Construction method of distributed flood forecasting and dispatching model based on sub-basin

Abstract

This invention discloses a method for constructing a distributed flood forecasting and scheduling model based on sub-basins, specifically relating to the field of flood forecasting model optimization technology. It addresses the problem of spatiotemporal heterogeneous local prediction failure in existing flood forecasting models due to neglecting parameter sensitivity. By dividing the target basin into sub-basins and extracting hydrogeographical feature parameters, a cross-basin hydraulic facility topology network is constructed to dynamically correct river evolution boundary conditions. The spatial heterogeneity of parameter sensitivity is quantified using the Sobol index method, and sensitive anomaly response zones are identified by combining historical flood event distributions. Spatial lag correction is applied through the river topology network to generate parameter sensitivity level classifications. A dynamic weight matrix is ​​constructed, and regional parameter calibration is performed using both global optimization and local adjustment strategies. The hydraulic balance constraints of the calibration results are verified based on river topology relationships, and a regionalized and graded flood discharge scheduling scheme is generated after iterative correction, significantly improving forecasting capabilities in sudden flood scenarios.

Description

Method for Constructing a Distributed Flood Forecasting and Scheduling Model Based on Sub-basins Technical Field

[0001] This invention relates to the field of flood forecasting model optimization technology, and more specifically, to a method for constructing a distributed flood forecasting and scheduling model based on sub-basins. Background Technology

[0002] Traditional distributed flood forecasting models often employ a global parameter calibration strategy, treating the watershed as a whole or optimizing parameters based on fixed partitions. The aim is to improve the overall accuracy of the model through unified calibration. They typically rely on historical flood data to construct the objective function and use intelligent optimization algorithms (such as genetic algorithms and particle swarm optimization) to globally optimize the model parameters. By default, the parameter sensitivity is uniform in the spatiotemporal dimensions, which can meet the basic forecasting requirements under normal flood scenarios.

[0003] Existing global parameter calibration methods ignore the spatiotemporal heterogeneity of parameter sensitivity, causing distributed flood forecasting models to fail in local areas or specific time periods. Specifically, errors in highly sensitive sub-basins (such as steep river channels) are overcorrected during optimization, while errors in low-sensitivity sub-basins (such as gentle river sections) are ignored by the system due to insufficient sensitivity. This makes it impossible to accurately capture the local evolution characteristics of sudden floods (such as missed downstream flood peaks), which seriously affects the spatiotemporal precision of flood forecasting. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a method for constructing a distributed flood forecasting and scheduling model based on sub-basins to address the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] The method for constructing a distributed flood forecasting and scheduling model based on sub-basins includes the following steps:

[0007] S1. Based on the digital elevation model and the boundary of the hydrological response unit, the target watershed is divided into multiple sub-watersheds, and the hydrological and geographical feature parameters of each sub-watershed are extracted.

[0008] S2. Construct a cross-basin water conservancy facility topology network to dynamically correct the current river channel evolution boundary conditions in the sub-basin;

[0009] S3. Based on hydrological and geographical characteristic parameters, the Sobol index method is used to analyze the sensitivity of hydrological model parameters and quantify spatial heterogeneity. Geographical consistency constraints are applied to sub-basins with significant spatial heterogeneity to generate parameter sensitivity level classification.

[0010] S4. Based on the corrected river evolution boundary conditions and parameter sensitivity level classification, construct a dynamic weight matrix by combining real-time geographic meteorological data.

[0011] S5. Based on the dynamic weight matrix, the target watershed is divided into a high-sensitivity optimization zone and a low-sensitivity optimization zone, and the partition parameters are calibrated for each zone.

[0012] S6. Verify whether the calibration results of the partition parameters meet the hydraulic balance constraints based on the river topology. If they do not meet the constraints, perform iterative corrections and generate flood forecasts and scheduling schemes.

[0013] In a preferred embodiment, the target watershed is divided into multiple sub-watersheds based on a digital elevation model and hydrological response unit boundaries, and hydrological and geographical characteristic parameters of each sub-watershed are extracted, including:

[0014] Acquire digital elevation model data and hydrological response unit boundary data for the target watershed;

[0015] Determine the minimum catchment area threshold for sub-basin division based on the boundary data of hydrological response units;

[0016] The D8 algorithm is used to calculate the flow direction of digital elevation model data, and the target watershed is divided into multiple sub-watersheds by combining the minimum catchment area threshold.

[0017] The topographic slope, river length, and land use type of each sub-basin were extracted as hydrogeographic feature parameters.

[0018] In a preferred embodiment, a cross-basin water conservancy facility topology network is constructed to dynamically correct the river evolution boundary conditions of the current sub-basin, including:

[0019] Obtain spatial location and scheduling records of reservoirs and sluice gate facilities in adjacent watersheds;

[0020] Constructing topological connections between cross-basin water conservancy facilities based on their spatial location;

[0021] Real-time access to flood discharge data and gate opening data of neighboring reservoirs;

[0022] Based on the current channel roughness coefficient and cross-sectional morphology parameters of the sub-basin, the channel evolution roughness coefficient and flow capacity parameters are dynamically corrected by combining the flood discharge data of neighboring reservoirs.

[0023] The revised river evolution boundary conditions are input into the hydraulic model to generate updated river evolution curves.

[0024] In a preferred embodiment, generating a parameter sensitivity level classification includes:

[0025] Based on hydrogeographic characteristic parameters, the global sensitivity coefficients of each hydrological model parameter are calculated using the Sobol index method.

[0026] Extract spatial distribution data of historical extreme flood events in the target watershed and calculate the spatial correlation index between the spatial distribution data of historical extreme flood events and the global sensitivity coefficient.

[0027] Sub-basins with spatial correlation indices exceeding a significance threshold are identified and marked as sensitive abnormal response areas;

[0028] Based on the river topology network, the upstream and downstream influence range of the sensitive anomaly response zone is determined, and spatial lag correction is applied to the global sensitivity coefficient of the sub-basins within the upstream and downstream influence range.

[0029] Based on the difference between the corrected global sensitivity coefficient and the original global sensitivity coefficient, the parameter sensitivity levels are classified.

[0030] In a preferred embodiment, a dynamic weight matrix is ​​constructed based on the corrected river evolution boundary conditions and parameter sensitivity level classification, combined with real-time geographic meteorological data, including:

[0031] Access real-time geographic meteorological data, including rainfall intensity, wind speed, and soil moisture data for each sub-basin within the target watershed;

[0032] Initial weight coefficients are assigned based on parameter sensitivity level classification.

[0033] The initial weighting coefficients are dynamically adjusted based on real-time rainfall intensity. When the rainfall intensity exceeds the preset rainfall threshold, the initial weighting coefficients of the high-sensitivity sub-basins are adjusted upwards.

[0034] Combining the flow capacity parameters in the boundary conditions of river evolution, the adjusted weight coefficients are hydraulically balanced and corrected. The correction rule is to constrain the physical relationship between the difference in weight coefficients between upstream and downstream sub-basins and the flow capacity of the river.

[0035] A dynamic weight matrix is ​​generated, and the elements of the dynamic weight matrix are the weight coefficients of each sub-basin after hydraulic balance correction.

[0036] In a preferred embodiment, the initial weighting coefficient of the high-sensitivity sub-basin is higher than that of the medium-sensitivity sub-basin, and the initial weighting coefficient of the medium-sensitivity sub-basin is higher than that of the low-sensitivity sub-basin.

[0037] In a preferred embodiment, the target watershed is divided into a high-sensitivity optimization zone and a low-sensitivity optimization zone based on a dynamic weight matrix, and the partitioning parameters are calibrated for each zone, including:

[0038] The threshold for dividing the high-sensitivity optimization region and the low-sensitivity optimization region is determined based on the weight coefficients in the dynamic weight matrix. The threshold is calculated and generated by the distribution of the weight coefficients in the dynamic weight matrix.

[0039] The target watershed is divided into high-sensitivity optimization zones and low-sensitivity optimization zones based on the division threshold.

[0040] For the sub-basins within the high-sensitivity optimization zone, a global optimization algorithm is used for parameter calibration. The optimization objective of the global optimization algorithm is to minimize the total error between the model output and the measured flood process.

[0041] For sub-basins within the low-sensitivity optimization zone, a local parameter adjustment method is used for parameter calibration. The adjustment range of the local parameter adjustment method is based on the calibration results of adjacent high-sensitivity optimization zones.

[0042] Generate the parameter calibration results for the high-sensitivity optimization zone and the low-sensitivity optimization zone, and output the calibrated hydrological model parameter set.

[0043] In a preferred embodiment, sub-basins with weight coefficients higher than the division threshold are classified as high-sensitivity optimization zones, and sub-basins with weight coefficients lower than the division threshold are classified as low-sensitivity optimization zones.

[0044] In a preferred embodiment, the calibration results of the zoning parameters are verified based on the river topology to ensure compliance with hydraulic balance constraints. If they do not comply, iterative corrections are performed, and flood forecasting and scheduling schemes are generated, including:

[0045] Obtain the zoning parameter calibration results and the river connectivity relationships in the cross-basin water conservancy facility topology network;

[0046] Verify whether the zoning parameter calibration results conform to hydraulic balance constraints based on river connectivity;

[0047] When the calibration results of the partition parameters do not meet the hydraulic balance constraints, the calibration parameters of the high-sensitivity optimization zone are iteratively corrected. The correction rule is to adjust the parameter values ​​according to the weight coefficient ratio in the dynamic weight matrix.

[0048] The parameter calibration results after iterative correction are input into the hydrological model to generate flood evolution simulation data. Based on the flood evolution simulation data, the peak arrival time, inundation range and water level change curves are extracted.

[0049] Generate flood forecasts and dispatching plans, including recommendations on the opening degree of flood discharge gates in different zones and levels, early warning information for risk areas, and emergency response time points.

[0050] In a preferred embodiment, the hydraulic balance constraint is determined when the simulated flow difference between the upstream and downstream sub-basins does not exceed the flow capacity parameter in the modified river evolution boundary conditions.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] 1. By employing a dynamic weight matrix and a regional and hierarchical optimization strategy, the spatiotemporal heterogeneity of parameter sensitivity in traditional flood forecasting models is effectively addressed, significantly improving the model's adaptability to complex geographical environments and extreme weather conditions. A dynamic weight matrix is ​​constructed based on parameter sensitivity level classification, and combined with real-time meteorological data and river evolution constraints, enabling dynamic adjustment of model weights: highly sensitive sub-basins automatically receive higher weight priority during critical periods such as heavy rainfall, ensuring refined correction of local errors; low-sensitivity areas avoid conflicts in physical laws caused by isolated optimization through neighborhood parameter correlation and hydraulic balance constraints. Through geographical consistency constraints and dynamic weight allocation, the model can capture the characteristics of sudden flood evolution in highly sensitive areas while maintaining the overall coordination of parameters in low-sensitivity areas, significantly improving the spatial resolution and temporal reliability of forecast results.

[0053] 2. Through hydraulic balance verification and iterative correction mechanisms, the parameter calibration results are ensured to strictly conform to the physical laws of river water conveyance capacity, solving the problem of the disconnect between the mathematical optimal solution and engineering feasibility of traditional models. Based on topological relationships, the difference in flow between upstream and downstream is constrained for verification. When the calibration parameters cause a sudden change in flow that exceeds the flow capacity, priority is given to iterative correction of parameters in highly sensitive areas driven by weights. This not only preserves the optimization accuracy of key areas, but also forces the parameter adjustment direction to conform to hydraulic laws through physical transmission mechanisms. This ensures that the flood forecast results can not only meet the goal of minimizing statistical errors, but also match the actual engineering scheduling capabilities. It provides highly reliable data support for decisions such as zoned and graded flood discharge and risk warning, and is especially suitable for complex flood control scenarios involving the joint operation and control of cross-basin water conservancy facilities. Attached Figure Description

[0054] Figure 1 is a flowchart of the method for constructing a distributed flood forecasting and scheduling model based on sub-basins according to the present invention;

[0055] Figure 2 is a flowchart of the parameter sensitivity level classification generated by the present invention. Detailed Implementation

[0056] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0057] Example: Figure 1 illustrates the method for constructing a distributed flood forecasting and scheduling model based on sub-basins according to the present invention, including the following steps:

[0058] S1. Based on the digital elevation model and the boundary of the hydrological response unit, the target watershed is divided into multiple sub-watersheds, and the hydrological and geographical feature parameters of each sub-watershed are extracted.

[0059] S2. Construct a cross-basin water conservancy facility topology network to dynamically correct the current river channel evolution boundary conditions in the sub-basin;

[0060] S3. Based on hydrological and geographical characteristic parameters, the Sobol index method is used to analyze the sensitivity of hydrological model parameters and quantify spatial heterogeneity. Geographical consistency constraints are applied to sub-basins with significant spatial heterogeneity to generate parameter sensitivity level classification.

[0061] S4. Based on the corrected river evolution boundary conditions and parameter sensitivity level classification, construct a dynamic weight matrix by combining real-time geographic meteorological data.

[0062] S5. Based on the dynamic weight matrix, the target watershed is divided into a high-sensitivity optimization zone and a low-sensitivity optimization zone, and the partition parameters are calibrated for each zone.

[0063] S6. Verify whether the calibration results of the partition parameters meet the hydraulic balance constraints based on the river topology. If they do not meet the constraints, perform iterative corrections and generate flood forecasts and scheduling schemes.

[0064] S1. Based on the digital elevation model and the boundaries of hydrological response units, the target watershed is divided into multiple sub-watersheds, and the hydrological and geographical characteristic parameters of each sub-watershed are extracted. The specific implementation is as follows:

[0065] Obtain digital elevation model (DEM) data and hydrological response unit boundary data for the target watershed. The DEM data originates from publicly available datasets generated by satellite remote sensing platforms or lidar measurement technology, such as the Space Shuttle Radar Topographic Mapping Mission Database. The data is in raster format, with each raster cell representing a surface elevation value. The spatial resolution is determined based on the target watershed area; for example, a resolution of 30 meters is used when the target watershed area is less than 1000 square kilometers, and a resolution of 90 meters is used when the target watershed area is greater than or equal to 1000 square kilometers.

[0066] The boundary data of the hydrological response units are derived from the watershed management planning documents issued by the water resources department. They are imported and converted into vector format using geographic information system software, and include the boundary coordinates of each hydrological unit and its connection relationship with adjacent units.

[0067] The minimum catchment area threshold for sub-basin division is determined based on the boundaries of hydrological response units. This threshold controls the fineness of sub-basin division, and its setting rule is based on classifying complexity levels according to the average slope value of the terrain. The average slope value is calculated by using the spatial difference method for each raster cell in the digital elevation model data; that is, the slope direction and magnitude are determined based on the elevation difference between adjacent raster cells, and the arithmetic mean of the slopes of all raster cells in the entire basin is taken. Terrain complexity is divided into three levels: for example, when the average slope is greater than 15 degrees, it is considered high-complexity terrain, and the minimum catchment area threshold is set at 0.5 square kilometers; when the average slope is between 5 and 15 degrees, it is considered medium-complexity terrain, and the threshold is set at 1 square kilometer; when the average slope is less than 5 degrees, it is considered low-complexity terrain, and the threshold is set at 2 square kilometers.

[0068] The D8 algorithm is used to calculate the flow direction of digital elevation model (DEM) data, and the target watershed is divided into multiple sub-watersheds based on the minimum catchment area threshold. In the preprocessing stage, a depression-filling operation is performed on the DEM data, using an iterative method to correct local low-point areas in the elevation data and eliminate false depressions caused by data acquisition errors. The flow direction calculation uses the D8 algorithm; specifically, for each grid cell, the elevation difference between it and its eight adjacent grid cells is calculated to determine the flow direction as the direction of maximum slope. Based on the flow direction, a runoff accumulation grid is generated, and the upstream runoff area of ​​each grid cell is counted. When the runoff accumulation reaches the number of grid cells corresponding to the minimum catchment area threshold, the area is marked as a sub-watershed. After division, the spatial consistency between the sub-watershed boundary and the hydrological response unit boundary is verified. If the sub-watershed boundary spans multiple hydrological units, the sub-watershed is divided based on the centroid of the hydrological unit, ensuring that each sub-watershed completely contains at least one hydrological unit.

[0069] The topographic slope, river length, and land use type of each sub-basin were extracted as hydrogeographic feature parameters. The topographic slope was calculated based on digital elevation model data, using the spatial difference method to calculate the slope value of each raster cell. The arithmetic mean of the slopes of all raster cells within the sub-basin was taken as the topographic slope parameter for that sub-basin. The river length was extracted by identifying the main river network paths generated by the D8 algorithm within the sub-basin boundary. The total length from the sub-basin outlet to the farthest source was calculated using line feature measurement tools in the geographic information system (GIS), for example, by measuring the length along the river centerline using the "Survey Line" function in ArcGIS software. Land use types were obtained by accessing remote sensing image interpretation data of the target watershed. The interpretation method employed supervised classification technology, classifying land cover types into four categories based on spectral characteristics: cultivated land, forest land, construction land, and water area. The area percentage of each type within the sub-basin was statistically analyzed, and the type with the highest area percentage was selected as the land use type parameter for that sub-basin.

[0070] When verifying the accuracy of digital elevation model data, at least 10 field-measured elevation points should be selected, and the root mean square error between these points and the model elevation values ​​should be calculated. The error should not exceed ±5%. When verifying sub-basin boundaries, the spatial overlap should be greater than 90% when comparing the sub-basin boundaries with those in historical hydrological maps provided by the water resources department.

[0071] When verifying topographic slope parameters, 5% of the sub-basins are randomly selected for on-site slope measurements. The average absolute error between the measured slope and the model value is calculated and is required to be no more than ±10%. When verifying river length, the measurements are compared with publicly available river length records from the water resources department, and the deviation is required to be no more than ±5%. When verifying land use types, the classification results are verified through drone aerial photography or on-site surveys, and the consistency is required to be no less than 85%.

[0072] Step S1 is used to divide the watershed into sub-basins and extract hydrogeographic parameters, ensuring that the physical boundaries and geographical features of each sub-basin correspond accurately, providing reliable input for subsequent parameter sensitivity analysis and model optimization.

[0073] S2. Construct a cross-basin water conservancy facility topology network to dynamically correct the current river channel evolution boundary conditions of the sub-basin. The specific implementation is as follows:

[0074] Spatial location and scheduling records of reservoirs and sluice gates in adjacent river basins were obtained. Reservoir spatial location data was sourced from the publicly available geographic information database of water conservancy projects by the water resources department. Vector-format reservoir coordinate data, including the latitude and longitude of the dam center point, reservoir capacity, and design water level parameters, was obtained by accessing the National Water Resources Data Sharing Platform. Sluice gate data was exported from the water conservancy facility management platform using geographic information system software, and ArcGIS software was used to extract gate location coordinates, design flow rates, and historical opening and closing records. Scheduling record data was accessed in real-time from the water situation monitoring system of the river basin management agency. Hourly discharge data, sluice gate opening data, and corresponding scheduling instruction numbers from adjacent reservoirs were obtained via API interfaces.

[0075] Based on the spatial location data of reservoirs and sluice gates, a topological connection relationship of inter-basin water conservancy facilities was constructed using Geographic Information System (GIS) tools. The rules for constructing the topological connection relationship included reservoir-river channel connections, sluice gate-river channel connections, and inter-basin water conveyance channel connections. The criteria for determining a reservoir-river channel connection were set as follows: the straight-line distance between the reservoir's spillway and the river inlet of the target sub-basin was less than a preset threshold, such as 500 meters; the criteria for a sluice gate-river channel connection were set as follows: the straight-line distance between the sluice gate outlet and the river inlet of the target sub-basin was less than another preset threshold, such as 300 meters; and the criteria for an inter-basin water conveyance channel connection were set as follows: the straight-line distance between the channel endpoint and the river confluence point of the target sub-basin was less than, for example, 1 kilometer. All connections were automatically generated using a spatial buffer analysis tool, and a topological network layer was generated in ArcGIS using the buffer analysis function.

[0076] The system receives real-time data on flood discharge and gate opening from adjacent reservoirs. Flood discharge data is acquired through the data interface of the hydrological monitoring system, using the Ministry of Water Resources' national hydrological information exchange protocol to obtain hourly updated values ​​in cubic meters per second. Gate opening data is collected in real-time by IoT sensors. Displacement sensors installed on the gates convert the opening height into a percentage value, such as 0%-100%, and transmit it wirelessly to the data platform. The data transmission protocol uses MQTT to ensure real-time performance and stability, with data latency controlled to within, for example, 5 minutes.

[0077] Based on the current river roughness coefficient and cross-sectional morphology parameters of the sub-basin, and combined with the flood discharge data of adjacent reservoirs, the boundary conditions for river evolution are dynamically adjusted. The adjustment rule for the river roughness coefficient is set as follows: when the flood discharge of the adjacent reservoir exceeds, for example, 30% of the design flow of the current sub-basin river, the roughness coefficient is reduced by, for example, 10%-15%; when the flood discharge is lower than, for example, 10% of the design flow, the roughness coefficient is increased by, for example, 5%-8%.

[0078] The correction rules for the cross-sectional morphology parameters are set to dynamically adjust the river channel flow capacity parameters based on the discharge volume data. For example, when the discharge volume continuously increases, the cross-sectional width increases by 2 meters and the water depth increases by 0.5 meters for an increase of, say, 100 cubic meters per second; when the discharge volume decreases, the parameters are reduced by the same proportion. The corrected roughness coefficient and flow capacity parameters are updated in real time through the parameter input interface of the hydraulic model, and the Manning coefficient and cross-sectional geometry data are updated in the HEC-RAS model.

[0079] The revised boundary conditions for river evolution are input into the hydraulic model to generate updated river evolution curves. The hydraulic model uses a set of one-dimensional unsteady flow equations to simulate changes in river level and flow. The initial conditions are the measured water level and flow data of the current sub-basin, and the boundary conditions are the revised roughness coefficient and cross-sectional morphology parameters. The model calculation step is set to, for example, 1 minute, and the simulation time range covers, for example, the next 6 hours. The output results include minute-by-minute water level change curves, flow change curves, and predicted peak arrival times. The model calculation results are used to generate a spatial distribution map through a geographic information system, and the evolution curves are overlaid onto the river network layer of the target sub-basin in QGIS software.

[0080] When verifying the accuracy of topological connections, compare the spatial location deviation with the facility connection drawings provided by the water resources department; for example, the deviation should be controlled within 50 meters. When verifying the reliability of real-time data access, check the data loss rate and transmission delay; the loss rate should be less than, for example, 1%, and the delay should not exceed, for example, 5 minutes. When verifying the effectiveness of the correction parameters, compare the model-predicted water level before and after correction with the measured water level in historical flood events; the root mean square error should be reduced by, for example, more than 20%. When verifying the accuracy of the river evolution curve, compare the actual flood peak arrival time with the model-predicted time; the deviation should be controlled within, for example, 15 minutes.

[0081] Figure 2 shows the flowchart of the parameter sensitivity level classification generated by this invention. Based on hydrological and geographical characteristic parameters, the Sobol index method is used to analyze the parameter sensitivity of hydrological models and quantify spatial heterogeneity. Geographical consistency constraints are applied to sub-basins with significant spatial heterogeneity to generate parameter sensitivity level classification. The specific implementation is as follows:

[0082] Based on the hydrological and geographical feature parameters extracted in step S1, the global sensitivity coefficients of each hydrological model parameter are calculated using the Sobol index method. The Sobol index method involves Monte Carlo sampling of the input parameters of the hydrological model for the target watershed to generate a parameter sample set. Specifically, Monte Carlo sampling generates random numbers based on the parameter value range. For example, for the river roughness coefficient, the value range is typically 0.03 to 0.05. During sampling, 1000 parameter combinations are generated according to a uniform distribution. The parameter sample set is then input into the hydrological model for simulation calculation, outputting target variables (such as peak flow and water level). The hydrological model can be a software tool such as SWMM (Suspended Flood Management Model) or HEC-HMS (Hydrological Engineering Center Hydrological Modeling System).

[0083] The global sensitivity coefficients of each parameter are calculated using the variance decomposition method. Specifically, the variance decomposition method decomposes the total variance of the model output into components representing the individual effects and interactions of each parameter. The sensitivity coefficient of a parameter is the percentage of that parameter component relative to the total variance, and its value ranges from, for example, 0 to 1. A larger value indicates a more significant impact of the parameter on the model output. Hydrogeographic features include topographic slope, river length, and land use type. The topographic slope is derived from the spatial difference calculation results based on the digital elevation model in step S1. The river length is obtained through line feature measurement tools in the geographic information system, and the land use type is obtained through remote sensing image interpretation and classification.

[0084] Spatial distribution data of historical extreme flood events in the target watershed are extracted. Data is sourced from flood disaster bulletins issued by water resources departments or from inundation extent data interpreted from remote sensing imagery. The time frame is extreme flood events within the past 10 years. Data selection criteria include events where the inundated area exceeds, for example, 5% of the total watershed area or the inundation depth exceeds, for example, 1.5 meters. The spatial distribution data format is vector polygon data, containing the inundation boundary coordinates, occurrence time, and inundation depth parameters for each flood event. Data processing involves overlaying the historical flood event vector polygon data onto the target sub-watershed boundaries defined in step S1 using geographic information system software (such as ArcGIS or QGIS). The frequency and maximum inundation depth of extreme floods within each sub-watershed are statistically analyzed. The statistical rule is that if the flood-inundated area covers more than, for example, 50% of the sub-watershed area, it is counted as one event.

[0085] The spatial correlation index between the spatial distribution data of historical extreme flood events and the global sensitivity coefficient is calculated. The spatial correlation index is calculated using the Pearson correlation coefficient method. The calculation steps for the Pearson correlation coefficient are as follows: The global sensitivity coefficient of the sub-basin is used as variable X, and the flood frequency or maximum inundation depth of the corresponding sub-basin is used as variable Y; the covariance and standard deviation of X and Y are calculated. The correlation coefficient formula is the covariance divided by the product of the standard deviations of X and Y, with a value ranging from, for example, -1 to 1. The larger the absolute value, the stronger the correlation. The significance threshold is set, for example, a p-value less than 0.05. The p-value is calculated by using a t-test or permutation test to determine the statistical significance of the correlation coefficient. When the absolute value of the Pearson correlation coefficient exceeds, for example, 0.3 and the p-value is less than 0.05, the spatial correlation is considered significant.

[0086] Sub-basins with spatial correlation indices exceeding a significance threshold are identified and marked as sensitive areas. The criteria for determining anomalous response areas are as follows: if the correlation coefficient between the frequency of flooding and the sensitivity coefficient of a sub-basin exceeds, for example, 0.3 and the p-value is less than, for example, 0.05, or the correlation coefficient between the maximum inundation depth and the sensitivity coefficient exceeds, for example, 0.25 and the p-value is less than, for example, 0.05, then the sub-basin is marked as an anomalous response area. The spatial distribution results of anomalous response areas are saved as vector point data, including the sub-basin number, correlation coefficient value, and significance marker. The marking rules must meet the following conditions: if the same sub-basin simultaneously satisfies both frequency and water depth significant correlations, it is prioritized for water depth correlation; if only one condition is met, it is marked separately.

[0087] Based on the river topology network, the upstream and downstream influence range of the sensitive anomaly response zone is determined, and a spatial lag correction is applied to the global sensitivity coefficient of the sub-basins within this range. The river topology network is an extension of the cross-basin hydraulic facility topology generated in step S2, including the river flow direction, distance, and hydraulic connectivity of the target sub-basins. The upstream and downstream influence range is determined by starting from the sub-basin of the anomaly response zone, tracing upstream along the river topology network to, for example, 5 kilometers, and downstream to, for example, 10 kilometers, covering all hydraulically connected sub-basins. The correction rule is to set a lag weight based on the river distance between the sub-basin and the anomaly response zone. The weight decay function can be linear decay or exponential decay, for example, the weight decays by 10% for every 1 kilometer increase in distance. The decay weight is multiplied by the original sensitivity coefficient to generate the corrected sensitivity coefficient. The correction formula can be described as follows: the corrected sensitivity coefficient equals the original sensitivity coefficient multiplied by (1 minus the decay rate multiplied by the distance).

[0088] Based on the difference between the corrected global sensitivity coefficient and the original coefficient, the parameter sensitivity levels are classified. The difference is calculated as the percentage of the absolute difference between the corrected coefficient and the original coefficient relative to the original coefficient. The classification rules are as follows: if the difference exceeds, for example, 20%, it is classified as high sensitivity; if the difference is between 10% and 20%, it is classified as medium sensitivity; and if the difference is less than, for example, 10%, it is classified as low sensitivity. The classification results are saved as an attribute table, and the attribute table fields include sub-basin number, parameter name, original sensitivity coefficient, corrected coefficient, and level label. The data format is compatible with the dynamic weight matrix construction in step S4.

[0089] When verifying the accuracy of the spatial correlation index calculation, the calculation is repeated using the same input data, comparing it with the Pearson correlation coefficient calculation example in statistics textbooks. The error should be controlled within, for example, ±5%. When verifying the sensitivity correction effect, independent data from historical flood events that were not used for training are selected, such as flood events from the last two years as the test set. The model's predicted peak flow and the measured values ​​are compared before and after correction. The root mean square error (RMSE) and Nash efficiency coefficient (NSE) are calculated. The corrected RMSE should decrease by, for example, more than 15%, and the NSE should increase to, for example, more than 0.7. When verifying the rationality of the classification, the consistency between highly sensitive sub-basins and actual disaster-prone areas is assessed through expert experience. For example, three hydrological experts are invited to independently label disaster-prone areas, and their Kappa coefficients with the model classification results are calculated. The consistency should reach, for example, more than 0.8.

[0090] Step S3 addresses the problem of local error accumulation caused by traditional hydrological models neglecting spatial differences in parameter sensitivity by quantifying the spatial heterogeneity of parameter sensitivity and imposing geographical consistency constraints. It identifies abnormal response areas based on the correlation between the spatial distribution of historical floods and sensitivity coefficients, and corrects the sensitivity coefficients through topological networks to match parameter sensitivity with actual geographical features and disaster patterns. Compared with existing technologies, it dynamically links historical disaster data with parameter sensitivity and forces spatial consistency through the river topology transmission mechanism, so that the sensitivity level classification reflects both the internal characteristics of the model and external geographical constraints, thereby improving the physical rationality of subsequent parameter calibration.

[0091] S4. Based on the revised river channel evolution boundary conditions and parameter sensitivity level classification, and combined with real-time geographic meteorological data, a dynamic weight matrix is ​​constructed. The specific implementation is as follows:

[0092] Obtain the parameter sensitivity level classification results generated in step S3 and the channel evolution boundary conditions corrected in step S2. The parameter sensitivity level classification results include labels for high sensitivity, medium sensitivity, and low sensitivity levels for each sub-basin, and the channel evolution boundary conditions include the corrected channel roughness coefficient and flow capacity parameters.

[0093] Real-time geographic meteorological data is integrated, including rainfall intensity, wind speed, and soil moisture data for each sub-basin within the target watershed. Rainfall intensity data is collected in real-time through a network of meteorological monitoring stations, spatially interpolated, and distributed to the central points of each sub-basin, with an update frequency of, for example, once per hour. Wind speed data is derived from meteorological satellite inversion results and spatially averaged according to the sub-basin scope. Soil moisture data is acquired through microwave remote sensing technology and decomposed to the sub-basin scale. All real-time data undergoes outlier filtering, for example, removing data points exceeding three standard deviations of the historical average for the same period.

[0094] Initial weight coefficients are assigned to each sub-basin based on its parameter sensitivity level. The allocation rule is as follows: the initial weight coefficient for high-sensitivity sub-basins is set higher than that for medium-sensitivity sub-basins, and the initial weight coefficient for medium-sensitivity sub-basins is set higher than that for low-sensitivity sub-basins. In practice, the initial weights are set by referring to the proportion of contribution of each level of sub-basin to the model error in historical flood events; for example, the proportion of error contribution from high-sensitivity sub-basins is greater than the corresponding proportion from medium-sensitivity sub-basins.

[0095] The initial weighting coefficients are dynamically adjusted based on real-time rainfall intensity. The dynamic adjustment rule is as follows: when the real-time rainfall intensity of a sub-basin exceeds a preset rainfall threshold, its initial weighting coefficient is adjusted upwards. The preset rainfall threshold is set based on the statistical values ​​of historical extreme precipitation events in the target basin, for example, taking 70% of the historical maximum hourly rainfall as the threshold; the adjustment range is determined according to the degree of deviation between the rainfall intensity and the threshold, for example, when the rainfall intensity exceeds the threshold by, for example, 20%, the weighting coefficient increases by, for example, 10%, and when it exceeds, for example, 50%, it increases by, for example, 30%.

[0096] By incorporating the flow capacity parameters in the river channel evolution boundary conditions, the adjusted weighting coefficients are hydraulically balanced. The correction rule constrains the physical relationship between the weighting coefficient differences between upstream and downstream sub-basins and the river's flow capacity. Specifically, when the flow capacity parameter of the upstream sub-basin is lower than that of the downstream sub-basin, the difference in their weighting coefficients is limited to no more than the modulus of the flow capacity ratio. For example, if the upstream flow capacity is, for instance, 80% of the downstream flow capacity, then the upstream weighting coefficient must not exceed, for instance, 120% of the downstream flow capacity. This rule is implemented through an iterative algorithm until the weight differences between all adjacent sub-basins satisfy the flow capacity constraint.

[0097] A dynamic weight matrix is ​​generated, with matrix elements representing the hydraulically balanced weight coefficients of each sub-basin. The row and column indices of the matrix correspond one-to-one with the sub-basin numbers. The diagonal elements of the matrix represent the weight coefficients of the current sub-basin, while the off-diagonal elements are generated based on the topological relationships between sub-basins. For example, the weight differences between adjacent sub-basins are mapped to the matrix through topological connection weight parameters. The generated matrix is ​​output in a sparse matrix storage format for the partitioning parameter calibration in step S5.

[0098] To verify the effectiveness of the dynamic weight matrix, a dataset from historical flood events that was not used for training is selected. The improvement in the agreement between the optimized model's predictions and the actual flood events is calculated, requiring, for example, an improvement of at least 15% in the prediction accuracy of peak flow over a given period. To verify the rationality of the weight correction rules, the physical relationship between the weight differences between upstream and downstream sub-basins before and after correction and the river's flow capacity is compared, requiring, for example, that the weight differences between more than 90% of the sub-basins meet the flow capacity constraints.

[0099] The above steps enable the construction of a dynamic weight matrix based on multi-source data fusion, ensuring that the model parameter optimization process considers both geographical heterogeneity and hydrophysical mechanisms, thus providing a reliable data foundation for subsequent zoning and calibration.

[0100] Step S4 addresses the limitations of traditional weight allocation methods by constructing a dynamic weight matrix, which is static and unable to respond to real-time meteorological changes. Initial weights are assigned based on parameter sensitivity levels, and the weight coefficients are dynamically adjusted in conjunction with real-time rainfall intensity. Hydraulic balance correction is then performed by coupling river flow capacity parameters, enabling the weight matrix to simultaneously integrate model sensitivity and real-time hydrological response characteristics. Compared to existing technologies, the weight coefficients dynamically evolve with meteorological conditions and river hydraulic states. For example, the weights in highly sensitive areas significantly increase during heavy rainfall, while weight differences are automatically limited in rivers with insufficient flow capacity, thereby improving the model's optimization accuracy and stability during extreme events.

[0101] S5. Based on the dynamic weight matrix, the target watershed is divided into high-sensitivity optimization zones and low-sensitivity optimization zones, and the partition parameters are calibrated for each zone. The specific implementation is as follows:

[0102] The threshold for dividing high-sensitivity and low-sensitivity optimization zones is determined based on the weight coefficients in the dynamic weight matrix generated in step S4. The threshold is calculated by analyzing the distribution characteristics of the weight coefficients of all sub-basins in the dynamic weight matrix. For example, the median or standard deviation in statistics can be used as the threshold benchmark. When the weight coefficient distribution exhibits a bimodal characteristic, the trough value between the two peaks is selected as the threshold; if the distribution is unimodal, the mean plus, for example, one standard deviation is selected as the threshold. After the threshold is generated, its rationality needs to be verified. For example, it is checked whether the threshold correctly divides the sub-basins with the highest and lowest weight coefficients, for example, the top 10%, into the corresponding optimization zones.

[0103] The target watershed is divided into high-sensitivity optimization zones and low-sensitivity optimization zones based on a threshold. The division rule is as follows: if the weight coefficient of a sub-watershed is higher than the threshold, it is classified into a high-sensitivity optimization zone; if the weight coefficient is lower than the threshold, it is classified into a low-sensitivity optimization zone. The division results are spatially visualized using geographic information system software, generating a vector layer containing sub-watershed boundaries, weight coefficients, and optimization zone labels. The layer data format strictly corresponds to the sub-watershed number in the dynamic weight matrix of step S4.

[0104] For sub-basins within the high-sensitivity optimization zone, a global optimization algorithm is used for parameter calibration. The global optimization algorithm may be a genetic algorithm or a particle swarm optimization algorithm. Specifically, the optimization objective is to minimize the total error between the model output and the measured flood process. The total error is calculated as a weighted sum of the root mean square error and the Nash efficiency coefficient. The parameter search space is limited to the range of parameters with a high sensitivity level in step S3, for example, the search range for the river roughness coefficient is 0.03 to 0.05. During the iteration process, the search step size is dynamically adjusted. When the error decreases by less than, for example, 1% over five consecutive iterations, convergence is determined, and the optimal parameters are output.

[0105] The method for adjusting sub-local parameters within the low-sensitivity optimization zone is used for parameter calibration. The specific implementation of the local parameter adjustment method is as follows: The calibration results of adjacent high-sensitivity optimization zone sub-basins are obtained, and the spatial weighted average of their parameter values ​​is calculated. The initial parameter values ​​of the low-sensitivity optimization zone sub-basins are set to, for example, 50% to 80% of this weighted average. The adjustment range is dynamically set according to the spatial distance between the low-sensitivity zone and the high-sensitivity zone; for example, for every 1 kilometer increase in distance, the adjustment range decreases by, for example, 5%. The adjusted parameters must meet the error tolerance range of the low-sensitivity level in step S3, for example, allowing parameter values ​​to fluctuate within, for example, 10% above or below the initial value.

[0106] Generate the parameter calibration results for high-sensitivity and low-sensitivity optimization zones. The calibration results are saved as a structured parameter set, including sub-basin number, parameter name, calibration value, and optimization zone label. The parameter set is stored as a relational database table, with the primary key being a dynamic weighted matrix index that associates the sub-basin number with foreign keys, ensuring data traceability.

[0107] When verifying the rationality of the partition parameter calibration results, a dataset from historical flood events that was not used for training is selected. The improvement in prediction accuracy of the optimized model is calculated. For example, the root mean square error of peak flow during a time period should be reduced by more than 15%, and the correlation between the parameter adjustment magnitude and its weight coefficient in the low-sensitivity optimization zone should reach, for example, 0.7 or higher. The verification process must exclude abnormal calibration results caused by data noise, such as parameter values ​​exceeding the physically reasonable range (e.g., roughness coefficient less than 0.02), and manual verification is required.

[0108] The above steps enable the calibration of partition parameters based on the dynamic weight matrix, ensuring that the fine-tuning of parameters in the high-sensitivity optimization zone is combined with the rapid convergence in the low-sensitivity optimization zone, thereby improving the overall accuracy and computational efficiency of the flood forecasting model.

[0109] Step S5 addresses the issues of high computational cost and overfitting in low-sensitivity regions by employing a partitioned parameter calibration strategy. The watershed is divided into high- and low-sensitivity optimization zones. The high-sensitivity zone is finely calibrated using a global optimization algorithm, while the low-sensitivity zone undergoes local adjustments based on the results of adjacent high-sensitivity zones. For example, the initial parameter values ​​in the low-sensitivity zone are set as a percentage of the calibration values ​​in adjacent high-sensitivity zones. Compared to existing weight-driven differential calibration strategies, this approach significantly reduces the computational cost in low-sensitivity zones while maintaining accuracy in high-sensitivity zones. Furthermore, spatial correlation constraints prevent isolated parameter adjustments, improving the overall coordination of the model parameter set.

[0110] S6. Verify whether the zoning parameter calibration results conform to hydraulic balance constraints based on river topology. If they do not conform, perform iterative corrections and generate flood forecasts and scheduling schemes. The specific implementation is as follows:

[0111] The data interface is used to obtain the zoning parameter calibration results generated in step S5 and the river connectivity relationships in the cross-basin water conservancy facility topology network generated in step S2. The zoning parameter calibration results include calibration parameter sets for high-sensitivity and low-sensitivity optimization zones. The river connectivity relationships include upstream and downstream flow directions, distances, and hydraulic connectivity weight coefficients between sub-basins. The data format is consistent with the output results of step S2.

[0112] The verification of whether the calibration results of the zoning parameters meet the hydraulic balance constraints is based on the river connectivity relationship. The criterion for determining the hydraulic balance constraint is that the simulated flow difference between the upstream and downstream sub-basins does not exceed the flow capacity parameter in the river evolution boundary conditions corrected in step S2. The specific verification method is as follows: input the calibration parameters into the hydrological model, simulate the flood evolution process of each sub-basin, and extract the outlet section flow of adjacent upstream and downstream sub-basins; calculate the percentage of the flow difference to the flow capacity parameter of the downstream sub-basin; if the difference exceeds, for example, 20%, it is determined that the hydraulic balance constraint is not met.

[0113] When the calibration results of the zoning parameters do not meet the hydraulic balance constraints, the calibration parameters of the high-sensitivity optimization zone are iteratively corrected. The correction rule is to adjust the parameter values ​​according to the weight coefficient ratio in the dynamic weight matrix of step S4. Specifically, the higher the weight coefficient of the sub-basin in the high-sensitivity zone, the greater the parameter adjustment range. For example, the parameter adjustment range of the top 10% of sub-basins in terms of weight coefficient is, for example, 1.5 times the baseline value, and subsequent sub-basins decrease according to the weight coefficient ratio. The corrected parameters need to be re-substituted into the hydrological model for verification until the flow differences of all sub-basins meet the hydraulic balance constraints.

[0114] The iteratively corrected parameter calibration results are input into the hydrological model to generate flood evolution simulation data. The hydrological model uses a one-dimensional unsteady flow equation set, with the initial condition being the current measured water level in the river channel and the boundary conditions being the corrected parameter set. The simulation data includes the peak arrival time, inundation extent, and water level change curves for each sub-basin, with a data output frequency of, for example, once per minute. The peak arrival time is extracted from the peak point of the simulated flow process curve, the inundation extent is generated by superimposing areas where the water level exceeds the levee crest elevation, and the water level change curves are recorded at the sub-basin outlet sections.

[0115] Generate flood forecasting and control plans. The plans are based on flood evolution simulation data and include recommendations for the opening degree of floodgates in different zones and levels, early warning information for risk areas, and emergency response timelines. The recommended floodgate opening degree is dynamically generated based on the arrival time and inundation extent of the flood peak in the sub-basin. For example, pre-discharge may be carried out on the upstream gates 2 hours in advance, with the opening degree set to, for example, 50%. Early warning information for risk areas is marked by the coordinates of residential areas or infrastructure where the inundation depth exceeds, for example, 1 meter. Emergency response timelines are accurate to the minute, for example, activating the evacuation plan 30 minutes before the flood peak arrives.

[0116] When verifying the rationality of flood forecasting and dispatching schemes, a dataset from historical flood events that was not used for training is selected. The consistency between the gate openings suggested in the scheme and the actual dispatching records is compared, requiring, for example, that the deviation between more than 80% of the suggested gate operations and the actual records be less than, for example, 10%. When verifying the accuracy of risk area early warnings, spatial overlay analysis is performed between the actual inundation range interpreted from remote sensing images and the simulation results, requiring, for example, a 90% overlap.

[0117] The above steps enable hydraulic balance verification and dynamic correction of the zoning parameter calibration results, generating high-precision flood forecasts and scheduling schemes, and ensuring the consistency between model output and physical mechanism.

[0118] Step S6 addresses the disconnect between parameter calibration results and physical laws through hydraulic balance constraint verification and iterative correction. Based on topological relationships, it verifies whether flow differences are within the flow capacity range. Parameters in highly sensitive areas that do not meet the constraints are iteratively corrected according to weighted proportions; for example, parameters in sub-basins with higher weights are adjusted more significantly until the simulated flow matches the river's water conveyance capacity. Compared to existing technologies, this approach uses hydraulic balance conditions as a rigid constraint for parameter optimization and employs weight-driven correction rules to ensure that the optimization direction aligns with the physical mechanism, enabling flood forecast results to simultaneously satisfy both mathematical accuracy and engineering feasibility.

[0119] The above formulas are all dimensionless calculations. The formulas are derived from software simulations using a large amount of collected data, and are the closest to the real situation. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0120] It should be noted that this invention can be deployed on the device itself to realize embedded applications, or it can run on a PC or other terminal with a user interface, thereby meeting various hardware environments and usage requirements.

[0121] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0122] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0123] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0124] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0125] In addition, the functional modules in the various embodiments of this application 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.

[0126] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0127] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0128] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for constructing a distributed flood forecasting and scheduling model based on sub-basins, characterized in that, The process includes the following steps: S1. Dividing the target watershed into multiple sub-watersheds based on the digital elevation model and hydrological response unit boundaries, and extracting hydrological and geographical characteristic parameters for each sub-watershed; S2. Constructing a cross-watershed water conservancy facility topology network to dynamically correct the river evolution boundary conditions of the current sub-watershed; S3. Based on the hydrological and geographical characteristic parameters, using the Sobol index method to analyze the sensitivity of hydrological model parameters and quantify spatial heterogeneity, and generating parameter sensitivity level classifications by applying geographical consistency constraints to sub-watersheds with significant spatial heterogeneity; S4. Constructing a dynamic weight matrix based on the corrected river evolution boundary conditions and parameter sensitivity level classifications, combined with real-time geographic meteorological data; S5. Dividing the target watershed into high-sensitivity optimization zones and low-sensitivity optimization zones based on the dynamic weight matrix, and calibrating the partition parameters for each zone, including: determining high-sensitivity zones based on the weight coefficients in the dynamic weight matrix. The threshold for dividing the optimization zone and the low-sensitivity optimization zone is calculated and generated through the weight coefficient distribution of the dynamic weight matrix. Based on the threshold, the target watershed is divided into high-sensitivity optimization zones and low-sensitivity optimization zones. For sub-watersheds within the high-sensitivity optimization zone, a global optimization algorithm is used for parameter calibration. The optimization objective of the global optimization algorithm is to minimize the total error between the model output and the measured flood process. For sub-watersheds within the low-sensitivity optimization zone, a local parameter adjustment method is used for parameter calibration. The adjustment range of the local parameter adjustment method is based on the calibration results of adjacent high-sensitivity optimization zones. The partition parameter calibration results for the high-sensitivity and low-sensitivity optimization zones are generated, and the calibrated hydrological model parameter set is output. S6: Based on the river topology, the partition parameter calibration results are verified to ensure compliance with hydraulic balance constraints. If not, iterative corrections are performed, and flood forecasting and scheduling schemes are generated.

2. The method for constructing a distributed flood forecasting and scheduling model based on sub-basins according to claim 1, characterized in that, The target watershed is divided into multiple sub-watersheds based on the digital elevation model and the boundaries of hydrological response units, and the hydrogeographic feature parameters of each sub-watershed are extracted. This includes: obtaining the digital elevation model data and hydrological response unit boundary data of the target watershed; determining the minimum catchment area threshold for sub-watershed division based on the hydrological response unit boundary data; calculating the flow direction of the digital elevation model data using the D8 algorithm, and dividing the target watershed into multiple sub-watersheds in combination with the minimum catchment area threshold; and extracting the topographic slope, channel length, and land use type of each sub-watershed as hydrogeographic feature parameters.

3. The method for constructing a distributed flood forecasting and scheduling model based on sub-basins according to claim 1, characterized in that, To construct a cross-basin water conservancy facility topology network to dynamically correct the river evolution boundary conditions of the current sub-basin, the following steps are taken: acquiring the spatial location and scheduling records of reservoirs and gate facilities in neighboring basins; constructing the topological connection relationship of cross-basin water conservancy facilities based on the spatial location of the facilities; accessing the flood discharge data of neighboring reservoirs and gate opening data in real time; dynamically correcting the river evolution roughness coefficient and flow capacity parameters based on the river roughness coefficient and cross-sectional morphology parameters of the current sub-basin, combined with the flood discharge data of neighboring reservoirs; and inputting the corrected river evolution boundary conditions into the hydraulic model to generate updated river evolution curves.

4. The method for constructing a distributed flood forecasting and scheduling model based on sub-basins according to claim 1, characterized in that, The generation of parameter sensitivity level classification includes: calculating the global sensitivity coefficient of each hydrological model parameter based on hydrogeographic feature parameters using the Sobol index method; extracting spatial distribution data of historical extreme flood events in the target watershed and calculating the spatial correlation index between the spatial distribution data of historical extreme flood events and the global sensitivity coefficient; identifying sub-watersheds whose spatial correlation index exceeds the significance threshold and marking them as sensitive anomaly response areas; determining the upstream and downstream influence range of the sensitive anomaly response area based on the river topology network, and applying spatial lag correction to the global sensitivity coefficient of the sub-watersheds within the upstream and downstream influence range; and classifying the parameter sensitivity level according to the difference between the corrected global sensitivity coefficient and the original global sensitivity coefficient.

5. The method for constructing a distributed flood forecasting and scheduling model based on sub-basins according to claim 1, characterized in that, Based on the revised river evolution boundary conditions and parameter sensitivity level classification, a dynamic weight matrix is ​​constructed using real-time geographic meteorological data. This includes: accessing real-time geographic meteorological data, including rainfall intensity, wind speed, and soil moisture data for each sub-basin within the target watershed; assigning initial weight coefficients based on parameter sensitivity level classification; dynamically adjusting the initial weight coefficients according to real-time rainfall intensity, with upward adjustment of the initial weight coefficients for high-sensitivity sub-basins when rainfall intensity exceeds a preset rainfall threshold; performing hydraulic balance correction on the adjusted weight coefficients using the flow capacity parameters in the river evolution boundary conditions, with the correction rule constraining the physical relationship between the weight coefficient differences between upstream and downstream sub-basins and the river's flow capacity; and generating a dynamic weight matrix, where each element is the hydraulically balanced corrected weight coefficient for each sub-basin.

6. The method for constructing a distributed flood forecasting and scheduling model based on sub-basins according to claim 5, characterized in that, The initial weight coefficients of sub-basins with high sensitivity are higher than those of sub-basins with medium sensitivity, and the initial weight coefficients of sub-basins with medium sensitivity are higher than those of sub-basins with low sensitivity.

7. The method for constructing a distributed flood forecasting and scheduling model based on sub-basins according to claim 1, characterized in that, Sub-basins with weight coefficients higher than the partitioning threshold are classified as high-sensitivity optimization zones, while sub-basins with weight coefficients lower than the partitioning threshold are classified as low-sensitivity optimization zones.

8. The method for constructing a distributed flood forecasting and scheduling model based on sub-basins according to claim 1, characterized in that, The process involves verifying whether the calibration results of zoning parameters conform to hydraulic balance constraints based on river topology relationships. If not, iterative corrections are performed, and flood forecasting and dispatching schemes are generated. This includes: obtaining the calibration results of zoning parameters and the river connectivity relationships in the cross-basin water conservancy facility topology network; verifying whether the calibration results of zoning parameters conform to hydraulic balance constraints based on river connectivity relationships; when the calibration results of zoning parameters do not conform to hydraulic balance constraints, iterative corrections are performed on the calibration parameters of the high-sensitivity optimization zone, with the correction rule being to adjust the parameter values ​​according to the weight coefficient ratio in the dynamic weight matrix; inputting the iteratively corrected parameter calibration results into the hydrological model to generate flood evolution simulation data; extracting the peak arrival time, inundation range, and water level change curves based on the flood evolution simulation data; and generating flood forecasting and dispatching schemes, including recommendations for the opening degree of flood discharge gates in different zones and levels, early warning information for risk areas, and emergency response time nodes.

9. The method for constructing a distributed flood forecasting and scheduling model based on sub-basins according to claim 8, characterized in that, The hydraulic balance constraint is determined when the simulated flow difference between the upstream and downstream sub-basins does not exceed the flow capacity parameter in the modified river evolution boundary conditions.

Images