percentage of the remaining 80% flow can be used for the identification of flood events, that is, it meets:

[0147]

[0148] wherein, is the flood peak threshold (flow threshold).

[0149] According to the test, this percentile is considered suitable to identify all flood events of watersheds with different characteristics.

[0150] The screened data points are then sorted and labeled with their indices. Data points above the flow threshold and with consecutive indices are merged into an event, and finally a set of intervals "groups" is returned, which contains all the flow process data belonging to flood events.

[0151] In this embodiment, the specific method of step S3 is as follows:

[0152] The flood set obtained in step S2 is iteratively processed by merging intervals that do not meet the threshold of independence test, to obtain a set of independent flood intervals. The core goal of this process is to ensure that the independence of flood events is effectively judged, avoiding false division due to data record missing, short-term fluctuations or other human factors. On this basis, through reasonable setting of the threshold of independence test, the real flood events can be more accurately reflected, providing reliable basis for subsequent water resources management and flood risk assessment.

[0153] Based on the consideration of data record gaps, filtering of small-scale fluctuations and maintenance of event continuity, the shortest interval time between two consecutive peaks (i.e. the maximum flow points of two flood events) is selected as the threshold of independence test.

[0154] The consideration of data record gaps includes the following:

[0155] During the recording of flood events, due to equipment failure, data transmission problems or gaps in manual recording, the actual flood event may be split into multiple events. If the actual interval time between two flood events is less than the preset threshold, it may only be due to these technical or recording problems, rather than a truly independent flood event. Therefore, setting a reasonable threshold can avoid misjudging these small-scale fluctuations as new events, and thus avoid false segmentation due to data missing or technical problems.

[0156] The filtering of small-scale fluctuations includes the following:

[0157] During the flood process, fluctuations or declines in flow do not necessarily mean the occurrence of a new flood event. Often due to weather or other factors, flow will fluctuate in the short term, and these fluctuations are usually only small-scale changes in a continuous flood event and should not be considered as independent events. Therefore, during the independence test, by setting an appropriate threshold, it is ensured that only when the peak interval exceeds this threshold, a new independent flood event is identified, avoiding misjudgment of small-scale fluctuations.

[0158] The maintenance of event continuity includes the following:

[0159] Flood events usually have a certain continuity, and the time interval between two peaks can reflect the duration of the flood event to some extent. If the threshold is too low, it may lead to fragmentation of the same event, making the flood event with strong continuity be misjudged as multiple independent events. Therefore, by setting a reasonable threshold, the continuity of flood events can be effectively maintained, ensuring that the overall nature of the event is not disturbed. Different purposes and standards will affect the setting of independence test. For example, if the purpose is to compile flood scenes in water resource management, it may be necessary to refer to the regulations of the Water Resources Commission or other relevant agencies. Therefore, when setting the threshold of independence test, it needs to be adjusted according to actual needs.

[0160] In this embodiment, an empirical formula based on watershed area is used to determine the threshold of independence test, and the calculation formula is as follows:

[0161]

[0162] wherein, is the shortest interval time (days), is the reference minimum interval time (days), is the area unit conversion coefficient, A is the watershed area (km²).

[0163] (reference minimum interval time): This parameter represents the minimum time interval between two flood events. Usually this value is determined according to the regulations of water resource management or historical flood data. For example, some water resource management agencies stipulate that consecutive flood events should be at least 5 days apart, so can be set to 5 days. In different scenarios, can be adjusted according to actual management needs to ensure compliance with specific management objectives.

[0164] (area unit conversion coefficient): This coefficient is used to convert the unit of watershed area. For example, if the unit of watershed area is square kilometers (km²), the value of is 1.609 (corresponding to 1 mile ≈ 1.609 kilometers). If the original area unit is other units, the value of needs to be adjusted according to the conversion relationship to ensure consistency with the unit of watershed area.

[0165] A (Basin Area): Basin area is a key factor in determining the minimum interval time θ. Studies have shown that there is a logarithmic relationship between basin area and flood event independence. The larger the basin area, the longer the interval between flood events. This is because the flood convergence range of larger basins is wider, and multiple flood events may be influenced by similar meteorological conditions and geographical environments, resulting in certain correlation. Therefore, the larger the basin area, the more significant the influence on flood event independence.

[0166] Basin area can be calculated and obtained through various methods, mainly including the following approaches:

[0167] Geographic Information System (GIS): GIS technology can provide accurate basin division and area calculation. By obtaining digital elevation model (DEM) data and vector data of basin boundaries, the basin is divided in GIS software, and the area of the basin is calculated using GIS tools after determining the hydrological boundary, usually in square kilometers. This method has the advantages of high precision and reliability, and can provide accurate area data for subsequent independence testing.

[0168] Online resources or existing data: In some cases, existing basin area data provided by national hydrological data centers, local governments or research institutions can be used. These data are usually calculated and audited by professionals and have high accuracy. If the basin area data is already public, using these resources can save calculation time and directly obtain basin area information.

[0169] Based on the fitting analysis of the independence data of multiple basin flood events, the natural logarithm (ln) in the formula is obtained. Through statistical analysis of historical flood data of different basins, it is found that there is a logarithmic relationship between basin area and flood convergence. The natural logarithm can reflect the influence of basin area on flood event independence. Specifically, the larger the basin area, the wider the catchment range, and the greater the time difference in the convergence process, which may lead to the extension of the interval between flood events. Specifically, in large basins, due to the convergence of precipitation in different places within the basin and gradually converging to the downstream, the propagation process of flood is relatively slow, which may lead to multiple flood events occurring at the same time with similar characteristics, thus the time interval between events is large.

[0170] Therefore, basin area has a significant impact on flood event independence. The larger the basin area, the more significant the mutual influence between flood events, and the smaller the basin area, the stronger the independence between events. Therefore, the formula using logarithmic relationship can effectively reflect this phenomenon.

[0171] ​In the formula, the natural logarithm (ln) plays a role in balancing the relationship between the basin area and the independence of flood events. Using the natural logarithm for calculation can compress the influencing factors of the basin area. Since the time interval between flood events increases with the increase of the basin area, using the natural logarithm can better quantify and standardize this growth rate.

[0172] In this step, based on the genetic algorithm, a multiple iteration loop is set. The current interval set "groups" is processed as follows:

[0173] Step 1, create a new list "coding" to store the coded representation of the interval set. If there are sets of original intervals, the list "coding" is a list of length , with values of 0 / 1. Each value in the list represents whether the adjacent two sets are merged or not. A value of "0" indicates that the current position is not merged, and a value of "1" indicates that the left and right intervals are merged at that position.

[0174] Step 2, initialize the list, which includes the following contents:

[0175] Randomly generate multiple 0 / 1 switch sequences, each of which is a "candidate merging scheme".

[0176] Ensure that the number of "on" in the switch does not exceed 30% of the total switch to prevent excessive merging at once.

[0177] Step 3, construct the objective function and constraints, the objective function is used to represent the deviation of the maximum value of the interval from the 99th percentile of the overall target value array , the calculation formula is as follows:

[0178]

[0179] Constraints include the following:

[0180] According to the regulations of water resources management agencies, consecutive flood events should be at least 5 days apart, so the length of the merged interval should be greater than or equal to 5; the difference between the two maximum value indexes corresponding to the merging point should be less than the index threshold `theta`; the relationship between the minimum value of the interval flow and the maximum value of the flow should satisfy the following expression:

[0181]

[0182] ​Step 4, cross operation is performed, that is, 1-2 cut points are randomly selected in the initial solution, the paternal gene fragment is exchanged to generate offspring of half the number of parents. Mutation operation is performed, that is, the 0 / 1 state of a bit is flipped with a small probability

[0183] Step 5, iterative update is performed, and a number of individuals with the highest fitness are directly entered into the next generation to prevent loss of good solutions.

[0184] A termination criterion is constructed, and the algorithm is stopped when the maximum number of generations is reached.

[0185] When the optimal fitness variation amplitude of the last 5 generations is less than a small threshold value, it is considered that the algorithm has converged, and the algorithm is terminated in advance.

[0186] Step 6, the binary sequence of the optimal individual is mapped back to the interval merging decision to obtain a final merging interval list "new_group". After iteration, the final interval set "new_group" is processed to ensure that all intervals are unique. Each interval is converted to a tuple form, and the set is used to remove duplicate intervals. Finally, the de-duplicated interval set "unique_new_group" is returned as the final result.

[0187] Step 7, output the unique interval set, along with the optimal fitness and evaluation index: the deviation of the maximum interval flow from the 99th percentile of the overall target value array.

[0188] In this embodiment, the specific method of step S4 is as follows:

[0189] S1. Based on the precipitation data, the starting position is adjusted, which includes the following contents:

[0190] To determine the new starting position, the starting position of each flood episode interval is searched forward until three consecutive zero precipitation values are found to determine the new starting position The expression is as follows:

[0191]

[0192] Wherein, represents the precipitation at time t, represents the precipitation at time t, represents the precipitation at time t.

[0193] ​​​​​​The rainfall amount of three consecutive zero values is selected as the judgment standard to ensure the starting time of the rainfall event, avoid identification errors caused by recording errors, and prevent short non-rainfall periods in the rainfall process from being misjudged as the end of rainfall.

[0194] S2. Based on the precipitation data, the end position is adjusted, which includes the following contents:

[0195] From the end position of the interval Start searching backward until a continuously decreasing flow value is found to determine the new end position , whose expression is as follows:

[0196]

[0197] Wherein, is the flow value at the moment, is the flow value at the moment, is the flow value at the moment.

[0198] S3. All screened, combined and adjusted intervals are collected to form an optimized interval set. The final optimized interval set is represented as multiple non-overlapping time intervals, such as , each interval corresponds to a specific flood event, wherein and represent the start and end times of the flood event, respectively.

[0199] A specific embodiment of applying the present application to divide the flood events of a certain basin:

[0200] A flood event division method based on flood characteristics is applied to divide the flood events of a certain basin, and the steps include:

[0201] Step S1, first, collect the flow and rainfall data sequences of each monitoring station in a specific basin (a certain basin). Specifically, it includes:

[0202] Rainfall data: each rainfall monitoring station provides a historical rainfall data sequence ;

[0203] Flow data: each flow monitoring station provides a historical flow data sequence ;

[0204] Time sequence: corresponding time point sequence .

[0205] Then, the rainfall and flow data of each monitoring site are respectively subjected to outlier detection. Since the rainfall and flow changes in natural processes should be continuous, the outliers are defined as the abrupt values appearing in the data sequence. For example, the flow data suddenly drops to 0 or reaches an obviously unreasonable high value.

[0206] In this embodiment, the moving window algorithm includes the following contents:

[0207] The window length is determined according to the time length of the flood in the historical flood data, the time length of the historical flood scene is counted, the interval range of the window selection is obtained, the cross-validation is performed according to the window value in the interval range, and finally the optimal window value is 5, that is, the previous and subsequent 5 data points are selected.

[0208] In this embodiment, the method for calculating the sliding window statistic is as follows:

[0209] The flow data in the window length n is intercepted, the median of the data in the window is obtained, the residual of the data in the window and the median is calculated, and the median absolute deviation (MAD) of the difference between all data points and the median is obtained.

[0210] Then, the data are subjected to outlier detection and missing value processing to obtain a plurality of data points meeting the requirements; and the plurality of data points are subjected to time consistency processing to unify the time format between the data points, thereby obtaining the flow data and rainfall data with unified time format.

[0211] The rainfall and flow data of all monitoring sites are resampled according to a predetermined time step (such as every hour or every day) to ensure that the data sequences have the same time frequency. The resampling includes downsampling and upsampling.

[0212] In this embodiment, the method for downsampling is as follows:

[0213] For the original data with a frequency higher than the target frequency, downsampling is performed. For example, the data every minute is converted into data every hour, and the sum or average value method is usually used.

[0214] In this embodiment, the method for upsampling is as follows:

[0215] For the original data with a frequency lower than the target frequency, the newly added data points are processed by upsampling and combining with the missing value completion method. For example, the data every day is converted into data every hour, and the linear interpolation method is used to fill in the newly added data points every hour.

[0216] In step S2, the method for threshold setting and preliminary screening is as follows:

[0217] Firstly, the flow data is arranged in ascending order, and then the cumulative frequency of each flow value is calculated .

[0218] By analyzing the cumulative frequency distribution, the flow value corresponding to the 80th percentile of the flow data of a certain watershed can be determined , that is, it meets the flood characteristics, and its expression is as follows:

[0219]

[0220] wherein, is the flood peak threshold (flow threshold).

[0221] The sorted data points are sorted and labeled with their indices. The data points above the threshold and with consecutive indices are merged into an event, and finally an interval set groups is returned, which contains all the flow process data belonging to the flood event.

[0222] Next, based on the genetic algorithm, the interval set of the flood is processed multiple times to obtain an interval set with optimal fitness and evaluation index, including the following steps:

[0223] First, create a new list , the original interval has sets, is a list of length , with values of 0 / 1. Randomly generate 60 0 / 1 switch sequences, each of which is a "candidate merging scheme". When setting the initial candidate solution, the number of "on" in the switch is not more than 30% of the total switch to prevent excessive merging at once.

[0224] According to the objective function and the constraint condition, the list is iterated and filtered, and the objective function is the deviation of the maximum value of the interval from the 99th percentile of the overall target value array , and its calculation formula is as follows:

[0225]

[0226] The constraint conditions include the length of the merged single interval, the index threshold, and the flow difference threshold.

[0227] The length of the single interval is greater than or equal to , the index threshold is the difference between the two maximum index values corresponding to the merging point , and the flow difference threshold is the relationship between the minimum flow value of the interval and the maximum flow value , and its expression is as follows:

[0228]

[0229] In the iteration process, the algorithm will perform crossover and mutation operations. Randomly select 1-2 cut points in the initial solution, exchange the parent gene fragments to produce 30 offspring. With a small probability , flip the 0 / 1 state of a certain bit. Keep the top 30 individuals with the highest fitness directly to the next generation to prevent loss of good solutions. When the preset maximum number of generations is reached or the optimal fitness changes by less than a small threshold for 5 consecutive generations, the algorithm will terminate.

[0230] After the iteration is terminated, the binary sequence of the optimal individual is mapped back to the interval merging decision to obtain the final merging interval list "new_group". De-duplicate the final interval set "new_group" to ensure that all intervals are unique. Convert each interval to a tuple form, remove duplicate intervals using set de-duplication, and finally return the de-duplicated interval set "unique_new_group" as the final result.

[0231] Return the unique interval set along with the optimal fitness and evaluation indicators: the maximum interval flow and the deviation of the 99th percentile of the overall target value array.

[0232] Step S3, independence test of events, independence test is to ensure that the identified flood events are independent of each other, to avoid repeated identification of events due to the continuity of data records. The key to this step is to reasonably define the boundaries of events and test the independence of adjacent events.

[0233] First, calculate the independence test threshold , the formula is:

[0234]

[0235] Where, is the basin area (unit: square kilometers).

[0236] Step 4: Adjust the starting position based on precipitation data, i.e. find the new starting position by searching forward from the starting position of each interval until three consecutive zero precipitation values are found , the calculation formula is as follows:

[0237]

[0238] Where, represents the precipitation at time k.

[0239] Adjust the end position: start from the end position of the interval , search backward until a consecutive decrease in the target value is found to determine the new end position , the calculation formula is as follows:

[0240]

[0241] All the screened, combined and adjusted intervals are collected to form an optimized interval set. The final optimized interval set is represented as a plurality of non-overlapping time intervals, such as , each interval corresponding to a specific flood event, wherein and represent the start and end times of the flood event, respectively.

[0242] As shown in Figure 2 , based on the flow data (unit: cubic meters / second) and rainfall (unit: millimeter), the present application can accurately identify and divide the flood scenes of a single-peak flood in a certain watershed, and obtain an accurate flood scene division result, which can be seen in the red area in Figure 2 .

[0243] The division of the entire flood scene presents a clear three-section structure, mainly divided into a rising period, a peak period and a falling period.

[0244] The rising period is the stage from the initial point of flood response to before the flow continuously rises and reaches the high threshold. The peak period is the stage after the flow exceeds the target percentile and the duration reaches the minimum length threshold. As can be seen from Figure 2 , the application can accurately mark the start and end times of the "peak period". The start time is the moment when the uplink curve first crosses the threshold, and the end time is when the downlink curve falls below the threshold. The falling period is the stage from the end point of the peak period to the end when the flow returns to the baseline, forming a continuous and non-missing "falling period".

[0245] The peak of the rainfall curve and the peak of the flow curve have obvious time correspondence. The flow curve peak appears after a certain time of the rainfall peak, indicating that in the division of the "rising-peak-falling" three sections, the rainfall input information is effectively utilized, verifying the accuracy of the present application method. In the rising period and the falling period, short and small amplitude oscillations are ignored by the present application model, maintaining the continuity of the sections, avoiding the analysis difficulty brought by excessive segmentation, and seamlessly connecting each section without missing or overlapping.

[0246] As shown in Figure 3 , based on the flow data and rainfall, the present application can accurately identify and divide the flood scenes of a multi-peak flood in a certain watershed, and obtain an accurate flood scene division result, which can be seen in the red area in Figure 3 .

[0247] The algorithm of the present application successfully distinguates each flow peak. The first peak, the second peak (and more) have independent "pre-rising-peak-falling" paragraphs, and the start and end points of the paragraphs are strictly determined according to the threshold crossing points of each paragraph, without merging two peaks into one, or artificially truncating the same peak.

[0248] In the trough period between the first falling and the next pre-rising, the division result is automatically identified as "interval period", and the length meets the minimum duration threshold, and is not misjudged as residual fluctuation. This ensures that different flood peak events do not interfere with each other, and facilitates subsequent separate extraction of each wave feature. Each rainfall pulse is mapped one-to-one with the corresponding flow peak. The division result shows that in the case of multiple peaks, a new pre-rising period enters soon after the rainfall peak appears. It can be seen that the present application is particularly sensitive to the switching of multiple rain-flow responses, and adapts to complex weather scenarios.

[0249] A system embodiment applying the method of the present application:

[0250] A flood event division system based on flood characteristics, comprising:

[0251] One or more processing units;

[0252] Storage device for storing one or more programs;

[0253] When the one or more programs are executed by the one or more processing units, the one or more processing units implement the above-mentioned flood event division method based on flood characteristics.

[0254] The storage device is an internal memory or an external memory or a cache memory or other special memory. The processing unit has signal processing capability, which can be a general-purpose processor, a digital signal processor, an application-specific integrated circuit, a ready-to-program gate array or other programmable logic device.

[0255] A device embodiment applying the method of the present application:

[0256] An electronic device is provided with a computer readable storage medium, and the computer readable storage medium stores a computer program, which, when executed by a processing unit, implements the above-mentioned flood event division method based on flood characteristics.

[0257] The computer readable storage medium refers to a physical carrier capable of storing computer-recognizable data, instructions or programs, which meets the core characteristic of "being readable by a computer" (i.e. the data exists in the form of electrical, magnetic, optical signals, which can be converted into binary information that can be processed by a computer through corresponding devices). The physical carrier is a magnetic storage medium, an optical storage medium, a semiconductor storage medium or other storage medium.

[0258] The model in the present application is an object that constitutes an objective description of a morphological structure by means of a physical or virtual representation. The object is not equal to the object, and is not limited to the physical and virtual. It can be a data processing function, a software program, a processing mode, a use method, an operation mode, a work flow, an application process, an electronic hardware, a circuit module, a processing system, a system imitation, or a simulation object.

[0259] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present application, which are used to illustrate the technical solutions of the present application, rather than limit the same. The protection scope of the present application is not limited thereto. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can modify the technical solutions recorded in the foregoing embodiments, or replace some of the technical features with equivalent replacements, within the technical scope disclosed by the present application. Such modifications or replacements do not cause the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and any modifications or replacements that do not deviate from the spirit and scope of the present application should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A flood episode division method based on flood characteristics, characterized in that it comprises the following steps: Step 1: Using a pre-constructed flood characteristic simulation model, obtain the flow data and rainfall data of a certain basin, and perform frequency analysis on the flow data to obtain the frequency distribution of the flow data; according to the frequency distribution, calculate the flood peak threshold value for representing the flood characteristics; The method for calculating the flood peak threshold value is as follows: Perform frequency analysis on the flow data, arrange the flow points in the flow data in ascending order to obtain the frequency distribution; According to the frequency distribution, calculate the cumulative frequency of each flow point to obtain the frequency data; Based on the frequency data and the flood characteristics, determine the cumulative threshold value; The flow point corresponding to the cumulative threshold value is taken as the flood peak threshold value; Step 2: Based on a pre-constructed event screening model, screen out several flow data points higher than the flood peak threshold value from the flow data according to the flood peak threshold value, and then mark the flow data points as events with potential flood risk; Step 3: Using a pre-constructed event merging model, find the maximum flow peak value and its occurrence time of each event, and obtain the peak interval time of adjacent events; Then, according to the peak interval time, process the several events to generate flood events; Step 4: Using a pre-constructed episode division model, obtain the rainfall start and end time according to the rainfall data; And according to the rainfall start and end time and the flow data, adjust the time boundary of the flood events to determine the flood episodes.

2. The flood episode division method based on flood characteristics according to claim 1, characterized in that: Step 1: The method for obtaining the flow data and rainfall data of a certain basin using a pre-constructed flood characteristic simulation model is as follows: Collect the original flow data sequence, original rainfall data sequence and corresponding time sequence of monitoring stations in a certain basin; Detect the mutation values of the original flow data sequence and the original rainfall data sequence respectively to obtain mutation data; Replace the mutation data using linear interpolation method, and complete the missing values in the original flow data sequence and the original rainfall data sequence to obtain standard flow sequence and standard rainfall sequence; Based on the time sequence, perform time consistency processing on the standard flow sequence and the standard rainfall sequence to obtain time stamps with unified format; According to the predetermined time step, resample the standard flow sequence and the standard rainfall sequence to obtain flow data and rainfall data, so that the flow data and the rainfall data have the same time frequency.

3. The flood episode division method based on flood characteristics according to claim 2, characterized in that: The method for detecting mutation values of the original flow data sequence and the original rainfall data sequence respectively to obtain mutation data is as follows: Step 11: According to historical flood data, statistics the duration of historical flood episodes to obtain the interval range selected by the window algorithm, cross-check the window values in the interval range, and finally obtain the optimal window value; Step 12: Obtain the original flow data sequence and the original rainfall data sequence, which respectively include several data points; ​ Step 13, according to the window value, intercept the data points within the window length to obtain the window data; take the median of the window data to obtain the median; then calculate the residual of the window data and the median to obtain the absolute deviation of all flow data points from the median; Step 14, based on the absolute deviation and the confidence interval of the outlier, the sensitive coefficient is calculated; Step 15, based on the sensitive coefficient and the absolute deviation, it is judged whether each data point exceeds the upper limit of the abnormal value or the lower limit of the abnormal value; When the data point does not exceed the upper limit of the abnormal value or the lower limit of the abnormal value, the data point is regarded as normal data; When the data point exceeds the upper limit of the abnormal value or the lower limit of the abnormal value, the data point is regarded as mutation data.

4. The flood event division method based on flood characteristics according to claim 3, wherein: The method for calculating the sensitive coefficient based on the absolute deviation and the confidence interval of the outlier is as follows: Set the value range of the sensitive coefficient and the calculation step; Based on the calculation step, generate a number of sensitive values; According to a number of sensitive values and absolute deviations, a number of abnormal threshold values are calculated; Based on a number of abnormal threshold values, a data quantile chart is drawn to obtain the confidence interval of each abnormal threshold value, and the confidence data is obtained; according to the confidence data, the sensitive value is selected as the sensitive coefficient.

5. The flood event division method based on flood characteristics according to claim 1, wherein: Step 2, based on the pre-constructed event screening model, a number of flow data points higher than the flood peak threshold are screened from the flow data according to the flood peak threshold, and then the flow data points are marked as events with potential flood risk, and the method is as follows: Step 21, according to the flood peak threshold, a number of flow data points higher than the flood peak threshold are screened from the flow data; Step 22, sort the screened flow data points and mark their indexes; Step 23, mark the flow data points with continuous indexes as an event to obtain an interval set, which includes the flow process data of each event; Step 24, according to the objective function and the constraint condition, the interval set is processed multiple times to obtain the optimal interval set; Step 25, based on the optimal interval set, the event with potential flood risk is determined.

6. The flood event division method based on flood characteristics according to claim 5, wherein: Step 24, the method for obtaining the optimal interval set by processing the interval set multiple times according to the objective function and the constraint condition is as follows: First, based on the interval number m of the interval set, create a set list with a length of m-1, and the values in the set list are 0 or 1; each value in the set list represents whether the adjacent two intervals are merged or not, when the value is "0", it means maintaining the current position and the interval is not merged; when the value is "1", it means merging the left and right intervals at this position; Then, randomly generate a sequence of 0 or 1 to assign values to the set list to form a candidate merging scheme; Based on the flood characteristics, the objective function and the constraint condition are constructed, the objective function is the deviation degree of the maximum flow value of a certain interval from the 99th percentile of the interval set, and the constraint condition includes the length of the merged single interval, the index threshold and the flow difference threshold; According to the objective function and the constraint condition, the candidate merging scheme is iterated and screened to obtain an optimal binary sequence; The binary sequence is mapped back to the set list to obtain a final merging interval list; The final merging interval list is de-duplicated to make all intervals unique, and each interval is converted into a tuple form to generate an optimal interval set.

7. The flood episode division method based on flood characteristics according to claim 1, characterized in that: Step three, using the pre-constructed event merging model, find out the maximum flow peak value and its occurrence time of each event, and get the peak interval time of adjacent events; then according to the peak interval time, process several events to generate the flood event as follows: Obtain the basic data of a basin; the basic data includes the basin area, the benchmark minimum interval time and the area unit conversion coefficient; Construct the time compression formula to quantify and standardize the growth rate of the time interval between flood events; Substitute the basin area and the area unit conversion coefficient into the time compression formula to get the flood interval time increment; Add the flood interval time increment and the benchmark minimum interval time to get the independence test threshold; Obtain a plurality of events with potential flood risk, and find out the maximum flow peak value and its occurrence time of each event; According to the occurrence time of the maximum flow peak value of each event, get the peak interval time of adjacent events; Compare the peak interval time with the independence test threshold, if the peak interval time between two adjacent events is less than the independence test threshold, then merge the two events into one event; The merged event is used as the flood event.

8. The flood episode division method based on flood characteristics according to claim 1, characterized in that: Step four, using the pre-constructed episode division model, according to the rainfall data, obtain the rainfall start and end time; and according to the rainfall start and end time and the flow data, adjust the time boundary of the flood event to determine the flood episode as follows: Step 41, according to the rainfall start and end time of the rainfall data, find from the starting position of each flood event to the front until finding N consecutive zero rainfall values to determine the new starting position; Step 42, according to the flow data, find from the end position of each flood event to the back until finding consecutive decreasing flow values to determine the new end position; Step 43, determine the time boundary of the flood event according to the new starting position and the new end position; Step 44, optimize the time boundary of the flood event so that the time intervals of each flood event do not overlap; Step 45, use the optimized flood event as the flood episode.

9. A flood episode division system based on flood characteristics, characterized in that: It includes: One or more processing units; Storage device for storing one or more programs; When the one or more programs are executed by the one or more processing units, the one or more processing units are caused to implement a flood event classification method based on flood characteristics as claimed in any one of claims 1-8.

Images