Power distribution facility risk management and control scheduling method and system based on urban flood forecasting

Abstract

The present application relates to the technical field of power system disaster prevention and reduction, and particularly relates to a power distribution facility risk management and control scheduling method and system based on urban flood forecasting. The method comprises the following steps: collecting multi-source static basic data of a target area, and constructing a flood risk ledger of power distribution facilities; integrating urban flood disaster-causing factors, coupling to form a flood forecasting mechanism model, and outputting a risk level; constructing a deep learning model, rolling forecasting short-term flood scenarios, and generating risk prediction results; obtaining Internet of Things monitoring data, identifying the water accumulation state of power distribution facilities, and forming facility monitoring results; using the risk prediction results and facility monitoring results to construct a risk list, and designing a scheduling scheme based on the risk list; mapping the scheduling scheme into scheduling instructions, and uploading to the management and control platform. The present application can effectively improve the disaster prevention and reduction level of power distribution facilities under extreme climate conditions, and provide technical support for the safe operation of power grids in typical tidal river network areas.

Description

Technical Field

[0001] This invention relates to the field of disaster prevention and mitigation technology in power systems, and in particular to a method and system for risk management and scheduling of power distribution facilities based on urban flood forecasting. Background Technology

[0002] For power distribution facilities located in tidal river network areas, their operating environment is complex. They are affected by the tide levels of external rivers and river channels, as well as by the limited capacity of stormwater runoff and drainage networks. Once flooded, they are highly susceptible to large-scale power outages, affecting urban public safety and socio-economic stability. Current technologies mainly rely on single-factor analysis or empirical formulas for risk assessment, which often fails to accurately reflect the flood risk of power distribution facilities in complex urban environments. In terms of emergency response, there are also common problems such as insufficient utilization of monitoring data, imprecise dispatch of emergency response forces, and separation of early warning and dispatch links, resulting in delayed prevention and control measures that fail to meet the requirements for real-time and precise control. Summary of the Invention

[0003] Therefore, it is necessary for the present invention to provide a method and system for risk management and scheduling of power distribution facilities based on urban flood forecasting, in order to solve at least one of the above-mentioned technical problems.

[0004] To achieve the above objectives, a risk management and dispatching method for power distribution facilities based on urban flood forecasting is provided. This method is applied to a management platform and includes the following steps:

[0005] Step S1: Collect multi-source static basic data of the target area and construct a flood risk ledger for power distribution facilities;

[0006] Step S2: Integrate urban flood-causing factors, couple them to form a flood forecasting mechanism model, and output the risk level;

[0007] Step S3: Based on the flood risk ledger and risk level, construct a deep learning model to make rolling predictions of short-term flood scenarios and generate risk prediction results;

[0008] Step S4: Obtain IoT monitoring data, identify the water accumulation status of power distribution facilities, and form facility monitoring results; construct a risk list using risk prediction results and facility monitoring results, design a dispatching scheme based on the risk list; map the dispatching scheme into dispatching instructions and upload them to the management and control platform.

[0009] Preferably, the present invention also provides a power distribution facility risk management and dispatching system based on urban flood forecasting, used to execute the above-described power distribution facility risk management and dispatching method based on urban flood forecasting, wherein the power distribution facility risk management and dispatching system based on urban flood forecasting includes:

[0010] The basic data filing module is used to collect multi-source static basic data of the target area and build a flood risk ledger for power distribution facilities;

[0011] The mechanism model evaluation module is used to integrate urban flood-causing factors, couple them to form a flood forecasting mechanism model, and output the risk level.

[0012] The intelligent prediction and analysis module is used to build a deep learning model based on the flood risk ledger and risk level, to make rolling predictions of short-term flood scenarios and generate risk prediction results.

[0013] The monitoring, early warning, and dispatching module is used to acquire IoT monitoring data, identify the water accumulation status of power distribution facilities, and generate facility monitoring results; it constructs a risk list using risk prediction results and facility monitoring results, designs dispatching schemes based on the risk list, maps the dispatching schemes into dispatching instructions, and uploads them to the management and control platform.

[0014] This invention establishes a flood risk ledger for power distribution facilities through surveys and GIS analysis, thus systematizing risk identification. It constructs a multi-factor coupled flood forecasting mechanism model and classifies risk levels based on the distribution of power grid facilities. By introducing deep learning methods and fusing them with the mechanism model, it develops a high-precision, rapid-response flood forecasting technology. Based on the Internet of Things and optimization algorithms, it enables intelligent scheduling of real-time monitoring and emergency drainage efforts. Finally, by developing an integrated flood forecasting and early warning system, it achieves integrated operation of flood forecasting, real-time monitoring and early warning, and emergency dispatch functions. Attached Figure Description

[0015] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0016] Figure 1 This is a flowchart illustrating the steps of a power distribution facility risk management and scheduling method based on urban flood forecasting according to the present invention.

[0017] Figure 2 This is a schematic diagram illustrating the method and system construction process for risk management and emergency dispatch of power distribution facilities based on urban flood forecasting.

[0018] Figure 3 Schematic diagram of the coupling method of storm surge-river-surface-pipeline-power grid model;

[0019] Figure 4 A diagram illustrating the refined allocation of resources for emergency response;

[0020] Figure 5 This is a schematic diagram of a flood forecasting and early warning system for power distribution facilities. Detailed Implementation

[0021] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0022] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0023] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0024] To achieve the above objectives, please refer to Figures 1 to 5 This invention provides a method for risk management and scheduling of power distribution facilities based on urban flood forecasting. The method is applied to a management platform and includes the following steps:

[0025] Step S1: Collect multi-source static basic data of the target area and construct a flood risk ledger for power distribution facilities;

[0026] Step S2: Integrate urban flood-causing factors, couple them to form a flood forecasting mechanism model, and output the risk level;

[0027] Step S3: Based on the flood risk ledger and risk level, construct a deep learning model to make rolling predictions of short-term flood scenarios and generate risk prediction results;

[0028] Step S4: Obtain IoT monitoring data, identify the water accumulation status of power distribution facilities, and form facility monitoring results; construct a risk list using risk prediction results and facility monitoring results, design a dispatching scheme based on the risk list; map the dispatching scheme into dispatching instructions and upload them to the management and control platform.

[0029] Of particular importance, step S1 includes:

[0030] Collect basic geographic information of the target area, including topography, administrative divisions, transportation network, land use and embankment location;

[0031] Integrate rainfall records, storm surge characteristics, river levels, flood flows, and historical rainstorm events in the target area to form hydrological and meteorological data;

[0032] Collect historical flood disaster records for the target area, identify historical inundation points, waterlogging areas, disaster losses, and typical cases, and generate disaster data;

[0033] Obtain information on the distribution of drainage pipe networks, the location of power distribution facilities and related main networks, measure the elevation of characteristic points, and compile the information into facility monitoring data;

[0034] A flood risk ledger for power distribution facilities in the target area is constructed using basic geographic information, hydrological and meteorological data, disaster data, and facility monitoring data.

[0035] In the first embodiment, a high-precision UAV aerial survey system is deployed within the target area. Combined with aerial imagery with a resolution better than 0.1m and lidar mapping results, topographic data is extracted. An administrative boundary layer is integrated using a geographic information platform, and traffic network data is accessed, including national highways, provincial highways, urban arterial roads, and secondary roads. Land use data is obtained through remote sensing image interpretation, with an accuracy controlled at a scale of 1:10000. The location of dikes is entered using surveying results from the water resources department, with data format using shapefile and coordinate system WGS84. All data forms basic geographic information, and a spatial mapping relationship is subsequently established with hydrological and meteorological data.

[0036] The system retrieves real-time rainfall data from meteorological monitoring stations with a time resolution limited to 5 minutes; it also retrieves storm surge characteristic data, including tide level, wind speed, and wind direction, from oceanographic tide gauges; it retrieves river water levels from hydrological stations with an accuracy controlled within ±1 cm; it uses flow meters to monitor flood flow with a sampling interval set to 10 minutes; and it compiles rainfall curves from historical rainstorm events, covering a time span of nearly 30 years. All data is stored in CSV format to form hydrological and meteorological data.

[0037] The system accesses the city's flood control database to retrieve the latitude and longitude coordinates of historical flooding points; extracts polygonal vector data of historical waterlogging areas; reads information on losses from past disasters, including power outage duration, number of damaged facilities, and economic losses; and retrieves typical case texts from disaster prevention archives, using natural language processing tools to extract keyword fields to form disaster data.

[0038] The locations of power distribution facilities are marked point by point in the target area, with the coordinate accuracy limited to ±0.5m; the information associated with the main grid is imported from the power dispatch center database; the elevation of feature points is obtained around each power distribution facility using RTK measurement, with an error range not exceeding 3cm; water level monitoring sensors are deployed at the entrance of the power distribution room and in low-lying areas, with the sensor range controlled between 0-5m and the accuracy controlled within ±0.5cm; all measurement results are archived according to the facility number to form facility monitoring information.

[0039] By layering basic geographic information, hydrological and meteorological data, disaster data, and facility monitoring information, and establishing related tables through spatial indexing, a flood risk ledger is formed. The ledger includes: facility number, geographical location, surface elevation, facility base elevation, historical inundation frequency, disaster loss level, and drainage network connection status.

[0040] In the second embodiment, basic geographic information is acquired by satellite remote sensing with a resolution of 0.5m; hydrological and meteorological data are automatically accessed through the API interface of the municipal hydrological bureau and meteorological bureau; historical disaster data are directly accessed from the digital archives of the emergency management platform; facility monitoring information is entered on-site through the geographic information system terminal; the flood risk ledger is generated in a relational database, and the data table structure contains 10 fields, including facility number, facility type, location, surface elevation, base elevation, surrounding drainage capacity, historical flooding frequency, maximum water depth, loss amount classification, and main network code.

[0041] Preferably, step S2, which integrates urban flood-causing factors and couples them to form a flood forecasting mechanism model, includes:

[0042] The boundary conditions for the one-dimensional hydrodynamic model of the river network and the two-dimensional flooding model of the surface were designed using the Xin'an River model with three water sources and full runoff generation.

[0043] The two-dimensional surface flooding model is vertically coupled with the one-dimensional drainage model of the pipe network through rainwater wells;

[0044] The two-dimensional surface flooding model is connected to the one-dimensional hydrodynamic model of the river network by means of riverbank embankments and water level boundaries.

[0045] The one-dimensional drainage model of the pipe network is coupled with the one-dimensional hydrodynamic model of the river network, and the boundary conditions of the one-dimensional and two-dimensional hydrodynamic models of the river channel are designed using the two-dimensional hydrodynamic model of storm surge.

[0046] By integrating urban flood-causing factors, simulating the propagation path and impact range of urban floods, and coupling them to form a flood forecasting mechanism model, the urban flood-causing factors include rainfall, storm surge, and river floods.

[0047] In the first embodiment, a three-source water storage-saturation runoff calculation method is adopted, decomposing the rainfall-runoff process into three components: surface runoff, interflow, and groundwater runoff. The runoff ratio for each component is set by control parameters, with the surface runoff threshold controlled at 10 mm, the interflow threshold at 30 mm, and the groundwater runoff threshold at 50 mm. The runoff results are input into a one-dimensional hydrodynamic calculation framework for the river network to establish boundary conditions for flow and water level. Simultaneously, the surface runoff is input into a two-dimensional surface flooding calculation framework as the inflow boundary.

[0048] Point-based connection units are established at the locations of storm drains in urban areas. For each storm drain, the manhole cover elevation, pipe diameter, and well depth are recorded, with elevation accuracy controlled within ±2cm and pipe diameter measurement error not exceeding 1cm. Using the flow exchange equation for the storm drains, the water depth in the two-dimensional surface flooding calculation framework is converted into the inflow rate at the wellhead, which is then transferred to the one-dimensional drainage calculation framework of the pipe network. Simultaneously, the return water level of the pipe network is back-calculated to the wellhead, feeding back to the surface water depth.

[0049] Boundary nodes are established at the riverbank and levee locations. Riverbank elevations are collected using RTK measurements with an error not exceeding 5 cm. Water level boundary conditions are used to transfer water levels from the one-dimensional hydrodynamic calculation framework of the river network to adjacent cells of the two-dimensional flooding framework on the surface, thereby describing the process of flood overflowing the levees or inundating the city. Simultaneously, the outflow from the surface cells is fed back to the river network cells, forming a two-way exchange.

[0050] A connection is established at the outlet of the pipeline, and the cross-sectional shape of the outlet is confirmed through video inspection and pipeline mapping results. The outflow of the one-dimensional drainage calculation framework of the pipeline network is transmitted to the downstream nodes of the one-dimensional hydrodynamic framework of the river network, while the water level of the river network affects the discharge capacity of the pipeline network in reverse.

[0051] Two-dimensional hydrodynamic calculation results of storm surges are introduced at the estuary location. The tidal level curve is used as the downstream boundary condition of the one-dimensional hydrodynamic framework of the river network, and also as the boundary of the two-dimensional shallow water calculation framework. The storm surge input data includes measured tidal levels, wind speed, and wind direction, with a temporal resolution controlled at 10 minutes and a spatial resolution controlled at 500m.

[0052] This system integrates three flood-causing factors—rainfall, storm surge, and river flood—and progressively transfers water volume and level through boundary conditions and coupling relationships to achieve continuous calculation of flood propagation paths in urban areas. The final output shows the flood impact range, including water depth distribution, inundated area, and spatial overlay results with the location of power distribution facilities.

[0053] Of particular importance is that the risk level output in step S2 includes:

[0054] The flood inundation status of each power distribution facility in the target area was calculated using a flood forecasting mechanism model, based on the rainstorm return period threshold standard.

[0055] The risk level of each power distribution facility is determined based on the calculation results.

[0056] In the first embodiment, the return period threshold for heavy rainfall is set at five levels: once every 1 year, once every 2 years, once every 5 years, once every 10 years, and once every 20 years. The input rainfall data is processed in 5-minute intervals, with rainfall intensity limited to 20–150 mm / h. Using the flood forecasting mechanism calculation framework established in the previous step, the distribution of water depth under rainfall conditions in each time period is calculated.

[0057] At each power distribution facility location, the calculated water depth is extracted and compared with the facility base elevation. When the water depth is less than 0.1m below the base elevation, it is marked as unflooded; when the water depth exceeds the base elevation by 0.1m but is less than 0.5m, it is marked as slightly flooded; when the water depth is in the range of 0.5m–1.0m, it is marked as moderately flooded; and when the water depth exceeds 1.0m, it is marked as severely flooded.

[0058] Using the flooding status as input, a risk level field is set for each power distribution facility in the ledger table. Slight flooding corresponds to low risk, moderate flooding to medium risk, and severe flooding to high risk. The risk level output is a structured data table with fields including facility number, geographic coordinates, recurrence interval, flooding status, and risk level.

[0059] Preferably, the one-dimensional hydrodynamic model of the river network is as follows:

[0060] The one-dimensional hydrodynamic model of the river network uses Saint-Venant's equations as the governing equations for the unsteady flow of the river channel. The specific calculation method is as follows:

[0061] Continuity equation for water flow:

[0062] ;

[0063] Equations of water flow:

[0064] ;

[0065] in, For water level, For time, The width of the water surface in the cross-section. For traffic, For mileage, This represents the lateral unit width flow rate; positive values ​​indicate inflow, and negative values ​​indicate outflow. It is the acceleration due to gravity. The cross-sectional area of ​​the water passage. For correction factors, The roughness coefficient is Manning's coefficient. The radius is the hydraulic radius.

[0066] In the first embodiment, the river channel is scanned using a drone lidar to extract the longitudinal and cross-sectional geometric parameters. The cross-sectional spacing is 30m, and the elevation accuracy is controlled within ±0.05m.

[0067] The upstream boundary condition is the flow rate curve with a time resolution of 5 minutes; the downstream boundary condition is the tide level curve, derived from actual tide gauge measurements.

[0068] The equations are discretized using the finite difference method. In the continuity equation, the flow term... Through the central difference approximation; in the momentum equation, the nonlinear term Dispersed by windward differential.

[0069] Manning roughness coefficient Correction coefficient The calculation step size is 30 seconds to ensure numerical stability; the wetted perimeter length is calculated by the measured cross-sectional area integral method.

[0070] The output results are the water level curve and flow rate curve for the entire river channel.

[0071] In the second embodiment, the flow sequence calculated by the three-source water storage and runoff generation method is input as the river inflow condition with a resolution of 10 minutes.

[0072] Calculation cross-sections are set every 100m in the main channel and every 200m in the tributaries. The calculation parameters for each cross-section include the water surface width, the cross-sectional area of ​​the water passage, and the wetted perimeter.

[0073] Using a time-by-time step-by-step approach, the water level increment is first calculated based on the continuity equation, and then substituted into the momentum equation to correct the flow rate until the residual is less than 1e-4.

[0074] The Manning roughness coefficient ranges from 0.028 to 0.033; the correction factor is fixed at 1.0; the wetted perimeter is input from the sonar cross-section mapping data when calculating the hydraulic radius.

[0075] The system generates water level-time and flow-time curves for the flood wave propagating through the river channel, spatially overlapping these curves with the location of the power distribution facilities. By comparing the elevation of the facility's base with the calculated water level, it is determined whether the facilities are flooded.

[0076] Preferably, the two-dimensional surface flooding model is as follows:

[0077] The two-dimensional surface flooding model uses a conservation-form two-dimensional shallow water equation, and the specific calculation method is as follows:

[0078] ;

[0079] in, For a conserved vector, For time, , They are respectively , The convective flux vector in the direction. , They are respectively , Diffusion flux vector caused by directional Reynolds stress , They are respectively , The diffusion flux vector caused by the directional secondary flow. This is the source term vector.

[0080] In this invention, the conservation vector and the convective flux vector are defined as follows:

[0081] ;

[0082] ;

[0083] ;

[0084] in, Water depth, in meters (m). and They are respectively and The velocity of water flow in a certain direction, in m / s; Let be the acceleration due to gravity, and take . .

[0085] The diffusion flux induced by Reynolds stress and the diffusion flux of secondary flow are expressed in the following form (the expanded form is used for numerical processing):

[0086] , ;

[0087] , ;

[0088] The turbulent viscosity coefficient Given by the following formula:

[0089] ;

[0090] Parameter value restrictions: , , The bed shear velocity is estimated using the bed shear stress and updated within the time step. Secondary flow diffusion coefficient. Take interval To reflect the differences in the strength of the secondary flow.

[0091] Source term vector The components, including the bottom slope term, friction term, wind stress, geostrophic force, and external inflow term, are expressed as follows:

[0092] ;

[0093] , (Riverbed or ground elevation gradient); Friction gradient is expressed using the Manning formula: and Manning roughness Take interval Based on surface material settings; wind stress , Calculated from surface wind speed input; Geostrophic force components , If applicable calculate; External inflow (unit: m / s) includes rainwater well inflow and surface runoff injection.

[0094] The numerical solution will ultimately output the following data files or datasets:

[0095] Submersion depth field: on a two-dimensional grid The time series; used for comparison with the surface elevation and base elevation in the power distribution facility ledger;

[0096] Water surface elevation field: Used to determine situations of overflowing or breaching the dike and as input for the river network coupling boundary;

[0097] Velocity field: Used to estimate the impact risk of flow velocity on facilities and to determine the flow regime's passability;

[0098] Table of flooding duration and maximum depth: Summarized by facility number for risk level classification.

[0099] In the first embodiment, it is applicable to urban central areas, with a grid resolution of [missing information]. The grid is established using regular rectangles to control the volume. Ground elevation is provided by a lidar DEM, with vertical accuracy... .

[0100] Read the DEM and initial water depth field, and establish Upstream and boundary inflows are injected according to the runoff sequence; water surface boundaries are set along the riverbanks, with one-dimensional hydrodynamic input from the river network; tidal time series are set at the estuary; velocity at the boundary is calculated using extrapolation or reflection conditions; for each unit surface, the convection flux (i.e., numerical flux) is calculated using the flux splitting method approximating the Riemann problem, with flux expression primarily based on unidirectional upwind flux; this step is the explicit flux calculation process; , , , Using second-order central difference discretization, the wet volume multiplier is... Weighted average; bottom slope term is handled separately using well balancing method to ensure hydrostatic balance; friction term is updated with explicit damping within time step; wind stress and inflow... Incorporate into the source term; explicitly update the conserved vector. Among them, the residual term Summarize the flux differences and source terms. Time step Determined by the stability conditions of CFL, ,Pick For example, if , ,but .

[0101] Manning roughness Secondary flow diffusion coefficient The turbulent viscosity coefficient is updated progressively according to the aforementioned expression, and the bed shear velocity... Estimated from near-bed velocity.

[0102] Parallel control volume is implemented using multi-core parallelism and GPU acceleration to shorten computation time. The output flood depth field and time series are written to geographic data formats (raster GeoTIFF and facility-level CSV).

[0103] Interpolate the inundation depth field based on the facility coordinates to generate "Inundation Depth" and "Maximum Inundation Depth" fields, and input them into the risk table.

[0104] Preferably, the one-dimensional drainage model of the pipe network is as follows:

[0105] The one-dimensional drainage model of the pipe network utilizes the narrow-slit method to unify the governing equations for open-flow and full-flow systems. The specific calculation method is as follows:

[0106] ;

[0107] ;

[0108] in, It is a width matrix vector. For a conserved vector, For time variables, It is a flow matrix vector. Here is the longitudinal coordinate along the pipeline direction. For the source term vector, The width of the water surface, when the pipe is an open channel. , For the wave velocity of open channel flow, For gravitational acceleration, when the flow inside the pipe is pressure flow. , For pressure wave velocity, This refers to the water level or pressure head. For traffic, For the water flow area, For friction loss along the path, , The roughness coefficient is Manning's coefficient. The radius is the hydraulic radius.

[0109] In this invention, pressure wave velocity Engineering expressions:

[0110] ;

[0111] in, Let be the bulk modulus of water, and take . ; ; Pipe diameter; Young's modulus of pipe (approximately) PVC ); This represents the pipe wall thickness. Example values ​​for reference: steel pipe. , hour PVC pipe , hour .

[0112] In the first embodiment, the pipeline network is divided into units according to pipe segments. Time step .

[0113] Upstream inflow is loaded step-by-step according to the runoff generation sequence; downstream head returns according to the one-dimensional hydrodynamics of the river network.

[0114] At each time step, the inertial flux and linearized friction term are calculated first, and a tridiagonal system of equations is constructed and solved. .

[0115] The source term is updated after five iterations of wellhead overflow and pump station inner layer.

[0116] The output file includes segment levels. Time series and instantaneous wellhead overflow. Output format is CSV and time series database records, with an inner layer iteration limit of 20 times.

[0117] Preferably, the two-dimensional hydrodynamic model of storm surge is as follows:

[0118] The storm surge two-dimensional hydrodynamic model uses an improved form of the two-dimensional shallow water equation as the governing equation. The specific calculation method is as follows:

[0119] ;

[0120] in, For a conserved vector, For time, and They are respectively and The convective flux vector in the direction. and They are respectively and The diffusion flux vector in the direction. For source terms.

[0121] In this invention,

[0122] ;

[0123] ;

[0124] ;

[0125] ;

[0126] ;

[0127] ;

[0128] in, Because of the water depth, and They are respectively and The average velocity perpendicular to the direction of flow. It is the acceleration due to gravity. The elevation of the riverbed. The turbulent viscosity coefficient in the horizontal direction. This is the proportionality coefficient. Karman coefficient, The bed surface shear velocity, and They are respectively and The bottom slope in the direction is expressed as follows: and , and They are respectively and The friction gradient in the direction of wave-current interaction, without considering wave-current interaction, is expressed as follows: and , The Cobb force coefficient, The angular velocity of Earth's rotation. The latitude is the local latitude. For water surface wind stress, and The densities of air and water are respectively. The wind stress coefficient, The wind speed is 10 meters above the water surface. , , , This represents the wave radiation stress component.

[0129] In the first embodiment, it is applicable to high-precision storm surge coupling in urban nearshore coastal areas.

[0130] Grid resolution to Time step to (CFL); meteorological input resolution is 10-minute wind and pressure fields; spectral data is provided by nearshore buoys, with detailed frequency resolution; radiation stress is handled using a spectral integration process; pump stations and gates are coupled in a step-by-step iterative manner.

[0131] Output: Hourly flooding depth field, maximum flooding depth, velocity field, and flooding duration table; these data are interpolated to the coordinates of power distribution facilities for risk level determination and emergency drainage path constraints.

[0132] In the second embodiment, it is applicable to regional-scale storm surge scenario forecasting.

[0133] Grid to Time step to (According to CFL); wave radiation stress is estimated using an engineering approximation formula based on significant wave height and main direction; numerical propulsion employs semi-implicit processing, with diffusion and friction implicitly coupled.

[0134] Output: Short-term (0–48h) tidal field and inundation range along the coast, for coupling use with upstream river network and distribution network.

[0135] In the third embodiment, it is applicable to engineering simulation of wave-tidal-river coupling.

[0136] The two-dimensional storm surge field and the one-dimensional river network exchange water levels in real time at the boundary nodes, with a time synchronization step of 30 seconds; the wave field first calculates the directional spectrum by an independent wave field module and then inputs it into the radiation stress calculation process in the form of time slices; parallel domain decomposition and GPU acceleration are adopted to meet the requirements of high-resolution large-scale simulation.

[0137] The output is a flood time series table generated by facility number and written into the flood risk ledger.

[0138] Preferably, step S3, which involves constructing a deep learning model based on flood risk records and risk levels, includes the following steps:

[0139] Construct a multidimensional input feature matrix using flood risk ledgers and risk levels;

[0140] The multidimensional input feature matrix is ​​divided into a training set, a validation set, and a test set.

[0141] Construct a spatiotemporal convolutional neural network and set the parameters of the convolutional and pooling layers to form the model structure;

[0142] Train the model weights using the training set, and output the trained model.

[0143] Use the validation set to evaluate the performance of the trained model and generate evaluation results;

[0144] Based on the evaluation results, the hyperparameters are adjusted and an early stopping strategy is triggered to form a modified model;

[0145] Load the test set to test the modified model, and select the modified model with the best performance as the deep learning model for rolling prediction.

[0146] In the first embodiment, the spatial coordinates, base elevation, density of adjacent drainage networks, and surrounding low-lying terrain indicators of each power distribution facility are extracted from the flood risk ledger. The corresponding rainstorm recurrence interval classification, tide level classification, and river water level classification are extracted from the risk level results. These static attributes are then combined with time-series rainfall intensity, real-time water level sensor curves, and pump station operating status to form a multi-dimensional input feature matrix.

[0147] Historical data was divided into training set (70%), validation set (15%), and test set (15%) according to time sequence. The division followed the principle of segmenting entire time periods to avoid confusion between adjacent time points.

[0148] A spatiotemporal convolutional neural network is used, with the kernel size of the convolutional layer fixed at [value missing]. The number of convolutional layers is set to 3; the number of convolutional channels in each layer are 32, 64, and 128 respectively; max pooling is used in the pooling layers, and the window size is... The sampled data is downsampled to 1 / 8 of the original input; finally, a fully connected layer is connected to output the predicted flood probability distribution, with three classification dimensions: low risk, medium risk, and high risk.

[0149] The training batch size is 64; the initial learning rate is set to 0.001; weight updates use gradient backpropagation; and the number of training rounds is limited to no more than 200.

[0150] The network structure is input into the validation set and the validation error curve is output. When the validation error no longer decreases after 20 consecutive rounds, early stopping is triggered. The number of convolutional kernels and the learning rate are adjusted based on the validation error to form a corrected model.

[0151] Input the test set into the modified model and output the accuracy, recall, and F1 score; select the model as the rolling prediction model when the F1 score reaches or exceeds 0.9. The rolling prediction results are output in 5-minute increments.

[0152] In the second embodiment, the input is limited to rainfall intensity, facility elevation, pipeline discharge capacity, and real-time water level; the time series length is 30 minutes, and the step size is 1 minute. , characteristic number .

[0153] The division ratio is 60% for training, 20% for validation, and 20% for testing.

[0154] Two convolutional layers, kernel size The number of channels is 16 and 32 respectively; one pooling layer, window size Output binary classification: whether facility flooding has occurred.

[0155] Batch size 32; initial learning rate 0.002; maximum number of training epochs 100; accuracy is calculated for each epoch on the validation set, and early stopping is triggered if there is no improvement for 10 consecutive epochs; hyperparameters can only be adjusted for learning rate and pooling layer window size.

[0156] The test results output the facility-level flood probability; when the predicted probability exceeds 0.7, it is marked as high risk and added to the risk list.

[0157] Preferably, step S3 involves rolling forecasts of short-term flood scenarios to generate risk prediction results, including:

[0158] Collect real-time meteorological data, real-time hydrological data, and real-time monitoring data of power distribution facilities;

[0159] Real-time meteorological and hydrological data are input into a deep learning model, and simulation results are output.

[0160] The real-time monitoring data of power distribution facilities is compared with the simulation results, and the prediction bias of the deep learning model is corrected to form a correction result;

[0161] The latest real-time meteorological data, real-time hydrological data, and correction results are overlaid on the time series to generate rolling data;

[0162] The rolling data is input into the deep learning model to update the spatiotemporal sequence of the deep learning model, forming a rolling prediction model, which then outputs the risk prediction results.

[0163] In the first embodiment, meteorological data is collected at a time resolution of 10 minutes using regional automatic rain gauges (5km apart) and satellite rainfall inversion products; hydrological data is collected at a sampling frequency of 10 minutes using water level monitoring stations in main and tributary rivers; and facility data is collected at a sampling frequency of 2 minutes using water immersion probes in substations and power distribution rooms within the region.

[0164] Input the time series length of the deep learning model T=48 (one frame every 30 minutes within 24 hours); output the regional flood distribution and facility inundation level for the next 6 hours.

[0165] By comparing observed river water levels with simulated outputs, a residual regression method is used for correction. , As a correction factor, it corrects the simulated water level of the same watershed unit.

[0166] Add a new frame of meteorological and hydrological input, overlay the corrected sequence to form rolling data; re-input into the network to output risk predictions updated every 30 minutes.

[0167] The risk prediction results include the water level field, list of inundated facilities, and cumulative inundation duration within the watershed; the data is transmitted to the management and control platform for multi-regional coordinated scheduling.

[0168] Preferably, step S4 includes the following steps:

[0169] Step S41: Collect real-time water level values, flow rate signals and pump station start / stop status from IoT sensors and integrate them into sensor data;

[0170] Step S42: Receive video surveillance equipment and drone images, and use YOLO-V8 edge computing to identify the outline of water accumulation in the image in real time to form image data;

[0171] Step S43: Fuse sensor data and image data, and extract the water depth, flooding range and facility base height difference from the fusion results as facility monitoring results;

[0172] Step S44: Compare the risk prediction results with the facility monitoring results, mark the facility flooding level and treatment priority, and form a risk list;

[0173] Step S45: Construct a mathematical model for emergency response force scheduling based on the risk list, and output the scheduling plan;

[0174] Step S46: Analyze the dispatch plan to match vehicle locations, pump truck power, and emergency response personnel teams, generate dispatch instructions, and upload the dispatch instructions to the control platform.

[0175] In the first embodiment, water level sensors, flow velocity sensors, and pump station monitoring units are deployed around sunken overpasses, power distribution room entrances, and low-lying substations in the target urban area.

[0176] The water level sensor has a range of 0–5m and an accuracy of 1mm; the flow velocity sensor has a range of 0–5m / s and an accuracy of 0.01m / s; the pump station monitoring unit collects start-up and shutdown status data with a sampling period of 30 seconds.

[0177] The above data is aggregated into sensor data and uploaded to the data transfer module of the management and control platform via the 5G network.

[0178] Fixed video surveillance devices are deployed along main roads and around substations, while drones cruise at altitudes of 50–100 meters to collect images. YOLO-V8 edge computing units are deployed on the drones and front-end servers, with a convolutional kernel count set to 80 and a confidence threshold limited to 0.6. The system identifies water accumulation contours in real time and outputs coordinate boundaries and area values, forming image data.

[0179] By overlaying sensor data and image data using a GIS platform, and comparing water level sensor readings with image recognition to identify water accumulation boundaries, the water depth can be extracted. Flooding range and the height difference of the facility base .

[0180] Formula for calculating water depth: The flooding area was calculated using the area of ​​the polygon in the image; the base elevation difference was obtained by comparing the base elevation recorded in the facility risk ledger with the ground elevation difference. These results were then integrated to form the facility monitoring results.

[0181] The risk prediction results are retrieved and compared item by item with the facility monitoring results. If the difference between the predicted water depth and the measured water depth is greater than 0.1m, the measured data shall prevail. The facilities are marked as three levels of risk: low (<0.3m), medium (0.3–0.8m), and high (≥0.8m). Additional handling priorities are added according to the importance to the main power grid (Level I, Level II, Level III). The integrated results are written into the risk list.

[0182] Using the risk list as input, a mathematical model for emergency response force scheduling is constructed. The objective function is set to minimize the cumulative downtime of high-risk facilities. Constraints include:

[0183] Vehicle capacity constraints: ;in For the first Pump truck pumping flow rate The maximum total flow rate of the vehicle-mounted pump is 200L / s.

[0184] Personnel constraints: At least 2 people must be assigned to each facility, and the number of personnel shall not exceed the total number of available personnel.

[0185] Time constraints: The length of the dispatch route shall not exceed 30km, and the one-way travel time of the vehicle shall not exceed 1 hour.

[0186] The model solves and outputs a scheduling scheme, which includes a vehicle dispatch table, a pump truck allocation table, and a personnel grouping table.

[0187] The dispatch plan compares vehicle locations with facility locations and parses them into navigation paths; it compares pump truck power with facility water depth and allocates corresponding pumping times; and it matches personnel shifts according to skill levels and facility risk levels. This generates standardized JSON-formatted dispatch instructions.

[0188] The dispatch instructions are uploaded to the management and control platform via the MQTT protocol, written into the real-time task database, and pushed to the emergency vehicle terminal and pump station controller.

[0189] It is worth noting that similar sensors and drones can be deployed in multiple urban areas to generate multi-source sensor data and image data. The risk list is summarized by facility zone, and the scheduling scheme can incorporate cross-regional vehicle scheduling constraints. The output scheduling instructions not only include single-facility tasks but also cross-regional route planning, with a total distance limit of less than 100km and a response time of less than 2 hours.

[0190] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is not limited by the foregoing description. Thus, all changes falling within the meaning and scope of the equivalents of the application are intended to be included within the scope of the invention.

[0191] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for risk management and dispatching of power distribution facilities based on urban flood forecasting, characterized in that, The method is applied to a management and control platform, and the method includes the following steps: Step S1: Collect multi-source static basic data of the target area and construct a flood risk ledger for power distribution facilities; Step S1 includes: Collect basic geographic information of the target area, including topography, administrative divisions, transportation network, land use and location of dikes; Integrate rainfall records, storm surge characteristics, river levels, flood flows, and historical rainstorm events in the target area to form hydrological and meteorological data; Collect historical flood disaster records for the target area, identify historical inundation points, waterlogging areas, disaster losses, and typical cases, and generate disaster data; Obtain information on the distribution of drainage pipe networks, the location of power distribution facilities and related main networks, measure the elevation of characteristic points, and compile the information into facility monitoring data; A flood risk ledger for power distribution facilities in the target area is constructed using basic geographic information, hydrological and meteorological data, disaster data, and facility monitoring data. Step S2: Integrate urban flood-causing factors, couple them to form a flood forecasting mechanism model, and output the risk level; Step S2, integrating urban flood-causing factors and coupling them to form a flood forecasting mechanism model, includes: The boundary conditions for a one-dimensional hydrodynamic model of the river network and a two-dimensional surface flooding model were designed using the Xin'an River model with three water sources and full runoff generation. The one-dimensional hydrodynamic model of the river network is as follows: The one-dimensional hydrodynamic model of the river network uses Saint-Venant's equations as the governing equations for the unsteady flow of the river channel. The specific calculation method is as follows: Continuity equation for water flow: ； Equations of water flow: ； in, For water level, For time, The width of the water surface in the cross-section. For traffic, For mileage, This represents the lateral unit width flow rate; positive values ​​indicate inflow, and negative values ​​indicate outflow. It is the acceleration due to gravity. The cross-sectional area of ​​the water passage. For correction factors, The roughness coefficient is Manning's coefficient. The hydraulic radius; The specific two-dimensional surface flooding model is as follows: The two-dimensional surface flooding model uses a conservation-form two-dimensional shallow water equation, and the specific calculation method is as follows: ； in, For a conserved vector, , They are respectively , The convective flux vector in the direction. , They are respectively , Diffusion flux vector caused by directional Reynolds stress , They are respectively , The diffusion flux vector caused by the directional secondary flow. The source term vector; The two-dimensional surface flooding model is vertically coupled with the one-dimensional drainage model of the pipe network through rainwater wells; The one-dimensional drainage model of the pipe network is as follows: The one-dimensional drainage model of the pipe network utilizes the narrow-slit method to unify the governing equations for open-flow and full-flow systems. The specific calculation method is as follows: ； ； in, It is a width matrix vector. This is a flow matrix vector, when the pipe is an open channel. , For open channel flow wave velocity, when the flow inside the pipe is pressure flow. , For pressure wave velocity, This refers to the water level or pressure head. For friction loss along the path, ; The two-dimensional surface flooding model is connected to the one-dimensional hydrodynamic model of the river network by means of riverbank embankments and water level boundaries. The one-dimensional drainage model of the pipe network is coupled with the one-dimensional hydrodynamic model of the river network, and the boundary conditions of the one-dimensional and two-dimensional hydrodynamic models of the river channel are designed using the two-dimensional hydrodynamic model of storm surge. The two-dimensional hydrodynamic model of storm surge is as follows: The storm surge two-dimensional hydrodynamic model uses an improved form of the two-dimensional shallow water equation as the governing equation. The specific calculation method is as follows: ； By integrating urban flood-causing factors, simulating the propagation path and impact range of urban floods, and coupling them to form a flood forecasting mechanism model, the urban flood-causing factors include rainfall, storm surge and river floods; Step S3: Based on the flood risk ledger and risk level, construct a deep learning model to make rolling predictions of short-term flood scenarios and generate risk prediction results; Step S4: Obtain IoT monitoring data, identify the water accumulation status of power distribution facilities, and form facility monitoring results; construct a risk list using risk prediction results and facility monitoring results, design a dispatching scheme based on the risk list; map the dispatching scheme into dispatching instructions and upload them to the management and control platform.

2. The method for risk management and dispatching of power distribution facilities based on urban flood forecasting as described in claim 1, characterized in that, Step S3, which involves building a deep learning model based on flood risk records and risk levels, includes the following steps: Construct a multidimensional input feature matrix using flood risk ledgers and risk levels; The multidimensional input feature matrix is ​​divided into a training set, a validation set, and a test set. Construct a spatiotemporal convolutional neural network and set the parameters of the convolutional and pooling layers to form the model structure; Train the model weights using the training set, and output the trained model. Use the validation set to evaluate the performance of the trained model and generate evaluation results; Based on the evaluation results, the hyperparameters are adjusted and an early stopping strategy is triggered to form a modified model; Load the test set to test the modified model, and select the modified model with the best performance as the deep learning model for rolling prediction.

3. The method for risk management and dispatching of power distribution facilities based on urban flood forecasting as described in claim 1, characterized in that, Step S3 involves rolling forecasts of short-term flood scenarios, generating risk prediction results including: Collect real-time meteorological data, real-time hydrological data, and real-time monitoring data of power distribution facilities; Real-time meteorological and hydrological data are input into a deep learning model, and simulation results are output. The real-time monitoring data of power distribution facilities is compared with the simulation results, and the prediction bias of the deep learning model is corrected to form a correction result; The latest real-time meteorological data, real-time hydrological data, and correction results are overlaid on the time series to generate rolling data; The rolling data is input into the deep learning model to update the spatiotemporal sequence of the deep learning model, forming a rolling prediction model, which then outputs the risk prediction results.

4. The method for risk management and dispatching of power distribution facilities based on urban flood forecasting according to claim 1, characterized in that, Step S4 includes the following steps: Step S41: Collect real-time water level values, flow rate signals and pump station start / stop status from IoT sensors and integrate them into sensor data; Step S42: Receive video surveillance equipment and drone images, and use YOLO-V8 edge computing to identify the outline of water accumulation in the image in real time to form image data; Step S43: Fuse sensor data and image data, and extract the water depth, flooding range and facility base height difference from the fusion results as facility monitoring results; Step S44: Compare the risk prediction results with the facility monitoring results, mark the facility flooding level and treatment priority, and form a risk list; Step S45: Construct a mathematical model for emergency response force scheduling based on the risk list, and output the scheduling plan; Step S46: Analyze the dispatch plan to match vehicle locations, pump truck power, and emergency response personnel teams, generate dispatch instructions, and upload the dispatch instructions to the control platform.

5. A power distribution facility risk management and dispatching system based on urban flood forecasting, characterized in that, For executing the power distribution facility risk management and dispatching method based on urban flood forecasting as described in claim 1, the power distribution facility risk management and dispatching system based on urban flood forecasting includes: The basic data filing module is used to collect multi-source static basic data of the target area and build a flood risk ledger for power distribution facilities; The mechanism model evaluation module is used to integrate urban flood-causing factors, couple them to form a flood forecasting mechanism model, and output the risk level. The intelligent prediction and analysis module is used to build a deep learning model based on the flood risk ledger and risk level, to make rolling predictions of short-term flood scenarios and generate risk prediction results. The monitoring, early warning, and dispatching module is used to acquire IoT monitoring data, identify the water accumulation status of power distribution facilities, and generate facility monitoring results; it constructs a risk list using risk prediction results and facility monitoring results, designs dispatching schemes based on the risk list, maps the dispatching schemes into dispatching instructions, and uploads them to the management and control platform.

Images