Reservoir flood forecasting method and device based on multi-source grid precipitation and distributed hydrological model

Abstract

The invention relates to an in-reservoir flood forecasting method and device based on multi-source grid precipitation and a distributed hydrological model. The in-reservoir flood forecasting method comprises the following steps: acquiring related data of a pre-selected key reservoir in a target area within a preset time range to form a typical flood process sample set; determining a reservoir object of the target area, selecting an on-dam position of the reservoir as a drainage basin outlet, and performing reservoir area hydrological analysis by using Arcgis to obtain analysis data; a pre-established CREST model is initialized, and localization application is carried out; performing system evaluation on the multi-source grid precipitation product of the target area; inputting a multi-source grid precipitation product into a localized CREST model to form an in-reservoir flood forecasting method suitable for a key reservoir; and carrying out reservoir flood forecasting based on a reservoir flood forecasting method, and carrying out visual display on a forecasting result in a preset mode. According to the method, the problems of single rainfall input data, difficulty in balancing model efficiency and precision, short flood forecasting period and system splitting in traditional flood forecasting are solved.

Description

Technical Field

[0001] This invention relates to the field of hydro-meteorological coupling technology, and in particular to a method and apparatus for forecasting inflow floods based on multi-source grid precipitation and distributed hydrological models. Background Technology

[0002] Inflow flood forecasting is a core component of reservoir flood control scheduling and water resource management. Its accuracy directly affects the safety of people's lives and property in downstream areas and the overall benefits of the reservoir. With the increasing frequency of extreme rainfall events under the background of global climate change, the spatiotemporal heterogeneity of inflow floods has been further aggravated, which puts forward higher requirements for forecasting technology.

[0003] Existing hydro-meteorological coupling studies have significant potential to improve the accuracy and lead time of inflow flood forecasts, but they still have the following limitations in practical applications: 1. Due to the variable structure of the initial model and the chaotic characteristics of the atmosphere itself, the quantitative precipitation forecast error of a single numerical model is relatively large. As the error accumulates gradually in the calculation of the hydrological model, the hydrological forecast results are not satisfactory, especially in extreme precipitation events where there are often large deviations.

[0004] 2. Distributed hydrological models can effectively capture the spatial variability of precipitation and watershed underlying surfaces, and can better characterize runoff generation and collection mechanisms, including processes such as rainfall infiltration, surface runoff, and groundwater runoff. However, they typically involve a large number of parameters, with the spatial dimension of these parameters increasing exponentially. In practical applications, this presents significant challenges in parameter calibration efficiency and model portability across watersheds.

[0005] 3. The existing hydro-meteorological coupling process is fragmented, and model parameter configuration, data input, and result visualization require switching between multiple software programs, which increases the complexity of operation and reduces the timeliness and operability of real-time rolling forecasts. Summary of the Invention

[0006] The purpose of this invention is to at least address one of the shortcomings of the prior art by providing a method and apparatus for forecasting inflow floods based on multi-source grid precipitation and distributed hydrological models.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: Specifically, a flood inflow forecasting method based on multi-source grid precipitation and a distributed hydrological model is proposed, including the following: Step 110: Obtain relevant data of key reservoirs in the target area within a preset time range. The relevant data includes typical flood process data and meteorological data for the corresponding time period. The relevant data is then preprocessed to obtain a typical flood process sample set. Step 120: Determine the reservoir objects in the target area, select the upstream location of the reservoir as the watershed outlet, and use ArcGIS to perform hydrological analysis of the reservoir area to obtain analysis data. The analysis data includes information on the cumulative runoff, flow direction, slope, river network and average elevation difference in a preset format. Step 130: Initialize the pre-established CREST model, and implement the localized application of the pre-established CREST model to reservoir objects in the target area based on the typical flood process sample set and analysis data; Step 140: Conduct a systematic evaluation of the multi-source grid precipitation products for the target area; Step 150: Input multi-source grid precipitation products with different forecast lead times into the localized CREST model, conduct a rolling simulation test of reservoir inflow, compare the runoff simulation effect driven by different lead times and different grid precipitation forecast products, and the impact of changes in the spatial and temporal distribution of rainfall on the watershed runoff process, and form a flood forecasting method for key reservoirs in the target area. Step 160: Develop inflow flood forecasts based on the inflow flood forecasting method applicable to key reservoirs in the target area, and visualize the forecast results in a preset manner.

[0008] Furthermore, specifically, the process of obtaining a sample set of typical flood processes includes, The typical flood process data includes flow rate, water level and warning water level data, and the meteorological data includes precipitation and evaporation data collected by national meteorological stations, regional meteorological stations and rain gauges in the target area, with a data interval of 1 hour. The preprocessing includes using piecewise linear interpolation to fill in missing rainfall, evaporation, and runoff data.

[0009] Furthermore, specifically, step 130 includes the following sub-steps: Step 131: Set up the various components of the CREST model. The components of the CREST model include the basic data module, parameter module, observation data module, rainfall and evaporation module, parameter and calibration module, and initial conditions module. Step 132: Determine the initial values ​​and parameter distribution range of the CREST model; Step 133: Perform parameter calibration on the initial parameters of the CREST model to optimize the initial parameters.

[0010] Furthermore, specifically, the various components of the CREST model are defined, including... The basic data module includes elevation (dem), cumulative discharge, flow direction, river network, and average elevation difference information, with data in both raster and text formats. The observation module includes information on hydrological stations within the watershed and hourly runoff observation data, in Shapefile and CSV formats. The precipitation and evaporation module includes hourly precipitation and evaporation data; the measured precipitation and evaporation data are interpolated into raster files with the same resolution as the demo data using the Kriging spatial interpolation method, and then input into the CREST model for conversion into MATLAB data format. The initial conditions module sets the initial values ​​for soil geology and the multilinear reservoir. Initial values ​​include initial values ​​for soil moisture content, surface linear reservoir, and underground linear reservoir. The parameter module sets relevant parameters related to runoff generation and runoff processes, including soil saturated hydraulic conductivity, precipitation conversion coefficient, average water storage capacity, variable permeability curve index, impermeable area ratio, coefficient for potential evapotranspiration to actual evapotranspiration, surface runoff velocity coefficient, index for surface runoff to river runoff, index for marked runoff to interflow, surface runoff outflow coefficient, interflow outflow coefficient, and surface runoff velocity index. The calibration module sets the calibration range for each relevant parameter.

[0011] Furthermore, specifically, determine the initial values ​​and parameter distribution range of the CREST model, including: Calculate the average water storage capacity WM of the basin:

[0012]

[0013] / F, Where Pi represents the rainfall at rain gauge station numbered i in the watershed; n is the total number of rain gauge stations in the watershed; F is the catchment area of ​​the watershed; Qi is the flow rate collected at time i; Δt is the time interval between two consecutive flow rate collections; R is the runoff depth; m is the number of flow rate collections; and Pa is the pre-holiday rainfall at 8:00 AM on the selected morning. Calculate the soil saturated hydraulic conductivity Ksat: Statistically analyze the names and area proportions of each soil type in the watershed, obtain the proportions of sand, silt, and clay soils in different soil types, query the initial values ​​of soil saturated hydraulic conductivity corresponding to different soil types, and discretize the initial values ​​of soil saturated hydraulic conductivity to watershed grid cells. Calculate the impermeability IM: Statistically analyze the names and area proportions of each land use type in the watershed, query the initial values ​​of impermeability for different land use types according to the pre-established table, and discretize the initial values ​​of impermeability to the watershed grid cells; The potential evapotranspiration coefficient KE is set to an empirical value of 0.7, the initial value of the variable permeability curve exponent is set to an empirical value of 0.1, and the other parameters are set to the default initial values ​​of the model.

[0014] Furthermore, specifically, parameter calibration is performed on the initial parameters of the CREST model to optimize the initial parameter values, including: The first step is to determine the initial values ​​of the model and the calibration range of the parameters; The second step is to perform sensitivity analysis on the model parameters, using different scalar values ​​'a' multiplied by the initial parameter values. Meanwhile, keeping other parameters constant, the sensitivity parameters are determined by comparing the measured flood hydrograph with the simulated flood hydrograph:

[0015] Where X is the sampled parameter, and a is the scaling factor. Initial values ​​for the q-dimensional parameters; The third step involves selecting several flood events with similar initial flows and using the SEC-UA algorithm built into the CREST model for calibration, thereby automatically optimizing the parameters. For other flood events with significantly different initial flows, different initial condition parameters are used, while other parameters remain unchanged. Based on the optimized parameters, the parameters are manually adjusted cyclically by comparing the simulated flood hydrograph and the measured flood hydrograph, combined with the results of parameter sensitivity analysis, until no better results can be obtained.

[0016] Furthermore, specifically, a systematic evaluation of multi-source grid precipitation products in the target area is conducted, including: Using pre-selected key reservoirs in the target area as the analysis object, multi-source grid precipitation forecast data was integrated. This data included national, provincial, CMA-TRAMS, CMA-GD, and EC model grid precipitation forecasts. A composite index was used to quantify forecast accuracy, including absolute error (AE), relative error (RE), fuzzy score (FS), and graded score (TS). The Thiessen polygon method was used to calculate the 24-hour areal rainfall in the basin based on station monitoring data, and this areal rainfall was taken as the true value. The 24-hour areal rainfall monitoring product of the tested object, based on the grid data, was calculated using the arithmetic mean method. The error variation patterns of the basin at different lead times (24-hour, 48-hour, 72-hour, and 96-hour) throughout 2024 were evaluated. By comparing the advantages and disadvantages of the multi-source grid precipitation forecast products, a multi-dimensional comprehensive evaluation report was generated, identifying the advantageous scenarios and error sources of each model.

[0017] Furthermore, specifically, step 150 includes, The measured precipitation from the stations and the multi-source grid forecast precipitation are interpolated onto the DEM grid and then stitched together to generate precipitation input data. The default evaporation rate for the entire grid is 0.1 mm / h. Precipitation forecast products with lead times of 0-96h and 1-hour intervals are provided for gift1km, gift5km, CMA-GD, ecmwf-s2s, and CMA-TRAMS. The forecast precipitation is retrieved from the latest available data from the Tianqing interface. For GIFT grid forecast precipitation, if the forecast time is ≥08:00, the forecast time is selected from 08:00 of the current day, and if the forecast time is <08:00, the forecast time is selected from 20:00 of the previous day. For precipitation forecasts from CMA-GD, CMA-TRAMS, ecmwf_s2s, and multi-source grid fusion models, if the forecast time is ≥14:00, the forecast time is selected from 08:00 of the current day, and if the forecast time is <14:00, the forecast time is selected from 20:00 of the previous day.

[0018] Furthermore, specifically, step 160 includes, The process of processing and analyzing meteorological and hydrological data is systematically integrated and encapsulated. The CREST hydrological model calling engine and automated plotting process, along with the visualization display components and structured data entry process, store key information such as model outputs, statistical data, and visualization products in an orderly manner in a pre-defined, queryable structured format in the database system. The backend provides a visual operation interface with options for historical and real-time modes. The default is real-time mode, which enables hourly rolling forecast updates. Users can manually switch to historical mode. The interface provides options for model forecast products, start time, and simulation duration, with past time options ranging from 0 to 12 days and future time options ranging from 0 to 5 days.

[0019] The present invention relates to an apparatus for predicting inflow floods based on multi-source grid precipitation and a distributed hydrological model, comprising the following: The data acquisition module is used to acquire relevant data of key reservoirs in the target area within a preset time range. The relevant data includes typical flood process data and meteorological data for the corresponding time period, and the relevant data is preprocessed to obtain a typical flood process sample set. The data analysis and calculation module is used to determine the reservoir objects in the target area, select the upstream location of the reservoir as the watershed outlet, and use ArcGIS to perform hydrological analysis of the reservoir area to obtain analysis data. The analysis data includes information on the cumulative runoff, flow direction, slope, river network and average elevation difference in a preset format. The model deployment module is used to initialize the pre-built CREST model and, based on a typical flood process sample set and analysis data, realize the localized application of the pre-built CREST model to reservoir objects in the target area. The system evaluation module is used to systematically evaluate the multi-source grid precipitation products of the target area; The simulation experiment module is used to input multi-source grid precipitation products with different forecast lead times into the localized CREST model to carry out rolling simulation experiments of reservoir inflow, compare the runoff simulation effects driven by different lead times and different grid precipitation forecast products, and the impact of changes in the spatiotemporal distribution of rainfall on the watershed runoff process, so as to form a flood forecasting method for key reservoirs in the target area. The inflow flood forecasting module is used to forecast inflow floods based on inflow flood forecasting methods applicable to key reservoirs in the target area, and to visualize the forecast results in a preset manner.

[0020] The beneficial effects of this invention are as follows: This invention proposes a method and apparatus for forecasting inflow floods based on multi-source grid precipitation and a distributed hydrological model. By integrating multi-source grid precipitation products and dynamically optimizing data sources based on real-time verification, the problem of single data bias is solved, significantly improving the reliability of flood control decisions.

[0021] Leveraging the advantages of the CREST model's "distributed grid + physical simplification," high-resolution simulation of complex underlying surfaces is achieved through parameter localization (calculation of soil saturated hydraulic conductivity, etc., and optimization using trial and error), effectively reducing peak flow error and extending the forecast period to over 72 hours.

[0022] Based on a modular architecture, the entire process of meteorological data reading, quality control, interpolation, model calling, plotting and storage is encapsulated, a standardized pipeline is built, and seamless connection with provincial business platforms (such as the Pearl River Basin system) is achieved, improving operational efficiency by 50% and eliminating cross-platform operation barriers.

[0023] This invention solves the problems of single precipitation input data, difficulty in balancing model efficiency and accuracy, short flood forecast period, and fragmented system in traditional flood forecasting. Attached Figure Description

[0024] The above and other features of this disclosure will become more apparent from the detailed description of the embodiments illustrated in conjunction with the accompanying drawings. In the accompanying drawings, the same reference numerals denote the same or similar elements. Obviously, the drawings described below are merely some embodiments of this disclosure. For those skilled in the art, other drawings can be obtained from these drawings without any creative effort. In the drawings: Figure 1 The diagram shown is a schematic of the inflow flood forecasting method based on multi-source grid precipitation and distributed hydrological model of the present invention. Figure 2 The diagram shows the principle of systematically evaluating multi-source grid precipitation products for a target area. Detailed Implementation

[0025] The following will provide a clear and complete description of the concept, specific structure, and technical effects of the present invention in conjunction with embodiments and accompanying drawings, so as to fully understand the purpose, solution, and effects of the present invention. It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The same reference numerals used throughout the accompanying drawings indicate the same or similar parts.

[0026] Reference Figure 1 Example 1, refer to Figure 1 This invention proposes a method for forecasting inflow floods based on multi-source grid precipitation and a distributed hydrological model, including the following: Step 110: Obtain relevant data of key reservoirs in the target area within a preset time range. The relevant data includes typical flood process data and meteorological data for the corresponding time period. The relevant data is then preprocessed to obtain a typical flood process sample set. Step 120: Determine the reservoir objects in the target area, select the upstream location of the reservoir as the watershed outlet, and use ArcGIS to perform hydrological analysis of the reservoir area to obtain analysis data. The analysis data includes information on the cumulative runoff, flow direction, slope, river network and average elevation difference in a preset format. Step 130: Initialize the pre-established CREST model, and implement the localized application of the pre-established CREST model to reservoir objects in the target area based on the typical flood process sample set and analysis data; Step 140: Conduct a systematic evaluation of the multi-source grid precipitation products for the target area; Step 150: Input multi-source grid precipitation products with different forecast lead times into the localized CREST model, conduct a rolling simulation test of reservoir inflow, and compare the results. The effects of runoff simulation driven by different lead times and different grid precipitation forecast products, as well as the impact of changes in the spatiotemporal distribution of rainfall on the watershed runoff process, form a flood inflow forecasting method for key reservoirs in the target area. Step 160: Develop inflow flood forecasts based on the inflow flood forecasting method applicable to key reservoirs in the target area, and visualize the forecast results in a preset manner.

[0027] In a preferred embodiment of the present invention, the specific process of step 120 includes, Collect digital elevation model (DEM) data with resolutions of 30m and 90m within Guangdong Province; collect land use type data and digital soil type maps within Guangdong Province with a resolution of at least 1km; after determining the research reservoir, select the location above the reservoir dam as the watershed outlet, and use ArcGIS to conduct hydrological analysis of the reservoir area to obtain data such as runoff accumulation, flow direction, slope, river network, and average elevation difference. The data format is uniformly raster format.

[0028] The specific implementation process is as follows: The approximate watershed area is determined using Google Earth software, and the coordinates of its lower left and upper right corners are obtained. The location of the Feilaixia Reservoir dam is then pinpointed within the software to obtain the coordinates of the watershed outlet. Based on these coordinates, 30m and 90m elevation data with varying rates of variation are downloaded from the Geospatial Data Cloud website (https: / / www.gscloud.cn / search). Taking the 90m resolution elevation data as an example, the data is imported into ArcGIS software. Preliminary cropping is performed based on the watershed area. Then, the ArcGIS-Spatial Analyst Tools-Hydrology tool is used for processing, including filling depressions, flow direction, runoff accumulation, slope, and sub-watershed generation, to obtain data such as elevation, runoff accumulation, flow direction, slope, river network, and average elevation difference. The data format is uniformly raster format. Obtain 1km resolution remote sensing monitoring data on land use in China and a 1:1,000,000 digital soil map of the whole country from the Resource, Environmental Science and Data Center website of the Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences (http: / / www.resdc.cn / Default.aspx), and crop it according to the watershed boundaries.

[0029] In a preferred embodiment of the present invention, specifically, the process of obtaining a sample set of typical flood processes includes: The typical flood process data includes flow rate, water level and warning water level data, and the meteorological data includes precipitation and evaporation data collected by national meteorological stations, regional meteorological stations and rain gauges in the target area, with a data interval of 1 hour. The preprocessing includes using piecewise linear interpolation to fill in missing rainfall, evaporation, and runoff data.

[0030] In a preferred embodiment of the present invention, step 130 specifically includes the following sub-steps. Step 131: Set up the various components of the CREST model. The components of the CREST model include the basic data module, parameter module, observation data module, rainfall and evaporation module, parameter and calibration module, and initial conditions module. Step 132: Determine the initial values ​​and parameter distribution range of the CREST model. The CREST model corresponds to 12 parameters. Based on the soil type, land use type, and previous precipitation and evaporation data of the study watershed, calculate the initial values ​​of model parameters such as soil saturated hydraulic conductivity, average water storage capacity, initial value of soil moisture content, and impervious area. The remaining parameters are referenced from watershed experience parameters. Finally, determine the initial values ​​and parameter distribution range of the model. Step 133: Perform parameter calibration on the initial parameters of the CREST model to optimize the initial parameters.

[0031] In a preferred embodiment of the present invention, specifically, the various component modules of the CREST model are defined, including, The basic data module includes files such as elevation (DM), cumulative discharge (FAC), flow direction (FDR), river network (Stream), and average elevation difference (outlet slope), in both raster and text formats. The observation module includes information on hydrological stations within the watershed and hourly runoff observation data, in both Shapefile and CSV formats. The precipitation and evaporation modules include hourly precipitation and evaporation data. Kriging spatial interpolation is used to interpolate the measured precipitation and evaporation data into raster files with the same resolution as the DM, which are then input into the CREST model and converted to MATLAB data format. The initial conditions module sets initial values ​​for soil geology and the multilinear reservoir. The initial values ​​for soil geology and multilinear reservoirs include initial values ​​for soil moisture content, surface linear reservoirs, and underground linear reservoirs. The parameter module sets relevant parameters related to runoff generation and runoff processes, including soil saturated hydraulic conductivity, precipitation conversion coefficient, average water storage capacity, variable permeability curve index, impermeable area ratio, coefficient for potential evapotranspiration to actual evapotranspiration, surface runoff velocity coefficient, index for surface runoff to river runoff, index for landmark runoff to interflow, surface runoff outflow coefficient, interflow outflow coefficient, and surface runoff velocity index. The calibration module sets the calibration range for each relevant parameter.

[0032] In a preferred embodiment of the present invention, specifically, determining the initial values ​​of the parameters and the parameter distribution range of the CREST model includes, Calculate the average water storage capacity WM of the basin:

[0033]

[0034] / F, Where Pi represents the rainfall at rain gauge station numbered i in the watershed; n is the total number of rain gauge stations in the watershed; F is the catchment area of ​​the watershed; Qi is the flow rate collected at time i; Δt is the time interval between two consecutive flow rate collections; R is the runoff depth; m is the number of flow rate collections; and Pa is the pre-holiday rainfall at 8:00 AM on the selected morning. Calculate the soil saturated hydraulic conductivity Ksat: Statistically analyze the names and area proportions of each soil type in the watershed, obtain the proportions of sand, silt, and clay soils in different soil types, query the initial values ​​of soil saturated hydraulic conductivity corresponding to different soil types, and discretize the initial values ​​of soil saturated hydraulic conductivity to watershed grid cells. Calculate the impermeability IM: Statistically analyze the names and area proportions of each land use type in the watershed, query the initial values ​​of the impermeability for different land use types according to the pre-established table (Table 1 below: Impermeability under different land use types), and discretize the initial values ​​of the impermeability to the watershed grid cells;

[0035] Table 1 The potential evapotranspiration coefficient KE is set to an empirical value of 0.7, the initial value of the variable permeability curve exponent is set to an empirical value of 0.1, and the other parameters are set to the default initial values ​​of the model.

[0036] In a preferred embodiment of the present invention, specifically, parameter calibration is performed on the initial parameter values ​​of the CREST model to optimize the initial parameter values, including: The first step is to determine the initial values ​​of the model and the calibration range of the parameters; The second step is to perform sensitivity analysis on the model parameters, using different scalar values ​​'a' multiplied by the initial parameter values. Meanwhile, keeping other parameters constant, the sensitivity parameters are determined by comparing the measured flood hydrograph with the simulated flood hydrograph:

[0037] Where X is the sampled parameter, and a is the scaling factor. Initial values ​​for the q-dimensional parameters; The third step involves selecting several flood events with similar initial flows and using the SEC-UA algorithm built into the CREST model for calibration, thereby automatically optimizing the parameters. For other flood events with significantly different initial flows, different initial condition parameters are used, while other parameters remain unchanged. Based on the optimized parameters, the parameters are manually adjusted cyclically by comparing the simulated flood hydrograph and the measured flood hydrograph, combined with the results of parameter sensitivity analysis, until no better results can be obtained.

[0038] Reference Figure 2 As a preferred embodiment of the present invention, specifically, a systematic evaluation of multi-source grid precipitation products of the target area is performed, including, This study uses pre-selected key reservoirs in the target area as the analysis object, integrating multi-source grid precipitation forecast data, including national, provincial, CMA-TRAMS, CMA-GD, and EC model grid precipitation forecasts. A composite index is used to quantify forecast accuracy, including absolute error (AE), relative error (RE), fuzzy score (FS), and graded score (TS). The Thiessen polygon method is used to calculate the 24-hour areal rainfall in the basin based on station monitoring data, and this areal rainfall is taken as the true value. The 24-hour areal rainfall monitoring product of the tested object based on grid data is calculated using the arithmetic mean method. The error variation patterns of the basin at different lead times (24-hour, 48-hour, 72-hour, and 96-hour) throughout 2024 are evaluated. By comparing the advantages and disadvantages of the multi-source grid precipitation forecast products, a multi-dimensional comprehensive evaluation report is generated, identifying the advantageous scenarios and error sources of each model. The aim is to provide reliable and refined gridded precipitation forecast products for reservoir flood forecasting and to provide a reference for the selection of hydrological model input fields.

[0039] Specifically, the accuracy of flood forecasts is assessed according to the "Specifications for Hydrological Information Forecasting" (GB / T22482-2008). The evaluation indicators include the coefficient of certainty (DC), relative bias (Bias), Nash efficiency coefficient (NSCE), relative peak error (RPE), and peak arrival time difference (PTD). Each indicator is calculated using the following formula: , , , , , in, To simulate flow rate values; This is the measured flow rate value; To simulate peak flood flow values; To measure the peak flow value This represents the average simulated flow rate. This represents the average measured flow rate. To simulate the timing of the flood peak; denoted by DC, where n is the measured peak time; n is the data sequence length. When DC=1, it indicates that the simulated value is exactly equal to the measured value; when DC>0.5, the simulated value is acceptable.

[0040] This study selects key reservoir basins as the analysis object, integrating multi-source grid precipitation forecast data, including national and provincial forecasts, CMA-TRAMS, CMA-GD, and EC model grid precipitation forecasts. Composite indicators are used to quantify forecast accuracy, including: absolute error (AE), relative error (RE), fuzzy score (FS), and graded score (TS). Using the Thiessen polygon method, the 24-hour areal rainfall of the basin is calculated based on station monitoring data, and this areal rainfall is taken as the true value. The 24-hour areal rainfall monitoring product of the tested object based on grid data is calculated using the arithmetic mean method. The error variation patterns of the basin at different lead times (24 hours, 48 ​​hours, 72 hours, and 96 hours) for the entire year of 2024 are evaluated. By comparing the advantages and disadvantages of multi-source grid precipitation forecast products, a multi-dimensional comprehensive evaluation report is generated, identifying the advantageous scenarios and error sources of each model.

[0041] The algorithms for absolute error AE and relative error RE are as follows: , Where RF is the forecast areal rainfall and RO is the observed areal rainfall. In the special case where RO is zero, .

[0042] The fuzzy scoring algorithm is as follows: , in It is the level of the areal rainfall forecast. It is the level of the actual rainfall. This is the maximum level error in the areal rainfall forecast level. The result of the maximum level error is as follows: When there is no report or a missed report, the score is directly 0; otherwise, refer to the formula above.

[0043] The maximum level error result (MAX) is illustrated in Table 2 below:

[0044] Table 2 The algorithm for forecast accuracy (FA) in tiered testing is as follows: , CS is the correct score (the correct score is the sum of the scores of all intervals within the evaluation period), NA is the number of correct forecasts (the number of times a forecast has a score), and NC is the number of false alarms and missed forecasts (the number of times a forecast has no score).

[0045] Where CS is the correct score (the sum of scores for all intervals within the evaluation period), NA is the number of correct forecasts (the number of times a forecast received a score), and NC is the number of false alarms and missed forecasts (the number of times a forecast did not receive a score). The forecast score CS is calculated using the following method: , Where FL and FH are the lower and upper forecast limits, respectively, MD is the median, and D is the difference between the median and the lower forecast limit.

[0046] Points will be assigned according to the following priority order: ① The upper limit of the areal rainfall forecast and the actual rainfall are both 0.0 mm, which do not participate in the scoring. At this time, CS, NA, NC have no score, and FA also has no score.

[0047] ② The actual rainfall was 0.0 mm, and the lower limit of the forecast is ≥0.0 mm. If |live-MD|≤2D, CS is 0.3 points, NA is 1 time, and NC is 0 times; otherwise, no points are awarded.

[0048] ③ Actual rainfall > 0.0 mm If |actual situation - MD| ≤ D (i.e., within the span), CS is worth 1 point, NA is worth 1 time, and NC is worth 0 times; otherwise, no points are awarded. If |actual-MD|≤2D, CS is 0.6 points, NA is 1 time, and NC is 0 times; If |live-MD|≤3D, CS is 0.3 points, NA is 1 time, and NC is 0 times; By comparing the grid precipitation products with the measured precipitation distribution at the stations during the flood process, the mean absolute error, root mean square error, and mean error of the grid precipitation relative to the measured precipitation are calculated. The cumulative amount of grid precipitation during the process is then compared and analyzed with the cumulative amount of measured precipitation during the process.

[0049] The precipitation grid forecast products are tested by watershed and by season. The whole year is divided by season according to the standards of pre-flood season, post-flood season and non-flood season. During the flood season, major precipitation events within the watershed are also assessed and analyzed.

[0050] In a preferred embodiment of the present invention, step 150 specifically includes... The measured precipitation from the stations and the multi-source grid forecast precipitation are interpolated onto the DEM grid and then stitched together to generate precipitation input data. The default evaporation rate for the entire grid is 0.1 mm / h. Precipitation forecast products with lead times of 0-96h and 1-hour intervals are provided for gift1km, gift5km, CMA-GD, ecmwf-s2s, and CMA-TRAMS. The forecast precipitation is retrieved from the latest available data from the Tianqing interface. For GIFT grid forecast precipitation, if the forecast time is ≥08:00, the forecast time is selected from 08:00 of the current day, and if the forecast time is <08:00, the forecast time is selected from 20:00 of the previous day. For precipitation forecasts from CMA-GD, CMA-TRAMS, ecmwf_s2s, and multi-source grid fusion models, if the forecast time is ≥14:00, the forecast time is selected from 08:00 of the current day, and if the forecast time is <14:00, the forecast time is selected from 20:00 of the previous day.

[0051] In a preferred embodiment of the present invention, step 160 specifically includes, This system systematically integrates and encapsulates a series of key steps in the meteorological and hydrological data processing and analysis workflow. Specifically, it includes: a meteorological data reading stage for seamless access and parsing of gridded precipitation data products in various mainstream formats (such as GRIB2, HDF5, NetCDF, etc.); an automated data quality control stage that uses preset rules and algorithms to perform real-time quality checks and error correction on input data, ensuring the accuracy of subsequent analysis; an intelligent dynamic spatiotemporal interpolation stage that adaptively performs spatial downscaling or upscaling transformations and refines time series data based on data distribution characteristics and spatiotemporal scale requirements; a CREST hydrological model calling engine and automated plotting stage; and visualization components and structured result storage, which stores key information such as model outputs, statistical data, and visualization products in a pre-set, queryable structured format in an orderly manner in the database system, facilitating subsequent querying, tracing, and in-depth analysis applications. The backend provides a visual operation interface with options for historical and real-time modes. The default mode is real-time, which can update forecasts hourly. Users can manually switch to historical mode. The interface provides options for model forecast products, start time, and simulation duration (accurate to the hour). Past time can be selected from 0 to 12 days, and future time can be selected from 0 to 5 days.

[0052] In application, based on a modular and layered architecture, core functions such as meteorological data reading (supporting multi-source grid precipitation products such as GRIB2 / HDF5 / NetCDF), automated quality control, dynamic spatiotemporal interpolation, high-resolution distributed hydrological model calling, automated plotting and display, and structured data storage are integrated and encapsulated to build an scalable pipelined processing framework. Forecast results are displayed in the form of "rainfall-runoff time curves" and synchronized to the "Pearl River Basin Meteorological Impact Analysis System - Forecasting and Judgment - Hydrological Prediction" module to achieve hourly rolling updates.

[0053] The present invention relates to an apparatus for predicting inflow floods based on multi-source grid precipitation and a distributed hydrological model, comprising the following: The data acquisition module is used to acquire relevant data of key reservoirs in the target area within a preset time range. The relevant data includes typical flood process data and meteorological data for the corresponding time period, and the relevant data is preprocessed to obtain a typical flood process sample set. The data analysis and calculation module is used to determine the reservoir objects in the target area, select the upstream location of the reservoir as the watershed outlet, and use ArcGIS to perform hydrological analysis of the reservoir area to obtain analysis data. The analysis data includes information on the cumulative runoff, flow direction, slope, river network and average elevation difference in a preset format. The model deployment module is used to initialize the pre-built CREST model and, based on a typical flood process sample set and analysis data, realize the localized application of the pre-built CREST model to reservoir objects in the target area. The system evaluation module is used to systematically evaluate the multi-source grid precipitation products of the target area; The simulation experiment module is used to input multi-source grid precipitation products with different forecast lead times into the localized CREST model to carry out rolling simulation experiments of reservoir inflow, compare the runoff simulation effects driven by different lead times and different grid precipitation forecast products, and the impact of changes in the spatiotemporal distribution of rainfall on the watershed runoff process, so as to form a flood forecasting method for key reservoirs in the target area. The inflow flood forecasting module is used to forecast inflow floods based on inflow flood forecasting methods applicable to key reservoirs in the target area, and to visualize the forecast results in a preset manner.

[0054] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0055] If the integrated module is implemented as a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or system capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0056] Although the description of the invention has been quite detailed and particularly of several described embodiments, it is not intended to limit it to any of these details or embodiments or any particular embodiment, but should be considered as providing a broad possible interpretation of the claims by referring to the appended claims and taking into account the prior art, thereby effectively covering the intended scope of the invention. Furthermore, the invention has been described above with respect to embodiments foreseeable by the inventors in order to provide a useful description, and non-substantial modifications to the invention that have not yet been foreseen may still represent equivalent modifications.

[0057] The above description is merely a preferred embodiment of the present invention. The present invention is not limited to the above-described embodiments. Any embodiment that achieves the technical effects of the present invention using the same means should fall within the protection scope of the present invention. Within the protection scope of the present invention, various modifications and variations can be made to the technical solutions and / or implementation methods.

Claims

1. A method for forecasting inflow floods based on multi-source grid precipitation and a distributed hydrological model, characterized in that, Including the following: Step 110: Obtain relevant data of key reservoirs in the target area within a preset time range. The relevant data includes typical flood process data and meteorological data for the corresponding time period. The relevant data is then preprocessed to obtain a typical flood process sample set. Step 120: Determine the reservoir objects in the target area, select the upstream location of the reservoir as the watershed outlet, and use ArcGIS to perform hydrological analysis of the reservoir area to obtain analysis data. The analysis data includes information on the cumulative runoff, flow direction, slope, river network and average elevation difference in a preset format. Step 130: Initialize the pre-established CREST model, and implement the localized application of the pre-established CREST model to reservoir objects in the target area based on the typical flood process sample set and analysis data; Step 140: Conduct a systematic evaluation of the multi-source grid precipitation products for the target area; Step 150: Input multi-source grid precipitation products with different forecast lead times into the localized CREST model, conduct a rolling simulation test of reservoir inflow, compare the runoff simulation effect driven by different lead times and different grid precipitation forecast products, and the impact of changes in the spatial and temporal distribution of rainfall on the watershed runoff process, and form a flood forecasting method for key reservoirs in the target area. Step 160: Develop inflow flood forecasts based on the inflow flood forecasting method applicable to key reservoirs in the target area, and visualize the forecast results in a preset manner.

2. The method for forecasting inflow floods based on multi-source grid precipitation and distributed hydrological models according to claim 1, characterized in that, Specifically, the process of obtaining a sample set of typical flood events includes: The typical flood process data includes flow rate, water level and warning water level data, and the meteorological data includes precipitation and evaporation data collected by national meteorological stations, regional meteorological stations and rain gauges in the target area, with a data interval of 1 hour. The preprocessing includes using piecewise linear interpolation to fill in missing rainfall, evaporation, and runoff data.

3. The method for forecasting inflow floods based on multi-source grid precipitation and distributed hydrological models according to claim 1, characterized in that, Specifically, step 130 includes the following sub-steps: Step 131: Set up the various components of the CREST model. The components of the CREST model include the basic data module, parameter module, observation data module, rainfall and evaporation module, parameter and calibration module, and initial conditions module. Step 132: Determine the initial values ​​and parameter distribution range of the CREST model; Step 133: Perform parameter calibration on the initial parameters of the CREST model to optimize the initial parameters.

4. The method for forecasting inflow floods based on multi-source grid precipitation and distributed hydrological model according to claim 3, characterized in that, Specifically, the various components of the CREST model are defined, including: The basic data module includes elevation (dem), cumulative flow, flow direction, river network, and average elevation difference information. The data is available in two formats: raster and text. The observation module includes information on hydrological stations within the watershed and hourly runoff observation data, in Shapefile and CSV formats. The precipitation and evaporation modules include hourly precipitation and evaporation data. Kriging spatial interpolation is used to interpolate the measured precipitation and evaporation data into raster files with the same resolution as the DEM, which are then input into the CREST model and converted to MATLAB data format. The initial conditions module sets the initial values ​​for soil geology and the multilinear reservoir, including initial values ​​for soil moisture content, surface linear reservoir, and other parameters. The initial values ​​of the linear reservoir are set; relevant parameters related to the runoff generation and runoff process are set through the parameter module, including soil saturated hydraulic conductivity, precipitation conversion coefficient, average water storage capacity, variable permeability curve index, impermeable area ratio, coefficient of potential evapotranspiration converted to actual evapotranspiration, surface runoff velocity coefficient, index of surface runoff converted to river runoff, index of landmark runoff converted to interflow, surface runoff outflow coefficient, interflow outflow coefficient, and surface runoff velocity index; the calibration module sets the calibration range for each relevant parameter.

5. The method for forecasting inflow floods based on multi-source grid precipitation and distributed hydrological models according to claim 4, characterized in that, Specifically, determine the initial values ​​and parameter distribution range of the CREST model. include, Calculate the average water storage capacity WM of the basin: / F， Where Pi represents the rainfall at rain gauge station numbered i in the watershed; n is the total number of rain gauge stations in the watershed; F is the catchment area of ​​the watershed; Qi is the flow rate collected at time i; Δt is the time interval between two consecutive flow rate collections; R is the runoff depth; m is the number of flow rate collections; and Pa is the pre-holiday rainfall at 8:00 AM on the selected morning. Calculate the soil saturated hydraulic conductivity Ksat: Statistically analyze the names and area proportions of each soil type in the watershed, obtain the proportions of sand, silt, and clay soils in different soil types, query the initial values ​​of soil saturated hydraulic conductivity corresponding to different soil types, and discretize the initial values ​​of soil saturated hydraulic conductivity to watershed grid cells. Calculate the impermeability IM: Statistically analyze the names and area proportions of each land use type in the watershed, query the initial values ​​of impermeability for different land use types according to the pre-established table, and discretize the initial values ​​of impermeability to the watershed grid cells; The potential evapotranspiration coefficient KE is set to an empirical value of 0.7, the initial value of the variable permeability curve exponent is set to an empirical value of 0.1, and the other parameters are set to the default initial values ​​of the model.

6. The method for forecasting inflow floods based on multi-source grid precipitation and distributed hydrological models according to claim 5, characterized in that, Specifically, parameter calibration is performed on the initial parameters of the CREST model to optimize the initial parameter values, including: The first step is to determine the initial values ​​of the model and the calibration range of the parameters; The second step is to perform sensitivity analysis on the model parameters, using different scalar values ​​'a' multiplied by the initial parameter values. Meanwhile, keeping other parameters constant, the sensitivity parameters are determined by comparing the measured flood hydrograph with the simulated flood hydrograph: Where X is the sampled parameter, and a is the scaling factor. Initial values ​​for the q-dimensional parameters; The third step involves selecting several flood events with similar initial flows and using the SEC-UA algorithm built into the CREST model for calibration, thereby automatically optimizing the parameters. For other flood events with significantly different initial flows, different initial condition parameters are used, while other parameters remain unchanged. Based on the optimized parameters, the parameters are manually adjusted cyclically by comparing the simulated flood hydrograph and the measured flood hydrograph, combined with the results of parameter sensitivity analysis, until no better results can be obtained.

7. The method for forecasting inflow floods based on multi-source grid precipitation and distributed hydrological model according to claim 1, characterized in that, Specifically, a systematic evaluation of multi-source grid precipitation products in the target area is conducted, including: Using pre-selected key reservoirs in the target area as the analysis object, multi-source grid precipitation forecast data was integrated. This data included national, provincial, CMA-TRAMS, CMA-GD, and EC model grid precipitation forecasts. A composite index was used to quantify forecast accuracy, including absolute error (AE), relative error (RE), fuzzy score (FS), and graded score (TS). The Thiessen polygon method was used to calculate the 24-hour areal rainfall in the basin based on station monitoring data, and this areal rainfall was taken as the true value. The 24-hour areal rainfall monitoring product of the tested object, based on the grid data, was calculated using the arithmetic mean method. The error variation patterns of the basin at different lead times (24-hour, 48-hour, 72-hour, and 96-hour) throughout 2024 were evaluated. By comparing the advantages and disadvantages of the multi-source grid precipitation forecast products, a multi-dimensional comprehensive evaluation report was generated, identifying the advantageous scenarios and error sources of each model.

8. The method for forecasting inflow floods based on multi-source grid precipitation and distributed hydrological model according to claim 1, characterized in that, Specifically, step 150 includes, The measured precipitation from the stations and the multi-source grid forecast precipitation are interpolated onto the DEM grid and then stitched together to generate precipitation input data. The default evaporation rate for the entire grid is 0.1 mm / h. Precipitation forecast products with lead times of 0-96h and 1-hour intervals are provided for gift1km, gift5km, CMA-GD, ecmwf-s2s, and CMA-TRAMS. The forecast precipitation is retrieved from the latest available data from the Tianqing interface. For GIFT grid forecast precipitation, if the forecast time is ≥08:00, the forecast time is selected from 08:00 of the current day, and if the forecast time is <08:00, the forecast time is selected from 20:00 of the previous day. For precipitation forecasts from CMA-GD, CMA-TRAMS, ecmwf_s2s, and multi-source grid fusion models, if the forecast time is ≥14:00, the forecast time is selected from 08:00 of the current day, and if the forecast time is <14:00, the forecast time is selected from 20:00 of the previous day.

9. The method for forecasting inflow floods based on multi-source grid precipitation and distributed hydrological model according to claim 1, characterized in that, Specifically, step 160 includes, The process of processing and analyzing meteorological and hydrological data is systematically integrated and encapsulated. The CREST hydrological model calling engine and automated plotting process, along with the visualization display components and structured data entry process, store key information such as model outputs, statistical data, and visualization products in an orderly manner in a pre-defined, queryable structured format in the database system. The backend provides a visual operation interface with options for historical and real-time modes. The default is real-time mode, which enables hourly rolling forecast updates. Users can manually switch to historical mode. The interface provides options for model forecast products, start time, and simulation duration, with past time options ranging from 0 to 12 days and future time options ranging from 0 to 5 days.

10. A device for predicting reservoir flooding based on multi-source grid precipitation and a distributed hydrological model, characterized in that, Including the following: The data acquisition module is used to acquire relevant data of key reservoirs in the target area within a preset time range. The relevant data includes typical flood process data and meteorological data for the corresponding time period, and the relevant data is preprocessed to obtain a typical flood process sample set. The data analysis and calculation module is used to determine the reservoir objects in the target area, select the upstream location of the reservoir as the watershed outlet, and use ArcGIS to perform hydrological analysis of the reservoir area to obtain analysis data. The analysis data includes information on the cumulative runoff, flow direction, slope, river network and average elevation difference in a preset format. The model deployment module is used to initialize the pre-built CREST model and, based on a typical flood process sample set and analysis data, realize the localized application of the pre-built CREST model to reservoir objects in the target area. The system evaluation module is used to systematically evaluate the multi-source grid precipitation products of the target area; The simulation experiment module is used to input multi-source grid precipitation products with different forecast lead times into the localized application's CREST model to conduct rolling simulation experiments of reservoir inflow and compare the results. The effects of runoff simulation driven by different lead times and different grid precipitation forecast products, as well as the impact of changes in the spatiotemporal distribution of rainfall on the watershed runoff process, form a flood inflow forecasting method for key reservoirs in the target area. The inflow flood forecasting module is used to forecast inflow floods based on inflow flood forecasting methods applicable to key reservoirs in the target area, and to visualize the forecast results in a preset manner.

Images