Water conservancy project flood four prediction data processing method and system based on edge computing

Abstract

The application discloses a water conservancy project flood four-pre data processing method and system based on edge calculation, relates to the technical field of water conservancy flood warning data processing, and comprises the following steps: arranging edge perception nodes on the upstream watershed of the water conservancy project, synchronously collecting real-time rainfall intensity values, multi-section water level continuous monitoring values and spatial distribution values of soil humidity saturation. The rainfall intensity of the whole watershed is analyzed in time and space correlation to generate a rainfall intensity dynamic evolution cloud picture, the multi-section water level is fused and calculated with a digital elevation model of the underlying surface to generate a unit confluence resistance delay surface. The flood risk channel is determined through the time and space coupling of the two, and is screened through safety checking, the early stage influence rainfall is calculated in combination with the soil humidity saturation, the partition runoff coefficient and the confluence starting time are determined, and the flood evolution scenario with soil water regime constraint is generated. The application can accurately quantify the runoff collection resistance delay effect, and improve the fine level and actual working condition matching degree of flood evolution simulation.

Description

Technical Field

[0001] This invention belongs to the field of water conservancy flood early warning data processing technology, specifically a water conservancy project flood four-prevention data processing method and system based on edge computing. Background Technology

[0002] The current flood forecasting and early warning data processing of water conservancy projects mostly adopts a centralized cloud processing mode. Hydrological information such as rainfall, water level, and soil moisture is collected through watershed monitoring stations. Rainfall data is only statistically summarized and aggregated at single points. The confluence process analysis is calculated separately using digital elevation models or single-section water level data. The flood evolution channel is directly delineated based on basic hydrological data. The early impact rainfall, runoff coefficient, and confluence time are calculated using fixed parameters. The multi-source data collected synchronously by edge sensing nodes have not been collaboratively integrated, and distributed data processing has not been carried out based on edge computing architecture.

[0003] Existing technologies cannot perform spatiotemporal correlation analysis of real-time rainfall intensity across the entire basin, cannot fully characterize the dynamic evolution of rainfall, and have not achieved fusion calculation of multi-section water level data and underlying surface topography data. It is difficult to quantify the obstruction and delay effects of different geographical units on surface runoff collection. Flood evolution channels have not been safety-checked in conjunction with river network connectivity and levee integrity, channel selection lacks effectiveness verification, and the amount of rainfall affecting the early stages has not been calculated in conjunction with the spatial distribution of soil moisture saturation. The runoff generation coefficient and runoff initiation time of different zones in the basin cannot be determined in detail, and the flood evolution scenario has not incorporated real-time soil water condition constraints.

[0004] This invention requires completing the spatiotemporal correlation analysis of rainfall intensity across the entire basin and generating a dynamic evolution cloud map of rainfall. It generates unit confluence delay surfaces by fusing multi-section water level monitoring values ​​with digital elevation models, conducts safety checks on flood evolution channels, calculates the impact rainfall in the early stages based on the spatial distribution of soil moisture, determines the zonal runoff generation coefficient and confluence start time, maps the parameters to risk channels, and generates a set of flood evolution scenarios with soil water condition constraints. Summary of the Invention

[0005] This invention aims to solve at least one of the technical problems existing in the prior art; Therefore, this invention proposes a flood pre-monitoring data processing method for water conservancy projects based on edge computing, including: In the upstream basin of the target water conservancy project, multiple edge sensing nodes are deployed to synchronously collect raw basin information, which includes real-time rainfall intensity values ​​across the entire basin, continuous monitoring values ​​of water levels at multiple cross sections, and spatial distribution values ​​of soil moisture saturation. The spatiotemporal correlation analysis of the real-time rainfall intensity values ​​collected synchronously across the entire watershed is performed to generate a dynamic evolution cloud map of rainfall intensity in the watershed where the target water conservancy project is located. By integrating the continuous monitoring values ​​of water levels at multiple cross sections with the digital elevation model of the underlying surface of the watershed, a unit confluence delay surface of the watershed where the target water conservancy project is located is generated. The unit confluence delay surface is used to reflect the obstruction and delay effect of different geographical units on the surface runoff collection process. Based on the spatiotemporal coupling relationship between the dynamic evolution cloud map of rainfall intensity and the unit confluence delay surface, a set of potential evolution channels for flood risk is determined, and the river network connectivity and dike integrity of the set of potential evolution channels are checked to screen out channels with risks. Based on the spatial distribution value of soil moisture saturation, the early impact rainfall is calculated, the runoff generation coefficient and runoff initiation time of different zones in the watershed are predicted, and the runoff generation coefficient and runoff initiation time are mapped to the channels with risks to generate a preliminary flood evolution scenario set with soil water conditions constraints.

[0006] Furthermore, the step of performing spatiotemporal correlation analysis on the synchronously collected real-time rainfall intensity values ​​across the entire watershed to generate a dynamic evolution cloud map of rainfall intensity in the watershed where the target water conservancy project is located includes: The dynamic evolution cloud map of rainfall intensity includes the movement trajectory of the rain cluster core, the center coordinates of the rainfall intensity peak, and the intensity change curve of rainfall intensity along the movement trajectory of the rain cluster core. The real-time rainfall intensity values ​​from all edge sensing nodes are uniformly aligned with their spatial location and acquisition timestamp to form a spatial distribution sequence of rainfall intensity under the same spatiotemporal reference. Rain cluster tracking processing is performed on the spatial distribution sequence of rainfall intensity over multiple consecutive time periods to extract the continuous movement path of the rain cluster in the spatiotemporal dimension and identify the core movement trajectory of the rain cluster. On the core movement trajectory of the rain cluster, detect the local peak points of rainfall intensity and the divergence points of the movement path itself, and record the geographical coordinates and timestamps of the local peak points and divergence points to form the center coordinates of the rainfall intensity peak. Along the movement trajectory of the rain cluster core, the spatial average rainfall intensity is calculated at fixed time step intervals, and the calculated spatial average rainfall intensity is connected in chronological order to form the intensity variation curve of the rainfall intensity along the movement trajectory of the rain cluster core. The core movement trajectory of the rain cluster, the center coordinates of the peak rainfall intensity, and the intensity variation curve of the rainfall intensity along the core movement trajectory of the rain cluster are coded into a dynamic evolution cloud map of the rainfall intensity.

[0007] Furthermore, by fusing the continuous monitoring values ​​of water levels at multiple cross sections with the digital elevation model of the underlying surface of the watershed, a unit confluence and delay surface of the watershed where the target water conservancy project is located is generated, including: Differential calculations are performed on the continuous water level monitoring values ​​of the multi-section sections to obtain the water level gradient between each hydrological section. Extract river channel longitudinal slope, river network density, ground slope and surface cover type codes from the pre-set digital elevation model of the watershed underlying surface; Based on the water surface gradient, river longitudinal slope, river network density, ground slope and surface cover type code, the roughness coefficient and confluence velocity reduction coefficient corresponding to each spatial grid cell are calculated by calling the hydraulic relationship function. The parameters of the distributed motion wave confluence model are assigned using the roughness coefficient and the confluence velocity reduction coefficient, and the theoretical runoff propagation time of each spatial grid cell is calculated. By interpolating the theoretical runoff propagation time of all spatial grid cells within the entire watershed using spatial surfaces, a unit runoff delay surface is constructed that characterizes the degree of obstruction and delay caused by underlying surface conditions during the runoff collection process.

[0008] Furthermore, based on the spatial distribution values ​​of the soil moisture saturation, the antecedent rainfall is calculated to predict the runoff generation coefficient and runoff initiation time of different zones within the watershed, including: Using the spatial distribution value of soil moisture saturation collected by each edge sensing node as input, and combined with the soil field water holding capacity parameter corresponding to the edge sensing node location, the anterior impact rainfall in the control area of ​​the edge sensing node is calculated in reverse. Spatial kriging interpolation is performed on the aforementioned antecedent rainfall to generate a spatial distribution map of antecedent rainfall covering the entire watershed; Based on the spatial distribution map of the preceding impact rainfall, the initial loss and infiltration capacity of each zone are calculated using the watershed runoff generation model, thereby determining the runoff generation coefficient of each zone; Based on the unit confluence delay surface and the rainfall intensity dynamic evolution cloud map, the time it takes for the runoff to converge from the farthest point of the watershed to the outlet section after the rainfall generates runoff on the watershed underlying surface is calculated. Combined with the runoff generation coefficient to calculate the runoff generation start time, the confluence start time of each zone is determined. The calculated flow generation coefficient for each partition is assigned a coupling identifier to the flow initiation time.

[0009] Furthermore, the runoff generation coefficient and the confluence start time are mapped to the risky channels to generate a preliminary flood evolution scenario set with soil water condition constraints, including: Establish a spatial topology network for the risky channel, and associate the runoff generation coefficient marked with a coupling identifier with the runoff initiation time to the corresponding watershed partition; For each of the aforementioned risky channels, the superposition calculation of multi-source runoff in the channel is performed based on the runoff generation coefficients of all upstream associated zones and the confluence start time. Based on the front-end rainfall input provided by the dynamic evolution cloud map of rainfall intensity and the confluence delay effect provided by the unit confluence delay surface, the flow process line and water level process line of the flood wave during its evolution in the channel are calculated. The calculated flow process line and water level process line of each channel, together with the information on the proportion of runoff contributed by its upstream zone, are packaged into a flood evolution scenario. The flood evolution scenarios corresponding to all the aforementioned channels with risks are summarized to form the preliminary flood evolution scenario set with soil water conditions constraints.

[0010] Furthermore, the method also includes the step of simulating and iteratively optimizing the reservoir regulation and storage of the preliminary flood evolution scenario set with soil water condition constraints: From the preliminary flood evolution scenario set with soil water conditions constraints, the flood evolution scenario with the highest threat level to downstream target water conservancy projects is selected as the primary analysis object. Based on the scheduling rules and flood control capacity of the target water conservancy project, the boundary conditions for reservoir scheduling calculation are set; The hydrodynamic-based flood evolution simulator is invoked to calculate the water level and flow rate process of the primary analysis object as it evolves to the front of the reservoir dam under the reservoir scheduling calculation boundary conditions; The calculated upstream water level and flow rate process lines are compared with the reservoir's flood control scheduling curve to assess the impact of floods on the reservoir's flood control safety. Based on the assessment results, scheduling operation instructions such as reservoir pre-discharge and flood control are generated, and the impact of the scheduling operation instructions on the downstream river flood process is fed back to the flood evolution simulator. The overall flood evolution process of the basin after the execution of the scheduling operation command is recalculated using the flood evolution simulator, the preliminary flood evolution scenario set with soil water condition constraints is updated, and an optimized flood control scheduling plan is generated.

[0011] Furthermore, the invocation of a hydrodynamic-based flood evolution simulator to calculate the water level and flow rate process of the primary analysis object evolving to the reservoir dam under the reservoir scheduling calculation boundary conditions includes: The channel hydrological cross section, river topography, initial flow and water level conditions, and the unit confluence delay surface defined in the primary analysis object are used as inputs to the hydrodynamic model. Based on the boundary conditions for reservoir scheduling calculation, the initial water level and outflow control rules of the reservoir are determined; The Saint-Venant equation solver was used to simulate the evolution of flood waves in the river network from upstream to in front of the reservoir dam, and to calculate the water level and flow rate at each calculation time step and at each calculation section. Real-time monitoring of the moment when the flood wave front reaches the reservoir dam, and recording the process of the water level in front of the dam starting to rise; The water level and flow rate changes at the dam front section during the entire simulation period are extracted from the simulation results to form the water level and flow rate process lines in front of the dam.

[0012] Furthermore, the method also includes a dynamic flood forecast correction step based on real-time feedback data from edge nodes: During the flood's evolution, the system receives real-time measured water level and flow data reported by each edge sensing node. The measured water level data is compared with the predicted water level data at the corresponding time and location in the preliminary flood evolution scenario set with soil water conditions constraints, and the water level prediction error sequence is calculated. The water level forecast error sequence is processed using an adaptive filtering algorithm to generate a dynamic correction field for water level forecasts. The water level forecast dynamic correction field is assimilated into the flood evolution model to drive the model to perform real-time rolling forecast calculations; Based on the results of rolling forecast calculations, the future time period portion of the preliminary flood evolution scenario set with soil water condition constraints is dynamically updated to generate dynamically corrected flood forecast results.

[0013] Furthermore, the step of processing the water level forecast error sequence using an adaptive filtering algorithm to generate a dynamic correction field for the water level forecast includes: For each edge sensing node location, statistically analyze its water level forecast error sequence over the most recent period; Calculate the average error value and error covariance of the water level forecast error sequence, and construct a function describing the spatiotemporal correlation structure of the forecast error; Based on the average error value, error covariance, and the relevant structure function, the error estimate of the area around the edge sensing node where no monitoring station is set up is calculated. The measured errors of each node are spatially fused with the error estimates of the surrounding area to generate a spatially continuously distributed error correction field. The error correction field is superimposed onto the original flood evolution model forecast field to generate the water level forecast dynamic correction field.

[0014] Furthermore, the present invention also includes a flood pre-data processing system for water conservancy projects based on edge computing. The system includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of the flood pre-data processing method for water conservancy projects based on edge computing as described above.

[0015] Compared with the prior art, the beneficial effects of the present invention are: By integrating continuous water level monitoring values ​​from multiple cross sections with the digital elevation model of the watershed's underlying surface, a unit confluence delay surface is generated to reflect the obstruction and delay effects of different geographic units on the surface runoff collection process. This enables the synergistic coupling of water level monitoring data and topographic spatial data, fully characterizing the combined effects of topographic conditions and real-time water levels on runoff transmission within the watershed. It accurately quantifies the resistance characteristics and delay effects of each geographic unit in the runoff collection process, refines the spatial differences in the watershed's confluence process, corrects numerical deviations caused by relying solely on single topographic or water level data to extrapolate the confluence process, fully restores the true transport patterns of runoff within the watershed, and improves the detail and accuracy of the confluence process representation.

[0016] Based on the spatial distribution of soil moisture saturation, preliminary rainfall impact calculations are performed to determine the runoff generation coefficient and confluence start time of different zones within the watershed. These parameters are then mapped to risk channels that have undergone safety checks for river network connectivity and levee integrity, generating a preliminary flood evolution scenario set with soil water condition constraints. By leveraging spatial differences in soil moisture, the regionalized and accurate calculation of runoff generation and confluence parameters can be achieved. Invalid paths in the flood evolution channels are eliminated through safety checks, improving the accuracy of risk channel identification. Real-time soil water storage status is integrated into the flood evolution scenario construction process, eliminating scenario biases caused by using uniform fixed parameters. This ensures that the flood evolution scenario matches the actual runoff generation and confluence conditions in different areas of the watershed, strengthening the fit between flood evolution simulation and real-time water conditions and soil moisture in the watershed, refining the spatiotemporal distribution characteristics of flood evolution, and improving the precision and rationality of flood four-prevention data processing. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the steps of the edge computing-based flood pre-data processing method for water conservancy projects described in this invention. Figure 2 A flowchart for generating cloud maps showing the dynamic evolution of rainfall intensity; Figure 3 This is a flowchart for generating the unit bus delay surface. Detailed Implementation

[0018] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. 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 skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] See Figure 1 This invention provides a method for processing flood forecasting data for water conservancy projects based on edge computing. The specific method includes: In the upstream basin of the target water conservancy project, multiple edge sensing nodes are deployed to synchronously collect raw basin information, including real-time rainfall intensity values ​​across the entire basin, continuous monitoring values ​​of water levels at multiple cross-sections, and spatial distribution values ​​of soil moisture saturation. The synchronously collected real-time rainfall intensity values ​​across the entire basin are analyzed for spatiotemporal correlation to generate a dynamic evolution cloud map of rainfall intensity in the basin where the target water conservancy project is located. By fusing the continuous monitoring values ​​of water levels at multiple cross-sections with the digital elevation model of the basin's underlying surface, a unit confluence delay surface is generated for the basin where the target water conservancy project is located. This unit confluence delay surface reflects the obstruction and delay effects of different geographical units on the surface runoff collection process. Based on the spatiotemporal coupling relationship between the dynamic evolution cloud map of rainfall intensity and the unit confluence delay surface, a set of potential flood risk evolution channels is determined. The safety of the potential evolution channel set is then checked for river network connectivity and levee integrity, and channels with potential risks are screened out. Based on the spatial distribution of soil moisture saturation, the early impact rainfall is calculated, the runoff generation coefficient and runoff initiation time of different zones in the watershed are predicted, and the runoff generation coefficient and runoff initiation time are mapped to channels with risks, generating a preliminary flood evolution scenario set with soil water conditions constraints.

[0020] In one embodiment of the present invention, see [reference] Figure 2The system performs unified alignment processing on the spatial location and acquisition timestamps of real-time rainfall intensity values ​​from all edge sensing nodes, forming a spatial distribution sequence of rainfall intensity under the same spatiotemporal reference. Rain cluster tracking is performed on multiple consecutive time-series spatial distribution sequences of rainfall intensity to extract continuous rain cluster movement paths in the spatiotemporal dimension, identifying the core movement trajectory of the rain cluster. Local peak points of rainfall intensity values ​​and divergence points along the core movement trajectory are detected, and the geographical coordinates and timestamps of the local peak points and divergence points are recorded to form the coordinates of the rainfall intensity peak center. Spatial average rainfall intensity is calculated at fixed time step intervals along the core movement trajectory of the rain cluster, and the calculated spatial average rainfall intensity is connected in chronological order to form an intensity variation curve of rainfall intensity along the core movement trajectory of the rain cluster. The core movement trajectory of the rain cluster, the coordinates of the rainfall intensity peak center, and the intensity variation curve of rainfall intensity along the core movement trajectory of the rain cluster are fused and encoded into a dynamic evolution cloud map of rainfall intensity. This cloud map includes the core movement trajectory of the rain cluster, the coordinates of the rainfall intensity peak center, and the intensity variation curve of rainfall intensity along the core movement trajectory of the rain cluster.

[0021] In practical implementation, the process of generating a dynamic evolution cloud map of rainfall intensity in the flood pre-monitoring data processing method for water conservancy projects based on edge computing involves a specific watershed application scenario. More than one hundred edge sensing nodes are deployed within the watershed to synchronously collect real-time rainfall intensity values. In a specific example scenario, the edge sensing nodes upload real-time rainfall intensity values ​​with precise geographic coordinates and Coordinated Universal Time (UTC) timestamps to the central processing unit at fixed time intervals of five minutes. The unit of the real-time rainfall intensity values ​​is millimeters per hour. In practice, the real-time rainfall intensity values ​​from all edge sensing nodes undergo unified alignment processing based on spatial location and collection timestamps. The processing, based on a preset watershed spatial grid system and standard time reference, distributes the real-time rainfall intensity values ​​uploaded by discrete nodes to each grid using a spatial interpolation algorithm, arranging them chronologically to form a spatial distribution sequence of rainfall intensity under the same spatiotemporal reference. In practice, rain cluster tracking is performed on a series of consecutive time-series rainfall intensity spatial distribution sequences. This process is based on image recognition and spatiotemporal correlation algorithms. In the continuous rainfall intensity spatial distribution sequence, the core region of rainfall intensity that is spatially continuous, exceeds a preset threshold, and moves over time is identified. The movement path of the rain cluster that is continuous in the spatiotemporal dimension is extracted, and the core movement trajectory of the rain cluster is identified.

[0022] It is understandable that, along the movement trajectory of the rain cluster core, it is necessary to detect local peak points of rainfall intensity and divergence points of the movement path itself. This detection process is accomplished by traversing the real-time rainfall intensity values ​​of all grid points along the rain cluster core's movement trajectory and calculating the rate of change of the movement path direction. Local peak points are grid points whose real-time rainfall intensity values ​​are greater than those of the eight adjacent grid points. Divergence points are inflection points where the direction of the rain cluster core's movement trajectory changes significantly. Recording the geographical coordinates and timestamps of the local peak points and divergence points constitutes the coordinates of the rainfall intensity peak center. In some embodiments, along the rain cluster core's movement trajectory, the spatial average rainfall intensity is calculated at fixed time step intervals of ten minutes. The calculation range for the spatial average rainfall intensity is a circular area with a radius of five kilometers centered on the current position on the rain cluster core's movement trajectory. Connecting the calculated spatial average rainfall intensities in chronological order forms a curve showing the intensity change along the rain cluster core's movement trajectory. In practice, the core movement trajectory of the rain cluster, the center coordinates of the peak rainfall intensity, and the intensity change curve of the rainfall intensity along the core movement trajectory of the rain cluster are integrated and encoded into a dynamic evolution cloud map of rainfall intensity. The encoding format adopts a multi-dimensional data array that includes a time layer, a spatial coordinate layer, and an attribute value layer. The core movement trajectory of the rain cluster is represented by a series of ordered point coordinate sequences, the center coordinates of the peak rainfall intensity are recorded by the point layer and its attribute table, and the intensity change curve of the rainfall intensity along the core movement trajectory of the rain cluster is recorded by a correspondence table between time and intensity values.

[0023] In a data comparison example, the raw data before processing is a discrete list of real-time rainfall intensity values ​​organized by nodes. Each entry contains a node number, latitude and longitude, timestamp, and real-time rainfall intensity value. After processing to generate a dynamic evolution cloud map of rainfall intensity, the data is organized into a continuous spatiotemporal field, which can intuitively display the movement process of the rain cluster core from upstream to downstream. For example, the trajectory shows movement from the southwest to the northeast of the basin. At the same time, the intensity change curve of rainfall along the movement trajectory of the rain cluster core shows that the peak rainfall intensity occurs in the middle of the movement trajectory, and the center coordinates of the peak rainfall intensity record the specific geographical location and time of the peak. In specific implementation, the final dynamic evolution cloud map of rainfall intensity contains three core types of information: the movement trajectory of the rain cluster core, the center coordinates of the peak rainfall intensity, and the intensity change curve of rainfall along the movement trajectory of the rain cluster core. These are used for subsequent spatiotemporal coupling analysis with the unit confluence delay surface.

[0024] In one embodiment of the present invention, see [reference] Figure 3Differential calculations were performed on continuous water level monitoring values ​​from multiple sections to obtain the water surface gradient between each hydrological section. River longitudinal slope, river network density, ground slope, and surface cover type codes were extracted from a pre-built digital elevation model of the watershed underlying surface. Based on the water surface gradient, river longitudinal slope, river network density, ground slope, and surface cover type codes, hydraulic relational functions were used to calculate the roughness coefficient and runoff velocity reduction coefficient corresponding to each spatial grid cell. The roughness coefficient and runoff velocity reduction coefficient were used to assign parameters to the distributed motion wave runoff model, and the theoretical runoff propagation time of each spatial grid cell was calculated. Spatial surface interpolation was performed on the theoretical runoff propagation time of all spatial grid cells throughout the entire watershed to construct a unit runoff delay surface characterizing the degree of obstruction and delay caused by underlying surface conditions during the runoff collection process.

[0025] In practical implementation, the process of generating unit confluence delay surfaces in the flood pre-monitoring data processing method for water conservancy projects based on edge computing involves a specific example of a mountainous watershed. Thirty hydrological monitoring stations along the main river channel and major tributaries are deployed as edge sensing nodes, continuously uploading multi-section water level monitoring values. In the specific implementation, differential calculation is performed on the multi-section water level monitoring values. The differential calculation selects the water level values ​​of two adjacent hydrological sections at the same time, and divides the water level difference by the river distance between the two sections to obtain the water surface gradient between each hydrological section. In the specific implementation, river longitudinal slope, river network density, ground slope, and land cover type codes are extracted from a pre-set digital elevation model of the watershed underlying surface. The spatial resolution of the digital elevation model of the watershed underlying surface is ten meters. The river longitudinal slope is calculated from the river system vector layer. The river network density is obtained by statistically analyzing the river length per unit area. The ground slope is generated through calculation using the digital elevation model. The land cover type codes are derived from a remote sensing interpretation land classification map of the same phase. In practice, based on the water surface gradient, river longitudinal slope, river network density, ground slope, and surface cover type code, the roughness coefficient and confluence velocity reduction coefficient corresponding to each spatial grid cell are calculated using a hydraulic relationship function. This hydraulic relationship function is an empirical function that integrates the principles of Manning's formula with the delaying effect of landscape ecology. A specific expression of the hydraulic relationship function can be understood as follows:

[0026] in: Show the first The velocity reduction factor for each spatial grid cell. This represents the Manning roughness coefficient obtained by looking up the surface cover type code in a table. This represents the combined slope value of the water surface gradient and the longitudinal slope of the river channel at this spatial grid cell. This represents the river network density value of the sub-basin where the spatial grid cell is located. This represents the ground slope angle of the spatial grid cell. This represents the landscape impedance coefficient associated with land cover type. These are the weighting parameters calibrated based on watershed characteristics. In practical implementation, the roughness coefficient and the runoff velocity reduction coefficient are used to assign parameters to the distributed kinematic wave runoff model. The distributed kinematic wave runoff model uses each spatial grid cell as a computational unit and simulates the flow of runoff between grids based on the kinematic wave equation. By substituting the roughness coefficient and the runoff velocity reduction coefficient, the theoretical runoff propagation time of each spatial grid cell is calculated. The theoretical runoff propagation time refers to the time required for a unit volume of water to flow from that spatial grid cell to its downstream adjacent outlet.

[0027] In some embodiments, a data comparison example is as follows: the input data before processing includes a series of discrete hydrological cross-section locations and corresponding continuous monitoring values ​​of multi-section water levels, as well as a digital elevation model of the underlying surface of the watershed covering the entire area and its derived attribute layers. After the fusion calculation, the output data is a continuous surface covering the entire watershed, with each spatial grid unit having a theoretical runoff propagation time. In a specific implementation, the theoretical runoff propagation time of all spatial grid units within the entire watershed is interpolated using a spatial surface. The spatial surface interpolation uses the Kriging interpolation algorithm to smooth out abrupt changes in calculation results caused by parameter spatial heterogeneity, constructing a unit confluence delay surface that characterizes the degree of obstruction and delay of the runoff collection process by underlying surface conditions. In the data display of the unit confluence delay surface, areas with longer theoretical runoff propagation times are darker in color, indicating greater confluence resistance and significant delay effects, while areas with shorter theoretical runoff propagation times are lighter in color, indicating rapid confluence. In some embodiments, the generated unit confluence delay surface can visually show that the theoretical runoff propagation time in mountainous forest-covered areas is significantly longer than that in urban construction areas, and the theoretical runoff propagation time in densely river-networked areas is shorter than that in sparsely river-networked areas. This surface can be directly used for subsequent spatiotemporal coupling analysis with the dynamic evolution cloud map of rainfall intensity.

[0028] In one embodiment of the present invention, the spatial distribution value of soil moisture saturation collected by each edge sensing node is used as input, and combined with the soil field water holding capacity parameter corresponding to the edge sensing node location, the antecedent rainfall in the control area of ​​the edge sensing node is calculated. Spatial kriging interpolation is performed on the antecedent rainfall to generate a spatial distribution map of antecedent rainfall covering the entire watershed. Based on the spatial distribution map of antecedent rainfall, the initial loss and infiltration capacity of each zone are calculated based on the watershed runoff generation model, thereby determining the runoff generation coefficient of each zone. Based on the unit confluence delay surface and the dynamic evolution cloud map of rainfall intensity, the time it takes for runoff to converge from the farthest point of the watershed to the outlet section after the rainfall generates runoff on the watershed underlying surface is estimated. Combined with the runoff start time calculated by the runoff generation coefficient, the confluence start time of each zone is determined. A coupling identifier is assigned to the runoff generation coefficient and the confluence start time calculated for each zone. A spatial topology network of channels with risks is established, and the runoff generation coefficient and the confluence start time marked with the coupling identifier are associated with the corresponding watershed zone. For each at-risk channel, multi-source runoff is superimposed and calculated based on the runoff generation coefficients and confluence start times of all upstream associated zones. Using the front-end rainfall input provided by the dynamic evolution cloud map of rainfall intensity and the confluence delay effect provided by the unit confluence delay surface, the flow rate and water level process curves during the evolution of the flood wave in the channel are calculated. The calculated flow rate and water level process curves for each channel, along with the runoff generation proportion information contributed by its upstream zones, are packaged into a flood evolution scenario. The flood evolution scenarios corresponding to all at-risk channels are summarized to form a preliminary flood evolution scenario set with soil water condition constraints.

[0029] In practical implementation, the process of calculating runoff parameters and generating a preliminary flood evolution scenario set based on soil moisture in the edge computing-based flood pre-monitoring data processing method for water conservancy projects involves an example scenario encompassing five sub-basin zones. In this implementation, the spatial distribution value of soil moisture saturation collected by each edge sensing node is used as input, expressed as a percentage. Combined with the soil field capacity parameter corresponding to the edge sensing node location (obtained from the watershed soil type database), the anterior impact rainfall in the control area of ​​the edge sensing node is calculated. This calculation is based on the principle of soil water balance, assuming that the difference between the current soil moisture saturation and field capacity reflects the loss from previous rainfall. In the practical implementation, spatial kriging interpolation is performed on the anterior impact rainfall to generate a spatial distribution map of the anterior impact rainfall covering the entire watershed. The spatial kriging interpolation considers topographic elevation and slope aspect as auxiliary variables. In practice, based on the spatial distribution map of rainfall impacts from previous periods, the initial losses and infiltration capacity of each zone are calculated using a watershed runoff generation model. The watershed runoff generation model employs an improved form of the SCS curve number method to determine the runoff generation coefficient for each zone. The runoff generation coefficient is defined as the ratio of net rainfall to total rainfall. The calculation can be expressed as:

[0030] in: Indicates the first The flow generation coefficient of each zone, Indicates falling to the first Total rainfall in each zone Indicates the first The initial impact rainfall (i.e., initial loss) of each zone. Indicates the first The maximum possible retention capacity of soil in each zone is determined by soil type and previous soil moisture conditions.

[0031] In practice, based on the unit runoff delay surface and the dynamic evolution cloud map of rainfall intensity, the time it takes for runoff to converge from the farthest point of the watershed to the outlet section after rainfall generates runoff on the watershed underlying surface is calculated. Combined with the runoff generation coefficient calculated to determine the runoff initiation time, the runoff initiation time for each zone is determined. In practice, a coupling identifier is assigned to the calculated runoff generation coefficient and runoff initiation time for each zone. This coupling identifier is generated by combining the zone number and the calculation timestamp. Table 1 shows a data comparison, illustrating the changes in key parameters before and after processing. Before processing, the data consisted of discrete soil moisture monitoring values; after processing, the runoff generation coefficient and runoff initiation time for each zone were obtained with spatiotemporal significance.

[0032] Table 1: Calculation Table of Zonal Runoff Parameters Partition Number Average anterior rainfall (mm) Production flow coefficient Convergence start time (UTC) Coupling identifier A1 15.2 0.63 2023-07-1208:20 A1_202307120800 A2 8.7 0.78 2023-07-1208:05 A2_202307120800 B1 22.1 0.45 2023-07-1208:35 B1_202307120800 B2 12.5 0.69 2023-07-1208:15 B2_202307120800 In practical implementation, a spatial topology network of channels at risk is established. This network consists of channel segments, nodes, and their contribution relationships with upstream sub-regions. Runoff generation coefficients marked with coupling identifiers are associated with the corresponding watershed sub-regions based on their confluence start times. In some embodiments, for each channel at risk, a multi-source runoff superposition calculation is performed based on the runoff generation coefficients and confluence start times of all upstream sub-regions. This superposition calculation employs a time-shifted linear superposition method, aligning the unit hydrograph processes generated by each sub-region according to their confluence start times before superposition. In practical implementation, based on the front-end rainfall input provided by the dynamic evolution cloud map of rainfall intensity and the confluence delay effect provided by the unit confluence delay surface, the flow process line and water level process line during the evolution of the flood wave in the channel are calculated. A one-dimensional hydrodynamic model is used for the calculation, and the theoretical runoff propagation time provided by the unit confluence delay surface is used to determine the propagation delay of the inflow at the upper boundary. In practice, the calculated flow and water level biomass of each channel, along with the runoff generation ratio information from its upstream sub-regions, are packaged into a flood evolution scenario. The data structure of this scenario includes channel identifiers, time series, flow series, water level series, and a list of contribution ratios from each sub-region. Essentially, summarizing the flood evolution scenarios corresponding to all channels at risk constitutes a preliminary flood evolution scenario set with soil hydrological constraints. This set is a list containing multiple flood evolution scenario objects. In a data comparison example, before processing, there was scattered soil moisture data, sub-regional parameters, and isolated channel topology information. After processing, a set of flood evolution scenarios, organized by channel, containing complete hydrological processes and upstream contribution sources, is formed. Each scenario is explicitly associated with runoff generation and confluence initiation conditions under soil hydrological constraints. Optionally, when packaging flood evolution scenarios, the flow and water level biomass are standardized and encoded to facilitate direct reading and use in subsequent scheduling simulations and dynamic correction steps.

[0033] In one embodiment of the present invention, the flood evolution scenario with the highest threat level to the downstream target water conservancy project is selected as the primary analysis object from the preliminary flood evolution scenario set with soil water condition constraints. Based on the scheduling rules and flood control capacity of the target water conservancy project, reservoir scheduling calculation boundary conditions are set. A hydrodynamic-based flood evolution simulator is invoked to calculate the water level and flow process of the primary analysis object evolving to the reservoir dam under the reservoir scheduling calculation boundary conditions. The calculated water level and flow process lines in front of the dam are compared with the reservoir's flood control scheduling curve to assess the impact of the flood on the reservoir's flood control safety. Based on the assessment results, scheduling operation instructions such as reservoir pre-discharge and flood interception are generated, and the impact of the scheduling operation instructions on the downstream river flood process is fed back to the flood evolution simulator. The overall flood evolution process of the basin after executing the scheduling operation instructions is recalculated using the flood evolution simulator, the preliminary flood evolution scenario set with soil water condition constraints is updated, and an optimized flood control scheduling plan is generated. The process involves using a hydrodynamic-based flood evolution simulator to calculate the water level and flow rate of the primary analysis object as it evolves towards the reservoir dam under reservoir scheduling calculation boundary conditions. This includes inputting the hydrological cross-sections, river topography, initial flow and water level conditions, and unit confluence delay surfaces defined in the primary analysis object into the hydrodynamic model. Based on the reservoir scheduling calculation boundary conditions, the initial water level and outflow control rules for the reservoir are determined. The Saint-Venant equations solver is used to simulate the evolution of the flood wave in the river network from upstream to the reservoir dam, calculating the water level and flow rate values ​​at each calculation time step and at each calculation cross-section. The moment the flood wave front reaches the reservoir dam is monitored in real time, recording the process of the water level beginning to rise in front of the dam. The water level and flow rate change sequences of the cross-sections in front of the dam throughout the entire simulation period are extracted from the simulation results to form the water level and flow rate process lines in front of the dam.

[0034] In practical implementation, the process of reservoir regulation simulation and iterative optimization of the preliminary flood evolution scenario set in the edge computing-based flood pre-monitoring data processing method for water conservancy projects involves a target water conservancy project example called "Qingshuihe Reservoir," whose upstream basin has three channels identified as having risks. In the specific implementation, from the preliminary flood evolution scenario set with soil water condition constraints, the peak flow, flood volume, and peak occurrence time at the reservoir section are calculated based on each scenario. The threat levels are then ranked according to the reservoir's flood control design standards, and the flood evolution scenario with the highest threat level to the downstream target water conservancy project is selected as the primary analysis object. In the specific implementation, based on the target water conservancy project's scheduling rules and flood control capacity, reservoir scheduling calculation boundary conditions are set. These boundary conditions include the initial regulation water level, flood control limit water level, the discharge flow rules corresponding to each level of flood control capacity, and the reservoir's discharge capacity curve.

[0035] In some embodiments, a hydrodynamic-based flood evolution simulator is invoked to calculate the water level and flow process of the primary analysis object evolving to the reservoir dam under reservoir scheduling calculation boundary conditions. In specific implementations, the channel hydrological cross-section, river topography, initial flow and water level conditions, and unit confluence delay surfaces defined in the primary analysis object are used as inputs to the hydrodynamic model. In specific implementations, based on the reservoir scheduling calculation boundary conditions, the initial water level and outflow control rules of the reservoir are determined. The initial water level is the flood control limit water level, and the outflow control rules are set in stages according to the inflow magnitude. The rules can be expressed as follows:

[0036] in: This represents the outflow rate of the reservoir at time t. This indicates the maximum discharge capacity of the reservoir at the current water level. This represents the inflow rate at time t-Δ, where Δ is the time delay of the floodwaters reaching the dam front. This represents the water level in front of the dam at time t. Indicates the flood control limit water level. and These coefficients are determined according to reservoir operation rules. In practice, the Saint-Venant equation solver is used to simulate the evolution of flood waves in the river network from upstream to the reservoir dam, calculating the water level and flow rate at each calculation time step and at each cross-section. The calculation time step is one minute. In practice, the moment the flood wave front arrives at the reservoir dam is monitored in real time, and the process of the water level in front of the dam starting to rise is recorded. In practice, the water level change sequence and flow rate change sequence at the cross-section in front of the dam are extracted from the simulation results throughout the entire simulation period, forming the water level and flow rate process lines in front of the dam. Refer to Table 2, which shows a comparison segment between the water level and flow rate process lines in front of the dam and the reservoir flood control operation curve in a single simulation calculation.

[0037] Table 2: Comparison of Water Level-Flow Rate Processes and Flood Control Scheduling Curves in Front of the Dam Simulated time Calculate the water level in front of the dam (m). Calculate the inflow rate (m³ / s) Flood control dispatch curve limit water level (m) Scheduling status 08:00 175.0 500 175.5 Safety 10:30 176.8 3200 177.0 Approaching the limit 12:15 178.5 5800 178.0 Exceeding limits In practice, the calculated upstream water level and flow rate hydrographs are compared with the reservoir's flood control scheduling curve. The flood control scheduling curve defines the highest permissible water level for flood control safety at different times. The impact of floods on the reservoir's flood control safety is assessed, focusing primarily on whether the upstream water level exceeds the flood control scheduling curve and the magnitude and duration of any exceedance. Based on the assessment results, scheduling instructions for reservoir pre-discharge and flood control are generated. For example, a pre-discharge instruction to release water in advance before the flood arrives to lower the initial regulating water level, or a flood control instruction to close some gates during the flood peak to reduce the outflow. The impact of these scheduling instructions on the downstream river flood process is fed back to the flood evolution simulator. This feedback is achieved by modifying the outflow process in the model's boundary conditions. It is understandable that the overall flood evolution process of the basin after the execution of the dispatch operation instructions is recalculated using a flood evolution simulator. During the recalculation, the outflow process after the execution of the dispatch operation instructions is used as the downstream boundary condition of the reservoir. The changes in water level and flow along the river and in the reservoir are recalculated, and the initial flood evolution scenario set with soil water condition constraints is updated to generate an optimized flood control dispatch plan. In some embodiments, the optimized flood control dispatch plan not only includes the updated flood evolution scenario, but also includes a detailed sequence of dispatch instructions and their expected effects for dealing with the scenario, forming a complete decision support scheme. Optionally, the iterative optimization process can be performed multiple times. By comparing the results under different dispatch strategies, the optimal dispatch plan that achieves a balance between ensuring reservoir safety and reducing downstream flood control pressure is finally selected. In a data comparison example, the scenario set before optimization shows that the flood evolution will cause the water level in front of the dam to exceed the limit by 1.2 meters. The scenario set after optimization shows that after the execution of pre-discharge and phased flood interception dispatch, the water level in front of the dam is controlled within the limit, but the flood process line of some downstream sections has changed.

[0038] In one embodiment of the present invention, during the flood evolution process, measured water level data and flow data reported in real time by each edge sensing node are received. The measured water level data are compared with the predicted water level data at the corresponding time and location in the preliminary flood evolution scenario set with soil water conditions constraints, and a water level prediction error sequence is calculated. An adaptive filtering algorithm is used to process the water level prediction error sequence to generate a dynamic correction field for water level prediction. The dynamic correction field for water level prediction is assimilated into the flood evolution model to drive the model to perform real-time rolling prediction calculations. Based on the results of the rolling prediction calculations, the future time period portion of the preliminary flood evolution scenario set with soil water conditions constraints is dynamically updated to generate dynamically corrected flood prediction results. Specifically, processing the water level prediction error sequence using an adaptive filtering algorithm to generate the dynamic correction field for water level prediction includes: statistically analyzing the water level prediction error sequence for each edge sensing node location over a recent period; calculating the average error value and error covariance of the water level prediction error sequence; and constructing a function describing the spatiotemporal correlation structure of the prediction error. Based on the average error value, error covariance, and related structure function, the error estimate for the area around the edge sensing node without monitoring stations is calculated. The measured errors of each node are spatially fused with the error estimates of the surrounding area to generate a spatially continuous error correction field. This error correction field is then superimposed onto the original flood evolution model forecast field to generate a dynamic correction field for water level forecasting.

[0039] In practical implementation, the dynamic correction step of flood forecasting based on real-time feedback data from edge nodes in the flood forecasting data processing method for water conservancy projects based on edge computing involves a specific example of a flood evolution process. Twenty edge sensing nodes are deployed within the basin, continuously reporting measured data. During the flood evolution process, the measured water level and flow data reported in real-time by each edge sensing node are received. The reporting frequency of measured water level and flow data is once every ten minutes, and the data format includes node number, timestamp, water level value, and flow value. In practical implementation, the measured water level data is compared with the predicted water level data at the corresponding time and location in the preliminary flood evolution scenario set with soil water condition constraints. The comparison is completed through temporal interpolation and spatial matching, extracting the predicted water level value with the same time and geographical coordinates as the measured data from the forecast scenario. A water level forecast error sequence is then calculated; the water level forecast error is the difference sequence obtained by subtracting the predicted water level value from the measured water level value. In practice, an adaptive filtering algorithm is used to process the water level forecast error sequence and generate a dynamic correction field for the water level forecast. The adaptive filtering algorithm adopts the Kalman filter framework, but it is extended to the correction of the spatial field.

[0040] The process of using an adaptive filtering algorithm to process the water level forecast error sequence and generate a dynamic correction field for water level forecasts can be understood as follows: For each edge sensing node location, the water level forecast error sequence over a recent period is statistically analyzed. The recent period typically refers to the past six hours, containing thirty-six consecutive data points. The average error value and error covariance of the water level forecast error sequence are calculated, and a function describing the spatiotemporal correlation structure of the forecast error is constructed. This correlation structure function uses a Gaussian variogram model to characterize the characteristic that the spatial correlation of the error decays with distance. In practice, based on the average error value, error covariance, and correlation structure function, the error estimate for the area around the edge sensing node without monitoring stations is calculated. The calculation process is based on the optimal interpolation principle to minimize the variance of the estimated error. Finally, the measured errors of each node are spatially fused with the error estimates of the surrounding area. Spatial fusion is achieved by solving a system of linear equations based on the error covariance and correlation structure function, generating a spatially continuously distributed error correction field. In practice, the error correction value field is superimposed onto the original flood evolution model forecast field, thus generating a dynamic correction field for water level forecasts. This superposition directly adds the forecast water level value at each spatial grid point to the corresponding error correction value. The key calculation steps for generating the dynamic correction field for water level forecasts can be described as follows:

[0041] in: Represents the spatial coordinates at time t. Correction value for water level forecast error at the location, This represents the total number of edge-sensing nodes. This indicates that at time t, the first... The measured water level prediction error of each edge sensing node Indicates the first The error of each node affects the target position. Corrected weighting coefficients, weighting coefficients It is determined by solving the Kriging equations that take into account the spatiotemporal correlation structure of the error.

[0042] In some embodiments, before processing, at a certain forecast time, the forecast water level at the edge sensing node S05 is 125.3 meters, while the measured water level received at that time is 126.1 meters, resulting in a +0.8-meter water level forecast error at node S05. After processing by the adaptive filtering algorithm, not only is a correction value of approximately +0.8 meters given for the location of node S05, but also an error estimate correction value of +0.5 meters is given for a location 5 kilometers upstream of node S05 without a monitoring station, thereby generating a spatially continuously distributed dynamic correction field for water level forecasts. In specific implementation, the dynamic correction field for water level forecasts is assimilated into the flood evolution model. The assimilation process is accomplished by directly adding the correction field to the initial field of the model at the next forecast time step, driving the model to perform real-time rolling forecast calculations. The lead time for the rolling forecast is still 24 hours, but the model is re-initialized and a forecast is performed every time a new batch of measured data is received. In practice, based on the results of rolling forecast calculations, the future time period portion of the preliminary flood evolution scenario set constrained by soil water conditions is dynamically updated. This update replaces the water level and flow process data from the current moment to the next 24 hours in the original scenario set, generating a dynamically corrected flood forecast. Optionally, the dynamically corrected flood forecast not only includes the updated scenario set but also records the timestamps of the measured data used for each assimilation correction and the generated correction field file, forming a complete record of forecast correction origination. In a complete flood process example, the initial forecast showed a peak water level of 127.5 meters. After multiple rounds of dynamic correction based on real-time data from edge nodes, the final corrected forecast updated the peak water level to 126.8 meters, with a more accurate peak time.

[0043] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A method for processing flood forecasting data for water conservancy projects based on edge computing, characterized in that: include: In the upstream basin of the target water conservancy project, multiple edge sensing nodes are deployed to synchronously collect raw basin information, which includes real-time rainfall intensity values ​​across the entire basin, continuous monitoring values ​​of water levels at multiple cross sections, and spatial distribution values ​​of soil moisture saturation. The spatiotemporal correlation analysis of the real-time rainfall intensity values ​​collected synchronously across the entire watershed is performed to generate a dynamic evolution cloud map of rainfall intensity in the watershed where the target water conservancy project is located. By integrating the continuous monitoring values ​​of water levels at multiple cross sections with the digital elevation model of the underlying surface of the watershed, a unit confluence delay surface of the watershed where the target water conservancy project is located is generated. The unit confluence delay surface is used to reflect the obstruction and delay effect of different geographical units on the surface runoff collection process. Based on the spatiotemporal coupling relationship between the dynamic evolution cloud map of rainfall intensity and the unit confluence delay surface, a set of potential evolution channels for flood risk is determined, and the river network connectivity and dike integrity of the set of potential evolution channels are checked to screen out channels with risks. Based on the spatial distribution value of soil moisture saturation, the early impact rainfall is calculated, the runoff generation coefficient and runoff initiation time of different zones in the watershed are predicted, and the runoff generation coefficient and runoff initiation time are mapped to the channels with risks to generate a preliminary flood evolution scenario set with soil water conditions constraints.

2. The flood pre-monitoring data processing method for water conservancy projects based on edge computing according to claim 1, characterized in that, The step of performing spatiotemporal correlation analysis on the synchronously collected real-time rainfall intensity values ​​across the entire watershed to generate a dynamic evolution cloud map of rainfall intensity in the watershed where the target water conservancy project is located includes: The dynamic evolution cloud map of rainfall intensity includes the movement trajectory of the rain cluster core, the center coordinates of the rainfall intensity peak, and the intensity change curve of rainfall intensity along the movement trajectory of the rain cluster core. The real-time rainfall intensity values ​​from all edge sensing nodes are uniformly aligned with their spatial location and acquisition timestamp to form a spatial distribution sequence of rainfall intensity under the same spatiotemporal reference. Rain cluster tracking processing is performed on the spatial distribution sequence of rainfall intensity over multiple consecutive time periods to extract the continuous movement path of the rain cluster in the spatiotemporal dimension and identify the core movement trajectory of the rain cluster. On the core movement trajectory of the rain cluster, detect the local peak points of rainfall intensity and the divergence points of the movement path itself, and record the geographical coordinates and timestamps of the local peak points and divergence points to form the center coordinates of the rainfall intensity peak. Along the movement trajectory of the rain cluster core, the spatial average rainfall intensity is calculated at fixed time step intervals, and the calculated spatial average rainfall intensity is connected in chronological order to form the intensity variation curve of the rainfall intensity along the movement trajectory of the rain cluster core. The core movement trajectory of the rain cluster, the center coordinates of the peak rainfall intensity, and the intensity variation curve of the rainfall intensity along the core movement trajectory of the rain cluster are coded into a dynamic evolution cloud map of the rainfall intensity.

3. The flood forecasting data processing method for water conservancy projects based on edge computing according to claim 1, characterized in that, By fusing the continuous water level monitoring values ​​from multiple cross-sections with the digital elevation model of the underlying surface of the watershed, a unit confluence and delay surface is generated for the watershed where the target water conservancy project is located, including: Differential calculations are performed on the continuous water level monitoring values ​​of the multi-section sections to obtain the water level gradient between each hydrological section. Extract river channel longitudinal slope, river network density, ground slope and surface cover type codes from the pre-set digital elevation model of the watershed underlying surface; Based on the water surface gradient, river longitudinal slope, river network density, ground slope and surface cover type code, the roughness coefficient and confluence velocity reduction coefficient corresponding to each spatial grid cell are calculated by calling the hydraulic relationship function. The parameters of the distributed motion wave confluence model are assigned using the roughness coefficient and the confluence velocity reduction coefficient, and the theoretical runoff propagation time of each spatial grid cell is calculated. By interpolating the theoretical runoff propagation time of all spatial grid cells within the entire watershed using spatial surfaces, a unit runoff delay surface is constructed that characterizes the degree of obstruction and delay caused by underlying surface conditions during the runoff collection process.

4. The flood pre-monitoring data processing method for water conservancy projects based on edge computing according to claim 1, characterized in that, Based on the spatial distribution values ​​of soil moisture saturation, the antecedent rainfall is calculated to predict the runoff generation coefficient and runoff initiation time of different zones within the watershed, including: Using the spatial distribution value of soil moisture saturation collected by each edge sensing node as input, and combined with the soil field water holding capacity parameter corresponding to the edge sensing node location, the anterior impact rainfall in the control area of ​​the edge sensing node is calculated in reverse. Spatial kriging interpolation is performed on the aforementioned antecedent rainfall to generate a spatial distribution map of antecedent rainfall covering the entire watershed; Based on the spatial distribution map of the preceding impact rainfall, the initial loss and infiltration capacity of each zone are calculated using the watershed runoff generation model, thereby determining the runoff generation coefficient of each zone; Based on the unit confluence delay surface and the rainfall intensity dynamic evolution cloud map, the time it takes for the runoff to converge from the farthest point of the watershed to the outlet section after the rainfall generates runoff on the watershed underlying surface is calculated. Combined with the runoff generation coefficient to calculate the runoff generation start time, the confluence start time of each zone is determined. The calculated flow generation coefficient for each partition is assigned a coupling identifier to the flow initiation time.

5. The flood forecasting data processing method for water conservancy projects based on edge computing according to claim 1, characterized in that, Mapping the runoff generation coefficient and the confluence start time to the at-risk channel generates a preliminary flood evolution scenario set with soil water condition constraints, including: Establish a spatial topology network for the risky channel, and associate the runoff generation coefficient marked with a coupling identifier with the runoff initiation time to the corresponding watershed partition; For each of the aforementioned risky channels, the superposition calculation of multi-source runoff in the channel is performed based on the runoff generation coefficients of all upstream associated zones and the confluence start time. Based on the front-end rainfall input provided by the dynamic evolution cloud map of rainfall intensity and the confluence delay effect provided by the unit confluence delay surface, the flow process line and water level process line of the flood wave during its evolution in the channel are calculated. The calculated flow process line and water level process line of each channel, together with the information on the proportion of runoff contributed by its upstream zone, are packaged into a flood evolution scenario. The flood evolution scenarios corresponding to all the aforementioned channels with risks are summarized to form the preliminary flood evolution scenario set with soil water conditions constraints.

6. The flood pre-monitoring data processing method for water conservancy projects based on edge computing according to claim 5, characterized in that, The method further includes the steps of reservoir regulation simulation and iterative optimization of the preliminary flood evolution scenario set with soil water condition constraints: From the preliminary flood evolution scenario set with soil water conditions constraints, the flood evolution scenario with the highest threat level to downstream target water conservancy projects is selected as the primary analysis object. Based on the scheduling rules and flood control capacity of the target water conservancy project, the boundary conditions for reservoir scheduling calculation are set; The hydrodynamic-based flood evolution simulator is invoked to calculate the water level and flow rate process of the primary analysis object as it evolves to the front of the reservoir dam under the reservoir scheduling calculation boundary conditions; The calculated upstream water level and flow rate process lines are compared with the reservoir's flood control scheduling curve to assess the impact of floods on the reservoir's flood control safety. Based on the assessment results, scheduling operation instructions such as reservoir pre-discharge and flood control are generated, and the impact of the scheduling operation instructions on the downstream river flood process is fed back to the flood evolution simulator. The overall flood evolution process of the basin after the execution of the scheduling operation command is recalculated using the flood evolution simulator, the preliminary flood evolution scenario set with soil water condition constraints is updated, and an optimized flood control scheduling plan is generated.

7. The flood pre-monitoring data processing method for water conservancy projects based on edge computing according to claim 6, characterized in that, The invocation of a hydrodynamic-based flood evolution simulator to calculate the water level and flow rate process of the primary analysis object evolving to the reservoir dam under the reservoir scheduling calculation boundary conditions includes: The channel hydrological cross section, river topography, initial flow and water level conditions, and the unit confluence delay surface defined in the primary analysis object are used as inputs to the hydrodynamic model. Based on the boundary conditions for reservoir scheduling calculation, the initial water level and outflow control rules of the reservoir are determined; The Saint-Venant equation solver was used to simulate the evolution of flood waves in the river network from upstream to in front of the reservoir dam, and to calculate the water level and flow rate at each calculation time step and at each calculation section. Real-time monitoring of the moment when the flood wave front reaches the reservoir dam, and recording the process of the water level in front of the dam starting to rise; The water level and flow rate changes at the dam front section during the entire simulation period are extracted from the simulation results to form the water level and flow rate process lines in front of the dam.

8. The method for processing flood forecasting data for water conservancy projects based on edge computing according to claim 1, characterized in that, The method also includes a dynamic correction step for flood forecasting based on real-time feedback data from edge nodes: During the flood's evolution, the system receives real-time measured water level and flow data reported by each edge sensing node. The measured water level data is compared with the predicted water level data at the corresponding time and location in the preliminary flood evolution scenario set with soil water conditions constraints, and the water level prediction error sequence is calculated. The water level forecast error sequence is processed using an adaptive filtering algorithm to generate a dynamic correction field for water level forecasts. The water level forecast dynamic correction field is assimilated into the flood evolution model to drive the model to perform real-time rolling forecast calculations; Based on the results of rolling forecast calculations, the future time period portion of the preliminary flood evolution scenario set with soil water condition constraints is dynamically updated to generate dynamically corrected flood forecast results.

9. The flood pre-monitoring data processing method for water conservancy projects based on edge computing according to claim 8, characterized in that, The step of processing the water level forecast error sequence using an adaptive filtering algorithm to generate a dynamic correction field for the water level forecast includes: For each edge sensing node location, statistically analyze its water level forecast error sequence over the most recent period; Calculate the average error value and error covariance of the water level forecast error sequence, and construct a function describing the spatiotemporal correlation structure of the forecast error; Based on the average error value, error covariance, and the relevant structure function, the error estimate of the area around the edge sensing node where no monitoring station is set up is calculated. The measured errors of each node are spatially fused with the error estimates of the surrounding area to generate a spatially continuously distributed error correction field. The error correction field is superimposed onto the original flood evolution model forecast field to generate the water level forecast dynamic correction field.

10. A flood pre-monitoring data processing system for water conservancy projects based on edge computing, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the edge computing-based flood pre-data processing method for water conservancy projects as described in any one of claims 1 to 9.

Images